A Case-Based Introduction to Modeling Occupational Inhalation Exposures to Chemicals 9781950286171

Citation preview

A Case-Based Introduction to

Modeling Occupational Inhalation Exposures to Chemicals A case-based approach that demonstrates how industrial hygienists can apply exposure modeling in daily practice. Edited by Chris Keil, PhD, CIH

A Case-Based Introduction to Modeling Occupational Inhalation Exposures to Chemicals

Published by AIHA® Falls Church, VA

Disclaimer This publication was developed by experts with background, training, and experience in various aspects of industrial hygiene (IH) and occupational and environmental health and safety (OEHS), working with information that was available at the time of publication. AIHA (as publisher) and the author(s) have been diligent in ensuring that the material and methods addressed in this publication reflect prevailing IH and OEHS practices. It is possible, however, that certain policies or procedures discussed will require modification because of changing federal, state, local, or international regulations. AIHA and the author(s) disclaim any liability, loss, or risk resulting directly or indirectly from use of the practices and/or theories presented in this publication. Moreover, it is the user’s responsibility to stay informed of any changing federal, state, local, or international regulations that might affect the material contained herein, as well as the policies adopted specifically in the user’s workplace. Specific mention of manufacturers and products in this book does not represent an endorsement by AIHA or the author(s).

Copyright © 2023 by AIHA All rights reserved. No part of this publication may be reproduced in any form or by any other means (graphic, electronic, or mechanical, including photocopying, taping, or information storage or retrieval systems) without written permission from the publisher. Book design by Jim Myers Editorial support provided by Lisa Lyubomirsky Stock Number: AEAB23-826 ISBN: 978-1-950286-17-1 AIHA 3120 Fairview Park Drive, Suite 360 Falls Church, VA 22042 Tel: (703) 849-8888 Fax: (703) 207-3561 Email: [email protected] aiha.org

Modeling Occupational Inhalation Exposures to Chemicals Chapter 1: Modeling in IH Practice

Table of Contents

FOREWORD.................................................................................................................................................................................................... ix References .................................................................................................................................................................................................. x PREFACE .......................................................................................................................................................................................................... xi ACKNOWLEDGMENTS ............................................................................................................................................................................. xiii CHAPTER 1: MODELING IN THE FRAMEWORK OF INDUSTRIAL HYGIENE PRACTICE ..................................................1 The Industrial Hygienist and Exposure Assessment....................................................................................................................1 Assessing Inhalation Exposures to Chemicals ...............................................................................................................................1 The Role of Modeling in Exposure Assessment .............................................................................................................................3 Physical-Chemical Concentration Modeling ...................................................................................................................................4 Current Applications of Air Concentration Modeling for Regulatory Purposes..................................................................5 How Close Is Close Enough? ................................................................................................................................................................6 Advancing the Use of Models in Exposure Assessment .............................................................................................................7 References ..................................................................................................................................................................................................7 CHAPTER 2: INFORMATION GATHERING FOR AIR CONCENTRATION MODELING .........................................................9 Information About the Chemical .........................................................................................................................................................9 Information About the Airspace ....................................................................................................................................................... 12 Worker Activity Information............................................................................................................................................................... 16 The Importance of the Information-Gathering Step ................................................................................................................. 17 References ............................................................................................................................................................................................... 17 CHAPTER 3: VOLATILE LIQUID SPILL ............................................................................................................................................... 21 Scenario .................................................................................................................................................................................................... 21 Information Gathering.......................................................................................................................................................................... 21 Modeling Approach 1: Zero-Ventilation Model ........................................................................................................................... 22 Interpreting the Modeled Concentration ....................................................................................................................................... 23 Modeling Approach 2: Well-Mixed Room Model ........................................................................................................................ 24 Conclusions .............................................................................................................................................................................................. 25 References ............................................................................................................................................................................................... 25

Copyright AIHA®. For personal use only. Do not distribute.

iii

Modeling Occupational Inhalation Exposures to Chemicals Chapter 1: Modeling in IH Practice CHAPTER 4: LOCAL EXHAUST VENTILATION FOR A SANDING OPERATION ................................................................ 27 Scenario .................................................................................................................................................................................................... 27 Information Gathering.......................................................................................................................................................................... 27 Modeling Approach: Well-Mixed Room ......................................................................................................................................... 30 Interpreting the Modeled Concentration ....................................................................................................................................... 30 Conclusions .............................................................................................................................................................................................. 32 References ............................................................................................................................................................................................... 32 CHAPTER 5: VOLATILE LIQUID CONTAINER LEFT OPEN ......................................................................................................... 33 Scenario .................................................................................................................................................................................................... 33 Information Gathering.......................................................................................................................................................................... 33 Modeling the Concentration in the Cabinet: Saturation Vapor Pressure Concentration .............................................. 35 Modeling Average Concentration in the Room: Zero-Ventilation Model............................................................................ 37 Modeling the Decrease of Concentration in the Room: Well-Mixed Room Purging Model ......................................... 38 Best Practice for Control of Fugitive Emissions From Volatile Liquid Containers in Laboratories ............................ 39 References ............................................................................................................................................................................................... 42 CHAPTER 6: BARREL FILLING INDOORS ........................................................................................................................................ 43 Scenario .................................................................................................................................................................................................... 43 Information Gathering.......................................................................................................................................................................... 43 Modeling Approach: Diffusion Model ............................................................................................................................................. 45 Interpreting the Modeled Concentrations ..................................................................................................................................... 47 Selecting Turbulent Diffusion Coefficients .................................................................................................................................... 49 References ............................................................................................................................................................................................... 50 CHAPTER 7: HISTORICAL EXPOSURES NEAR A TRICHLOROETHYLENE DEGREASER.............................................. 51 Scenario .................................................................................................................................................................................................... 51 Information Gathering.......................................................................................................................................................................... 51 Modeling Approach: Near-Field/Far-Field Model ....................................................................................................................... 53 Interpreting the Modeled Concentration ....................................................................................................................................... 55 References ............................................................................................................................................................................................... 58 CHAPTER 8: PRESSURIZED GAS RELEASE .................................................................................................................................... 59 Scenario .................................................................................................................................................................................................... 59 Information Gathering.......................................................................................................................................................................... 59 Modeling Approach 1: Zero-Ventilation Model ........................................................................................................................... 61 Interpreting the Modeled Concentration – 1 ................................................................................................................................ 62 Modeling Approach 2: Two-Box Model of a Slow Release..................................................................................................... 62 Interpreting the Modeled Concentration – 2 ................................................................................................................................ 63 References ............................................................................................................................................................................................... 64

iv

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 1: Modeling in IH Practice CHAPTER 9: VENTILATION REQUIREMENTS FOR A FLAVORING PROCESS ................................................................... 65 Scenario .................................................................................................................................................................................................... 65 Information Gathering.......................................................................................................................................................................... 65 References ............................................................................................................................................................................................... 71 Appendix .................................................................................................................................................................................................. 72 CHAPTER 10: CLEANING WITH PRODUCT CONTAINING ACETIC ACID ........................................................................... 75 Scenario .................................................................................................................................................................................................... 75 Information Gathering.......................................................................................................................................................................... 75 Modeling Approach 1: Well-Mixed Box With Rapid Evaporation of Acetic Acid ........................................................... 78 Modeling Approach 2: Time-Varying Well-Mixed Box Concentration With Repeated Mass Applications and Decreasing Emission Rates ....................................................................................................................................................... 79 Modeling Approach 3: Steady-State Well-Mixed Box Concentration With Repeated Mass Applications and Decreasing Emission Rates ....................................................................................................................................................... 82 Interpreting the Modeled Concentration ....................................................................................................................................... 83 References ............................................................................................................................................................................................... 83 CHAPTER 11: EXPERIMENTAL DETERMINATION OF MODEL PARAMETERS .................................................................. 85 Case 1: Dust Emissions From Mixer Loading ............................................................................................................................... 85 Case 2: Ventilation Rate in a Room Without Mechanical Ventilation ................................................................................. 87 References ............................................................................................................................................................................................... 89 CHAPTER 12: ORGANIC CHEMISTRY TEACHING LAB EXPOSURE AND A FIRST EXPLORATION OF MONTE CARLO MODELING.................................................................................................................................................................... 91 Scenario .................................................................................................................................................................................................... 91 Information Gathering.......................................................................................................................................................................... 91 Steady-State Well-Mixed Room Model ......................................................................................................................................... 92 Capturing Variability in Parameters: Best Case and Worst Case ........................................................................................ 93 Capturing Variability in Parameters: An Introduction to Monte Carlo Approaches ....................................................... 94 Applying a Monte Carlo Approach to the Chemistry Lab ....................................................................................................... 96 Using Excel to Do the Monte Carlo Simulation of the Scenario ............................................................................................ 97 References ............................................................................................................................................................................................... 98 CHAPTER 13: FILLING A CONTAINER WITH A MIXTURE AND USING IHMOD 2.0 FOR MODELING CALCULATIONS .......................................................................................................................................................................................... 99 Scenario .................................................................................................................................................................................................... 99 Using IHMOD 2.0 .................................................................................................................................................................................103 Conclusion ..............................................................................................................................................................................................105 References .............................................................................................................................................................................................105

Copyright AIHA®. For personal use only. Do not distribute.

v

Modeling Occupational Inhalation Exposures to Chemicals Chapter 1: Modeling in IH Practice CHAPTER 14: OXYGEN DEFICIENCY HAZARD IN A STORAGE TANK ...............................................................................107 Scenario ..................................................................................................................................................................................................107 Information Gathering........................................................................................................................................................................108 Modeling Approach: Near-Field/Far-Field Model .....................................................................................................................108 Interpreting the Modeled Outputs (Concentration and Partial Pressure) ........................................................................111 Conclusions ............................................................................................................................................................................................113 Appendix A: Oxygen Deficiency Hazard .....................................................................................................................................114 Appendix B: Mass Balance Equations .........................................................................................................................................115 References .............................................................................................................................................................................................117 CHAPTER 15: USE OF AEROSOL SOLVENTS IN AN AUTO SHOP ......................................................................................119 Scenario ..................................................................................................................................................................................................119 Information Gathering........................................................................................................................................................................119 Modeling Approach.............................................................................................................................................................................121 Interpreting the Modeled Concentrations ...................................................................................................................................126 An Additional Consideration for Evaporation Rates ...............................................................................................................126 References .............................................................................................................................................................................................127 CHAPTER 16: FURNITURE STRIPPING WITH METHYLENE CHLORIDE: INTRODUCING THE ADVANCED REACH TOOL (ART V1.5) ......................................................................................................................................................................129 Scenario ..................................................................................................................................................................................................129 Information Gathering........................................................................................................................................................................129 Modeling Approach: The Advanced Reach Tool ......................................................................................................................130 Using ART: A Walk-Through of Entering Inputs .......................................................................................................................132 Running the Art V1.5 Model and Interpreting the Results ....................................................................................................137 Appendix A: More About ART .........................................................................................................................................................139 Appendix B: Bayesian Module ........................................................................................................................................................140 Appendix C: Activity 1 Configuration Screens ..........................................................................................................................142 References .............................................................................................................................................................................................144 CHAPTER 17: RECONSTRUCTION OF A LABORATORY SPILL .............................................................................................145 Introduction ............................................................................................................................................................................................145 Background............................................................................................................................................................................................146 Initial Modeling Approaches ............................................................................................................................................................146 Probabilistic Modeling ........................................................................................................................................................................148 Summary of Modeling Results ........................................................................................................................................................154 Discussion ..............................................................................................................................................................................................154 Conclusion ..............................................................................................................................................................................................157 Acknowledgments ..............................................................................................................................................................................157 Appendix A: Estimating the Generation Rate ...........................................................................................................................158 Appendix B: Estimating the Near-Field Volume and Flow Rate (β)...................................................................................160 References .............................................................................................................................................................................................161

vi

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 1: Modeling in IH Practice CHAPTER 18: UNDERSTANDING THE LIMITATIONS OF MATHEMATICAL MODELS AND PROACTIVE COMMUNICATION OF MODELING RESULTS.................................................................................................................................163 Knowing the Limitations of Mathematical Models ..................................................................................................................163 Proactive Communication of Modeling Results ........................................................................................................................164 Conclusions ............................................................................................................................................................................................168 References .............................................................................................................................................................................................169 CHAPTER 19: FEEDING MODELS USING SIMPLE AND SOME NOT-SO-SIMPLE MEANS ..........................................171 Introduction ............................................................................................................................................................................................171 Models Are Beasts That Need to Be Fed ....................................................................................................................................171 Trading Conservatism for Data ......................................................................................................................................................172 Chamber Case Studies ......................................................................................................................................................................173 The “Leaky” Chamber ........................................................................................................................................................................175 The Disappearing Biocide ................................................................................................................................................................175 Simple Stuff With Big Payouts .......................................................................................................................................................176 Postscript ................................................................................................................................................................................................176 References .............................................................................................................................................................................................176

Copyright AIHA®. For personal use only. Do not distribute.

vii

Modeling Occupational Inhalation Exposures to Chemicals Chapter 1: Modeling in IH Practice

Foreword John Mulhausen, PhD, CIH, CSP, FAIHA Exposure models are critically important tools for effective and efficient exposure risk management. They offer important advantages over other approaches to evaluating exposure risks. First, they are inexpensive, typically providing results without the need for sample collection and analysis. Second, they provide results quickly, as there is no need to wait for sample results from the laboratory. Third, they can be used to characterize risks associated with exposure scenarios that are absent, either retrospectively for past scenarios or prospectively in anticipation of future scenarios. And fourth, once a model exposure estimate is calculated, it can provide useful prior knowledge to feed a Bayesian analysis to enhance information provided by monitoring results in order to reduce uncertainty in our exposure judgments.1,2 Given its many advantages and the fact that the models, training resources, and free computer tools to facilitate its use have been available to us for decades, why is exposure modeling still so rarely used in the industrial hygiene profession? It could be fear of the need to correctly apply the math required. Of course, the freely available computer tools allow for use of the models without an extensive understanding of mathematics, so that is not really an excuse. It could be a misunderstanding of modeling itself―a feeling that models are somehow pencil-whipping “hocuspocus” that is disconnected from the reality of the workplace and its associated exposures. But that could not be further from the truth. The science and art behind a well-validated model are quite extensive. Additionally, the detailed understanding of the exposure scenario needed to collect and record the information required to “feed” a model’s inputs demands more systematic exploration and documentation of the exposure situation than is typically performed for qualitative exposure judgments, our most frequently used approach to assessing exposures. Perhaps this need to collect the critical information required to choose and “feed” the model most appropriate to the exposure scenario of interest may in itself be a crucial reason for the lack of model use in our routine practice. We may lack confidence in our understanding of how, exactly, to approach an exposure situation and appropriately apply the powerful modeling tools at our disposal to make an accurate exposure risk decision. That is why this book is so important. Its approach―an emphasis on case studies and advice for collecting critical exposure determinant information―is designed to address people’s discomfort about how to approach an exposure scenario and choose the appropriate model, gather needed model inputs, and interpret and communicate the results. And it comes not a moment too soon. Studies have shown that the accuracy of exposure judgments is often poor,3,4,5 sometimes not statistically different from random chance,4,5 and tends to be biased low.3,5 Exposure modeling offers many benefits for improving the accuracy of our exposure risk decisions. The deliberate approach to understanding the workplace, work activities and patterns, exposure sources, and the geometry of work and controls demanded by exposure modeling provides critical inputs to making accurate exposure risk decisions― whether the decision is based on the model’s output or includes analysis of monitoring results. When we make qualitative exposure risk assessments―those made without the benefit of monitoring data―we are, in essence, performing exposure modeling in our heads. We consider the workplace, we review the physical, chemical, and toxicological properties of the materials being used, we watch the way people work and interact with exposure sources and controls, and we integrate all of that information into an exposure risk decision. Exposure modeling does the same thing. But it does it using a more robust, reproducible, and well-documented

Copyright AIHA®. For personal use only. Do not distribute.

ix

Modeling Occupational Inhalation Exposures to Chemicals Chapter 1: Modeling in IH Practice approach than is typically seen with qualitative judgments. Borrowing the decision-making framework described by Daniel Kahneman,6 modeling shifts our approach to exposure risk decision making from Fast Thinking, with its inherent biases and inaccuracies, to the deliberate and logical Slow Thinking most appropriate for these important decisions about exposure risks. Accurate decisions about exposure risks are critical to the effective and efficient protection of workers and communities. When we underestimate exposures, we leave people at risk. When we overestimate exposures, we waste precious control resources and often introduce unnecessary constraints for employees and production. We need ready access to all available tools if we are to practice effectively and efficiently. The unique approach taken by this book will give us all the understanding and confidence we need to move exposure modeling up to the top of our toolboxes.

References 1. Hewett P, Logan P, Mulhausen J, Ramachandran G, Banerjee S. Rating exposure control using Bayesian decision analysis. Journal of Occupational and Environmental Hygiene, 3(10):568–581, 2006. https://doi. org/10.1080/15459620600914641. 2. Vadali M, Ramachandran G, Mulhausen J. Exposure modeling in occupational hygiene decision making. Journal of Occupational and Environmental Hygiene, 6(6):353–362, 2009. https://doi. org/10.1080/15459620902855161. 3. Logan P, Ramachandran G, Mulhausen J, Hewett P. Occupational exposure decisions: can limited data interpretation training help improve accuracy? Annals of Occupational Hygiene, 53(4):311–324, 2009. https://doi.org/10.1093/annhyg/mep011. 4. Vadali M, Ramachandran G, Mulhausen J, Banerjee S. Effect of training on exposure judgment accuracy of industrial hygienists. Journal of Occupational and Environmental Hygiene, 9(4):242–256, 2012. https://doi.or g/10.1080/15459624.2012.666470. 5. Arnold SF, Stenzel M, Drolet D, Ramachandran G. Using checklists and algorithms to improve qualitative exposure judgment accuracy. Journal of Occupational and Environmental Hygiene, 13(3):159–168, 2016. https://doi.org/10.1080/15459624.2015.1053892. 6. Kahneman D. Thinking, Fast and Slow. Macmillan, 2011.

x

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 1: Modeling in IH Practice

Preface Chris Keil, PhD, CIH I’ve worked in the field of occupational exposure modeling since the early 1990s. As a young industrial hygienist and later as an academic, I was excited by the possibilities of using modeling to get a more complete picture of current, past, and potential worker exposures. Since then, I’ve worked with exposure modeling in a myriad of ways. I used it to practice industrial hygiene. I’ve taught modeling as part of university and professional development courses (PDCs). I’ve participated in modeling conferences and committees. I’ve performed research and contributed to books on modeling. Thirty years after my introduction to modeling workers’ exposures, I look back and see that modeling has come a long way in terms of the scope of its application and the science supporting it. However, as I look around, I am a little saddened that there has been a slow spread in the application of modeling throughout the breadth of the profession. I believed then—and still believe now—that exposure modeling can be a valuable part of every industrial hygienist’s exposure assessment toolbox. I also believe that every industrial hygienist is capable of using models. Why aren’t more industrial hygienists regularly modeling as part of their exposure assessments? Those of us who “live” in the modeling world tend to get excited about the mathematics and chemistry involved. Mathematics and chemistry are integral to modeling. For me, the more I work with modeling, the deeper into the weeds of calculations and chemicals I can happily go. Yet for a plant-level hygienist, math and chemistry— although core areas of industrial hygiene practice—may not be explicitly used on a day-to-day basis in their health and safety office. When I taught modeling by leading with derivation of the equations, I might have looked smart, but perhaps I outpaced my audience. To clarify, I was not outpacing them in their capacity to “get it,” as I’m sure they can understand, but outpacing the speed at which they could access and apply principles that may be a bit buried in pathways of their brains. By the time I got to example applications, the students may have been so focused on remembering and doing the math and chemistry that they couldn’t appreciate the context of the example and see how else to apply the model. I hypothesized that industrial hygienists needed a better “on-ramp” to the application of exposure modeling. I also thought that part of that on-ramp needed to be to spark their imaginations about how modeling can be used. To try to address this, I developed a PDC with a “case-based” approach to modeling. Exposure assessment scenarios were presented in a way they might be encountered in a workplace. The students spent time gathering information about the chemical, the workspace, and the work practices around the exposure. Then, potential models were introduced based on the conditions of the scenario and the information available. I pointed toward the primary literature for full treatments of the models and their derivations. But I focused on getting to an exposure estimate for the scenario or case. The class discussed “the answer” in terms of how to interpret the results considering the assumptions of the model, the pertinent exposure limits, and the uncertainty of the exposure estimate. I think the case-based approach was successful. Fewer students headed straight to the door at the end of the PDC; rather, they stuck around to talk about their questions, scenarios, and ideas. I worked with other instructors and delivered PDCs of similar formats several times in the following years. After the PDC, I started getting more interaction from students bringing follow-up questions or what-if scenarios.

Copyright AIHA®. For personal use only. Do not distribute.

xi

Modeling Occupational Inhalation Exposures to Chemicals Chapter 1: Modeling in IH Practice This book grows out of the case-based approach of those PDCs. The core of the book is a series of case studies that lead the reader through an introduction to the scenario, how information might be gathered, and how models might be applied and interpreted. Some modeling applications are very basic, and others illustrate more advanced approaches. Also included are chapters that try to set the context for model use, from the approaches to modeling to gathering data and communicating the results. The hope and intent are to spark readers’ imaginations about the use of modeling and provide an on-ramp to welcome more industrial hygienists into the use of models in their daily practice.

This book is the latest in a series of modeling resources available from AIHA. Modeling’s role in exposure assessment is introduced in A Strategy for Assessing and Managing Occupational Exposures, 4th edition. Mathematical Models for Estimating Occupational Exposures to Chemicals, 2nd edition provides a technical discussion of different modeling approaches. IHMOD 2.0 is a spreadsheet-based tool that helps with the calculations for a number of models. This current book gives insights into the practical application of various models. Additionally, AIHA professional development courses offer interactive training on model use. These materials provide an excellent grounding in and set of resources for exposure modeling that can benefit all industrial hygienists.

xii

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 1: Modeling in IH Practice

ACKNOWLEDGMENTS This book is dedicated to Rick Wadden, who opened my eyes to the possibilities of modeling; John Franke, who took me to my first Exposure Assessment Strategies Committee meeting; Peter Scheff, who saw potential in me; Lorraine Conroy, who mentored me along the path of modeling. A special thanks to the team that provided review of the chapters: T. Renee Anthony Stephanie Battista Kent Candee Kang Chen Quincy Coleman Joe Dartt Daniel Drolet Matthew Ferreri Joselito Ignacio Josh Maskrey Ephraim Massawe Enrique Medina Mark Nicas Mike Sellitto Mark Stenzel

Copyright AIHA®. For personal use only. Do not distribute.

xiii

Modeling Occupational Inhalation Exposures to Chemicals Chapter 1: Modeling in IH Practice

Chapter 1

Modeling in the Framework of Industrial Hygiene Practice Chris Keil, PhD, CIH A standard and useful definition of industrial hygiene is “the anticipation, recognition, evaluation, control, and management of hazards arising in and from the workplace.” Evaluating the level of risk posed by hazards is rooted in the process of exposure assessment. In exposure assessment, the industrial hygienist characterizes exposure to an environmental agent in terms of the route, intensity, duration, and frequency of the exposure. The exposure assessment is then used to make a judgment about the acceptability of the exposure and whether the exposure needs to be controlled.

The Industrial Hygienist and Exposure Assessment Best practice exposure assessment will include information on the full distribution of exposures experienced by a worker or group of workers.1 This “exposure profile” reflects the between-worker exposure variability exhibited within a group of similarly exposed workers. Additionally, each individual worker will have a within-worker variability of exposures from day to day and even from task to task. A group of workers that have the same general exposure profile based on similarity of jobs and tasks is called a similar exposure group (SEG).1 Once a good exposure profile has been developed for an SEG, the industrial hygienist then compares the exposure profile to a benchmark decision-making value to judge whether the exposure is acceptable, uncertain, or unacceptable. Exposures judged as acceptable should be scheduled for regular reassessment. Unacceptable exposures need to be controlled and then reassessed to assure acceptability. Sometimes the first pass through exposure characterization does not provide enough information to make a judgment of “acceptable” or “unacceptable,” leading to an uncertain exposure profile. When uncertain exposure profiles arise, more information is gathered to refine the exposure characterization and move toward being able to confidently judge the acceptability of the exposure. In all these cases, the exposure assessment and decisions should be documented.

Assessing Inhalation Exposures to Chemicals As is the case for all exposure characterizations, inhalation exposures are expressed in terms of intensity and time components. The intensity of air pollutants is the concentration of the chemical that a person is breathing. To fully understand risks from an exposure, more information than just the concentration is needed. How long and how often the concentrations are inhaled are required to fully characterize an exposure profile. Once an exposure profile is determined by air sampling, modeling, or surrogate data, it is compared to some decision criteria to judge the acceptability of the exposure. The starting point for establishing a decision criterion, or “acceptable level,” is an occupational exposure limit (OEL). A number of organizations produce OELs. Table 1.1 describes some of the most common OELs. Organizations involved in developing new chemical compounds or using chemicals without an OEL often establish internal OELs for use in worker protection. An exposure assessor then establishes their own/company’s “acceptable exposure” based on all available information. This acceptable exposure is often less than the regulatory limit, as many regulations may be decades old and not reflect current science regarding risks from the chemical.

Copyright AIHA®. For personal use only. Do not distribute.

1

Modeling Occupational Inhalation Exposures to Chemicals Chapter 1: Modeling in IH Practice Table 1.1: Common Occupational Exposure Limits

Organization Occupational Safety and Health Administration (OSHA)

Occupational Exposure Limit Permissible Exposure Limit (PEL)

Notes Legal limits in United States. Majority were established in 1970.

National Institute for Occupational Safety and Health (NIOSH)

Recommended Exposure Limit (REL)

American Conference of Governmental Industrial Hygienists (ACGIH)

Threshold Limit Value (TLV®)

American Industrial Hygiene Association (AIHA) until 2013, Occupational Alliance for Risk Science Scientific Committee on Occupational Exposure Limits (SCOEL)

Workplace Environmental Exposure Level (WEEL)

European Agency for Safety and Health at Work Deutsche Forschungsgemeinschaft (DFG)

Indicative Occupational Exposure Limit Value (IOELV) Maximale Arbeitsplatz-Konzentration (MAK)

Binding Occupational Exposure Limit Value (BOELV)

Many do not reflect current scientific information and are “inadequate for ensuring protection of worker health.”2 Not a regulatory requirement, but a recommended guideline. Updated as new information on chemicals becomes available. Guidelines based on committee review of information on the health of a substance. Many consider these OELs the most reflective of current information. Guidelines for substances that were not being addressed by other organizations. European Community minimum level of protection for workers. Considers socioeconomic factors as well as technological feasibility. European Community health-based limits. German health-based maximum worker exposure levels.

An OEL is more than just a concentration: it is almost always written to include a time frame for which the OEL concentration is acceptable. The time frame of an OEL reflects the toxicological action of the chemical. A chemical that produces health effects after long-term exposure will typically have the OEL expressed in terms of an 8-hr time-weighted average (TWA-8). Chemicals that produce effects quickly may have OELs expressed in terms of a short-term exposure limit (STEL), a ceiling limit (C), or peak concentration. These OELs also have a time frame for which the OEL concentration is acceptable. Common time frames are 15 and 10 min.

Characterizing Exposure When it comes to determining concentrations for characterizing inhalation exposure profiles, often the first thought is to make a measurement. Industrial hygiene sampling and analysis science is quite advanced and, when properly performed, can provide an excellent measurement of air concentrations for thousands of chemicals. Yet a single air sample is only one data point in developing an exposure profile. With only a few air samples, making confident statistical conclusions about the acceptability of exposure is challenging. For example, consider four samples taken randomly across an SEG. Suppose that each of the concentrations measured is below the predetermined OEL. A data set of four samples is not a large enough sample size to use parametric statistics to draw conclusions. Parametric statistics are the inferential statistics that almost everyone learns first: t-tests, 95% confidence intervals, etc. Parametric statistics assume that the data are normally distributed or can be transformed to be normally distributed. With only four samples in our example, a normal distribution cannot be assumed, and nonparametric statistics must be used to draw statistically based inferences. With four samples all below the OEL, a nonparametric tolerance limit statement can be made. With these results, one can be:

2

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 1: Modeling in IH Practice • 26% confident that 75% of all similar exposures will be below the OEL • 5% confident that 90% of all similar exposures will be below the OEL • 1% confident that 95% of all similar exposures will be below the OEL Refer to Mulhausen and Damiano Appendix VIII for more detail on this approach.3 As more exposure measurements are made on an SEG and assumptions about the distribution of the exposure profile can be justified, parametric statistics will be useful in judging the acceptability of exposure. But the point of this thought experiment is to illustrate that one or two samples, even when correctly made, are of only limited use in judging the acceptability of an entire exposure profile. Certainly, there are times when a single air measurement can shed great light on the acceptability of exposure. Consider a measurement made right at the edge of a metal plating bath. Or consider a measurement made at the edge of a degreasing station. These locations could be considered “reasonable worst case” air concentrations because they are as close to the source of the pollutant as a worker is likely to get. If the TWA-8 concentration at these locations is a small fraction of the TWA-8 OEL, it could be appropriate to judge typical exposures from these processes as acceptable based on that one sample. However, these data cannot be categorized as a substitute for a full exposure assessment. Modern sensor technology greatly improves the ability to collect measurements when these resources are available. Although the cost is dropping and accessibility is increasing, no one sensor can measure all the potential exposures in a workplace. Also, an industrial hygienist still needs to think carefully about how, where, and when to use these resources to best characterize exposure profiles.

The Role of Modeling in Exposure Assessment Workplaces today are complex environments. A regularly increasing number of chemical, physical, and biological agents are entering production processes. Additionally, new processes are using agents in innovative ways. The sheer number of exposure assessments required in modern workplaces can be daunting. There are limits on resources (both time and money) needed to complete exposure assessments. The AIHA exposure assessment strategy recommends 6 to 10 measurements to characterize each SEG.1 Although this number of samples can provide statistical rigor to judge the exposure profile acceptability, collecting that many measurements Past Exposures for all SEGs at a workplace can be a substantial • Litigation undertaking. Even when resources are available to execute statistically well-designed sampling • Epidemiology campaigns, a priori decisions are needed to develop • Follow-up of accidental exposures sampling plans. Which exposures should be measured Current Exposures first? Which exposures could have a lower priority for assessment? Ultimately, industrial hygienists • Prioritizing sampling are faced with the dilemma that it is impossible to • Supporting professional judgment of measure all exposures everywhere at all times. It has acceptable exposures been estimated that professional judgment is used for • Supplementing exposure profiles with small more than 90% of all exposure judgments without any 4,5 sample sizes measurement data. There are also exposures of concern that cannot be measured. A new process may be proposed, or process engineers may suggest a shift to a new material in an existing process. Before changes are made, it would be helpful to know what changes in exposures could occur. Planning for an emergency response to unplanned accidental chemical releases also requires some understanding of potential exposures.

Copyright AIHA®. For personal use only. Do not distribute.

Future Exposures • Planning for process change • Anticipating exposures from process disruptions • Emergency preplanning Figure 1.1: The range of application of modeling in exposure assessment.

3

Modeling Occupational Inhalation Exposures to Chemicals Chapter 1: Modeling in IH Practice In addition to looking to future exposures, exposures that occurred in the past may not have been measured. Yet these past exposure levels can be of interest for multiple reasons. For example, a spill of a volatile liquid may have been quickly cleaned up, but the exposures around the spill were not measured. Or there may be concern that exposures in the past are linked to health effects that manifest later. These exposure assessments can be of great importance for epidemiology or litigation. Modeling exposures can be an integral part of a comprehensive exposure assessment program (Figure 1.1). Modeling can support professional judgment in the many cases where judgment alone is used to characterize exposures. Research has shown that professional judgment is inaccurate and underpredicts actual exposures.6,7 Qualitative professional judgment that combines an industrial hygienist’s education, experience, and intuition can be checked and supplemented with modeling and other tools such as checklists.7 Modeling can also provide documentation supporting professional judgment. When preparing a sampling campaign, modeling can help prioritize exposures to measure. A limited number of exposure measurements can be supplemented by modeling. Exposures from process changes can be projected using modeling, and historical exposures can be estimated with models.

Physical-Chemical Concentration Modeling There are a variety of approaches to exposure modeling. This book will primarily provide examples of physicalchemical concentration modeling. There are also statistical or knowledge-based models that use databases of previously collected exposure data and exposure parameters to estimate likely exposure levels for SEGs. One of these approaches will be covered in Chapter 16. Physical-chemical models use the properties of the release and transport of an agent to estimate exposure intensity at a location. Modeled concentrations can then be used in conjunction with what the industrial hygienist knows about the time components of the exposure to develop an exposure characterization. In physical-chemical modeling, information on the chemical released into the air and information on the airspace into which the chemical is released are used to calculate air concentration estimations. The next chapter will expand on the types of information that are needed in modeling and approaches to gathering the information. In general, information on the chemical includes its properties, such as molecular weight, density, and vapor pressure, as well as the quantities of the chemical used and released. Airspace characteristics that often play a role in determining exposures’ intensity are volume and geometry, general and local exhaust ventilation design and layout, net airflow rate, local airflow patterns, and air mixing. The models themselves provide techniques to calculate air concentrations. Depending on the models and the inputs selected, the calculated concentrations could be “worst case,” “reasonable worst case,” or “representative” concentrations. A quickly calculated model with conservative, overestimating assumptions may yield a “worst case” concentration estimation sufficiently below an OEL to judge an exposure as acceptable. When these estimates exceed an OEL, more precise models may be needed. Although a physical-chemical model can provide estimates of concentration, the industrial hygienist needs to interpret the concentration alongside the time frame of the worker’s exposure to judge the acceptability of the exposure. For example, consider a case where modeling determines that the concentration around a process is 520 mg/m3. The relevant OEL TWA-8 is 410 mg/m3. The modeled concentration is higher than 410 mg/m3, but the industrial hygienist also knows that the worker operates the process a maximum of 2 hr and has very little or no exposure to the chemical when not operating the process. This concentration can be used with the time frame of the exposure to calculate the modeled TWA-8 exposure. mg mg 520 —— · 2 hr + 0 —— · 6 hr m3 m3 mg TWA-8 = ———————————————— = 130 —— 8 hr m3

4

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 1: Modeling in IH Practice When the modeled concentration is considered in light of the time components of the exposure, the TWA-8 is 32% of the OEL. The industrial hygienist can then judge within the context of their organization whether this exposure is acceptable or unacceptable or if more data are needed.

Current Applications of Air Concentration Modeling for Regulatory Purposes For decades, the United States Environmental Protection Agency (U.S. EPA) has been using air pollution modeling for a variety of regulatory purposes. Like indoor air chemical concentration modeling, outdoor dispersion models use data on the release of the pollutant and information about the airspace into which the pollutant is being released (the meteorology) to calculate estimates of pollutant concentrations. The EPA uses models as part of its permitting process to determine if a new source will contribute to ambient air pollution at levels that produce unacceptable levels of risk. Models are also used for forecasting air quality levels that would result from regulatory initiatives. Dispersion models, such as the Gaussian plume model (Figure 1.2), are commonly used for these applications.

Figure 1.2: Illustration of the Gaussian Plume Model. Mass emission models like those found in the EPA document AP-42: Compilation of Air Emissions Factors provide estimates of pollutant release rates for use in the Gaussian plume model.8 AP-42 includes a range of reliability in the emission estimates, reporting a reliability ranking for each emission factor. The EPA accepts that there is emission variability at a source and between sources. Yet it also recognizes the need for information to assess exposures while at the same time considering the limitations of the data and models when interpreting the results. In the early years of exposure assessment science, OSHA developed a simplistic approach to the enforcement of permissible exposure limits (PELs) when judging the acceptability of exposures. If the measured value minus the sampling and analytical error (SAE) is above the OEL for a single sample, OSHA can bring an enforcement action. This approach only considers the single exposure being measured and is a legal compliance test rather than a true exposure characterization.

Copyright AIHA®. For personal use only. Do not distribute.

5

Modeling Occupational Inhalation Exposures to Chemicals Chapter 1: Modeling in IH Practice However, in a number of OSHA standards, the use of modeling is permitted for various aspects of exposure assessment. In the butadiene standard, determining cases where products containing butadiene can be exempted can be done using “objective data” [29 CFR 1910.1051(a)(2)(i and ii)].9 Objective data are defined in the standard as “monitoring data, or mathematical modelling or calculations based on composition, chemical and physical properties of a material, stream or product.” The formaldehyde standard also allows the use of objective data when conducting exposure assessments in a number of situations [29 CFR 1910.1048].10 The methylene chloride standard requires initial exposure monitoring except “where objective data demonstrate that methylene chloride cannot be released in the workplace in airborne concentrations at or above the action level or above the STEL.”11 OSHA’s respiratory protection standard requires that employers make a “reasonable estimate of employee exposures to respiratory hazard(s).”12 A letter to OSHA in October 2011 asked for clarification regarding the need for air testing. The OSHA response stated: Although the most reliable and accurate method to determine exposure is to conduct personal air monitoring, it is not explicitly required by the Respiratory Protection standard. Instead, other means can be used to estimate workplace exposures. These methods include, but are not limited to, the use of objective data, application of mathematical approaches, and others.13 These instances illustrate examples of when even OSHA accepts modeling as part of exposure assessment.

How Close Is Close Enough? A common objection to using modeling as part of exposure assessment is that there are uncertainties around the model selection and assumptions of the model and the inputs to the model, which result in uncertainty in the modeled concentration. This certainly is true. However, when considering this objection, there are several factors to keep in mind. One aspect to consider is what OEL an organization decides to use. Regulatory OELs often allow higher exposures than OELs developed by nonregulatory entities. In addition to selecting the OEL to use, there are different approaches to defining acceptable exposures relative to the OEL. An organization could decide that an acceptable exposure profile is one in which the long-term average exposure of an SEG is less than 10% of the OEL. Another organization could state that “acceptable exposure” means that the 95th percentile exposure within an SEG is some fraction of the OEL. AIHA’s A Strategy for Assessing and Managing Occupational Exposures, 4th edition covers the many aspects of exposure rating.1 Once these decision criteria are determined, uncertainty around the modeled exposure can be considered.

Analogous to the case where an industrial hygienist uses their professional judgment to make a “worst case” concentration measurement, “reasonable worst case” models can be used with inherently overestimating assumptions to model concentrations. In both measured and modeled “reasonable worst cases,” if the concentrations measured or modeled are a fraction of the OEL, an industrial hygienist may be satisfied with judging an exposure as acceptable. 6

0.12

OELs 0.1

Proportionofvalues

Although modeled air concentrations have uncertainty associated with them, air concentration measurements have uncertainty as well. There is sampling and analytical error, which tends to be small. Additionally, when characterizing the exposure profile of an SEG, there is between-worker and withinworker variability, which can be quite high. When judging the acceptability of an exposure profile using a small number of samples, there can be tremendous uncertainty, as has been shown above.

0.08

0.06 Distributionfrommodel 0.04 Distributionfromsamples 0.02

0 0

20

40

60

80

100

Concentration(mg/m3)

Figure 1.3: Comparing modeled and measured exposure profiles to occupational exposure limits (OELs). Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 1: Modeling in IH Practice There are also techniques for quantifying the uncertainty associated with exposure modeling. From propagation of error calculations to probabilistic “Monte Carlo” models, the variability and uncertainty of modeled exposure estimates can be tracked. Some of these techniques will be touched on within case studies presented in various book chapters. At the root of addressing the question “How close is close enough?” is the consideration of how close the exposure assessment is to the OEL. Consider an exposure scenario where a simple modeled concentration is 15 mg/m3 ± 10 mg/m3. This estimate was arrived at using a wide range of plausible input values for pollutant emission and ventilation assumptions. This uncertainty of ±67% seems at face value to be extremely limiting to the use of this modeled exposure information. For the sake of illustration, we will consider the estimated range of values to be uniformly distributed. That is, all values along the range are equally likely to occur. Now consider six measurements taken to characterize the concentration: 14, 18, 10, 20, 22, and 8 mg/m3. These measurements have a mean of 15 mg/m3, a geometric mean (GM) of 14 mg/m3, and a 95% confidence interval on the GM of 9–22 mg/m3. This confidence interval is almost as wide as the uncertainty in the modeled concentrations. In both cases, if the OEL were 100 mg/m3, the exposure could probably be judged as acceptable. If the OEL were 30 mg/m3, it is likely that with either approach (model or measurement), more information gathering would be needed prior to concluding on the exposure acceptability. Figure 1.3 illustrates this decisionmaking process.

Advancing the Use of Models in Exposure Assessment The science of exposure assessment modeling is steadily advancing. Authors in this volume and others are leading the development of occupational exposure modeling in theory and in practice. Many of the examples presented are based on peer-reviewed research studies. However, large gains have yet to be made in the integration of modeling into routine industrial hygiene practice. The scientific literature may seem daunting to facility-level industrial hygiene and safety professionals. The potential benefits of the scientific advances still need to be translated into regular practice. To that end, this volume presents a series of cases to demonstrate the application of modeling to exposure scenarios that could be encountered in industrial hygiene practice. Some of the examples are based on actual exposures, whereas others are posited as illustrations of modeling applications. One of the goals is to spark the imagination of the reader regarding the use of models: “If modeling can be used in this case presented, perhaps it could be used in my situation.” With some experience with worked applications, such as those in this book, hopefully an industrial hygienist will be emboldened to think creatively about how modeling can be used in their practice. Advancing both the science and practice of modeling in occupational exposure assessment will benefit occupational and public health professions. These cases are presented in the hope that they will nudge the industrial hygiene community into making full use of the opportunities presented by modeling.

References 1. Jahn SD, Bullock W, Ignacio JS. (Eds.). A Strategy for Assessing and Managing Occupational Exposures, 4th edition. AIHA: Falls Church, VA, 2015. 2. Occupational Safety and Health Administration (OSHA). Permissible Exposure Limits – Annotated Tables. https://www.osha.gov/dsg/annotated-pels/ (accessed August 7, 2019). 3. Mulhausen JR, Damiano J. (Eds.). A Strategy for Assessing and Managing Occupational Exposures, 2nd edition. American Industrial Hygiene Association: Fairfax, VA, 1998. 4. Vadali M, Ramachandran G, Mulhausen J. Exposure modeling in occupational hygiene decision making. Journal of Occupational and Environmental Hygiene, 6(6):353–362, 2009. https://doi. org/10.1080/15459620902855161.

Copyright AIHA®. For personal use only. Do not distribute.

7

Modeling Occupational Inhalation Exposures to Chemicals Chapter 1: Modeling in IH Practice 5. Vadali M, Ramachandran G, Mulhausen JR, Banerjee S. Effect of training on exposure judgment accuracy of industrial hygienists. Journal of Occupational and Environmental Hygiene, 9(4):242–256, 2012. https://doi. org/10.1080/15459624.2012.666470. 6. Logan P, Ramachandran G, Mulhausen J, Hewett P. Occupational exposure decisions: Can limited data interpretation training help improve accuracy? Annals of Occupational Hygiene, 53(4):311–324, 2009. https://doi.org/10.1093/annhyg/mep011 7. Arnold SF, Stenzel M, Drolet D, Ramachandran G. Using checklists and algorithms to improve qualitative exposure judgment accuracy. Journal of Occupational and Environmental Hygiene, 13(3):159–168, 2016. https://doi.org/10.1080/15459624.2015.1053892. 8. United States Environmental Protection Agency (EPA). AP-42: Compilation of Air Emissions Factors, 5th Edition. Volume I: Stationary and Point Sources. Office of Air Quality Planning and Standards, 1995. 9. Occupational Safety and Health Administration. 29 CFR 1910.1051 – 1,3-Butadiene Standard. 1996. 10. Occupational Safety and Health Administration. 29 CFR 1910.1048 – Formaldehyde Standard. 1991. 11. Occupational Safety and Health Administration. 29 CFR 1910.1052 – Methylene Chloride Standard. 1997. 12. Occupational Safety and Health Administration. 29 CFR 1910.134 – Respiratory Protection Standard. 1998. 13. Occupational Safety and Health Administration. Respiratory Protection Standard Interpretation. 2011.

8

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 2: Information Gathering

Chapter 2

Information Gathering for Air Concentration Modeling Chris Keil, PhD, CIH The models in this book can be used to calculate the concentration of chemicals in air. Air concentrations are only one component of a personal exposure profile. Inhalation exposures, like all exposures, include both the intensity of the exposure (the concentration) and the time components of the exposure (the duration and frequency). Mathematical models provided in this text will offer repeatable methods to estimate air concentrations. Duration and frequency components of exposure are part of worker activity profiles. The modeled air concentrations and worker activity patterns are used together to characterize the exposure profile. The acceptability of the modeled exposure profile is then evaluated against known information about the chemical’s toxicity, typically in the form of an occupational exposure limit (OEL), which includes both concentration and time components. Air concentration modeling relies on two categories of information: 1) information about the substance, such as the quantity released and its chemical properties, and 2) information about the airspace into which the airborne chemical is released, such as room volume and ventilation characteristics. Gathering and applying necessary information to model concentrations can be an iterative process. That is, readily available information can be used to make coarse estimates of concentration. If these coarse estimates are adequate for decision making, as will sometimes be the case, the method of modeling the concentration can be documented and the industrial hygienist can move on to the next exposure assessment. In other cases, more information will be needed about the pollutant emission and/or its movement throughout the airspace before a judgment regarding the exposure can be made.

Information About the Chemical When beginning a concentration modeling project, a good starting point is gathering basic physical and chemical properties of the substance. Common useful properties are the chemical formula, molecular weight, density, and vapor pressure. Additional properties may include the melting point, boiling point, and explosive limits. It is important to know that some properties change as other model inputs change. For example, vapor pressure changes as temperature changes. Ultimately, the modeling tool chosen to characterize emissions will dictate the necessary properties. Simplistic models may require few inputs, whereas more sophisticated models will require additional inputs. As a general rule, gather as many properties as possible prior to implementing the model. In the event that the simple model provides inconclusive results, the user can quickly integrate the data into a more sophisticated model. The NIOSH Pocket Guide to Chemical Hazards1 is a useful resource for this information, as are safety data sheets (SDSs). PubChem is an even more robust online source of chemical information.2 The mass of the chemical released into the air is a key parameter for modeling air concentrations and is often quite challenging to determine. There are a myriad of industrial processes that produce air emissions. Grinding, sanding, welding, combustion, misting, evaporation of liquids from open surfaces or leaks, pressurized releases, and off-gassing are just some of the mechanisms that create mass releases into the air. The rate of the release of chemical mass into the air, also called generation, emission, or source rates, can be relatively constant over time or can vary with time as processes change. Depending on the information available, the average emission rate may be determined, while recognizing that there is variation of the emission rate within the averaging time. For example, a worker may be able to estimate that they use 400 g of a volatile chemical over the course of an 8-hr shift. An 8-hr average emission rate of 50 g/hr can be calculated using the following values:

Copyright AIHA®. For personal use only. Do not distribute.

9

Modeling Occupational Inhalation Exposures to Chemicals Chapter 2: Information Gathering 400 g g ——— = 50 — 8 hr hr This is the average emission rate over the course of the shift. The average emission rate does not offer information regarding peak or zero emission generation events during the time period. The emission generation rate will typically increase or decrease in correlation with process activities.

Mass Balance Emission Rates There are a variety of approaches to obtaining emission information. As illustrated in the previous calculation, material inventory information is a useful source of emission data. Mass balance emission rate calculations can be as detailed as the inventory information available. For example, a company purchases one 55-gal drum of cleaning solvent each month. In the cleaning process, some of the solvent is recovered. The rest evaporates into the workplace air. About 30 gal per month are recovered and shipped off site as hazardous waste. The rest evaporates into the workplace air. The difference between the solvent volume purchased and the solvent volume shipped off site is the volume of solvent that evaporated into the workplace air. 55 galin 30 galout 25 galevap ———— – ————— = ————— month month month In this example, the solvent’s SDS indicates that the solvent density is 0.81 g/mL. This particular facility operates about 22 days per month, 16 hr each day. Using these values and some conversion factors, the average evaporation rate of the solvent can be calculated. 25 gal month day hr 3,785 mL 0.81 g 1,000 mg 3,629 mg ———— · ———— · ———— · ———— · ————— · ———— · ————— = ————— month 22 day 16 hr 60 min gal mL g min The average emission rate is 3,629 mg/min over the course of the month. There will likely be peaks and troughs around this average based on process variations. If there is local exhaust ventilation (LEV) associated with the processes, a portion of this mass would be captured by the LEV and not enter the workplace air. In another example, a copper electroplating process operates 8 hr/day and plates approximately 6,500 parts. An air sparger is used to keep the plating solution mixed but also creates bubbles that burst on the surface of the plating solution, aerosolizing some of the liquid. Copper rods are consumed in the process, with 1.5 kg of copper rods consumed in the process daily. This value of 1.5 kg is equivalent to 1,500 g and 1,500,000 mg. Each part has an area of 125 cm2 and is plated to a thickness of 0.0002 cm in accordance with manufacturing specifications. The area multiplied by the thickness gives a total volume of 0.025 cm3 copper plated onto each part. This volume can be used with the density of copper (8.96 g/cm3, a physical property of copper available in many chemical databases) to give the mass plated onto each part. The mass of copper rods used daily minus the mass of the copper deposited on plated surfaces can give an estimate of the copper released. Assuming that the copper concentration in the tank does not change, the mass plated onto the parts can be calculated. 6,500 parts 0.025 cm3 8.96 g 1,000 mg 1,456,000 mgonto parts —————— · ————— · ———— · ————— = —————————— day part cm3 g day 1,500,000 mginto tank 1,456,000 mgonto parts 44,000 mgreleased —————————— – —————————— = ———————— day day day

10

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 2: Information Gathering Knowing that the process operates 8 hr/day, the copper emission rate (in mg/min) can be calculated. day hr 92 mg 44,000 mgreleased ———————— · ——— · ———— = ———— day 8 hr 60 min min The calculated emission rate assumes that all copper particulate is released into the workplace air. It does not account for LEV removal and/or any form of recovery mechanisms that may capture mist and return it to the plating bath. This simply is presented as an example of a mass balance approach to estimating emissions.

Emission Models There are various types of models available to estimate emissions. Some, such as evaporation rate models, use chemical properties (e.g., molecular weight and vapor pressure) to predict emissions. Other models are entirely empirical and experimentally developed. There are even models that combine both. A number of models describing evaporation rates from liquid surfaces are available. Lennert et al.3 and Arnold and Engel4 provide good reviews of the available models.

Emission Factors Emission factors have been used to describe ambient air emissions for decades and can be useful in occupational environments as well. An emission factor links the mass emission of a pollutant to a process variable. For example, an emission factor could link the mass of pollutant released to the number of parts produced during manufacturing processes. Knowing the rate of parts production and the emission factor, an emission rate can be calculated. AP-42 is the EPA document that lists ambient air emission factors.5 Depending on the exposure scenario, some of these emission factors could also be used for occupational exposure estimates. A number of studies provide emission factors specifically for workplace contexts. Some of the processes that have published emission factors include: • 3D printing6 • Candy glazing using ethanol7 • Chromium plating8,9 • Freon degreasing10 • Offset printing11,12 • Painting with diisocyanate containing product13 • Small caliber ammunition firing14 • Toluene parts washing15 • Trichloroethylene vapor degreasing9,10 • Welding16 Published emission rates are less generalizable than emission factors because they are often associated with the specific production conditions under which they are measured. Nevertheless, there are context-specific emission rates published for various processes that can be useful for air concentration modeling. Some of the available studies include: • 3D printing17 • Floor cleaning with acetic acid containing product18 • Formaldehyde in anatomy labs19 • Medical lasers20 • Paint and resin manufacture21 • Plasma cutting of stainless steel22

Copyright AIHA®. For personal use only. Do not distribute.

11

Modeling Occupational Inhalation Exposures to Chemicals Chapter 2: Information Gathering • Styrene resin application23 • Terpene at sawmills24 • Thermal spraying of metals25

Information About the Airspace Room Volume The first component of the airspace that is often useful is the room volume (Vroom or V). Many rooms can be approximated as cuboid, and the volume is calculated as shown in Equation 2.1: Length ∙ Width ∙ Height = Volume

(2.1)

Sometimes the geometry of an indoor airspace is more complex. An approach to determining the volume of a complex-shaped room is to break the space into simpler geometric pieces, calculate the volume of each “piece,” and sum the volumes. Sometimes spaces within the room are occluded, and a simpler approach would be to subtract the volume of the occluded space. Consider the space illustrated in Figure 2.1.

Figure 2.1: Example room geometry. The volume of the room could be determined two different ways. One could calculate the volume of three spaces as shown in inset (a): (21 ft ∙ 6 ft ∙ 9 ft) + (27 ft ∙ 8 ft ∙ 9 ft) + (24 ft ∙ 4 ft ∙ 9 ft) = 1,134 ft3 + 1,944 ft3 + 864 ft3 = 3,942 ft3 Or, as shown in inset (b), the volume of the large cuboid could be calculated and the volumes of the occluded parts subtracted: (27 ft ∙ 18 ft ∙ 9 ft) – (6 ft ∙ 6 ft ∙ 9 ft) – (3 ft ∙ 4 ft ∙ 9 ft) = 4,374 ft3 – 324 ft3 – 108 ft3 = 3,942 ft3

12

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 2: Information Gathering Objects inside the room occupy space inside the room where air cannot reside; therefore, their volumes must be estimated and subtracted from the overall room volume. If the modeling scenario is in an extremely full room, the volume of the objects can be calculated and subtracted from the room volume. Figure 2.2 is a photograph of an extremely crowded lab space. The room is 10 ft × 8 ft with a 9-ft ceiling. In the room are two desks, a workbench along the side, a laboratory hood, a gas chromatograph, and multiple compressed gas cylinders. The volume displaced by all these objects is about 15% of the room volume, which reduces the air volume in the room from 720 ft3 to 612 ft3. (8 ft ∙ 10 ft ∙ 9 ft ) ∙ (100% – 15%) = 720 ft3 ∙ 85% = 612 ft3

Figure 2.2: A crowded lab space with 15% of the volume displaced. (Photo: Chris Keil).

One could also evaluate a production room that might be considered crowded with three reaction vessels. Refer to Figure 2.3. The room is 25 m × 10 m with a 6-m ceiling, giving a room volume of 1,500 m3. The three reaction vessels are 35,000 L each, giving a total displaced air volume of 105,000 L (which is equivalent to 105 m3). This represents approximately 7% of the room volume.

Ventilation To perform very simple modeling, or when there is no information on ventilation for a workspace, zeroventilation models can be used. Zero-ventilation models are the simplest models because they only consider the mass emission and the volume Figure 2.3: An example of a crowded production of the space. Zero-ventilation models are usually space. Displaced volume is approximately 7%. overestimating and are not representative of reality because they assume that a chemical can accumulate in room air but cannot escape. Typically, zeroventilation models are used to perform screening-level decision making. Most occupational airspaces have air entering and exiting the room. Understanding how the airspace is ventilated is critically important to modeling air concentrations more accurately than is possible using zeroventilation models. There are usually two principal questions to address. The first question is: “Does the room have mechanical ventilation, or does it rely on natural ventilation?” The second question is: “What sort of air mixing occurs in the room?” Many industrial workspaces have mechanical ventilation. Air is supplied to the room, often through ceiling or wall diffusers. This air may be outdoor air or a mix of outdoor and recirculated air. Air can be exhausted through return grills or through the LEV. Ventilation through a room is described in terms of volumetric flow rate (Q) in dimensions of volume per time (with typical units being m3/sec or ft3/min). The volumetric flow rate of air entering a room (Qin) will be equal to the flow rate out of the room (Qout ). If the mechanical supply rate and exhaust rate are not exactly balanced, there will also be a positive or negative pressure in the room. This will result in airflow into or out of the room through open doors or windows, spaces along walls or closed doors, drop ceiling tiles, etc. Sometimes only one side of the air balance, Qin or Qout, is mechanically driven.

Copyright AIHA®. For personal use only. Do not distribute.

13

Modeling Occupational Inhalation Exposures to Chemicals Chapter 2: Information Gathering Room ventilation is sometimes described in terms of “air change rate” or “air exchange rate.” An air change rate is the volumetric flow rate of air divided by the volume of the space (Q/V), which results in air change rate having the dimensions of “per time” (1/time or time–1). This variable is commonly referred to as “air changes per hour” (ACH). For modeling, the air change rate is converted into a volumetric flow rate. There are several ways to obtain values for Q. The most straightforward way is to measure Q for the airspace of interest. With this method, all mechanical air supply and exhaust points are identified and Q through each is measured using standard industrial hygiene methods. A Balometer is useful for measuring airflow through diffusers and return grills. When air is flowing in to and out of a room through a plain opening such as a door, duct, or similar space, air velocity measurement with an anemometer can be used with the cross-sectional area to calculate volumetric flow rate. Q=v∙A

(2.2)

Where v: airspeed (length/time) A: cross-sectional area of opening, duct, etc. (length2) Because we do not have rooms collapsing from negative pressure or exploding from positive pressure, Qin = Qout. However, measuring airflow in and airflow out does not always result in a perfectly measured air balance. When one side of the air balance is more certain, that measured value is often the most useful for describing Qroom. For example, a room has four mechanical diffusers supplying air. There is no mechanical exhaust of the air. Air either goes out open doors or up through the drop ceiling into the plenum between the ceiling panels and the structural ceiling. In this case, a good measurement of Qin through the supply diffusers with a Balometer is a good estimate of Qroom. In another room there may be air supplied through some diffusers, but there are also multiple LEV systems that create a negative pressure in the room. Air leaks into the room through doors, cracks in walls, etc. In this case, quantifying the flow rate through each LEV system and summing the total provides a good measurement of Qout, which is a good estimate of Qroom. When looking to use models, there may be other resources available that provide insight into ventilation rates. Building engineers may be able to provide estimated ventilation rates using HVAC and LEV system specifications. Building blueprints often include design ventilation rates. Care must be taken if just relying on blueprints because a system may not be installed as designed or may not be currently operating as designed or installed. The modeler should document which estimate was used in the model in the event that future surveys are necessary so that future modelers can understand the assumptions made in the original assessment. When measurements of Q are not possible, there are some published values that can be used. Some of these values are typical values, and actual ventilation rates in specific airspaces may not match these typical rates. The Engineering Toolbox lists common air change rates in factories as 2–4 ACH for ordinary factories and 10–15 ACH for factory buildings with fumes or moisture.26 In standard 62.1, the American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE) recommends outside air supply rates for general manufacturing of 10 ft3/min/person (5 L/sec/person) occupying the airspace.27 On a floor area basis, ASHRAE recommends 0.18 ft3/min/ ft2 (0.9 L/sec/m2). In ASHRAE 62.1, there are also recommended ventilation rates for a variety of indoor airspaces. In its 2011 Exposure Factors Handbook, the U.S. EPA states that “Few air exchange rates for commercial buildings are provided in the literature.”28 In the Exposure Factors Handbook, the EPA proposes typical air change values for nonresidential buildings as a mean of 1.5 (0.87 standard deviation) with a range of 0.3–4.1 and a 10th percentile of 0.60. Older residential and work environments may not have mechanically provided outdoor air. In these buildings, air is introduced to the indoor airspace through the driving force of wind and negative pressures created by convective heat rise through the building. This type of ventilation is often referred to as natural ventilation. There are some published values for air change rates for buildings with natural ventilation. Table 2.1 presents some values for naturally ventilated residential housing.

14

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 2: Information Gathering Table 2.1: Reported Air Change Rates in Naturally Ventilated Buildings

Source Jayjock and Havics29

ACH Mean: 0.39

Note Residential buildings

SD: 0.25 Median: 0.33

Sparks

30

75th percentile: 0.49

Winter values

90th percentile: 0.70 Tight construction: 0.3

Residential buildings

Typical energy efficient construction: 0.5 Construction over 30 years old: 1.0 Mean: 0.45

28

U.S. EPA

Residential buildings

10th percentile: 0.18

Bartlett et al.

31

Mean: 0.8

Commercial buildings

Range: 0.6–0.9 Closed doors

Portable classrooms

Mean: 0.208 Median: 0.156 Range: 0.018–0.738 Open doors Mean: 3.65 Median: 2.82 Range: 0.612–14.4 ACH, air changes per hour; SD, standard deviation.

Air Mixing As chemicals are released into the room, air movement mixes the chemicals throughout the space to varying degrees. Variations in chemical concentration will always exist in an airspace. However, effective judgments of the acceptability of exposures can be made even when simplifying assumptions about air mixing are made. In general, there are three approaches to considering air mixing in a room: a well-mixed room, two or more well-mixed “boxes” or zones within the room, or continuous concentration gradients created in the room by turbulent diffusion. The simplest models treat a room as a well-mixed box, or single zone. Although this will never be completely the case, it can be a reasonable assumption in a number of conditions. For example, there may be multiple points of air supply and exhaust coupled with lots of motion in the room such that the air is very well mixed. Figure 2.4 illustrates a room with two ceiling diffusers, a Figure 2.4: A room with good air mixing. ceiling fan, and two air returns on opposite walls.

Copyright AIHA®. For personal use only. Do not distribute.

15

Modeling Occupational Inhalation Exposures to Chemicals Chapter 2: Information Gathering A space such as this could likely be modeled as a well-mixed room. In most cases, a well-mixed zone assumption will provide a reliable estimate of the average concentration in the room. Worker tasks, activities, and position in the room relative to the source of the pollutant must be considered when interpreting the air concentrations modeled. In many cases, the relative position of air supply and exhaust locations results in different degrees of mixing throughout a room. If the air mixing at a location of concern is likely to be poor, the approaches that follow should be considered. When designing dilution ventilation, ACGIH recommends using mixing factors to account for less-thanideal mixing.32 Although this can provide a valuable safety measure when designing ventilation systems, the use of mixing factors when modeling a concentration is not encouraged. Applying a mixing factor to an entire room violates conservation of mass principles. A better option is to consider one of the following approaches. Rooms typically have concentration gradients rather than perfectly well-mixed conditions. Diffusion models aim to approximate the concentration gradients that often actually exist in a room. Workers move through these concentration gradients as they work and accumulate exposure to the chemical. Figure 2.5 illustrates a worker standing at a worktable with their breathing zone moving throughout a gradient of chemical concentrations as they perform their tasks.

Figure 2.5: A worker in a concentration gradient.

Diffusion processes are complex, and modeling concentrations with this approach may overly complicate the modeling process. However, there will be situations where diffusion modeling will be appropriate, particularly when modeling concentrations near pollutant sources. The two-zone (or two-box or near-field/far-field) method has emerged as a suitable modeling approach that reflects a concentration gradient within a room while not overly complicating the calculations. The space around a pollutant source is divided into a “near zone” close to the point of chemical release, and the rest of the space is a “far zone.” The near zone will include the space occupied by the worker’s breathing zone. Each zone is considered well mixed, and the two zones are assumed to exchange air with each other at a fixed rate. Figure 2.6 illustrates the model conceptually. The near versus far zone modeling may appear simple in concept, but Figure 2.6: A worker in a concentration it often can closely represent actual conditions. At a distance “far” gradient. from the source, bulk airflow combines with turbulent diffusion to even out the concentration gradient. In the zone “near” the source, recent work suggests that a worker actively performing tasks within the near zone induces good mixing within the zone.33 The two-zone model has been used effectively in a number of published exposure assessments.34

Worker Activity Information The time the worker spends within modeled concentration zones is needed to appropriately interpret the modeled exposure. A short-term exposure to a high concentration of a chemical that has an 8-hr time-weighted average (TWA-8) OEL may be acceptable depending on the worker’s other exposures. Interpreting modeled or measured concentrations requires the industrial hygienist to understand the temporal patterns of the exposures and the time frames of the OELs.

16

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 2: Information Gathering As an example, two-zone modeling results predict that the concentration of a chemical is 50 mg/m3 near a production process, while the average concentration is 5 mg/m3 in the rest of the room. A worker spends their entire 8-hr workday in the room. They generally spend 15 min/hr working near the process, for a total of 120 min near the source each day. They have a 30-min lunch break, which they spend at a location with no exposure to the chemical. The remainder of the workday, 330 min, they spend “far” from the chemical source. Their TWA-8 exposure to the chemical can be calculated as follows: mg mg mg 50 —— · 120 min + 5 —— · 330 min + 0 —— · 30 min m3 m3 m3 mg TWA-8 = ——————————————————————————— = 16 —— 480 min m3 Another worker in the same room is more experienced and efficient when working near the chemical source. They spend 5 min/hr near the process, for a total of 40 min near the source daily. They too have a 30-min lunch break. Thus, this worker spends 410 min in the “far zone” from the chemical source. This worker’s TWA-8 exposure is as follows: mg mg mg 50 —— · 40 min + 5 —— · 410 min + 0 —— · 30 min m3 m3 m3 mg TWA-8 = ——————————————————————————— = 8.4 —— 480 min m3 This example illustrates how work practices and activity patterns can influence exposures of workers who may even be considered part of the same similar exposure group. The role of time and activity must always be considered in addition to the modeled concentration when characterizing exposure profiles.

The Importance of the Information-Gathering Step This chapter described methods for gathering information for air concentration modeling. Two key types of information exist: substance properties and airspace properties. This information represents the inputs to the models used to estimate air concentration. The selection of model inputs is often the most influential component of a modeling assessment. It is critical to understand that the context of a modeling result depends on how the information was gathered. If the information gathered was incorrect, the result will not be useful in assessing exposures (“garbage in, garbage out”). If the information gathered was designed to produce an overestimate, the result will be a higher concentration than actually exists in the workplace air. It is important to think through the information-gathering process as part of your exposure assessment. As we move through the case examples in this book, consider the methods used to gather information in each case and how the gathered information might influence the result.

References 1. National Institute for Occupational Safety and Health (NIOSH). NIOSH Pocket Guide to Chemical Hazards [DHHS (NIOSH) Publication Number 2005-149]. Department of Health and Human Services Centers for Disease Control and Prevention. September 2007. 2. National Institutes of Health. PubChem. https://pubchem.ncbi.nlm.nih.gov/ (accessed March 1, 2022). 3. Lennert A, Nielsen F, Breum NO. Evaluation of evaporation and concentration distribution models—A test chamber study. Annals of Occupational Hygiene, 41(6):625–641, 1997. https://doi.org/10.1016/S00034878(97)00032-X.

Copyright AIHA®. For personal use only. Do not distribute.

17

Modeling Occupational Inhalation Exposures to Chemicals Chapter 2: Information Gathering 4. Arnold FC, Engel AJ. Evaporation of pure liquids from open surfaces. In: Modelling of Environmental Chemical Exposure and Risk (Vol. 2, pp. 61–71), edited by Linders JB. NATO ASI Series, Springer Netherlands: Dordrecht, 2001. https://doi.org/10.1007/978-94-010-0884-6_6. 5. U.S. Environmental Protection Agency. AP-42, Compilation of Air Emissions Factors, 5th Edition. Volume I: Stationary and Point Sources. Office of Air Quality Planning and Standards: Research Triangle Park, NC, 1995. 6. Stefaniak AB, Bowers LN, Knepp AK, Luxton TP, Peloquin DM, Baumann EJ, Ham JE, Wells JR, Johnson AR, LeBouf RF, Su F-C, Martin SB, Virji MA. Particle and vapor emissions from vat polymerization desktop-scale 3-dimensional printers. Journal of Occupational and Environmental Hygiene, 16(8):519–531, 2019. https://doi.org/10.1080/15459624.2019.1612068. 7. Wadden RA, Baird DI, Franke JE, Scheff PA, Conroy LM. Ethanol emission factors for glazing during candy production. American Industrial Hygiene Association Journal, 55(4): 343–351, 1994. https://doi. org/10.1080/15428119491018998. 8. Conroy LM, Wadden RA, Scheff PA, Franke JE, Keil CB. Workplace emission factors for hexavalent chromium plating. Applied Occupational and Environmental Hygiene, 10(7):620–626, 1995. https://doi.org/1 0.1080/1047322X.1995.10387655. 9. Wadden RA, Hawkins JL, Scheff PA, Franke JE. Characterization of emission factors related to source activity for trichloroethylene degreasing and chrome plating processes. American Industrial Hygiene Association Journal, 52(9):349–356, 1991. https://doi.org/10.1080/15298669191364866. 10. Wadden RA, Scheff PA, Franke JE. Emission factors for trichloroethylene vapor degreasers. American Industrial Hygiene Association Journal, 50(9):496–500, 1989. https://doi.org/10.1080/00028894.1989.1041 1467. 11. Keil CB, Wadden RA, Scheff PA, Franke JE, Conroy LM. Determination of multiple source volatile organic compound emission factors in offset printing shops. Applied Occupational and Environmental Hygiene, 12(2):111–121, 1997. https://doi.org/10.1080/1047322X.1997.10389470. 12. Wadden RA, Scheff PA, Franke JE, Conroy LM, Javor M, Keil CB, Milz SA. VOC emission rates and emission factors for a sheetfed offset printing shop. American Industrial Hygiene Association Journal, 56(4):368–376, 1995. https://doi.org/10.1080/15428119591016999. 13. Jarand CW, Akapo SO, Swenson LJ, Kelman BJ. Diisocyanate emission from a paint product: a preliminary analysis. Applied Occupational and Environmental Hygiene, 17(7):491–494, 2002. https://doi. org/10.1080/10473220290035705. 14. Wingfors H, Svensson K, Hägglund L, Hedenstierna S, Magnusson R. Emission factors for gases and particle-bound substances produced by firing lead-free small-caliber ammunition. Journal of Occupational and Environmental Hygiene, 11(5):282–291, 2013. https://doi.org/10.1080/15459624.2013.858821. 15. Keil CB. The development and evaluation of an emission factor for a toluene parts-washing process. American Industrial Hygiene Association Journal, 59(1):14–19, 1998. https://doi. org/10.1080/15428119891010280. 16. Keane MJ, Siert A, Chen BT, Stone SG. Profiling mild steel welding processes to reduce fume emissions and costs in the workplace. Annals of Occupational Hygiene, 58(4):403–412, 2014. https://doi.org/10.1093/ annhyg/meu007. 17. Steinle P. Characterization of emissions from a desktop 3D printer and indoor air measurements in office settings. Journal of Occupational and Environmental Hygiene, 13(2):121–132, 2016. https://doi.org/10.1080/ 15459624.2015.1091957. 18. Arnold S, Ramachandran G, Kaup H, Servadio J. Estimating the time-varying generation rate of acetic acid from an all-purpose floor cleaner. Journal of Exposure Science & Environmental Epidemiology, 30(2):374– 382, 2019.

18

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 2: Information Gathering 19. Keil CB, Akbar-Khanzadeh F, Konecny KA. Characterizing formaldehyde emission rates in a gross anatomy laboratory. Applied Occupational and Environmental Hygiene, 16(10):967–972, 2001. https://doi. org/10.1080/104732201300367227. 20. Lippert JF, Lacey SE, Jones RM. Modeled occupational exposures to gas-phase medical laser-generated air contaminants. Journal of Occupational and Environmental Hygiene, 11(11):722–727, 2014. https://doi.org/1 0.1080/15459624.2014.916810. 21. Räisänen J, Niemelä R. On-line monitoring of solvent emission rates using an open path FTIR analyser. Annals of Occupational Hygiene, 46(5):501–506, 2002. https://doi.org/10.1093/annhyg/46.5.501. 22. Wang J, Hoang T, Floyd EL, Regens JL. Characterization of particulate fume and oxides emission from stainless steel plasma cutting. Annals of Work Exposures and Health, 61(3):311–320, 2017. https://doi. org/10.1093/annweh/wxw031. 23. Säämänen A, Skrifvars M. The effect of spraying and rolling process factors on styrene emission during the application of unsaturated polyester resins. AIHA Journal, 63(4):474–481, 2002. https://doi. org/10.1080/15428110208984736. 24. Welling I, Mielo T, Räisänen J, Hyvärinen M, Liukkonen T, Nurkka T, Lonka P, Rosenberg C, Peltonen Y, Svedberg U, Jäppinen P. Characterization and control of terpene emissions in Finnish sawmills. AIHA Journal, 62(2):172–175, 2001. https://doi.org/10.1080/15298660108984620. 25. Bémer D, Régnier R, Subra I, Sutter B, Lecler MT, Morele Y. Ultrafine particles emitted by flame and electric arc guns for thermal spraying of metals. Annals of Occupational Hygiene, 54(6):607–614, 2010. https://doi. org/10.1093/annhyg/meq052. 26. The Engineering ToolBox. Air Change Rates in Typical Rooms and Buildings https://www. engineeringtoolbox.com/air-change-rate-room-d_867.html (accessed August 15, 2019). 27. ASHRAE. ANSI/ASHRAE Standard 62.1-2016: Ventilation for Acceptable Indoor Air Quality. 2016. 28. U.S. EPA. Exposure Factors Handbook: 2011 Edition (Final Report) EPA/600/R-09/052F. U.S. Environmental Protection Agency, Washington, DC, 2011. 29. Jayjock M, Havics AA. Residential inter-zonal ventilation rates for exposure modeling. Journal of Occupational and Environmental Hygiene, 15(5):376–388, 2018. https://doi.org/10.1080/15459624.2018.14 38615. 30. Sparks, LE. IAQ Model for Windows, RISK Version 1.5. Research Triangle Park, NC, 1996. 31. Bartlett KH, Martinez M, Bert J. Modeling of occupant-generated CO2 dynamics in naturally ventilated classrooms. Journal of Occupational and Environmental Hygiene, 1(3):139–148, 2004. https://doi. org/10.1080/15459620490424393. 32. American Conference of Governmental Industrial Hygienists (ACGIH). Industrial Ventilation: A Manual of Recommended Practice for Design, 30th edition. ACGIH Signature Publications: Cincinnati, OH, 2019. 33. Keil CB. Experimental measurements of near-source exposure modeling parameters. Journal of Occupational and Environmental Hygiene, 12(10):692–698, 2015. https://doi.org/10.1080/15459624.2015.1 029619. 34. Sahmel J, Unice K, Scott P, Cowan D, Paustenbach D. The use of multizone models to estimate an airborne chemical contaminant generation and decay profile: Occupational exposures of hairdressers to vinyl chloride in hairspray during the 1960s and 1970s. Risk Analysis: An International Journal, 29(12):1699– 1725, 2009. https://doi.org/10.1111/j.1539-6924.2009.01311.x.

Copyright AIHA®. For personal use only. Do not distribute.

19

Modeling Occupational Inhalation Exposures to Chemicals Chapter 3: Volatile Liquid Spill

Chapter 3

Volatile Liquid Spill Chris Keil, PhD, CIH

Scenario You are working in the environmental health and safety (EHS) department in a medium-sized production facility. One morning a worker calls in sick. While on the phone, the worker tells their supervisor that they think a spill of liquid that occurred the previous day gave them the severe stomach cramps they were experiencing that morning. The worker knocked over a bottle of a chemical at their workstation, then wiped the spill up with some shop rags. At the time, they noticed the strong smell of the chemical, but it did not bother them then. The delayed nature of the symptoms makes you wonder about the connection between the exposure and the illness. A nontypical exposure occurred, and a follow-up investigation should be done. You want to arrive at an estimate of the worker’s exposure from the incident. This exposure estimate will both document what occurred and provide input as to whether the worker’s abdominal pain is related to the exposure. You set out to determine what concentration the worker was exposed to because of the spill described.

Information Gathering Information About the Chemical You check records in the EHS office and find a short report that indicates that 18 months ago a detector tube measurement was made in the room for toluene. The measurement was made during typical operating procedures. The concentration measured was 25 mg/m3. You then go to the worker’s workstation and talk to the area supervisor. It turns out that toluene is still used intermittently for parts cleaning at the station and there is always a 500-mL bottle of toluene on the table. You give the worker a call to get more information about the incident. From that conversation, you find that the spill took place about 5 hr into their 8-hr shift. The worker estimates that the bottle was about half-full at the time it was knocked over. Most of the liquid spilled onto the work surface. The worker wiped it up with some shop rags, then set them to the side to “dry out” about 2 m from their workstation. The shop rags were dry enough to use for another purpose about 2 hr later. The worker says they noticed the smell of toluene was stronger than usual, but it was not unbearable or concerning at the time. Some readily available information on toluene from the NIOSH Pocket Guide1 and the safety data sheet provided by the toluene vendor are shown in Table 3.1. Because this case involves symptoms, information on health effects is also included.

Copyright AIHA®. For personal use only. Do not distribute.

21

Modeling Occupational Inhalation Exposures to Chemicals Chapter 3: Volatile Liquid Spill Table 3.1: Information on Toluene

Physical Properties

Release Characteristic Health Effects

Exposure Guidelines

Description Molecular weight Liquid density Vapor pressure Lower explosive limit Volume Time Target organs Effects

PEL TWA-8 PEL C PEL peak REL TWA-8 REL STEL TLV TWA IDLH

Colorless liquid; sweet, pungent, benzene-like odor 92.1 g/mol 0.87 g/mL 21 mmHg 1.1% = 11,000 ppm 250 mL (half of the 500-mL container) All the liquid evaporated within 120 min Eyes, skin, respiratory system, central nervous system, liver, kidneys Irritation of the eyes and nose; confusion, euphoria, dizziness, headache, dilated pupils, lacrimation, anxiety, muscle fatigue, insomnia; paresthesia, dermatitis, liver or kidney damage 200 ppm = 750 mg/m3 300 ppm = 1,130 mg/m3 500 ppm = 1,880 mg/m3 100 ppm = 375 mg/m3 150 ppm = 560 mg/m3 20 ppm = 75 mg/m3 500 ppm = 1,880 mg/m3

C, ceiling limit; IDLH, immediately dangerous to life or health; PEL, permissible exposure limit; REL, recommended exposure limit; STEL, short-term exposure limit; TLV, threshold limit value; TWA-8, 8-hr time-weighted average.

Information About the Room The room where the spill took place has dimensions of approximately 10 m × 8 m, with a 3-m ceiling. This gives a room volume of approximately 240 m3. Air is supplied through a diffuser roughly in the center of the ceiling. Information on the supply air rate is not immediately available. There is no mechanical exhaust, and air leaves the room through open doors and other gaps in the room envelope. The room layout is illustrated in Figure 3.1.

Modeling Approach 1: Zero-Ventilation Model

Figure 3.1: Room layout illustrating relative position of workbench, rag drying location (basket), and ventilation (ceiling diffuser in center).

Because you do not have ventilation information, you decide to start with a zero-ventilation model. Zero-ventilation models usually overestimate the average concentration in a room. They are quick and easy models that can help with a screening-level decision to determine whether more information and/or more advanced modeling is needed to make a better judgment about exposures. The zero-ventilation model, Equation 3.1, assumes that the workspace is a sealed space and all of the pollutant is immediately released and mixed throughout the space. In reality, concentrations may be higher near the point

22

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 3: Volatile Liquid Spill of pollutant release. If the worker were close to the source, a different approach might be selected. For example, a virtual box could be drawn around the evaporating source and the worker in this smaller volume could be modeled. In this scenario, however, the worker is at a distance from the toluene evaporating from the rags, so we can use the room volume in the zero-ventilation model. M Czero vent = — V

(3.1)

Where Czero vent: concentration expressed as mass/volume (often in mg/m3) M: mass of pollutant released (often in mg) V: volume of the room expressed as length3 (often in m3) In this case, 250 mL of toluene evaporates into the 240 m3 room. The mass of the evaporating liquid can be calculated using the volume spilled and the liquid density of toluene along with a conversion factor. 250 mL 0.87 g 1,000 mg ———— · ——— · ————— = 217,500 mg mL g If all 217,500 mg of the liquid evaporated into the unventilated 240 m3 room, the average concentration throughout the room can be calculated with Equation 3.1. 217,500 mg 910 mg C = —————— = ———— 240 m3 m3

Interpreting the Modeled Concentration The average room concentration if all the toluene evaporated and thoroughly mixed with the air in the unventilated room would be approximately 910 mg/m3. This concentration is above the permissible exposure limit (PEL) 8-hr time-weighted average (TWA-8) of 750 mg/m3 but below the PEL ceiling (1,130 mg/m3) and peak (1,880 mg/m3). The spill occurred 5 hr into the shift, so the longest duration the exposure could have lasted was 3 hr. Three hours is longer than the relevant short-term exposure limits. Thus, comparing the concentration to the TWA-8 is more appropriate than comparison with the ceiling or peak PELs. If the worker’s toluene exposure for the last 3 hr of their shift were the zero-ventilation model calculated concentration of 910 mg/m3 and the other 5 hr were 25 mg/m3 that had been measured previously, the worker’s TWA-8 exposure could be calculated as follows: mg mg 25 —— · 5 hr + 910 —— · 3 hr m3 m3 mg TWA-8 = ———————————————— = 360 —— 8 hr m3 The result, 360 mg/m3, is below the PEL. If the PELs are the benchmark occupational exposure limits (OELs) for your organization, the exposure may not be of concern.

Copyright AIHA®. For personal use only. Do not distribute.

23

Modeling Occupational Inhalation Exposures to Chemicals Chapter 3: Volatile Liquid Spill However, if your organization uses the threshold limit value (TLV), this estimate of the TWA-8 exposure is about 5 times higher than the TLV. What additional considerations might factor into your judgment about the exposure? One point is that this zero-ventilation model is likely an overestimate. In actuality, the room is mechanically ventilated. Additionally, the toluene did not all evaporate immediately. Therefore, if you are concerned about exceeding the TLV with this simple model, you may want to consider a slightly more sophisticated model such as a well-mixed room model.

Modeling Approach 2: Well-Mixed Room Model At its most basic level, the well-mixed room model, assumes that air enters a room, passes through the space, mixes throughout the room in the process, and exits the room, carrying with it pollutant mass that was released into the room. This model requires an estimate of the ventilation rate as well as the mass release rate of the pollutant. At steady state, the concentration in a well-mixed room is modeled as: G Css = — Q

(3.2)

Where Css: steady-state concentration expressed as mass/volume (often in mg/m3) G: pollutant release rate expressed as mass/time (often in mg/min) Q: ventilation rate expressed as volume/time (often in m3/min) You can make some quick estimates of G and Q. The worker reported that the shop rags were dry after 2 hr. That means the 217,500 mg evaporated over 120 min, which is an average of about 1,815 mg/min (G). A typical air change rate in industrial buildings is often 3–4 air changes per hour (ACH). If we assume 3 ACH in the 240 m3 room, that means that the airflow is 720 m3/hr, or 12 m3/min (Q). Using these estimates of G and Q in the wellmixed room model, we can calculate the following: 1,815 mg min mg Css ————— · ——— = 150 —— min 12 m3 m3 This estimate is twice the TLV of 75 mg/m3. If the exposure for the other 5 hr of the shift were 25 mg/m3, the TWA-8 exposure would have been the following: mg mg 25 —— · 5 hr + 150 —— · 3 hr m3 m3 mg TWA-8 = ———————————————— = 72 —— 8 hr m3 If the assumptions about the rate of evaporation, the ventilation rate, and the concentration for the remaining 5 hr are considered reasonable, the exposure estimate was close to the TLV.

24

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 3: Volatile Liquid Spill

Conclusions From the modeling, it seems that the spill resulted in an exposure that likely did not exceed the PEL but may have been close to the TLV. TLV TWA-8s represent exposure levels to which most workers can be exposed daily without experiencing health effects. This case was a one-time exposure. You might want to do more planning and training on small spill response, but the health risks from the exposure are likely to be small. The effects that the worker is reporting do not particularly align with the acute effects normally associated with toluene. Followup on the health of the worker would be prudent. You have done your due diligence as an industrial hygienist by assessing and documenting the exposure.

References 1. National Institute for Occupational Safety and Health (NIOSH). NIOSH Pocket Guide to Chemical Hazards. Publication No. 2005-149. Department of Health and Human Services (DHHS), September 2007.

Copyright AIHA®. For personal use only. Do not distribute.

25

Modeling Occupational Inhalation Exposures to Chemicals Chapter 4: LEV for a Sanding Operation

Chapter 4

Local Exhaust Ventilation for a Sanding Operation Chris Keil, PhD, CIH

Scenario A woodshop department in a facility operates two shifts a day, six days a week. At one of the workstations, sanding is done with an electric hand tool on a downdraft table. In this facility, dust has not been reported as a problem previously. Business has been good, and more and more often workflow backs up at the sanding station. There is a need to do more sanding to keep up with the demand for the products. There is another worktable nearby that could be used for the same task, but it is not equipped with any local exhaust ventilation (LEV). Adding another downdraft table is a considerable expense and will take a bit more time. However, because business is good, another downdraft table could be put into the financial plans within a couple months. Until then, the shop manager asks if occasional use of that table for sanding would be alright. Doing so would improve the workflow and improve business even more to facilitate the purchase of another downdraft table. There are additional sanders available that could be used at this possible new sanding station. Figure 4.1 illustrates the general layout of the room. LEV ducting, hoods, fans, air cleaners, and stacks are not shown. The dark-topped table is the downdraft table. The middle of the three white-topped tables is the location proposed as an additional sanding station.

Figure 4.1: Layout of the room.

Would the use of a hand sander on the unventilated table for 2 hr/day produce unacceptable wood dust concentrations?

Information Gathering Information About the Substance The wood being sanded is a soft wood. Various types of pine are used. Other types of wood could potentially be used in the future. A cordless orbital hand sander is used for the sanding process. The tool does not have an integrated dust collector because it is used on a downdraft table. There are several ways to go about estimating the wood dust generation rate. You can look at published information on hand sander exposures. You can also use mass balance approaches to determine emission rates to use in concentration modeling. A review of the literature can provide surrogate data to help in your decision making. A 2018 study reports dust concentrations from hand sanders inside a test chamber.1 The study was done to compare the relative dustiness from four different hand sanders. The hand sanders had power cords, and all had local exhaust attached by flexible tubing. These sanders are not like the one in this shop, and your process is not in a test chamber.

Copyright AIHA®. For personal use only. Do not distribute.

27

Modeling Occupational Inhalation Exposures to Chemicals Chapter 4: LEV for a Sanding Operation Therefore, the direct utility of this study for your case is low. You can also locate a National Toxicology Program document that reports total dust concentrations from 0.42–8.01 mg/m3 for electrical orbital sanders without integrated dust collection bags.2 However, process conditions are not indicated in the report. Dust generated from the sanding operations are also dependent on the type of wood (e.g., pine wood, etc.) and sandpapers being used.3 One mass balance approach to determine emission rates is to estimate the amount of wood removed in the process of sanding each workpiece. The thickness of the wood removed multiplied by the area that was sanded gives the volume removed. The volume removed multiplied by the density of the wood provides an estimate of the mass removed as sawdust. This, coupled with process rate information, can provide a mass emission rate. In your shop, no one is able to provide an estimate of the thickness removed during sanding; therefore, you would need some additional information to use this approach. Another approach is to determine the mass of sawdust captured by the LEV system. Because dust is not reported to be a problem around the downdraft table, you assume the system captures nearly 100% of the wood dust generated. The filter on the downdraft table LEV system collects sawdust in a 55-gal (0.20 m3) drum. The drum is emptied approximately every 4 weeks. The density of pine sawdust varies, but a value of approximately 250 kg/m3 is about the middle of the reported ranges. You could also weigh a known volume sample of your sawdust to calculate the density. With information on the volume of sawdust produced, an estimate of the density of the sawdust, and information on the length of the workweek, the average rate of generation of sawdust can be calculated using the following equation: 0.20 m3 week day hr 250 kg 106 mg mg G = ————— · ——— · ——— · ———— · ———— · ———— = 2,170 —— 4 weeks 6 day 16 hr 60 min m3 kg min It is always prudent to determine whether values you model make physical sense for the particular scenario you are modeling. Is it reasonable that 2,170 mg (2.17 g) of pine wood could be removed from a part each minute? The density of dried white pine wood (not sawdust) is 0.35–0.50 g/cm3. Using the average of this range, calculate the volume of wood removed as follows: 2.17 g cm3 5.1 cm3 ———— · ———— = ———— min 0.425 g min Can we envision 5.1 cm3 of wood being removed from parts in a minute? If the sanding removed half a millimeter of wood (0.5 mm = 0.05 cm), that means the area sanded is: 5.1 cm3 102 cm2 ———— · ———— = ———— min 0.05 cm min With these assumptions, 102 cm2 of surface area would need to be sanded each minute to produce the calculated emission rate. That is equivalent to an area of 10 cm × 10 cm being sanded every minute. That does not seem unreasonable. Therefore, we can proceed with the calculated emission rate of 2,170 mg/min. Table 4.1 summarizes the information gathered on wood dust.

28

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 4: LEV for a Sanding Operation Table 4.1: Information on Wood Dust

Physical Properties

Health Effects

Exposure Guidelines

Description Density Safety hazard Target organs Effects PEL TWA-8 PNOR Wood dust (proposed)

Light brown to tan fibrous powder 250 (200–300) kg/m3 Wood dust poses explosion danger, no concentration identified Skin, eyes, respiratory system Dermatitis, allergic responses, mucous tissue irritation, decreased lung capacity, nasopharyngeal cancers 15 mg/m3 total 5 mg/m3 respirable 5 mg/m3 softwood 1 mg/m3 hardwood

TLV TWA Wood dust

0.5 mg/m3 red cedar 1.0 mg/m3 other species

PEL, permissible exposure limit; PNOR, particles not otherwise regulated; TLV, threshold limit value; TWA-8, 8-hr time-weighted average. An important part of modeling particle concentrations is the role of particle size in the transport of the pollutant from the source to the receptor. A study of hand sanding tools reported that the size distribution of sanding dust centered around 3 μm with about 95% of the particles being smaller than 10 μm.1 The density of an individual particle of wood will be different than bulk wood dust because bulk dust includes the airspaces between the particles. A typical density of oak, which this shop uses extensively, is 0.75 g/cm3. The settling velocity (Vs) of these particles (in m/sec) can be calculated as follows:4 Vs = 26.4 · 10–6 · ρ · D2

(4.1)

Where ρ: specific gravity of the particle D: particle diameter (μm) A 3-μm oak particle with a density of 0.75 g/cm3 has a settling velocity of 0.0002 m/sec. Because of the slow settling rate of these dust particles, we will assume that particle settling—although it may decrease airborne concentrations slightly—is small compared to the air change rate in the room. All particles will be assumed to stay aloft for this modeling approach.

Information About the Room The sanding process is in a room that measures 20 m wide, 26 m long, and 6 m high, for a volume of 3,120 m3. There are two entrances to the room, and neither has a door. There is some mechanically supplied air from drop ceiling diffusers. There are several workstations (including the downdraft table) that have fixed LEV, which provides air exhaust from the room. There are no air exhaust locations other than the LEV. The room is under negative pressure, which can be observed at each entryway by a slight inward flow of air. The building engineer tells you that the performance of each LEV system is monitored with a manometer indicating duct static pressure. The duct static pressure is correlated to flow rate by regular flow checks with a pitot tube and manometer. The total LEV exhausted through the various systems is 240 m3/min. This is equivalent to 14,400 m3/hr, or about 4.6 air changes per hour (ACH).

Copyright AIHA®. For personal use only. Do not distribute.

29

Modeling Occupational Inhalation Exposures to Chemicals Chapter 4: LEV for a Sanding Operation

Modeling Approach: Well-Mixed Room The well-mixed room model is not an exact representation of the concentration that will be generated in this scenario. There will likely be a higher concentration near the point of use of the hand sander, and the dust will disperse throughout the room. However, a well-mixed room model can give an initial approximation of an average room concentration. Experience tells us that when a pollutant releasing process begins, concentrations start low and then build up to higher concentrations. If the emission and ventilation rates are constant, the concentration will reach a steady state. The steady-state average concentration in a well-mixed space can be calculated as follows: G CSS = — Q

(4.2)

Where Css: concentration at steady state (mg/m3) G: generation rate (mg/min) Q: ventilation rate (m3/min) Using the emission rate and ventilation rate determined during the information-gathering stage, the steady-state concentration in the room while using the unventilated hand sander will be: 2,170 mg min mg CSS = ————— · ———— = 9.0 ——— min 240 m3 m3

Interpreting the Modeled Concentration The result, 9.0 mg/m3, is the average concentration in the room at steady state if the room is truly well mixed. This concentration is almost twice the OSHA permissible exposure limit (PEL) for respirable particulates not otherwise classified and the proposed softwood PEL. It is 9 times higher than the threshold limit value (TLV). There are some considerations to keep in mind while interpreting these modeled results. One is that the proposal was to use the sander without ventilation for 2 hr/day. The occupational exposure limits (OELS) are 8-hr timeweighted averages (TWA-8s). If the sander were used 2 hr/day and the rest of the day the concentration was essentially zero, the TWA exposure of the well-mixed room modeled concentration would be 2.2 mg/m3, which is still above the TLV. Additionally, while the room may be well mixed, it will not be perfectly mixed. The concentration near the operation of the sander will be higher. This further reinforces that unventilated use of the sander will create unacceptable exposures. However, the proposed sander is the same type that is currently being used on the downdraft table. This current sander model does not have any dust control. You have read that orbital sanders with fitted dust collection bags reduced exposures anywhere from 84% to 98%.2 The modeled steady-state concentrations could be lowered from 9.0 mg/m3 to somewhere in the range of 0.2–1.4 mg/m3. The upper end of that range is still above the OELs, but if used for 2 hr daily, the TWA might be acceptable. This is still assuming well-mixed conditions and does not take into account that concentrations near the sander operation will be higher.

30

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 4: LEV for a Sanding Operation The rate that the concentration builds up is governed by the air change rate in the room. If the air change rate in the room is high, the concentration will approach the steady-state concentration quickly. Conversely, if the air change rate is low, the build of concentrations will be slower. Figure 4.2 illustrates the effect of air change rate on the relative rate of concentration build up to a steady-state concentration. In the case we are currently considering, the air change rate is about 4.6 ACH, so the build up to a maximum concentration will be fairly rapid. The concentration build up in a well-mixed room with an initial concentration of zero is described by Equation 4.3:5 G Ct = — · 1 – e Q

–Q · t ——— V

(4.3)

Where t: time from the initiation of pollutant release (min) Ct: concentration at time t (mg/m3) V: room volume (m3) For example, in our scenario with a generation rate of 2,170 mg/min, a room ventilation rate of 240 m3/min, and a room volume of 3,120 m3, the concentration after 10 min will be as follows:

2,170 mg min Ct = ————— · ———— · min 240 m3

1–e

m3 – 240 —— · 10 min min —————————— 3,120 m3

mg = 4.85 —— m3

Equation 4.3 can be rearranged to calculate the time it takes to reach a specific concentration. Ct · Q 1 – ———— · V G t = ————————————— Q – ln

(4.4)

This can be used to calculate how long after the start of sanding the concentration will exceed the TLV of 1 mg/m3. The calculation is as follows: mg m3 1 —— · 240 —— m3 min – ln 1 – ————————— · 3,120 m3 mg 2,170 —— min t = —————————————————————— = 1.5 min m3 240 —— min

Copyright AIHA®. For personal use only. Do not distribute.

31

Modeling Occupational Inhalation Exposures to Chemicals Chapter 4: LEV for a Sanding Operation Figure 4.2 illustrates concentration buildup in rooms based on the number of air changes per hour (ACH). If the emission rate is the same, all air change rates will approach 100% of the maximum concentration, Css. The time it takes to reach that value is dependent on the ACH.

Conclusions Using the current model of sander on an unventilated table will produce unacceptable concentrations of wood dust. A temporary solution could be to look for an orbital sander model with integrated dust collection that also meets the performance requirements for the task. Recognizing that the well-mixed box model does not account for locally high concentrations near the source, measurements to confirm concentrations would be prudent if this option is considered. Ultimately, a new downdraft table is what is needed.

Figure 4.2: Relative rate of concentration buildup by air change rate. ACH, air changes per hour; Css, concentration at steady state.

References 1. Keller F-X, Chata F. Characterization of wood dust emission from hand-held woodworking machines. Journal of Occupational and Environmental Hygiene, 15(1):13–23, 2018. https://doi.org/10.1080/15459624. 2017.1368526. 2. Technology Planning and Management Corporation. Report on Carcinogens Background Document for Wood Dust. National Toxicology Program (NTP). Meeting of the NTP Board of Scientific Counselors Report on Carcinogens Subcommittee. Research Triangle Park, NC, December 13–14, 2000. 3. Sydor M, Mirski R, Stuper-Szablewska K, Rogoziński T. Efficiency of machine sanding of wood. Applied Sciences, 11(6):2860, 2021. https://doi.org/10.3390/app11062860. 4. Burton DJ. Industrial Ventilation Workbook, 3rd edition. IVE, Inc: Bountiful, UT, 1995. 5. Hewett P, Ganser GH. Models for nearly every occasion: Part I - One box models. Journal of Occupational and Environmental Hygiene, 14(1):49–57, 2017. https://doi.org/10.1080/15459624.2016.1213392.

32

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 5: Volatile Liquid Container

Chapter 5

Volatile Liquid Container Left Open Chris Keil, PhD, CIH and Kang Chen, PhD, CIE, CHMM

Scenario A worker opens a chemical storage cabinet after a long weekend. A wide-mouth container of a volatile liquid had been left open. There is still liquid in the container, and there is a very strong solvent odor in the cabinet. Figure 5.1 illustrates the open cabinet and jar. The worker caps the container but then exits the room because of the intensity of the odor, leaving the cabinet open. They report the incident to the health and safety team. The team gathers to respond to the incident and quickly collects some information and data. Questions that need to be answered include the following: • To what concentration was the worker exposed when they opened the cabinet? • What is the concentration in the room? • When will it be safe to enter the room again?

Figure 5.1: Container left open in cabinet (photo: Chris Keil).

Information Gathering Information About the Pollutant The chemical is identified as n-hexane. The 500-mL container was reported to be fairly full at the end of the last workweek. One team member gathers information on the chemical from the NIOSH Pocket Guide.1 Table 5.1 contains information on n-hexane.

Copyright AIHA®. For personal use only. Do not distribute.

33

Modeling Occupational Inhalation Exposures to Chemicals Chapter 5: Volatile Liquid Container Table 5.1: Information on n-Hexane

Physical Properties

Health Effects

Description Molecular weight Liquid density Vapor pressure Lower explosive limit Upper explosive limit Target organs Effects

Exposure Guidelines

PEL TWA-8 TLV TWA IDLH

Colorless liquid with a gasoline-like odor 86.2 0.66 g/mL 124 mmHg 1.1% = 11,000 ppm = 38,800 mg/m3 7.5% = 75,000 ppm = 264,000 mg/m3 Eyes, skin, respiratory system, central nervous system, peripheral nervous system Irritation of the eyes and nose; nausea, headache; peripheral neuropathy: numb extremities, muscle weakness; dizziness 1,760 mg/m3 176 mg/m3 3,880 mg/m3

IDLH, immediately dangerous to life or health; PEL, permissible exposure limit; TLV, threshold limit value; TWA-8, 8-hr time-weighted average. The cabinet has interior dimensions estimated to be 85 cm × 40 cm × 40 cm. This gives a cabinet volume of 136,000 cm3, or 0.136 m3.

Information About the Room The dimensions of the room are estimated as 20–25 ft long, 15–20 ft wide, and 8.5–10 ft high. A sketch of the room layout is shown in Figure 5.2. Using the estimates of the room dimensions, the volume can be calculated as length × width × height. The room volume could be as small as 2,550 ft3 (72 m3) using the low end of the dimension estimates or as large as 5,000 ft3 (142 m3) using the high end. Using the average of each dimension’s estimate range, the room volume would be 3,642 ft3 (or 103 m3), as calculated below.

Figure 5.2: Room layout.

1 m3 22.5 ft · 17.5 ft · 9.25 ft = 3,642 ft3 · ————— = 103 m3 35.31 ft3 The team knows that because chemicals are used in the room, it is kept under negative pressure. The negative pressure is achieved by an air return on one of the walls that is exhausted outside of the building. The exhaust port to the outside is a 6-inch diameter duct. The team looks up the results from the last time the airflow through the system’s fan was tested and finds that the velocity through the 6.0-inch (0.50-ft) diameter duct was measured as 591 ft/min. The volumetric airflow through the exhaust duct and thus the room can be calculated as follows: Q = v · A

34

(5.1)

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 5: Volatile Liquid Container Where Q: volumetric flow rate (volume/time) v: velocity through the duct (length/time) A: cross-sectional area of the duct (area) The cross-sectional area of the 0.50-ft diameter (0.25-ft radius) duct is calculated using Equation 5.2: (5.2)

A = π · r2 A = π · 0.252 = 0.196 ft2 591 ft 0.196 ft2 116 ft3 m3 3.28 m3 Q = ——— · ———— = ———— · ———— = ———— min min 35.31 ft3 min

Modeling the Concentration in the Cabinet: Saturation Vapor Pressure Concentration In this scenario, it will be useful to consider the phenomenon of the saturation vapor pressure concentration. When an open surface of a volatile liquid is in an unventilated space, there will be a net evaporation of the liquid until the air above the liquid is saturated with the chemical vapor. Air at normal temperatures and pressures can only hold so much of a vapor. Consider humidity in air. When relative humidity is 100%, it does not mean that 100% of the air is water vapor. It means that for the given temperature, the air is “full” of water vapor. The air is saturated with water vapor. The concept is the same for volatile liquids. If there is enough volatile chemical in a sealed system, the air above the liquid will become saturated with the vapors of the chemical. This situation can exist in the headspaces of sealed containers such as bottles of liquids, storage tanks, mixing vessels, airspaces above liquids flowing through pipes, and the like. Vapor pressure is very dependent on the temperature. When using this approach, be sure to use a vapor pressure value that is reported for temperatures as close to those in your scenario as possible. The saturation vapor pressure concentration of a chemical can be calculated with Equation 5.3: Pv C(ppm) = —— · 106 Patm

(5.3)

Where C: concentration in parts per million Pv: vapor pressure of the liquid Patm: atmospheric pressure above the liquid The vapor pressure and atmospheric pressure need to be in consistent units. For modeling, the concentration expression in mass per volume is often desired. The conversion from ppm to mg/m3 can be added to Equation 5.3 to get Equation 5.4: MW Pv C(mg/m3) = ——— · —— · 106 24.45 Patm Copyright AIHA®. For personal use only. Do not distribute.

(5.4)

35

Modeling Occupational Inhalation Exposures to Chemicals Chapter 5: Volatile Liquid Container For the case at hand, if enough of the n-hexane in the open container evaporated to saturate the air in the cabinet, the concentration in the cabinet when the worker opened the cabinet could be calculated with Equation 5.3. 124 mmHg —————— · 106 = 163,000 ppm 760 mmHg For mass/volume concentration units, we use Equation 5.4: 86.2 124 mmHg mg ——— · —————— · 106 = 575,000 —— 24.45 760 mmHg m3 The concentration that existed in the cabinet when it was opened was extremely high—well above occupational exposure limits (OELs). It would be prudent to have the worker’s health checked. If the worker’s exposure to this concentration were only 2 min and the concentration were zero for the 478 min of the rest of their 8-hr shift, their 8-hr time-weighted average (TWA-8) exposure would be 2,400 mg/m3. This exceeds the TWA-8 permissible exposure limit (PEL). Remember, a TWA-8 is intended to be protective of long-term exposure effects, not acute effects. Not only is the concentration well above all health guidelines, it is also a safety hazard. In volume ratio expressions of concentration, 10,000 ppm is equivalent to 1%. Thus, the saturation vapor pressure concentration of 163,000 ppm is equivalent to 16.3% and is above the upper explosive limit of 7.5%. As the vapor spread throughout the room, the concentration decreased and passed through the explosive range. Although it is likely that the concentration quickly decreased below the lower explosive limit (LEL), there was an explosion hazard as the concentrations passed through the explosive range. As an additional exercise, this initial pulse release could also be modeled using turbulent diffusion principles (refer to Chapter 6) to establish when the concentration dropped below the LEL. When considering the possibility of a saturation vapor pressure concentration existing, it is worthwhile to determine whether there is enough liquid available to evaporate to reach that concentration. For our example, in order to saturate the 0.136 m3 interior airspace of the cabinet, the mass of n-hexane that needs to evaporate can be calculated by multiplying the saturated mass concentration in the air by the volume of the air. 575,000 mg 0.136 m3 ——————— · ————— = 78,200 mg m3 To get that mass into terms of liquid volume, the density of the liquid is needed and the mass needs to be in the appropriate units. 78,200 mg g mL —————— · ————— · ———— = 118 mL 1,000 mg 0.66 g This volume of 118 mL is less than the amount of n-hexane previously reported in the jar. Thus, saturation vapor pressure concentrations in the cabinet are possible.

36

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 5: Volatile Liquid Container Generalizing this approach, the liquid volume of a chemical that needs to evaporate to reach saturation vapor pressure concentration can be calculated using Equation 5.5: Vair MW Pv Vliq = ——— · ——— · 106 · ————— 24.45 Patm 1,000 · ρ

(5.5)

Where Vliq: volume of liquid that needs to evaporate to saturate air volume (mL) Vair: volume of air above the liquid (m3) MW: molecular weight (g/mol) Pv: vapor pressure of the liquid (units consistent with Patm) Patm: atmospheric pressure above the liquid (units consistent with Pv) ρ: density of the liquid (g/mL)

Modeling Average Concentration in the Room: Zero-Ventilation Model If the interior volume of the cabinet (0.136 m3) is saturated with n-hexane, once the cabinet is opened, the n-hexane mass is released into the larger room airspace. The mass of n-hexane vapor in the cabinet is the concentration in the cabinet multiplied by the cabinet volume. 0.136 m3 575,000 mg ————— · —————— = 78,000 mg m3 As this mass is released into the room, the average room concentration can be calculated using the low, middle, and high estimates of the room volume. 78,000 mg mg —————— = 1,100 —— 72 m3 m3 78,000 mg mg —————— = 760 —— 103 m3 m3 78,000 mg mg —————— = 550 —— 142 m3 m3

This range of potential average room concentrations is below immediately dangerous to life or health (IDLH) levels and below the PEL TWA-8. However, these values are greater than the TLV TWA-8, perhaps up to 6 times higher. There are a couple of considerations to factor into the interpretation of these results as potential exposures. Regarding the intensity of the exposure, these values are average room concentrations. Shortly after opening

Copyright AIHA®. For personal use only. Do not distribute.

37

Modeling Occupational Inhalation Exposures to Chemicals Chapter 5: Volatile Liquid Container the cabinet, concentrations are likely higher closer to the cabinet and lower further away. Considering the time component of the exposure, recall that the OELs for n-hexane are TWA-8s. This exposure scenario would result in the short-term exposure of a worker. Also, the room is ventilated; concentrations would begin decreasing as a result of the ventilation.

Modeling the Decrease of Concentration in the Room: Well-Mixed Room Purging Model The worker left the cabinet open but had capped the bottle before they left the room. In this scenario, there will be no further release of n-hexane vapor into the room. When there is no additional release of pollutant into an airspace, the average concentration in the room will decrease according to Equation 5.6:2 , as shown below. Ct = C0 · e

–Q · t ——— V

(5.6)

Where C0: concentration at time = 0 Ct: concentration after time t Q: room volumetric airflow rate (vol/time) V: room volume (volume) t: time that has elapsed (time) For our scenario, the average concentration in the room after 30 min can be calculated using an estimate of room volume and initial concentration. The following is the calculation assuming the lowest room volume, which resulted in the highest initial concentration:

mg C30 = 1,100 —— · e m3

– 3.28 m3 30 min ———— · ———— · ———— min 72 m3

mg = 280 —— m3

Equation 5.6 can easily be plugged into a spreadsheet to calculate the concentration decrease over time. Figure 5.3 illustrates the concentration decay over a 1-hr period for all three of the potential room volumes and initial concentrations. Note that although the smaller room size results in a higher estimate of initial concentration, the smaller room volume results in a faster purging of the vapor. Within 50 min, the concentration estimates for all the possible room sizes are below the TLV TWA-8. Also, although the room average concentration is above the TLV following the initial vapor release and mixing, the 8-hr average concentration in the room would be approximately 50 mg/m3 (assuming no further release). Once the initial risk of acute exposure and explosion has passed, entry and activity in the room will be fairly low risk.

38

Figure 5.3: Concentration decay in the room. TLV, threshold limit value.

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 5: Volatile Liquid Container

Best Practice for Control of Fugitive Emissions From Volatile Liquid Containers in Laboratories There are multiple scenarios that can generate fugitive emissions of a volatile solvent in laboratories. Organic solvents used at high-performance liquid chromatography (HPLC) workstations, liquid transfer and distribution, and hazardous waste collection and storage are but a few of the processes that can release organic vapors in laboratories. The necessity of understanding the potential risk of fugitive emissions is widely neglected because laboratory technicians tend to consider that the release rate is low and the quantity evaporating into the air is negligible from fugitive emissions. Unfortunately, they fail to be aware of their extra-long working hours in the laboratory and the involvement of highly flammable/toxic materials during the experiments.3–6 Fugitive emissions, if not treated properly, can definitely bring significant fire, explosion, and occupational health concerns.7–9 Without good safety awareness and practice, it is commonplace for technicians to store solvent bottles temporarily in the fume hood or wrap the bottle cover with PVC tapes or cloths. Figure 5.4 illustrates some typical examples of this kind of poor safety practice. The result is degraded fume hood performance and an electrostatic accumulation, which can lead to high potential of fire, explosion, and occupational exposures as well as unnecessary energy consumption.10–12 Although local exhaust ventilation can be properly designed to capture and remove airborne contaminants, source control is considered the “golden rule” and most effective method for reduction of fugitive emissions.

Figure 5.4: Poor practice of fugitive emissions control in the laboratory (photo: Kang Chen).

Copyright AIHA®. For personal use only. Do not distribute.

39

Modeling Occupational Inhalation Exposures to Chemicals Chapter 5: Volatile Liquid Container So, what does good safety practice for solvent dispensing and storage look like? There are various specialized equipment and devices available that provide effective fugitive emissions control. For example, in HPLC solvent supply and waste collection systems, an air valve blocks solvent vapor release and ensures safe pressure equalization in the container. Furthermore, an integrated active carbon filter prevents any direct emission of the volatile solvent in the container and adsorbs fine particles from the ambient environment. Equipment such as this will minimize both occupational exposures and fire and explosion risk. A chamber test was conducted to evaluate the effectiveness of specialized equipment to control fugitive emissions from laboratory solvent storage practices. A mixture of methanol/water (volumetric ratio: 80/20), pure acetonitrile, and pure methanol was used as a test liquid. One-liter bottles containing the solvents were fitted with a safety cap. An open 1-L bottle of each solvent was used for a comparison. The volume of solvent and the air concentrations of solvent in the test chamber were checked regularly over a period of up to 28 days.13 The test chamber was operated under conditions based on DIN/EN/ISO 16000-9. The test results are summarized below, and more details can be explored in Figure 5.5, Figure 5.6, and Table 5.2. Key results include the following: • The initial density of the methanol/water mixture (0.855 g/cm3) in the bottle sealed with the safety cap kept constant throughout the entire period of testing, while there was an increase in the density of the mixture in the bottle without a safety cap (0.858 g/cm3). This phenomenon indicates that there has been a greater loss of the volatile solvent (methanol) than water from the mixture. • Significant reduction in volume was observed for both acetonitrile and methanol bottles without a safety cap after 28 days (9.3% and 12.2%, respectively), whereas the corresponding change in volume of the bottles with a safety cap was negligible (0.1%). • The concentrations of methanol and acetonitrile in the chamber were incredibly high (630–660 mg/m3 and 730–800 mg/m3, respectively) when the safety cap was not fitted, even with a continual air change rate of 0.5 air changes per hour (ACH). In contrast, the highest concentration of acetonitrile with a safety cap equipped only reached 5 mg/m3, which is equal to 15% of its recommended exposure limit.

Figure 5.5: Comparison of the change in volume of a methanol/water mixture.

40

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 5: Volatile Liquid Container

Figure 5.6: Comparison of the change in volume of pure acetonitrile and methanol.

To summarize, application of the safety cap plays a dominant role in control of fugitive emissions from volatile chemical containers. Such good practice is beneficial to occupational exposure control, indoor air quality enhancement, and experimental quality assurance.14,15 It is prudent that a volatile liquid container never be left open in any case. If this is not reasonably practical, a safety cap or an equivalent safety device is highly recommended. As shown in the example above, the fugitive emissions of hazardous chemicals could create a lethal environment in the test cabinet. The concentration of acetonitrile could reach 800 mg/m3 or higher, approaching its IDLH value of 840 mg/m3, which poses an unacceptable risk to laboratory technicians when they stand just in front of the safety cabinet and open it. What is worse, many safety cabinets for hazardous chemical storage are not ventilated because this may not be required by regulation and is energy consuming. This means that a much higher vapor concentration could be reached in the cabinet because of zero ventilation. The experiments given in this section can lead to a better understanding of the good practice of volatile liquid fugitive emissions control for industrial hygienists. Table 5.2: Comparison of the Test Chamber Concentration of Pure Acetonitrile and Methanol

Time (Day) 1 3 7

Test Chamber Concentration (mg/m3) Acetonitrile, Acetonitrile, Methanol, Without Safety Cap With Safety Cap Without Safety Cap 800 5 660 770 3 640 730 1 630

Copyright AIHA®. For personal use only. Do not distribute.

Methanol, With Safety Cap 2 1 1

41

Modeling Occupational Inhalation Exposures to Chemicals Chapter 5: Volatile Liquid Container

References 1. National Institute for Occupational Safety and Health. NIOSH Pocket Guide to Chemical Hazards. Publication No. 2005-149. Department of Health and Human Services (DHHS), September 2007. 2. Keil CB, Simmons CE, Anthony TR (Eds.). Mathematical Models for Estimating Occupational Exposure to Chemicals, 2nd edition. AIHA Press: Fairfax, VA, 2009. 3. ASHRAE. ASHRAE Guideline 10-2016: Interactions Affecting the Achievement of Acceptable Indoor Environments. ASHRAE Inc.: Atlanta, GA, 2016. 4. Chen K, Yang JL, Zhang HB, Zhang WJ. Low level noise analysis in laboratory fume hood. Journal of Chemical Health and Safety, 24(1):2–7, 2017. https://doi.org/10.1016/j.jchas.2016.03.002. 5. Klepeis N, Nelson W, Ott W, Robinson JP, Tsang AM, Switzer P, Behar JV, Hern SC, Engelmann WH. The National Human Activity Pattern Survey (NHAPS): a resource for assessing exposure to environmental pollutants. Journal of Exposure Science & Environmental Epidemiology, 11(3): 231–252, 2001. https://doi. org/10.1038/sj.jea.7500165. 6. Al horr Y, Arif M, Katafygiotou M, Mazroei A, Kaushik A, Elsarrag E. Impact of indoor environmental quality on occupant well-being and comfort: A review of the literature. International Journal of Sustainable Built Environment, 5(1):1–11, 2016. https://doi.org/10.1016/j.ijsbe.2016.03.006. 7. Dow Corning Corporation. Safe Handling of Silicon Hydride Containing Polysiloxanes. Midland, MI: Dow Corning Corporation, 2008. 8. CES Operating Safety Committee. Safe Handling of SiH Products. Belgium, Brussels: Centre Européen des Silicones (CES), 2003. 9. Chen K, Zhang HB, Zhang WJ. Process safety analysis in manufacturing process of silicone rubber compounding. Modern Chemical Industry, 38(9):188–191, 2018. 10. Greenley PL, Billings CE, DiBerardinis LJ, Edwards RW, Barkley WE. Containment testing of laboratory hoods in the as-used condition. Applied Occupational and Environmental Hygiene, 15(2):209–216, 2000. https://doi.org/10.1080/104732200301719. 11. Ahn K, Ellenbecker MJ, Woskie SR, DiBerardinis LJ. Effects of work practices and upper body movements on the performance of a laboratory fume hood. Journal of Chemical Health and Safety, 23(6):2–9, 2016. https://doi.org/10.1016/j.jchas.2015.10.022. 12. Chen K, Wang WZ, Zhang WJ. Investigation of influential factors on laboratory fume hood containment performance. Science and Technology for the Built Environment, 26(3):387–399, 2020. https://doi.org/10.10 80/23744731.2019.1637192. 13. Spark J. Testing the reduction of emissions when using SCAT SafetyCaps on solvent bottles in laboratories [Test Report 1500915-2]. SGS Institut Fresenius GmbH, October 2010. https://www.esslab.com/material/ S.C.A.T.%20Europe%20-%20Test%20Report%20SGS.pdf. 14. S.C.A.T. Europe GmbH. Test Report: Efficiency of S.C.A.T. SafetyCaps. https://www.scat-europe.com/media/ pdf/7a/b5/7e/S-C-A-T-Europe-Test-Report-SafetyCaps-EN.pdf (accessed July 8, 2021). 15. Chen K. Hazardous waste management practice and case study in pharmaceutical industry [Presentation]. Asia Hazardous Waste Treatment Congress (AHWTC). Shanghai, China: September 27, 2016.

42

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 6: Barrel Filling Indoors

Chapter 6

Barrel Filling Indoors Chris Keil, PhD, CIH

Scenario A worker reports an odd, strong, “fish-like” odor in a room where they were dropping materials off. They said the odor made them feel slightly nauseous, so they wanted to let you know. The room in question has an enclosed liquid degreasing process. The solvent n-methyl-2-pyrrolidone (NMP) is used in the degreaser. Around the degreaser the distinctive smell is absent. However, there is a waste container against the wall. NMP vapor may be released from the waste drum. How much and what air concentrations are being created near the drum?

Information Gathering Information About the Pollutant NMP is not listed in the NIOSH Pocket Guide, so you turn to PubChem1 and a safety data sheet to get the physical and chemical properties of NMP. NMP has a distinctive ammonia-like odor that is sometimes reported as “fishy.” The only occupational exposure limit (OEL) you can identify is an AIHA 8-hr time-weighted average (TWA-8) Workplace Environmental Exposure Level (WEEL) of 60 mg/m3 with a skin notation.2 A short-term exposure limit of 120 mg/m3 is listed without a time frame. In talking with the process operators, you find out that every 2–3 days, 10 L of used NMP is pumped out of the degreaser through a hose and into a waste drum and 10 L of new NMP is added to the unit. You learn that the process was performed earlier in the day. The addition of new NMP to the unit’s solvent reservoir is from a barrel that is rolled into place near the degreaser. A pump and hose transfer fresh NMP into the degreaser through tight fittings on the filling and emptying connections. The local exhaust ventilation (LEV) on the degreaser appears to be operating well when tested with smoke tubes; however, the waste barrel is not ventilated. You hypothesize that when the waste NMP is pumped to the waste barrel, vapor is released into the room. The waste NMP drum is normally sealed. During the liquid transfer, the plug of a 2-inch diameter filling port is unscrewed and the transfer line lowered into the barrel for filling. The workers say that usually the end of the hose does not touch the surface of the liquid in the barrel and that they can hear the pumped liquid splashing into the barrel. They tell you it is important to listen to the splashing because they can tell when the barrel is almost full by the sound it makes. The transfer of the 10 L of used NMP into the barrel occurs at a rate of 2 L/min, and the pumping operation lasts 5 min. At the end of the process, the hose is removed and the cap replaced onto the opening. The headspace of a normally sealed liquid-holding container can become saturated with the vapor phase of the chemical. As discussed in Chapter 5, the saturation vapor concentration for a chemical (in units of mg/m3) can be determined with Equation 6.1: MW Pv Csat(mg/m3) = ——— · ——— · 106 24.45 Patm

(6.1)

When a volume of liquid is being added to a container, the headspace air is displaced into the surrounding environment at the same rate the liquid is being added. Displaced headspace air carries vapor into the space

Copyright AIHA®. For personal use only. Do not distribute.

43

Modeling Occupational Inhalation Exposures to Chemicals Chapter 6: Barrel Filling Indoors around the container. If the vapor concentration in the displaced air is saturated, a general form of an equation to describe vapor emissions from container filling (in mg/min) is shown in Equation 6.2:3 MW Pv G (mg/min) = F · ——— · ——— · 103 24.45 Patm

(6.2)

Where F: fill rate in L/min If the hose transferring the liquid into the container is above the surface of the liquid in the container, splashing may occur. This can generate a mist of the volatile liquid, which could increase the mass emission rate. NMP has a molecular weight of 99.1 g/mol. Vapor pressure values for NMP at 25°C are reported to range from 0.24–0.46 mmHg. Using the upper end of the range, 0.46 mmHg, the saturation concentration of NMP vapors (from Equation 6.1) is as follows: 99.1 0.46 mmHg mg Csat, NMP = ——— · —————— · 106 = 2,450 —— 24.45 760 mmHg m3 With a fill rate of 2.0 L/min, the vapor emission rate from the process (using Equation 6.2) is as follows: 2L 99.1 0.46 mmHg 4.9 mg G = —— · ——— · —————— · 103 = ——— min 24.45 760 mmHg min

Information About the Room The room is about 12 m × 12 m with a 4-m ceiling, giving a room volume of 576 m3. The degreaser is positioned toward the middle of the room. The enclosed chamber of the degreaser is ventilated to the outdoors through an activated carbon filter. As shown in Figure 6.1, the barrel for the waste NMP is along the wall in a corner of the room with the degreasing process. The location of the barrel is away from traffic patterns. There are two air diffusers supplying clean air to the room. Measurements with a Balometer determine that the mechanically supplied airflow through each diffuser is 12 m3/min, for a total of 24 m3/min. The layout of the room is illustrated in Figure 6.1.

44

Figure 6.1: Room layout showing position of degreaser and waste barrel.

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 6: Barrel Filling Indoors The information gathered is summarized in Table 6.1. Table 6.1: Information Gathered on the Exposure Scenario

n-methyl-2-pyrrolidone Physical Properties

Release Characteristic Health Effects Exposure Guidelines Room

Description Molecular weight Liquid density Vapor pressure Lower explosive limit Rate Time Target organs Symptoms WEEL

Clear to light yellow liquid with an ammonia-like odor 99.1 g/mol 1.03 g/mL 0.34 mmHg 1.3% 13,000 ppm 4.9 mg/min 5 min Central nervous systems, mucous membranes Headache, eye irritation, potential reproductive effects 60 mg/m3 TWA-8, 120 mg/m3 short-term

Volume Mechanical ventilation

576 m3 24 m3/min

TWA-8, 8-hr time-weighted average; WEEL, Workplace Environmental Exposure Level.

Modeling Approach: Diffusion Model In a scenario such as this, diffusion modeling can be a good approach to modeling concentrations very close to the source. The pollutant release is in a location away from factors that can enhance air mixing, such as largescale air motion or worker or equipment movement and activity. A diffusion model will calculate a pollution concentration gradient and eliminate the discontinuities in concentrations that a two-zone approach would model between the near field and far field. In occupational environments, the microscale molecular diffusion of gas phase pollutants is dominated by the relatively larger scale turbulent random motion of parcels of air. Turbulent diffusion coefficients are two to four orders of magnitude greater than molecular diffusion coefficients. In Fickian molecular diffusion, the concentration gradient itself is the driving force for chemical mass movement. In turbulent diffusion, while a concentration gradient is created, the driving force of mass movement is the turbulent motion of air. Equation 6.3 describes the concentration gradients that turbulent diffusion will create around a source.4 G Cr,t = ———————— · Kg · π · DT · r

1 – erf

r ——————— √ 4 · DT · t

(6.3)

Where Cr,t: concentration (mg/m3) at a distance r (m) from source t (min) after release starts G: pollutant generation rate (mg/min) Kg: geometry factor for the shape of diffusion region Sphere = 4 Hemisphere = 2 Quarter sphere = 1 Copyright AIHA®. For personal use only. Do not distribute.

45

Modeling Occupational Inhalation Exposures to Chemicals Chapter 6: Barrel Filling Indoors DT: turbulent diffusion coefficient (m2/min) r: radial distance away from the source (m) t: time from the start of pollutant release (min) erf: the error function After the start of the pollutant release, concentrations build up until a steady-state concentration is reached. As time (t) increases, the error function approaches zero, the equation reduces, and the steady-state concentration (Cr,ss) can be calculated with Equation 6.4: G Cr,ss = ———————— Kg · π · DT · r

(6.4)

For decision making regarding acceptability of exposures, steady-state concentrations are useful because they represent a maximum concentration that can occur with that modeling approach. If a steady-state concentration is deemed an acceptable exposure, lower concentrations during build up to and decay from steady state should be acceptable as well.

Turbulent Diffusion Coefficient Values for DT have been determined experimentally in factories,5–7 residences,8 and chamber studies.9 Values measured in factory settings generally ranged from 0.1–1.0 m2/min with a central tendency of about 0.3 m2/min. Residential values of DT ranged from 0.06–0.80 m2/min. The chamber study results ranged from 0.07–0.60 m2/ min. Because concentrations are inversely related to the diffusion coefficient, conservative modeling for worker protection will use smaller values. A DT of 0.1 m2/min is at the low end of the range in factory, residential, and chamber studies. The central tendency of 0.3 m2/min will also be used in the model for this case.

Geometry Factor The geometry of the airspace into which the pollutant is released impacts the concentrations around a source. Mass spreading into the air in all directions will dilute more quickly than if all the mass is spreading in one direction. This is accounted for in the term Kg in Equations 6.3 and 6.4. If the pollutant is released into a completely open space, the molecules are free to move in all directions, creating a spherical concentration gradient. However, if the source is on the floor or against a wall, the mass will diffuse into a hemisphere. The mass spreads into one-half the space, increasing the concentration by a factor of two compared to spherical diffusion at the same distances. Similarly, release into a quarter sphere will increase the concentration by a factor of 4. This is accounted for with a geometry factor and assumes that floor or wall surfaces are nonabsorptive and reflect the pollutant molecules. In our current case, the barrel positioning against the wall and the top surface of the barrel and the nearby table suggest a roughly quarter spherical pattern, as shown in Figure 6.2, which will have a geometry factor of 1.

46

Figure 6.2: Illustration of quarter-spherical diffusion around the emission of vapor from the filling port.

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 6: Barrel Filling Indoors The parameters, whose selection is described previously, will be used to calculate concentrations. G = 4.9 mg/min Kg = 1 (quarter sphere) DT = 0.1 m2/min and 0.3 m2/min r = 0.25 m, 0.50 m, and 1.0 m At the end of this chapter is a short discussion of published values of DT. The values selected for this scenario are typical for similar types of industrial spaces. Equation 6.4, the steady-state concentration equation, will give the maximum concentration that will occur at the selected distances. Using the parameters above in Equation 6.4 gives the results shown in the steady-state concentration column of Table 6.2. Equation 6.3 can be used with a time set to 5 min to determine the concentration at three distances at the end of the filling process. Determining the average concentration at each location over the 5 min can be approached in a couple of ways. Integrating the equation with respect to time over the 5 min can give an exact solution. Integrating the error function is moderately complex. As an alternative to resorting to calculus, a spreadsheet can be set up to calculate Ct at times from 0–5 min. By setting the time steps very small (0.01 min = 0.6 sec), the concentration can be calculated for every 0.6 sec during the 5-min filling process. The average of the 0.6 sec concentrations over the 5 min can be averaged to approximate the 5-min average. The concentrations at 5 min, the average over the 5 min, and the potential steady-state concentration are shown in Table 6.2. The results are illustrated in Figure 6.3. Table 6.2: Concentrations (mg/m3) During the 5-Min Filling Process

Distance (m) 0.25 0.50 1.00

DT = 0.1 m2/min C5 Cavg 50.2 41.2 19.3 13.1 5.0 2.4

Css 62.5 31.2 15.6

C5 18.4 8.0 2.9

DT = 0.3 m2/min Cavg 16.4 6.4 1.9

Css 20.8 10.4 5.2

Concentration at 5 min (C5), 5-min average concentration (Cavg), and steady-state concentrations (Css; in mg/m3) at distances around the pollutant release for turbulent diffusion coefficient (DT) = 0.1 and 0.3 m2/min. The average value for each condition was determined using 0.01-min time steps in a spreadsheet, averaging the results for the 500 time steps. To check the uncertainty introduced by using a computational approach rather than an analytical solution, the concentrations were also calculated with a 0.001-min time step. The smaller time step changes the 5-min average concentration by only 0.1%. Once the process stops, the continued diffusion of the chemical will quickly facilitate concentration decrease. Removal of polluted air by mechanical ventilation will also decrease concentrations.

Interpreting the Modeled Concentrations The TWA-8 WEEL for NMP is 60 mg/m3. At 0.25 m (using the lower value for DT), the concentration reached at 5 min is 50.2 mg/m3, which is slightly below this OEL. Some organizations will strive for exposures that are a fraction of the OEL. This OEL is a TWA-8, and the exposure from this operation is short term. The process currently occurs once per day, at the most, for a duration of 5 min. After filling stops, the concentration at 0.25 m will decrease NMP as mass disperses throughout the rest of the room volume. Even if the 0.25 m steady-state concentration of 62.5 mg/m3 were the average concentration for 1 hr, the TWA-8 for a shift would be 7.8 mg/m3. Even this estimate is likely higher than any actual exposure would be for several reasons: • Workers are unlikely to be 0.25 m from the filling port for 60 min. • The DT for this modeled concentration is at the low end of reported values. • This TWA-8 assumes a steady-state concentration for 1 hr. In actuality, concentrations are increasing up to the end of the filling process and decrease thereafter. Copyright AIHA®. For personal use only. Do not distribute.

47

Modeling Occupational Inhalation Exposures to Chemicals Chapter 6: Barrel Filling Indoors

Figure 6.3: Concentrations around the barrel port during filling using a turbulent diffusion coefficient of 0.1 m2/min and 0.3 m2/min. DT, turbulent diffusion coefficient (m2/min); r, radial distance away from the source (m).

48

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 6: Barrel Filling Indoors Given these results, it seems reasonable to judge this NMP exposure as “acceptable.” The assessment should be documented and communicated as appropriate. Having gone through this exercise once, a spreadsheet set up to do these calculations will be ready and available for another process. This spreadsheet can also be used if this process changed parameters, such as filling rate or frequency, or switched to a different chemical.

Selecting Turbulent Diffusion Coefficients As early as 1981, S.A. Roach proposed equations including turbulent diffusion for modeling concentrations around a workplace source of air pollution.10 In that work, however, he does not detail how to select values for the turbulent diffusion coefficient when pollutant is being released into air with no overall directional velocity. Rather, in demonstrating the approach, Roach posits DT values of 0.001, 0.01, and 0.1 m2/sec (0.06–6.0 m2/min). In the same work, Roach reports experimentally measured DT around a ventilation hood positioned within a larger flow field in the same direction as the exhaust direction. DT ranged from 0.04–0.47 m2/min, with an average of 0.18 and median of 0.15 (n = 16). DT tended to decrease with increased velocity in the external flow field. A series of studies done at the University of Illinois at Chicago (UIC) developed indoor emission factors for sources and reported turbulent diffusion coefficients at those indoor environments.4–7 Cheng et al.8 measured turbulent diffusion in a living room and a family room in different houses in northern California. A general increase in DT with increasing air changes per hour (ACH) was observed. Shao et al.9 measured DT in an environmental chamber, also noting a general increase in DT with ACH. The measured values for turbulent diffusion are summarized in Table 6.3. Table 6.3: Turbulent Diffusion Values (m2/min) Reported in the Literature

Environment Laboratory10 Industrial trichloroethylene degreasing5 Industrial freon degreasing6

Range 0.04–0.47 0.097–0.722

Mean 0.18 0.44

Median 0.15 0.47

n 16 9

V = 967 m3

0.11–0.23

0.16

0.16

7

V = 1,324 m3 Q = 232 m3/min

Industrial chrome electroplating7

0.085–0.884

Residential living room8

0.114–0.413

Residential family room 9

Chamber

8

Room Info

0.118–0.410 0.060–0.583

0.35 0.27 0.31 0.21

0.28 0.27 0.22 0.16

9

Local air velocities 6.7–12 m/min V = 21,890 m3

5

Q = 2,080 m3/min Floor area = 39.5 m2

6

ACH = 0.17–1.25 Floor area = 26.4 m2

27

ACH = 0.19–5.40 V = 11.9 m3 ACH = 0.43–2.89

ACH, air changes per hour; Q, volumetric flow rate; V, volume of the space.

Copyright AIHA®. For personal use only. Do not distribute.

49

Modeling Occupational Inhalation Exposures to Chemicals Chapter 6: Barrel Filling Indoors

References 1. National Institutes of Health. PubChem. 1-Methyl-2-pyrrolidinone https://pubchem.ncbi.nlm.nih.gov/ compound/13387 (accessed October 6, 2021). 2. Occupational Alliance for Risk Science. OARS WEEL Table https://www.tera.org/OARS/PDF_documents/ OARS_WEEL_Table.pdf (accessed October 6, 2021). 3. Keil CB, Simmons CE, Anthony TR (Eds.). Mathematical Models for Estimating Occupational Exposure to Chemicals. AIHA Press: Fairfax, VA, 2009. 4. Wadden RA, Hawkins JL, Scheff PA, Franke JE. Characterization of emission factors related to source activity for trichloroethylene degreasing and chrome plating processes. American Industrial Hygiene Association Journal, 52(9):349–356, 1991. https://doi.org/10.1080/15298669191364866. 5. Wadden RA, Scheff PA, Franke JE. Emission factors for trichloroethylene vapor degreasers. American Industrial Hygiene Association Journal, 50(9):496–500, 1989. https://doi.org/10.1080/00028894.1989.1041 1467. 6. Scheff PA, Friedman RL, Franke JE, Conroy LM, Wadden RA. Source activity modeling of Freon® emissions from open-top vapor degreasers. Applied Occupational and Environmental Hygiene, 7(2):127–134, 1992. https://doi.org/10.1080/1047322X.1992.10388033. 7. Conroy LM, Wadden RA, Scheff PA, Franke JE, Keil CB. Workplace emission factors for hexavalent chromium plating. Applied Occupational and Environmental Hygiene, 10(7):620–626, 1995. https://doi.org/1 0.1080/1047322X.1995.10387655. 8. Cheng K, Acevedo-Bolton V, Jiang R, Klepeis N, Ott W, Fringer O, Hildeman L. Modeling exposure close to air pollution sources in naturally ventilated residences: association of turbulent diffusion coefficient with air change rate. Environmental Science and Technology, 45(9):4016–4022, 2011. 9. Shao Y, Ramachandran S, Arnold S, Ramachandran G. Turbulent eddy diffusion models in exposure assessment – determination of the eddy diffusion coefficient. Journal of Occupational and Environmental Hygiene, 14(3):195–206, 2017. https://doi.org/10.1080/15459624.2016.1238476. 10. Roach SA. On the role of turbulent diffusion in ventilation. Annals of Occupational Hygiene, 24(1):105–132, 1981. https://doi.org/10.1093/annhyg/24.1.105.

50

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 7: Exposures Near a TCE Degreaser

Chapter 7

Historical Exposures Near a Trichloroethylene Degreaser Chris Keil, PhD, CIH

Scenario In the 1990s, an open-topped trichloroethylene (TCE) vapor degreaser was used in a manufacturing facility. Workers who were in the facility at the time have developed adverse health effects, which suggests that employees were exposed to TCE in levels beyond occupational exposure limits. A legal case was subsequently triggered by the potentially exposed employees. An estimate of historic TCE concentrations around the vapor degreaser will be helpful for understanding whether the TCE exposure likely led to the adverse health effects. Each day workers performed multiple tasks in the department that involved either direct interaction with or work near the open-top degreaser. Vapor degreasing batches ranged from single large parts to small part batches. Workers used a bench a few meters from the degreaser to prep parts, load small part baskets, and perform minor assembly steps prior to and following degreasing. Sparse records suggest that the company did not conduct a systematic exposure assessment. Workers do not recall any air sampling occurring. The company provided documentation of three TCE samples that reported concentrations of 22 mg/m3, 25 mg/m3, and 40 mg/m3. The records fail to document if these three samples represent personal samples or area samples. Furthermore, the sample duration and location were unknown.

Information Gathering Information About the Process TCE was widely used as a metal degreaser throughout the 1990s, prior to mounting evidence of its carcinogenicity. Open-top vapor degreasers, as used in this case, were not uncommon at the time of the suspected exposure. Open-top vapor degreasers operate by heating a liquid solvent in a “sump” to create vapors. Cooling coils around the top of the degreaser create a cool zone where the solvent vapors condense and the liquid phase is recycled into the system. A vapor zone exists between the heated liquid surface and the top of the degreaser. Parts are cleaned by being lowered in a basket to the vapor zone where the solvent vapors condense on the parts and drip off, cleaning the parts by carrying away oil, grease, and particles. Figure 7.1 is a schematic representation of how a basic vapor degreaser works. The basket of parts can also be lowered into the heated liquid sump and then hung in the vapor zone. Some degreasers have an additional clean liquid solvent sump in the bottom, providing an additional immersive cleaning step. Spraying liquid Copyright AIHA®. For personal use only. Do not distribute.

Figure 7.1: Schematic drawing of a basic vapor degreaser.

51

Modeling Occupational Inhalation Exposures to Chemicals Chapter 7: Exposures Near a TCE Degreaser onto the parts while they hang in the vapor zone is also a potential step in the process. Careful movement of baskets and parts through the vapor layer is needed to allow the cooling coils to operate correctly to condense and control vapor emissions. Lateral exhaust hoods can be used near the lip of the degreaser as ventilation control of vapors. In this case, observations and measurements are not possible. The facility suspended the open-top TCE degreaser operations long before the referenced legal case. Interviews with employees indicate that the degreaser had a single heating sump and baskets of parts were hung in the vapor zone for approximately 10 min. Parts were not immersed in clean liquid solvent sump or sprayed with liquid solvent. The open surface area of the degreaser was approximately 0.65 m × 1.75 m (1.14 m2). The sump contained approximately 65 gal (246 L) of TCE. With current information technology, old manuals for equipment can sometimes be found as references for equipment that is no longer in place. Local exhaust ventilation was not installed on the degreaser.

Information About TCE TCE is a colorless, sweet-smelling, highly volatile liquid. It has historically been widely used as a metal part cleaner because of its excellent performance in removing oil, grease, fats, waxes, and tar.1 Table 7.1 summarizes some of the key information on TCE. Table 7.1: Information on Trichloroethylene

Physical Properties

Health Effects

Description Molecular weight Liquid density Vapor pressure Lower explosive limit Target organs Symptoms

Exposure Guidelines

PEL TWA-8 PEL C PEL peak REL TWA TLV TWA TLV STEL IDLH

Colorless liquid with a chloroform-like odor 131.4 1.46 g/mL 58 mmHg 8% = 429,600 mg/m3 Eyes, skin, respiratory system, heart, liver, kidneys, central nervous system Irritation of eyes and skin; headache, visual disturbance, weakness, exhaustion, dizziness, tremor, nausea, vomiting; dermatitis; cardiac arrhythmias, paresthesia; liver injury; potential occupational carcinogen 537 mg/m3 1,074 mg/m3 1,611 mg/m3 5-min maximum peak in 2 hr 135 mg/m3 10-hr TWA 54 mg/m3 134 mg/m3 5,370 mg/m3

C, ceiling limit; IDLH, immediately dangerous to life or health; PEL, permissible exposure limit; REL, recommended exposure limit; STEL, short-term exposure limit; TLV, threshold limit value; TWA-8, 8-hr time-weighted average. Wadden et al.2 published emission information on open-top TCE degreasers. The total TCE emission rate was described in terms of the open surface area of the degreaser as 2.9 g/min/m2. The open surface area of the degreaser in this case was 1.14 m2. An estimate of the emission rate of TCE can be calculated as follows: 2.9 g 1.14 m2 3.306 g 3,300 mg ———— · ———— = ———— ≈ ————— m2 min min min As always, it is wise to check whether the calculated values are physically reasonable. If 3.3 g/min of TCE vapor 52

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 7: Exposures Near a TCE Degreaser are being lost to the room air, how much liquid TCE does that represent over an 8-hr shift? 3.306 g mL 480 min 1,087 mL 1L ———— · ———— · ———— = ————— ≈ ——— min 1.46 g shift shift shift Losing 1 L per shift from a 246-L reservoir of liquid seems reasonable. Workers from the time period agree that adding an average of 1 L of TCE per shift seems “about right.”

Information About the Room The room dimensions where the degreasing was performed were 15 m × 21 m × 5 m, giving a volume of 1,575 m3. At the time of the exposures, air was supplied to the room by four diffusers that dropped from a duct that ran down the middle of the room. Old blueprints were found indicating that the design airflow through each diffuser was 650 ft3/min. This gives a total air supply rate (Q) of 2,600 ft3/min, which is equivalent to 73.6 m3/min. Figure 7.2 illustrates the room layout.

Figure 7.2: Layout of the degreaser and ventilation in the room.

Modeling Approach: NearField/Far-Field Model Because workers spent time both near and at a distance from the TCE vapor source, there were periods when they were in areas of higher concentrations close to the degreaser and times when they experienced lower concentrations. In scenarios such as this, it can be useful to model the concentrations in the room as a zone near to the source (a well-mixed “near field”), and the rest of the room as a well-mixed “far field.” This two-zone model (i.e., near field/far field) was conceptually developed by Hemeon in the 1950s.3 Nicas presented a form of this model in the 1990s, which has been used in a number of studies4–8 and was recently presented in a form that accounts for varying ventilation configurations and emission rates.9,10 Figure 7.3 conceptually illustrates the two-zone model. The large box represents a room with an airflow rate Q (volume/time, often in m3/min) entering and exiting the room. Within that room, there is a point source of air contaminant releasing the pollutant at a rate G (mass/volume, mg/min). Around the source is a zone of high pollutant concentration, designated as the near field. A portion of the air moving through the room moves in and out of the near field. This interzonal airflow between the two zones is designated β (volume/ time, m3/min). Pollutant mass emitted into the near field mixes with air entering the field (β). The pollutant mass from G and pollutant mass carried into the near field with the interzonal airflow are considered to be fairly well mixed in the near field, and the concentration in the field is

Figure 7.3: Conceptual illustration of the two-box model. β, interzonal airflow rate (volume/time); CN, concentration in the near field (mass/volume); CF, concentration in the far field (mass/volume); G, pollutant emission rate (mass/ time); Q, room ventilation rate (volume/time).

Copyright AIHA®. For personal use only. Do not distribute.

53

Modeling Occupational Inhalation Exposures to Chemicals Chapter 7: Exposures Near a TCE Degreaser CN (mass/volume, mg/m3). The rest of the room, the far field, is also considered to be relatively well mixed and has a pollutant concentration of CF (mass/volume, mg/m3). This is a result of the mass carried by β from the near field mixing with the room ventilation rate, Q. A full derivation of and expansion on the model can be found in the references cited. For the case under consideration, the steady-state forms of the general ventilation two-box model will be used. G CF,SS = — Q

(7.1)

G G CN,SS = — + — Q β

(7.2)

Where CF,SS: steady-state concentration in the far field (mg/m3) CN,SS: steady-state concentration in the near field (mg/m3) G: pollutant emission rate (mg/min) Q: room ventilation rate (m3/min) β: interzonal airflow rate (m3/min) As calculated earlier, the emission rate, G, was 3,300 mg/min. The room general ventilation rate, Q, was estimated to be 73.6 m3/min. Using Equation 7.1, the concentration in the far field was 45 mg/m3. However, workers also spent time in close proximity to the degreaser. What about the concentration in the near field? Calculating the concentration in the near field requires a value for the interzonal airflow rate, β. Conceptually, air is moving into the near field through half of the surface area of the field and out of the near field through half of its surface area. When the room air is generally “still” (that is, net movement of air in the vicinity of the pollutant source is negligible), the random omnidirectional speed of the air of the room describes the velocity of the air moving into and out of the near field. Illustrations of the concept of β often look like Figure 7.3, but such illustrations imply directional air movement in one side of the near-field volume and out the other. Rather, consider Figure 7.4, in which the near zone is defined as a cube surrounding the vapor degreaser 0.75 m above and to each side of the machine, projected out from the bottom of the chilling coils. Selecting the geometry of the near field is an area of continuing research, but for our example, the selected geometry is acceptable. Air is moving in and out of the near field, but with relatively still air, random air motion is moving air in and out all over the surface area of the near field. Imagine in Figure 7.4 that random air motion is moving air into the near field through the darker squares and out through the lighter squares. The surface area available for air movement is called the “free surface area” (FSA). Half of the FSA has an inward flux of air. The other half has an outward flux of air. The volumetric flow through the near field can be calculated as one-half of the FSA multiplied by the random airspeed in the room, as shown in Equation 7.3:

54

Figure 7.4: The near field around the source.

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 7: Exposures Near a TCE Degreaser 1 β = — · FSA · u 2

(7.3)

Where β: interzonal airflow rate (m3/min) FSA: free surface area of the near field (m2) u: random airspeed in the room (m/min) The FSA of the near field in this case consists of the following: • Two 1.15 m by 2.25 m sides (2 × 1.15 m × 2.25 m = 5.18 m2) • Two 1.15 m by 3.30 m sides (2 × 1.15 m × 3.30 m = 7.59 m2) • One 3.30 m by 2.25 m side (3.30 m × 2.25 m = 7.42 m2) • One 3.30 m by 2.25 m side with the cross section of the degreaser subtracted out (3.30 m × 2.25 m – 0.65 m × 1.75 m = 6.29 m2) This results in a total surface area of 26.5 m2. A published survey of indoor airspaces indicated that the median random airspeed in indoor spaces is 3.7 m/min.11 Using this value with the FSA calculated above, the interzonal airflow rate (β) in the current case is the following: 1 26.5 m2 3.7 m m3 β = — · ———— · ——— = 49.0 —— 2 min min This value for β can be used with the model parameters determined previously to calculate the steady-state concentration in the near field using Equation 7.2:

CN,SS

mg mg 3,300 —— 3,300 —— min min 112 mg = —————— + —————— = ————— m3 m3 m3 73.6 —— 49.0 —— min min

Interpreting the Modeled Concentration These results indicate that concentrations near the degreaser (112 mg/m3) were about 2.5 times higher than concentrations in the rest of the room (45 mg/m3). We know from intuition and experience that concentrations are higher near sources. The room layout shown in Figure 7.2 demonstrated that the degreaser is away from the wall and fairly near the source of general ventilation. Thus, it seems plausible that although the concentration near the degreaser was higher than the rest of the room, it was not extraordinarily higher. The modeled concentrations represent estimates of average concentrations within each zone and over time. Short-term concentrations could have been lower or higher. The near-field and far-field concentrations can now be used with worker activity patterns to estimate worker exposures. For example, consider a worker who generally spent her entire shift in the room with the degreaser. She worked from 8 a.m. to 4 p.m., which includes a 30-min lunch period and two 15-min breaks. There was

Copyright AIHA®. For personal use only. Do not distribute.

55

Modeling Occupational Inhalation Exposures to Chemicals Chapter 7: Exposures Near a TCE Degreaser essentially zero concentration of TCE in the lunch/break area. While in the room (420 min), she estimates she spent 20% of her time (84 min) near the degreaser. The remainder of the time (336 min), she was in the far field. This worker’s time-weighted average (TWA) exposure can be calculated to be 51 mg/m3, which is right about at the threshold limit value (TLV) TWA-8. mg mg mg 112 —— · 84 min + 45 —— · 336 min + 0 —— · 60 min m3 m3 m3 mg TWA = —————————————————————————————— = 51 —— 480 min m3 This exposure estimate is based on a number of estimated inputs: the emission rate, the room ventilation rate, the surface area of the near field, the random airspeed in the room, and the amount of time a worker spent in each location. With all the inputs clearly laid out, any parameter can be changed if new information arises to give a new exposure estimate. Spreadsheet programs are particularly useful for exploring the effect of changing input values. Table 7.2 illustrates what a spreadsheet might look like for this case. The first column identifies the variable. The second column is the value for the variable. The third column identifies the units for the variable. The fourth column would not be in the spreadsheet but is included for illustrative purposes and shows the source of the values in Column B, either input to the cell or calculated from input values. The formulas use programming common to many spreadsheet programs, such as those found in Microsoft products, Libre Office, Google Suite, etc.

56

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 7: Exposures Near a TCE Degreaser Table 7.2: Example Spreadsheet Approach to the Case

A 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

Variable Emission factor Degreaser opening width Degreaser opening length Degreaser opening area Emission rate Emission rate

B Value 2.9 0.65 1.75 1.14 3.298 3,300

C Units g/m2/min m m m2 g TCE/min mg TCE/min

D Source Input Input Input =B3*B4 =B2*B5 =B6*1000

Q Q

2,600 73.6

ft3/min m3/min

Input =B9/35.31

FSA μ beta

26.5 3.7 49.0

m2 m/min m3/min

Input (calculated externally) Input =0.5*B12*B13

Cfar Cnear

44.8 112

mg/m3 mg/m3

=B7/B10 =B7/B10+B7/B14

min in far min in near min in 0

336 84 60

min min min

Input Input Input

TWA

51.0

mg/m3

=(B16*B19+B17*B20+0*B21)/=sum(B19:B21)

Beta (β), interzonal airflow rate; μ, random airspeed in the room; Cfar, steady-state concentration in the far field; Cnear, steady-state concentration in the near field; FSA, free surface area of the near field; TCE, trichloroethylene; TWA, time-weighted average. With the rationale for the model set out and a spreadsheet established, one could easily do calculations to respond to different inputs. For example, someone might contend that the parameters used are not correct but actually the following are true: • The cooling coils did not operate effectively, resulting in 30% higher emissions. • Only three of the diffusers operated, reducing Q by 25%. • The random airspeed in the room was closer to 2.7 m/min. These values can be quickly entered into the spreadsheet to calculate a new estimate of the TWA, 89 mg/m3.

Copyright AIHA®. For personal use only. Do not distribute.

57

Modeling Occupational Inhalation Exposures to Chemicals Chapter 7: Exposures Near a TCE Degreaser

References 1. Agency for Toxic Substances and Disease Registry. Toxicological Profile for Trichloroethylene. U.S. Department of Health and Human Services, 2019. 2. Wadden RA, Scheff PA, Franke JE. Emission factors for trichloroethylene vapor degreasers. American Industrial Hygiene Association Journal, 50(9):496–500, 1989. https://doi.org/10.1080/00028894.1989.1041 1467. 3. Hemeon WCL, Burton DJ. Hemeon’s Plant & Process Ventilation, 3rd edition. Lewis Publishers: Boca Raton, FL, 1999. 4. Arnold SF, Shao Y, Ramachandran G. Evaluating well-mixed room and near-field–far-field model performance under highly controlled conditions. Journal of Occupational and Environmental Hygiene, 14(6):427–437, 2017. https://doi.org/10.1080/15459624.2017.1285492. 5. Arnold SF, Shao Y, Ramachandran G. Evaluating well-mixed room and near-field–far-field model performance under highly controlled conditions. Journal of Occupational and Environmental Hygiene, 14(6):427–437, 2017. https://doi.org/10.1080/15459624.2017.1285492. 6. Jones RM, Simmons CE, Boelter FW. Comparing two-zone models of dust exposure. Journal of Occupational and Environmental Hygiene, 8(9):513–519, 2011. https://doi.org/10.1080/15459624.2011.598 762. 7. Sahmel J, Unice K, Scott P, Cowan D, Paustenbach D. The use of multizone models to estimate an airborne chemical contaminant generation and decay profile: occupational exposures of hairdressers to vinyl chloride in hairspray during the 1960s and 1970s. Risk Analysis: An International Journal, 29(12):1699– 1725, 2009. https://doi.org/10.1111/j.1539-6924.2009.01311.x. 8. Nicas M, Neuhaus J. Predicting benzene vapor concentrations with a near field/far field model. Journal of Occupational and Environmental Hygiene, 5(9):599–608, 2008. https://doi. org/10.1080/15459620802282375. 9. Ganser GH, Hewett P. Models for nearly every occasion: Part II – Two box models. Journal of Occupational and Environmental Hygiene, 14(1):58–71, 2017. https://doi.org/10.1080/15459624.2016.1213393. 10. Ganser GH, Hewett P. Models for nearly every occasion: Part IV – Two-box decreasing emission models. Journal of Occupational and Environmental Hygiene, 14(11):919–930, 2017. https://doi.org/10.1080/154596 24.2017.1339167. 11. Keil C, Zhao Y. Interzonal airflow rates for use in near-field far-field workplace concentration modeling. Journal of Occupational and Environmental Hygiene, 14(10):793–800, 2017. https://doi.org/10.1080/154596 24.2017.1334903.

58

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 8: Pressurized Gas Release

Chapter 8

Pressurized Gas Release Chris Keil, PhD, CIH

Scenario A new vice president of operations at a research institution is touring the lab facilities. In one research lab, she notices that there are three compressed gas cylinders secured to a wall. Two of them have regulators attached. The executive turns and sternly addresses the hapless research technician, asking, “If that regulator were to fail, would you die?” The technician (young me) answers, “I don’t know, I guess I’d run.” This was not the answer the vice president was looking for. And while the manner of the inquiry might have been startling to the technician, it is a good question. Pressurized gasses are common in certain workplace settings. If there were a regulator or gas line failure or a valve were accidentally left open, what concentrations would result? The cylinders in question contain a chlorine-nitrogen mixture and are used as part of experiments being done in enclosed chambers, which are also in the room. Two cylinders have a regulator attached, and metal tubing is used to plumb the gas mixture to the experimental cabinets where the chlorine-nitrogen is consumed in the experimental process. There are valves so that one of the cylinders is in operation and another is ready to use when needed. The third cylinder is either waiting to be used or is empty and waiting to be replaced with a new cylinder. The worker usually spends their time at the worktable with an analytical instrument. They periodically check the conditions in the experimental cabinets. Hourly during the shift, the worker checks the pressure reading on the regulators on the cylinders. If the pressure in the cylinder being used falls below 250 psig, the worker closes the valve and opens the valve on the full cylinder waiting to be used. The empty cylinder is then switched out with the third cylinder, and a new cylinder is purchased to replace the empty one. This process repeats itself about every 3 days.

Information Gathering Information About the Chemical In this case, the potential source of the pollutant is a compressed gas cylinder containing 75 ppm chlorine (Cl2) in nitrogen. The 75 ppm is equivalent to 217 mg/m3 at normal temperature and pressure (NTP). When new, the cylinder contains the NTP equivalent of 300 ft3 (8,500 L) of the gas mixture pressurized at approximately 2,500 psig. If the regulator were to catastrophically fail on a new cylinder, what mass of Cl2 would be instantaneously released into the room? A full cylinder with 8,500 L (NTP) of gas mixture at a concentration of 217 mg/m3 (NTP) contains 1,850 mg of Cl2. 8,500 L m3 217 mg ———— · ———— · ———— = 1,850 mg 1,000 L m3 The project runs 8 hr/day. Cylinder changeover occurs about every 3 days after pressure goes from the initial level of 2,500 psig to 250 psig. Operational time between cylinder changes is as follows: 8 hr —— · 3 day = 24 hr day

Copyright AIHA®. For personal use only. Do not distribute.

59

Modeling Occupational Inhalation Exposures to Chemicals Chapter 8: Pressurized Gas Release Thus, 90% of the contents of a cylinder (7,650 L) are released over a total of 24 hr (1,440 min). This gives a normal volumetric use rate of the following: 90 Lreleased 320 L 8,500 Lcylinder —————— · —————— · ——— = ——— 100 Lcylinder 24 hr hr

320 L 1,000 mL hr 5,300 mL ——— · ————— · ———— = ————— hr L 60 min min

Information About the Room The laboratory space for this scenario is approximately 7.5 m long and 5.0 m wide with a ceiling height of 3.0 m, giving a volume of about 112.5 m3. The tanks are secured against one wall, and a work desk with an analytical instrument is on an adjacent wall between two doors. Three cabinets containing the research work are on another wall. The cabinets are large, so this may be a case where subtracting the volume they displace from the room volume would be important. Each experimental cabinet is 2.00 m × 0.84 m × 1.05 m = 1.76 m3. The three of them together occupy 5.29 m3, or 4.7% of the room volume. The room volume excluding the space displaced by the cabinets is 107.2 m3. There is one ceiling diffuser in the room and an air return grill in a wall. The layout of the room is illustrated in Figure 8.1.

Figure 8.1: Layout of laboratory with chlorine cylinders.

A Balometer is used to measure the air through the diffuser. The supply rate is 515 ft3/min. The flow rate out through the return grill is 530 ft3/min. The experimental cabinets have tight-fitting glass doors. They are exhausted outside the room at a rate of about 5 L/min. The total exhaust through the experimental cabinets (15 L/min ≈ 0.5 ft3/min) is low compared to the general ventilation of the room. Smoke tubes used at the doorways verify that the room is under negative pressure. Because the room is under negative pressure, the airflow rate (Q) for the room will be estimated as the airflow rate out of the return grill, 530 ft3/min (15 m3/min). This is an air change rate of 8.4 air changes per hour (ACH) using the volume of the room that subtracts the volume of the cabinets. Refer to Table 8.1 for information about this exposure scenario.

60

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 8: Pressurized Gas Release Table 8.1: Information Gathered on the Exposure Scenario

Chlorine Gas (Cl2) Physical Properties1

Health Effects1

Exposure Guidelines

Room Characteristics

Description Molecular weight Liquid density Vapor pressure Lower explosive limit Target organs Symptoms

PEL TWA-8 PEL C PEL peak REL TWA-8 REL C TLV TWA TLV STEL IDLH Volume Ventilation rate

Greenish-yellow gas with a pungent, suffocating odor 70.9 g/mol Gas phase at NTP (pure Cl2 can compress to liquid) 7,600 mmHg Nonexplosive Eyes, skin, respiratory system Burning of eyes, nose, mouth; lacrimation, rhinorrhea, cough, choking, substernal pain; nausea, vomiting; headache, dizziness; syncope; pulmonary edema; pneumonitis None 3 mg/m3 None None 1.45 mg/m3 [15 min] 1.45 mg/m3 3 mg/m3 [15 min] 30 mg/m3 112.5 m3, 107.2 m3 excluding occluded volume 15 m3/min, 8.0 ACH, 8.4 ACH excluding occluded volume

National Institutes of Health. PubChem. Chlorine.

1

ACH, air changes per hour; C, ceiling limit; IDLH, immediately dangerous to life or health; NTP, normal temperature and pressure; PEL, permissible exposure limit; STEL, short-term exposure limit; TLV, threshold limit value; TWA-8, 8-hr time-weighted average.

Modeling Approach 1: Zero-Ventilation Model If the entire contents of a full cylinder were instantaneously released into the room, the zero-ventilation model can be used to calculate the average concentration in the room at the time of release. Shown here again as Equation 8.1, the zero-ventilation model assumes that the workspace is a sealed box and all of the pollutant is released and mixes in the space.2 M Czero vent = — V

(8.1)

Where Czero vent: concentration in dimensions of mass/volume (often in mg/m3) M: mass of pollutant released rate in dimensions of mass (often in mg) V: volume of the room in dimensions of length3 (often in m3)

Copyright AIHA®. For personal use only. Do not distribute.

61

Modeling Occupational Inhalation Exposures to Chemicals Chapter 8: Pressurized Gas Release In the case of a full cylinder, 1,850 mg of Cl2 are released into the 107.2 m3 of open space in the room. The average concentration in the room can be calculated at the time of instantaneous release and mixing as follows: 1,850 mg mg Czero vent = ————— = 17.2 —— 107.2 m3 m3 This is the average concentration in the room at the point in time of the instantaneous release of all the contents of a full cylinder. Although the jet-type release that would be characteristic of an instantaneous release of highpressure gas would induce good mixing in the region of the release, the concentration would likely be higher closer to the point of release.

Interpreting the Modeled Concentration – 1 This concentration is much higher than any of the occupational exposure limits (OELs) and is a little over half of the immediately dangerous to life or health (IDLH) level. If a regulator failure occurred on a full cylinder, it would likely create a significant exposure. The likelihood of such an occurrence should be considered when contemplating controls that might be put in place.

Modeling Approach 2: Two-Box Model of a Slow Release An exposure scenario that may be more likely is a slow leak of the gas resulting from a problem with the cylinder changeover. Consider the case where the worker changes over cylinders as described above and records the starting pressure of the new cylinder at 2,500 psig, as is normal. While doing the pressure checks the next hour, the worker notes that the pressure in the active cylinder is 1,800 psig. The pressure dropped by nearly one-third in an hour—clearly there is a leak somewhere. The worker finds where a tube fitting was shaken loose during the cylinder changeover. They quickly tighten the fitting, stopping the release. But what were the concentrations of chlorine around the cylinder during the period of release? The pressure in the cylinder dropped 28% (2,500 psi to 1,800 psi). During the 60 min between checks, 28% of the full cylinder (8,500 L) contents were released from the cylinder; 28% of 8,500 L is 2,380 L. Recall that the experiment normally uses 320 L of the gas each hour. This means 2,060 L of the 2,380 L of gas released from the cylinder during that hour was released into the room. The chlorine mass release rate into the room can be calculated with the volumetric release rate and the concentration. 2,060 L m3 217 mg 7.45 mg ———— · ———— · ———— = ———— 60 min 1,000 L m3 min The cylinders are against the wall. A near-field/far-field approach (introduced in Chapter 7) can be used to model concentrations close to the leaky valve and for the rest of the room.3 A hemispherical near field with a radius of 1.0 m can be designated a near field around the release point at or near the top of the cylinder, as illustrated in Figure 8.2. The remainder of the space within the room would be considered the far field. 62

Figure 8.2: Near field around top of cylinder.

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 8: Pressurized Gas Release Looking at the layout of the room, the relative position of the supply air diffuser and return grill create a general path of air past the worker toward the corner of the room through the return grill. If the worker is usually at the worktable or by the experimental cabinets, they will be in the far field. With a relatively high air change rate in the room (8.6 ACH), steady-state concentrations will be approached fairly quickly. The steady-state concentration in the far field can be calculated with Equation 7.1: G 7.45 mg min 0.50 mg Cfar field, SS = — = ———— · ——— = ———— Q min 15 m3 m3 The steady-state concentration in the near field can be calculated using Equation 7.2: G G Cnear field, SS = — + — Q β To determine the steady-state near-field concentration, an estimate of the interzonal airflow rate (β) is needed. Refer to Chapter 7 for a full discussion on the interzonal airflow rate. This can be calculated from the random airspeed in the room and the free surface area (FSA) of the near field. A value of 3.7 m/min has been reported as a typical random airspeed indoors.4 The surface area of a hemisphere is 2πr2. Beta can be calculated with Equation 7.3: 1 β = — · FSA · u 2 1 m m3 β = — · 2 · π · (1 m)2 · 3.7 —— = 11.6 —— 2 min min This interzonal airflow rate can then be used with the room general ventilation rate and the mass release rate to calculate the near-field concentration. 7.45 mg min 7.45 mg min 1.14 mg Cnear, SS = ———— · ——— + ———— · ———— = ———— min 15 m3 min 11.6 m3 m3

Interpreting the Modeled Concentration – 2 The ceiling permissible exposure limit (PEL) and the 15-min short-term exposure limit (STEL) threshold limit value (TLV) for Cl2 are both 3 mg/m3. The modeled near-field steady-state concentration is about one-third of these OELs. The near-field concentration is also below both the ceiling recommended exposure limit (REL) and the 8-hr time-weighted average (TWA-8) TLV of 1.45 mg/m3. A leak as described in the scenario above does not appear to pose significant health risks. However, in the modeled scenario, only 28% of the contents of the cylinder leaked in 1 hr. If the rate of release were higher, the resulting concentrations would be proportionately higher. This would potentially result in exposures above the OEL, particularly if you use the ceiling REL of 1.45 mg/m3 as your guideline. Having gone through the calculations to model exposure with the posited conditions, it will be easy to model other conditions, especially if the calculations are done in a spreadsheet.

Copyright AIHA®. For personal use only. Do not distribute.

63

Modeling Occupational Inhalation Exposures to Chemicals Chapter 8: Pressurized Gas Release Health and safety procedures should be reviewed considering these modeling results. A simple practice such as implementing a soap water leak check after each cylinder change might be adequate. Installing pressure alarms or enclosing the cylinders and ventilating the enclosure are more resource-intensive options.

References 1. National Institutes of Health. PubChem. Chlorine. https://pubchem.ncbi.nlm.nih.gov/compound/24526 (accessed March 31, 2022). 2. Keil CB, Simmons CE, Anthony TR (Eds.). Mathematical Models for Estimating Occupational Exposure to Chemicals. AIHA Press: Fairfax, VA, 2009. 3. Nicas M. Estimating exposure intensity in an imperfectly mixed room. American Industrial Hygiene Association Journal, 57(6):542–550, 1996. https://doi.org/10.1080/15428119691014756. 4. Keil C, Zhao Y. Interzonal airflow rates for use in near-field far-field workplace concentration modeling. Journal of Occupational and Environmental Hygiene, 14(10):793–800, 2017. https://doi.org/10.1080/154596 24.2017.1334903.

64

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 9: Requirements for a Flavoring Process

Chapter 9

Ventilation Requirements for a Flavoring Process Chris Keil, PhD, CIH

Scenario A specialty snack food manufacturer is adding a new item to its product line. The product was developed in a test kitchen and has been approved for full-scale manufacturing. Engineers are working to develop the equipment and processes for mass production. One step of the production is to add a coating of flavoring to the small snack items. The process engineers plan to add the flavoring to the items in a heated and ventilated tumbler. The items will be loaded into the tumbler and heated to complete cooking. Then, the flavoring will be added as a slurry that is a mix of sugar, other flavors, ethanol, and water. The heated tumbler will both complete the cooking process and speed the evaporation of the liquid carrier from the slurry. The evaporation leaves the flavoring on the product. Figure 9.1 illustrates the tumbler. The design of the tumbler draws air from the workspace through the open face of the tumbler and exhausts the air through the rear of the unit. The engineer working on the process has done some test runs, and the product is turning out well. She says that face velocity into the tumbler seems to be controlling vapor emissions into the room very well. Little or no odor of ethanol is reported near the tumbler drier. However, the engineer realizes that because ethanol is part of the slurry, there may be a safety concern associated with the concentration of evaporated ethanol vapors in the exhaust duct. She asks members of the environmental health and safety (EHS) department if they think there might be an explosion hazard from the ethanol vapors. Measurements of the concentration in the duct can help determine this, but some initial time spent on modeling the potential duct concentrations will help the department understand the system and what concentrations might be expected in the duct.

Information Gathering Information About the Pollutant Ethanol is part of the liquid mix that carries the solid flavoring. A slurry is made from 50 lbs of liquid, which is a 45% ethanol and 55% water mixture, and 20 lbs of solid flavorings. The slurry is manually poured into the operating tumbler that contains 300 lbs of food item in small pieces. The process temperature is 300°F. The tumbling action spreads the slurry, coating all the food pieces. The heat evaporates the ethanol and water into vapor, which are exhausted out of the back of the tumbler. The liquid should be entirely evaporated from the product at the end of the coating

Copyright AIHA®. For personal use only. Do not distribute.

Figure 9.1: The heated, ventilated tumbler.

65

Modeling Occupational Inhalation Exposures to Chemicals Chapter 9: Requirements for a Flavoring Process process. The process engineer says that based on tests thus far, the liquid is completely evaporated after 20 min. Table 9.1: Information on Ethanol

Physical Properties

Room Characteristic

Description Molecular weight Liquid density Vapor pressure Lower explosive limit Upper explosive limit Mass Time

Clear, colorless liquid with a weak, ethereal, “wine-like” odor 46.1 g/mol 0.79 g/mL 44 mmHg 3.3% = 33,000 ppm 19% = 190,000 ppm 22.5 lbs (45% of the 50 lbs of liquid in the slurry) 20 min

The initial question in this case is related to explosion hazard. The lower explosive limit (LEL) for ethanol is 3.3%, which is a gas phase volume-to-volume ratio expression of concentration. “Percent” is “parts per hundred,” so 3.3% is equivalent to 33,000 ppm. Because we are working with gas volume-to-gas volume ratio concentration units, it will be useful to consider the emission rate in terms of moles and gas phase liters. The mass and number of moles of ethanol used in the production of one batch of the product can be calculated as follows: 50 lbliquid 45 lbethanol 453.6 g 10,206 gethanol ————— · —————— · ———— = ——————— batch 100 lbliquid lb batch

10,206 g mol 221 mol ————— · ———— = ————— batch 46.1 g batch If the evaporation of ethanol over the 20-min period is constant, the emission rate can be determined by dividing the amount of ethanol used and evaporated in a batch by 20 min. 10,206 g batch 510 g ————— · ———— = ——— batch 20 min min

221 mol batch 11.1 mol ————— · ———— = ————— batch 20 min min These emission rates are the emission rates if the evaporation is constant. If the evaporation rate is not constant, it is at least the average emission rate over the 20 min.

66

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 9: Requirements for a Flavoring Process For 300°F (422 K) and 1 ATM pressure, the volume of 11.1 moles of vapor can be calculated as follows: P · V = n · R · T

(9.1)

Where P: atmospheric pressure (ATM) V: volume (L) n: moles R: ideal gas constant (0.082057 L ATM mol–1 K–1) T: temperature (K)

n · R · T 11.1 mol 0.082057 L ATM 422 K V = ————— = ————— · ————————— · ———— · ———— = 384 L P mol K 1 ATM If the emission rate over 20 min is constant, 11.1 mol/min are evaporated. This is equivalent to 384 L/min at the process temperature.

Information About the Process The tumbler has a 20-inch diameter circular open face for loading. The tumbler is exhausted from the rear. The velocity in the duct exhausting the process is measured with a traverse using a pitot tube and manometer, giving a result of 3,850 ft/min. The duct is 6 inches (0.5 ft) in diameter, which has a cross-sectional area of 0.1964 ft2. The area of the duct and the velocity in the duct are used to calculate the duct volumetric airflow rate. Q=v · A

(9.2)

Where Q: volumetric flow rate (vol/time, e.g., ft3/min) v: airspeed in the duct (length/time, e.g., ft/min) A: cross-sectional area of the duct (length2, e.g., ft2) 3,850 ft 0.1964 ft2 756 ft3 Q = ———— · ————— = ———— min min Converting this to liters will be helpful as we calculate volume-to-volume concentrations (% or ppm): 756 ft3 28.32 L 21,410 L Q = ———— · ———— = ————— min ft3 min

Copyright AIHA®. For personal use only. Do not distribute.

67

Modeling Occupational Inhalation Exposures to Chemicals Chapter 9: Requirements for a Flavoring Process

Modeling Approach 1: Constant Emission Rate The airstream in the duct is 21,410 L/min. Of that airstream, 381 L/min are ethanol vapor. The action of the tumbler and the turbulence of the air entering the tumbler through the face then exhausting into the duct will mix the vapor and the ventilation air very well. The volume fraction of ethanol vapor in the duct is calculated as follows: 381 L/min —————— = 0.018 = 1.8% 21,410 L/min The result of 1.8% is more than half of the ethanol LEL of 3.3%. So, if the emission rate is constant, the concentration in the duct is likely to be over 50% of the LEL. This is well over 10% of the LEL and enough to be concerned about.

Modeling Approach 2: Exponentially Decreasing Emission Rate However, evaporation from surface coating often starts fast and then slows down. This can be approximated as an exponentially decreasing emission rate. Exponentially decreasing emission rates can be described by Equation 9.3:, shown below.1 Gt = α · M0 ∙ e–α

(9.3)

· t

Where M0: initial mass at time 0 (e.g., mg, g, mol) Gt: emission rate at time t (e.g., mg/min, g/sec, mol/min) α: a constant (e.g., sec–1, min–1) t: time elapsed (e.g., sec, min) The constant for this type of evaporation process is generally determined experimentally. In this current case, we get some information in that the process engineer estimates that “all” of the liquid is evaporated after 20 min. “All” is a bit of a vague term. Is there 1% of the liquid left? 5%? 0.1%? We can explore each of these possibilities. With an exponentially decreasing emission rate, the unevaporated liquid mass remaining can be described as follows: Mt = M0 ∙ e–α

· t

(9.4)

Where M0: initial mass at time 0 Mt: mass remaining at time t α: a constant t: time elapsed

68

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 9: Requirements for a Flavoring Process Taking the logarithm of both sides of the equation and rearranging the terms results in the following equation: Mt —— M0 α = —————— t –ln

(9.5)

If after 20 min, 99% of the liquid evaporated and only 1% of the liquid is left, the calculation is as follows: –ln(0.01) α = ————— = 0.23 min–1 20 min By the same process, if “all evaporated” means there is 5% of the ethanol left, α = 0.15 min–1. If “all evaporated” means there is 0.1% of the ethanol left, α = 0.34 min–1 The evaporation rate constant α can be used with Equation 9.3 and the initial amount of ethanol, 221 moles, to get the emission rate over the 20-min period. Figure 9.2 shows the emission rate over time. The different lines on the graphs are for the different values for α depending on what “all” evaporated means. The spreadsheet used to do the calculations to produce Figure 9.2 is illustrated at the end of the chapter.

90.0

1%leftafter20min

80.0

70.0

Emissionrate(mole/min)

5%leftafter20min 60.0

50.0

0.1%leftafter20min

40.0

30.0

20.0

10.0

0.0 0

2

4

6

8

10

12

14

16

18

Time(min)

Figure 9.2: Decreasing emission rates of ethanol.

Copyright AIHA®. For personal use only. Do not distribute.

69

Modeling Occupational Inhalation Exposures to Chemicals Chapter 9: Requirements for a Flavoring Process

14.0

12.0

DuctConcentration(%)

10.0

8.0

6.0 LEL 4.0

2.0

0.0 0

2

4

6

8

10

12

14

16

18

Time(min)

Figure 9.3: Concentration in duct over 20 min. LEL, lower explosive limit.

The ethanol emission rate in moles per minute can be converted to liters per minute at the operating temperature using Equation 9.1. The emission rate in liters of ethanol per minute can be used with the total volumetric flow rate through the duct (21,410 L/min) to get the fraction and percentage of ethanol in the exhaust airstream at various times during the 20-min period. The concentration over time is shown in Figure 9.3. The spreadsheet calculations done to produce the figure are illustrated at the end of the chapter. One can see that regardless of whether “all evaporated” means 5% left, 1% left, or 0.1 % left, the concentration in the duct is above the LEL for the first 3–4 min. The process engineer and EHS department should work together to address this serious safety concern. Some options and their limitations can be brainstormed. • Increasing the airflow through the system to dilute the concentration below the LEL. – Increasing the airflow could also increase the evaporation rate. Changing the batch time might also change the quality of the product. • Changing the slurry coating formulation to use less ethanol and more water. – This would decrease the evaporation rate and again has unknown effects on the product quality. • Decrease the heating temperature to slow the evaporation rate. – Again, this is an option with unknown impact on the product.

70

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 9: Requirements for a Flavoring Process In any case, a critical safety hazard has been identified. It can be verified with air sampling in the exhaust stream. This operation should cease until the production team and the facilities team find a combination of formulation and ventilation to keep the vapor concentration in the duct below 10% of the LEL.

Modeling Approach 3: Control of Ethanol Emissions Into the Room There is another aspect of this process that is worth checking on as well. The engineer states that she thinks the airflow into the system through the open face of the tumbler is doing a good job containing vapors and minimizing ethanol concentrations in the room. This approach to emissions control is related to the idea in ventilation hood design of “capture velocity.” Capture velocity is an old notion but can still be useful for quick hood design evaluation. The idea is that a high enough velocity of air at the point of pollutant release is needed to overcome forces associated with the release (buoyancy, release at a speed, etc.) and air currents in the room to draw the pollutant into the ventilation system. For the system we are considering, a sufficient air velocity into the tumbler across its open face needs to be great enough to prevent vapors escaping into the room. The ACGIH industrial ventilation manual recommends a capture velocity of 200–500 ft/min for pollutants that are actively generated into a zone of rapid air motion.2 The heated evaporation of ethanol in a tumbler seems like it would meet that description. If this is the desired velocity at the tumbler face, what is the inward air velocity across the face of the tumbler? Again, this can be measured depending on the availability of an anemometer, but some quick calculations with available information can provide initial insight. We know from previous information that 756 ft3/min of air is exiting the tumbler at the process temperature of 300°F (149°C, 422 K). The volumetric flow rate of room air into the tumbler will be lower because the air is at a lower temperature. The ideal gas law tells us that if the pressure is the same, the volume of a parcel of air at a second temperature is proportional to the ratio of the second absolute temperature over the first temperature. If the room air is at 20°C (293 K), the volumetric airflow of room air into the tumbler can be calculated using Equation 9.1 on the same molar volume of air at two temperature conditions: T2 756 ft3 293 K 525 ft3 Q2 = Q1 · —— = ———— · ——— = ———— T1 min 422 K min The 20-inch diameter opening (radius: 10 inches, 0.83 ft) has a face area of 2.18 ft2. The velocity of air moving across the face of the tumbler is as follows: 525 ft3 241 ft ———— · ———— = ———— min 2.18 ft2 min The face velocity of 241 ft/min is toward the low range of the suggested capture velocity. There may be vapors from evaporation in the tumbler escaping into the room. Ethanol vapors have a relatively low toxicity. The permissible exposure limit (PEL), recommended exposure limit (REL), and threshold limit value (TLV) 8-hr timeweighted average (TWA-8) are 1,000 ppm (1,880 mg/m3). Even so, after any changes are made to address the potentially explosive concentrations in the duct, this issue of ethanol vapors escaping into the room should be revisited with modeling and measurement.

References 1. Keil C, Nicas M. Predicting room vapor concentrations due to spills of organic solvents. AIHA Journal, 64(4):445–454, 2003. https://doi.org/10.1080/15428110308984838. 2. ACGIH. Industrial Ventilation: A Manual of Recommended Practice for Design, 29th edition. ACGIH, Cincinnati, OH, 2016.

Copyright AIHA®. For personal use only. Do not distribute.

71

Modeling Occupational Inhalation Exposures to Chemicals Chapter 9: Requirements for a Flavoring Process

Appendix Figure 9.4 is an example of the spreadsheet used to do the emission rate and concentration calculations.

A

B

t (min) 0 0.1 0.2 0.3 0.4 0.5 0.6

α=0.23 51.1 49.9 48.8 47.7 46.6 45.5 44.5

1 2 3 4 5 6 7 8 9

C G (mol/min) α=0.15 33.2 32.7 32.2 31.8 31.3 30.8 30.4

D

E

α=0.34 76.6 74.0 71.5 69.1 66.7 64.5 62.3

α=0.23 1779 1738 1699 1660 1622 1585 1549

F G (L/min) α=0.15 1157 1140 1123 1106 1090 1074 1058

G

H

α=0.34 2668 2577 2490 2405 2324 2245 2169

α=0.23 8.3 8.1 7.9 7.8 7.6 7.4 7.2

I C (%) α=0.15 5.4 5.3 5.2 5.2 5.1 5.0 4.9

J

α=0.34 12.5 12.0 11.6 11.2 10.9 10.5 10.1

Figure 9.4: Screenshot of spreadsheet results of the model.

Figure 9.5 is an example of the spreadsheet used to do the emission rate and concentration calculations showing the formulas used. An explanation of the approach is provided below the table .

A

B

1

t (min) 3 0 2

α=0.23

C G (mol/min) α=0.15

D

E

α=0.34

α=0.23

F G (L/min) α=0.15

G

H

α=0.34

α=0.23

I C (%) α=0.15

J

α=0.34

=0.23*221* =0.15*221* =0.34*221* =b3*0.08206* =c3*0.08206* =d3*0.08206* =e3/21410* exp(-0.23*A3) exp(-0.15*A3) exp(-0.34*A3) 422 422 422 100

=f3/21410* =g3/21410* 100 100

4

0.1

=0.23*221* =0.15*221* =0.34*221* =b4*0.08206* =c4*0.08206* =d4*0.08206* =e4/21410* exp(-0.23*A4) exp(-0.15*A4) exp(-0.34*A4) 422 422 422 100

=f4/21410* =g4/21410* 100 100

5

0.2

=0.23*221* =0.15*221* =0.34*221* =b5*0.08206* =c5*0.08206* =d5*0.08206* =e5/21410* exp(-0.23*A5) exp(-0.15*A5) exp(-0.34*A5) 422 422 422 100

=f5/21410* =g5/21410* 100 100

Figure 9.5: Formulas used in spreadsheet to calculate model results.

72

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 9: Requirements for a Flavoring Process

Columns B, C, and D These columns are programmed to calculate the emission rate in moles per minute for the three different estimates of α using Equation 9.3: Gt = α · M0 ∙ e–α

· t

The 221 is the moles per batch calculated previously. The α values for the column, shown in Row 2 of the sheet, are inserted in the cells for the calculation.

Columns E, F, and G These columns take the emission rate in moles per minute from Columns B, C, and D and convert them to liters per minute. The cells calculate liters from moles using the ideal gas law in Equation 9.1 rearranged to the following: n · R · T V = ————— P The atmospheric pressure is 1 ATM, so the cell is programmed to refer to the moles per minute column corresponding to the particular α and multiply by the ideal gas law constant and the temperature in K. Columns H, I, and J take the liters of ethanol vapor generated per minute and divide by the total volumetric flow rate to get the fraction of ethanol in the air mix. This is multiplied by 100 to get the percentage of ethanol concentration.

Copyright AIHA®. For personal use only. Do not distribute.

73

Modeling Occupational Inhalation Exposures to Chemicals Chapter 10: Cleaning With Acetic Acid

Chapter 10

Cleaning With Product Containing Acetic Acid Chris Keil, PhD, CIH

Scenario The environmental health and safety (EHS) department for a medium-sized manufacturing facility is in the process of implementing a comprehensive exposure management system. The team has developed similar exposure groups and prioritized information-gathering activities. For several exposure profiles, the team deemed the exposures “acceptable” based on their professional judgment. Their organization considers a long-term average exposure that is 10% of the occupational exposure limit (OEL) as acceptable. However, some of the team members are aware of literature that indicates that professional judgment is often flawed.1,2 They decide to apply some modeling techniques to provide documentation for their “acceptable” judgment of exposures. One of the tasks initially judged to produce an acceptable level of exposure is a parts cleaning process. A panel component of the final product moves down a production line where one side of the panel is cleaned. The side of the panel that is cleaned has a surface area of about 1.5 m2. A worker uses a hand spray bottle to apply a liquid cleaner to the panel. The cleaner is then wiped off the panel using a paper towel. The paper towels are reused for a few panels, then placed in either a trash can or on the nearby bench. Figure 10.1 illustrates the process. Workflow moves from left to right (foreground to background) in the figure. The worker starts their shift at 8:00 a.m., has a 15min break at 10:00 a.m., a 30-min break at 12:00 p.m., and a 15-min break at 2:30 p.m. The workday ends at 4:00 p.m. During the breaks, the worker is away from the workstation and has zero exposure to the liquid cleaner.

Figure 10.1: Diagram of panel cleaning process.

Information Gathering Information About the Chemical The cleaning liquid is a 4% acetic acid solution in water. Table 10.1 summarizes the key properties of neat (100%) acetic acid and a 4% acetic acid solution. This information was retrieved from PubChem, which is a valuable repository of information on chemicals.4 When looking for information on chemical properties, it is wise to check a few sources and compare them against one another. For properties that are sensitive to temperature, such as vapor pressure and solubility, be particularly mindful that the values you obtain and use are appropriate for the conditions you are modeling.

Copyright AIHA®. For personal use only. Do not distribute.

75

Modeling Occupational Inhalation Exposures to Chemicals Chapter 10: Cleaning With Acetic Acid Table 10.1: Information on Acetic Acid (CAS No. 64-19-7) and 4% Aqueous Solution

Physical Properties

Description

Neat acetic acid Colorless liquid or crystals with a sour, vinegar-like odor

Molecular weight Liquid density

60.05 g/mol 1.051 g/mL (20ºC)

4% acetic acid solution in water Description Colorless liquid or crystals with a sour, vinegar-like odor Liquid density

≈ 1.002 g/mL both

Vapor Pressure

temperatures, close to that of water 0.44 mmHg (20°C)a

1.045 g/mL (25ºC)

Health Effects Exposure Guidelines

Vapor Pressure

11 mmHg (20°C)

Lower explosive limit Target organs Symptoms PEL TWA-8 REL TWA-8 REL STEL TLV TWA-8 TLV STEL IDLH

15 mmHg (25ºC) 4.0% Eyes, skin, respiratory system Irritation of eyes, skin, nose, throat 25 mg/m3 25 mg/m3 37 mg/m3 25 mg/m3 37 mg/m3 123 mg/m3

0.60 mmHg (25ºC)a

Vapor pressure in dilute solution by Raoult’s law using UNIFAC method activity coefficient.

a

IDLH, immediately dangerous to life or health; PEL, permissible exposure limit; REL, recommended exposure limit; STEL, short-term exposure limit; TLV, threshold limit value; TWA-8, 8-hr time-weighted average.

Process Description Twelve “sprays” from the bottle, each of about 1.0 mL, are applied to the panel for a total of 12 mL per panel. The total 12 mL is applied in about 10 sec. This is followed by wiping for 30 sec with an absorbent towel. One panel is cleaned every 3 min, and the process is repeated as panels move along the production line. A new absorbent towel is used every couple of panels, as needed. Used wipe towels are placed on or near the worktables. All used wipe towels are removed from the area of the task immediately before break periods. The worker frequently moves back and forth between the worktables and the parts line. The cleaning solution evaporates as it is applied to the panel and evaporates from the towels that are used to wipe the parts. The density of the solution and the weight fraction of acetic acid in the solution can be used to determine the mass of acetic acid applied to each panel—and the average mass application rate—as follows: Mass applied/ panel:

12 mL 1.002 g 1,000 mg 4 mgacetic acid 481 mgacetic acid ——— · ———— · ————— · ——————— = ——————— panel mL g 100 mgsolution panel

Average mass application rate:

481 mgacetic acid 1 panel 160 mgacetic acid ——————— · ————— = ——————— panel 3 min min

76

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 10: Cleaning With Acetic Acid The evaporation rate of acetic acid mass, however, is slower than the liquid mass application rate. The low vapor pressure of acetic acid in the mixture results in a slow rate of evaporation. Not all the acetic acid will evaporate by the time the next panel comes along to be cleaned. So, although the mass application rate is (from a mass balance perspective) the maximum possible generation rate of acetic acid vapors, in practice, the emission rate will be lower. This is due to acetic acid’s low vapor pressure, the movement of the panel out of the workstation, and the periodic removal of used towels. A conservative approach to calculating the emission rate would be to assume all the acetic acid evaporates quickly in the immediate vicinity of the application task. This rate is calculated as shown previously (160 mg/min). Another approach to calculating the emission is to model the slow evaporation rate of the acetic acid. An empirical study assessed acetic acid vapor emission rates from mopping with a 4% acetic acid solution.3 The process evaluated in the study used small volumes of the solution applied to a surface followed by wiping with an absorbent device. That process is similar to the spraying and wiping in our current case. The mopping study reports an exponentially decreasing emission rate from the process, with an emission rate constant of 0.002– 0.006 min–1.

Information About the Room The panel cleaning process takes place in a very large room measuring 35 m wide × 60 m long with an 8 m high ceiling, for a total volume of 16,800 m3. General ventilation is provided by drop diffusers at various locations in the room. Some of the equipment throughout the room has local exhaust ventilation. Smoke tubes show that in the area where this process takes place, the airflow is primarily horizontal and in the direction of the production flow. The smoke tubes also indicate that the movement of the worker mixes the air and diffuses the smoke rather quickly. The movement of the assembly line combines with the rest of the air movement in the room to create the flow along the line at a rate of about 4.0 m/min. The boundaries of the panels, the floor, and the worktables create a virtual tunnel of lateral air movement past the worker. The space between the panel to be cleaned and the front portion of the workbench is 2.5 m. From the worker’s knees to approximately an arm’s reach overhead is 2.0 m. As the panel moves along the line, the worker moves with it laterally across a distance of approximately 3.75 m. Figure 10.2 illustrates how air movement through this workstation will be modeled. The box has a volume of 18.75 m3. The movement of the worker between the panels and the workbenches induces mixing of the air throughout the cross section of the virtual tunnel. The volumetric flow rate through the box is as follows: 2.5 m 2.0 m 4.0 m 20.0 m3 Q = ——— · ——— · ——— = ———— min min The concentration in the workstation airspace will be modeled as a well-mixed box. The emissions of evaporating acetic acid come from the surface area of the panel as well as used rags that are on the other side of the box. This pattern of emissions and the motion of the worker throughout the box, supported by the smoke tube observations, make good mixing a reasonable (although not perfect) assumption. Figure 10.2: General airflow through the workstation.

Copyright AIHA®. For personal use only. Do not distribute.

77

Modeling Occupational Inhalation Exposures to Chemicals Chapter 10: Cleaning With Acetic Acid

Modeling Approach 1: Well-Mixed Box With Rapid Evaporation of Acetic Acid The key assumption for this approach is that the acetic acid evaporates completely and at a steady rate in the 3-min interval between applications while the panel is in the virtual box. The concentration can be modeled as a well-mixed box (V = 18.75 m3, Q = 20 m3/min) with a constant emission rate (160 mg/min). G Ct = — · Q

1– e

–Q · t ——— V

(10.1)

Where Ct: concentration at time t (mg/m3) G: mass emission rate (mg/min) Q: ventilation rate (m3/min) V: box volume (m3) t: time since initiation of emissions (min) As time increases, the concentration builds up to a steady-state concentration (Css), as shown by Equation 10.2: G Css = — Q

(10.2)

In the current case, the air exchange rate in the box is very high (20 m3/min in an 18.75 m3 box). The concentration quickly builds up to a steady-state concentration of 8 mg/m3, as shown in Figure 10.3. 9.0

8.0

Concentration(mg/m3)

7.0

6.0

5.0

4.0

3.0

2.0

1.0

0.0 0

2

4

6

8

10

12

14

Time(min)

Figure 10.3: Concentration buildup for the well-mixed box with rapid evaporation model.

78

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 10: Cleaning With Acetic Acid If the worker’s exposure for the 420 min at their workstation were at 8 mg/m3 and their exposure during their 60 min on break were at 0 mg/m3, the 8-hr time-weighted average (TWA-8) exposure would be 7 mg/m3. The calculation is as follows: mg mg 8 —— · 420 min + 0 —— · 60 min min3 min3 mg TWA-8 = —————————————————— = 7 —— 480 min m3 Assuming rapid acetic acid evaporation, the modeled maximum (steady-state) concentration and the TWA-8 are about one-third of the permissible exposure limit (PEL) and threshold limit value (TLV) TWA-8. This model probably overestimates the concentration because of the assumption of complete rapid evaporation. But because this is above the company’s goal of 10% of the OEL, which would be 2.5 mg/m3, additional modeling will be done to try to capture the slower evaporation rate that likely is the actual condition. Two modeling approaches that can be applied—a time-varying model and a steady-state model—are described below.

Modeling Approach 2: Time-Varying Well-Mixed Box Concentration With Repeated Mass Applications and Decreasing Emission Rates In this approach, the virtual box will still be modeled as a well-mixed space. However, the emission rate will be modeled as a series of individual releases, each having an exponentially decreasing emission rate. The emission rate from a single exponentially decreasing evaporation source is described by Equation 10.3, as shown below.5 Gt = α · M0 ∙ e–α

(10.3)

· t

Where M: mass applied (e.g., mg, g, mole) Gt: emission rate at time t (e.g., mg/min, g/sec, mol/min) α: a constant (e.g., sec–1, min–1) t: time elapsed (e.g., sec, min) Arnold et al.6 report that the emission rate constant (α) for acetic acid from a process similar to the one in our case is in the range of 0.002–0.006 min–1. The concentration at points in time in a well-mixed box from the application of mass M can be calculated from Equation 10.4, which combines the exponentially decreasing emission rate in Equation 10.3 with a well-mixed box model:5 α · M Ct = ————— · α · V–Q

–Q · t

e ——— – e–α V

· t

(10.4)

Where Q: volumetric airflow rate (e.g., m3/min) V: volume of the box (e.g., m3)

Copyright AIHA®. For personal use only. Do not distribute.

79

Modeling Occupational Inhalation Exposures to Chemicals Chapter 10: Cleaning With Acetic Acid For this case, α = 0.006 min–1 M = 481 mg V = 18.75 m3 Q = 20.0 m3/min For each mass application onto a panel, the concentration will build up and then slowly decrease as the emission rate from that mass application diminishes. The next application of mass onto a panel occurs while the previous application is still evaporating. The concentrations resulting from each of these two applications are added together, thus increasing the total concentration in the box. For additional sequential applications of a mass, as in our current case, the individual event concentration created over time for each application can be calculated. Then, these concentration contributions from each application can be summed to get the total concentration for a point in time. This approach is shown in Equation 10.5. Table 10.4 illustrates a tabular approach to solving for concentrations at 3-min intervals.

 n

Ct =

i=1

α · M ————— · α · V–Q

– Q · ti

e ——— – e–α V

(10.5)

· (t – ti)

Where i: application number ti: time at application i Table 10.4: Tabular Approach to Solving Concentrations From Each Application and Resulting Total Concentration in Box

Time Into Shift (min) 0 3 6 9 12 15 18 21 24 27 … 480

80

Concentration (mg/m3) Contributed From Each Application No. Concentration in Box (mg/m3) 0.000 0.137 0.276 0.414 0.549 0.681 0.812 0.940 1.065 1.189 …

1

2

3

4

5

6

7

8

9

10

0.000 0.137 0.140 0.137 0.135 0.133 0.130 0.128 0.126 0.123 …

0.000 0.137 0.140 0.137 0.135 0.133 0.130 0.128 0.126 …

0.000 0.137 0.140 0.137 0.135 0.133 0.130 0.128 …

0.000 0.137 0.140 0.137 0.135 0.133 0.130 …

0.000 0.137 0.140 0.137 0.135 0.133 …

0.000 0.137 0.140 0.137 0.135 …

0.000 0.137 0.140 0.137 …

0.000 0.137 0.140 …

0.000 0.137 …

0.000 …

…n

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 10: Cleaning With Acetic Acid Figure 10.4 illustrates what is happening on a small time scale by solving the equation at 3-min intervals for the first 30 min. The concentration resulting from each application quickly builds up, then the mass emissions from the next application add to the existing concentration. 1.40

1.20

Concentration(mg/m3)

1.00

0.80

0.60

0.40

0.20

0.00 0

5

10

15

20

25

Time(min)

Figure 10.4: Small-scale concentration buildup by application. Over the length of the shift, the shape of the curve is smoothed out when plotted on a longer time scale x-axis. The concentration pattern throughout and following the shift is illustrated in Figure 10.5. You can see the effect of the two 15-min breaks and the 30-min lunch when no new mass is applied (as shown by the notches on the curve at approximately 4, 5, and 6 mg/m3). This graph was generated with the full spreadsheet that is illustrated in Table 10.4 with zero mass emission taking place during the break periods. The concentration at the workstation can be averaged for the entire shift or for periods of the shift using basic spreadsheet formulas.

10.00 9.00

Concentration(mg/m3)

8.00 7.00 6.00 5.00 4.00 3.00 2.00 1.00 0.00 0

100

200

300

400

500

600

Time(min)

Figure 10.5: Concentration pattern throughout the workday.

Copyright AIHA®. For personal use only. Do not distribute.

81

Modeling Occupational Inhalation Exposures to Chemicals Chapter 10: Cleaning With Acetic Acid The concentration reaches a maximum of 6.6 mg/m3 at the end of the shift. Extending the calculations to the beginning of the shift the next day, the concentration decreases to less than 0.05 mg/m3. This shows that the concentration will start at zero the next time the process starts and that day-to-day concentration buildup is not expected. The TWA concentration at the workstation over the course of the shift is 4.7 mg/m3. The worker normally leaves the workstation during the break periods and has essentially zero exposure to acetic acid during the breaks. The average concentration during the work periods can be determined using the spreadsheet developed for the case. The concentration of the different time periods is shown in Table 10.5. Table 10.5: Information for Calculating Worker’s TWA-8

Period Start to break 1 Break 1 – away from workstation End of break 1 to break 2 Break 2 – away from workstation End of break 2 until break 3 Break 3 – away from workstation End of break 3 to end of shift

Time (min) 120 15 105 30 120 15 75

Average Concentration (mg/m3) 2.3 0 4.8 0 5.7 0 6.3

TWA-8, 8-hr time-weighted average.

Modeling Approach 3: Steady-State Well-Mixed Box Concentration With Repeated Mass Applications and Decreasing Emission Rates Hewitt and Ganser7 presented an approach that describes the steady-state average concentration that will result in a well-mixed box with sequential applications of an exponentially decreasing vapor source. This is shown in Equation 10.6: n · M · T–1 C̅ : = —————— Q

(10.6)

Where C̅ : steady-state average concentration in well-mixed box n: number of applications M: mass per application T: time interval of interest Q: volumetric airflow rate (e.g., m3/min) For the 480 min of the shift, we assume 140 applications of 480 mg each. At a box ventilation rate of 20 m3/min, Equation 10.6 predicts a steady-state concentration of 7.0 mg/m3. This approximates a similar concentration to that of Equation 10.5, as shown in Figure 10.5. Although the concentration will build up, the steady-state average concentration can be useful for judging exposures without having to resort to determining the TWA of an ever-changing concentration.

82

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 10: Cleaning With Acetic Acid

Interpreting the Modeled Concentration All the models applied in this case used a well-mixed box approach. The difference between the approaches is in how the emissions are treated. The models estimate concentrations that occur in the box. This information can be combined with the worker’s work and break pattern to calculate the worker’s TWA-8. The rapid evaporation model estimated a steady-state concentration of 8 mg/m3 in the box and a TWA-8 for the worker of 7 mg/m3. The second approach treated the emissions as a series of exponentially decreasing discrete releases. These resulted in a concentration that slowly builds up in the box over the course of the shift, reaching a concentration of 6.6 mg/m3. The TWA-8 for the worker using this approach is 4.7 mg/m3. The Hewett and Ganser model indicates that if the similar pattern were to continue, the steady-state average concentration in the box would reach 7.0 mg/m3. The TWA-8 OELs from OSHA, NIOSH, and ACGIH are all 25 mg/m3. The rapid evaporation model predicts a TWA-8 exposure of about 7 mg/m3, less than one-third of the OEL. The model that accounts for a slower evaporation rate predicts a TWA-8 of 4.7 mg/m3, which is about 16% of the OEL. Both of these estimates are above the 10% of the OEL benchmark set by this organization. Another consideration is that the emission estimates used are likely high in all these models. This is due to the assumption that all the applied cleaning solution remains in the box until completely evaporated. In actuality, residual solution on the panels moves out of the box with each panel and the worker removes used wipe towels from the box at the beginning of each break. If the concentration contributions from previous applications are removed from Equation 10.6 at the beginning of each break, the TWA-8 of the worker drops from 7.0 mg/m3 to 1.6 mg/m3. An argument could be made that the airflow through the box might be lower than that proposed, resulting in higher concentrations. When using a well-mixed box approach, if the actual airflow was 5 m3/min rather than 20 m3/min, the modeled concentration would be 4 times higher. This change would increase the TWA-8 exposure of 1.6 mg/m3 to 6.4 mg/m3, which is still only 25% of the OEL but higher than the organizational acceptable longterm exposure of 10% of the OEL. These modeling approaches tend to confirm the professional judgment that the exposure from this particular task is low but not below 10% of the OEL. The EHS group decides to schedule some monitoring for this location but at a lower priority level than other exposure scenarios where their estimate of exposure is higher.

References 1. Arnold SF, Stenzel M, Drolet D, Ramachandran G. Using checklists and algorithms to improve qualitative exposure judgment accuracy. Journal of Occupational and Environmental Hygiene, 13(3):159–168, 2016. https://doi.org/10.1080/15459624.2015.1053892. 2. Vadali M, Ramachandran G, Mulhausen JR, Banerjee S. Effect of training on exposure judgment accuracy of industrial hygienists. Journal of Occupational and Environmental Hygiene, 9(4):242–256, 2012. https://doi. org/10.1080/15459624.2012.666470. 3. Arnold S, Ramachandran G, Kaup H, Servadio J. Estimating the time-varying generation rate of acetic acid from an all-purpose floor cleaner. Journal of Exposure Science & Environmental Epidemiology, 30(2):374– 382, 2019. https://doi.org/10.1038/s41370-019-0142-5. 4. National Institutes of Health. PubChem. https://pubchem.ncbi.nlm.nih.gov/ (accessed March 1, 2022). 5. Keil CB, Nicas M. Predicting room vapor concentrations due to spills of organic solvents. AIHA Journal, 64(4):445–454, 2003. https://doi.org/10.1080/15428110308984838. 6. Arnold S, Ramachandran G, Kaup H, Servadio J. Estimating the time-varying generation rate of acetic acid from an all-purpose floor cleaner. J Expo Sci Environ Epidemiol, 30(2):374–382, 2020. https://doi. org/10.1038/s41370-019-0142-5. 7. Hewett P, Ganser GH. Models for nearly every occasion: Part III – One box decreasing emission models. Journal of Occupational and Environmental Hygiene, 14(11):907–918, 2017. https://doi.org/10.1080/154596 24.2017.1339166. Copyright AIHA®. For personal use only. Do not distribute.

83

Modeling Occupational Inhalation Exposures to Chemicals Chapter 11: Model Parameters

Chapter 11

Experimental Determination of Model Parameters Chris Keil, PhD, CIH Obtaining inputs for modeling airborne concentrations of chemicals can be difficult for certain exposure scenarios. Using ingenuity and some “tricks of the trade” pioneered by earlier industrial hygienists, the selection of model parameters can be strengthened and improve the overall model. Two such situations will be illustrated with the cases in this chapter. In the first case, a particle emission rate model is necessary. Models exist for liquid evaporation rates.1,2 Pressurized gas releases can be modeled using the ideal gas law, as shown in Chapter 8 of this book and covered in Chapter 3 of Mathematical Models for Estimating Occupational Exposure to Chemicals, 2nd edition.3 Large-scale outdoor evaporative and pressurized gas releases can be modeled with the Areal Locations of Hazardous Atmospheres (ALOHA) hazard modeling program.4 Particle emissions are more difficult to characterize due to the variety of physical processes that generate them. The sanding case study of Chapter 4 presented some mass balance approaches. An approach by López Lilao et al.5 presents a model to describe dustiness. But overall, particle emission rates are harder to generalize. This chapter will illustrate a simplified experimental approach to estimating particle emissions from the emptying of powder-filled bags. In the second case, an adaptation of the well-mixed box model is used to determine ventilation rates. A variety of indoor spaces have conditions that make determining ventilation rates difficult: open windows, large doorways, out-of-reach air diffusers and returns, partial walls, etc. Although there are reported values of typical air change rates (refer to Chapter 2), there is a fairly straightforward method for getting more specific ventilation information for an indoor space if one has access to data-logging sensors.

Case 1: Dust Emissions From Mixer Loading You are asked to estimate the historical exposure of a worker to a dust as part of evidence in a lawsuit. The plaintiff claims that they suffer negative health effects from an occupational exposure 15 years prior. The plaintiff worked in a large room where one of the operations included the filling of an unventilated mixer with 50-lb bags of powder. The plaintiff was not the mixer operator. They worked across the room from the mixing process. Powder loading created “clouds of dust,” as described by the plaintiff. The mixer operator wore a respirator when loading. Although the dust was low toxicity, the plaintiff has some potentially exacerbating health conditions that are part of the complaint. You are asked to estimate the plaintiff’s exposure to the dust that is generated when he dumps the bag into the mixer. General ventilation present in the room is well documented by the company. On the other hand, information on the mass emission rate of the dust is not readily available. One way to obtain this information, if resources are available, is to experimentally simulate the process and take some measurements.

Information Gathering The plaintiff had a workstation across the room from the mixer. They would be the “far field” if a two-box model were used to calculate concentrations. Figure 11.1 illustrates the relative positions of the plaintiff’s workstation to the powder loading and mixing operation.

Copyright AIHA®. For personal use only. Do not distribute.

85

Modeling Occupational Inhalation Exposures to Chemicals Chapter 11: Model Parameters

Figure 11.1: Room where exposure took place. The cylindrical-shaped mixer is to the right, and the plaintiff was across the room to the left. Powder loading into the mixer occurred 2 or 3 times each hour. During loading, five 50-lb sacks of a dry ingredient were added to the mixer through a raised access hatch about the size of a 55-gal drum. The overall loading process took about 5 min. The plaintiff and other workers describe a faint but visible plume of dust rising from the mixer hatch during powder loading. In talking with various workers who worked in the room, none of them mention excessive dust on horizontal surfaces even when specifically asked about it. This provides some evidence that whatever the specific size distribution of the particles, removal from the air by settling was not significant. Bags of the powder material are still commercially available and used in industry. Attorneys will fund a simple test chamber experiment to measure the dust emission rate from emptying the sacks. Test chambers have been used to develop exposure models for several scenarios.6–9

The Test Chamber An 8 ft × 8 ft × 8 ft chamber is constructed (volume = 512 ft3) from readily available lumber materials. There is a 2 ft × 6 ft door for entering the chamber. An open-topped drum is positioned against the wall opposite the door. Immediately above the barrel top is a 6-inch-tall, 3-ft-wide slot (face area: 1.5 ft2). The slot is exhausted by a centrifugal fan. In line with the ductwork is a minimum efficiency reporting value (MERV) 16 air filter. MERV 16 filters remove particles down to 0.3 μm with >90% efficiency. The filter is 2 ft × 2 ft and 6 inches deep. The chamber set up is illustrated in Figure 11.2. The fan is rated to move up to 2,300 ft3/min. When the system is assembled, the average face velocity of the slot is 1,100 ft/min. With a slot area of 1.5 ft2, this gives an in-operation volumetric flow rate of 1,650 ft3/min.

86

Figure 11.2: Set up of the experimental chamber. Copyright AIHA®. For personal use only. Do not distribute..

Modeling Occupational Inhalation Exposures to Chemicals Chapter 11: Model Parameters This flow rate is undersized when compared to ACGIH ventilation manual hood design procedures for similar processes.10 However, this is an experimental set up to determine generation rate and not a workstation where worker protection is paramount. In this small chamber, the airflow rate represents greater than three air changes per minute. The chamber is under negative pressure, and the slot is the only air outlet. Almost all dust generated in the chamber will be pulled out of the chamber and through the filter.

The Experimental Procedure Five 50-lb bags of the material are placed into the chamber. A new MERV 16 filter is weighed on a 9 inch × 9 inch weighing platform (maximum load: 3,000 g, readability: 0.2 g, certified to 1 g accuracy). The filter is loaded into the filter holder in the ventilation system. A technician in a Tyvek® suit and full-face respirator enters the chamber. The fan is turned on. The 5 bags of material are opened and emptied into the barrel, 1 per minute. The fan is left on for an additional 5 min. The fan is turned off, and the technician leaves the chamber. The filter is removed from the system and reweighed. The mass gain is calculated. The procedure is repeated 6 times using a new filter each time.

Experimental Results The results for the six trial runs are presented in Table 11.1. Table 11.1: Results of Experimental Measurement of Mass Emissions From Bag Emptying

Trial 1 2 3 4 5 6

Filter Initial Mass (g) 681 679 684 680 681 682

Filter Final Mass (g) 688 688 697 686 690 692

Mass Gain/Release Into Air (g) 7 9 13 6 9 10

The average mass gain on the filter (and particle release into the air) is 9 g, or 9,000 mg. For a 5-min process, that is an emission rate of 1,800 mg/min. This emission information can also be generalized into an emission factor for the material. A total of 9 g being released from 5 sacks gives an emission factor of about 1.8 g per sack dumped. This emission factor can be used with different emptying rates and numbers of sacks to provide emission estimates for other rates of use. Is this emission rate estimate of 1,800 mg/min perfect? No. But it provides an empirically demonstrable starting point for selecting an emission rate to use in a well-mixed room, two-zone, or diffusion model. The six data points from the experiment pass a Shapiro-Wilk test for normal distribution and could be modeled in a Monte Carlo approach with a normal or perhaps triangular distribution (refer to Chapter 12).

Case 2: Ventilation Rate in a Room Without Mechanical Ventilation Your company would like to install a process in a vacant room in an older part of the facility. Prior to installing the process, you want to get an idea about what airborne chemical concentrations will occur in the room. The process will be largely automated, but employees will pass through the room en route to other parts of the facility. The production engineers can estimate the chemical release rate, but little is known about the ventilation rate in the room. The space is 75 ft × 100 ft with 16-ft ceilings, giving a total volume of 120,000 ft3. It is currently entirely empty. There is no mechanical ventilation. Large doors are open to adjacent areas, some of which are mechanically ventilated. There are also windows on the roof that can be opened with mechanical chains. There is no easy way to directly measure the airflow in and out of the large open doors, through small gaps in the roof and walls, and via other points where natural ventilation will occur.

Copyright AIHA®. For personal use only. Do not distribute.

87

Modeling Occupational Inhalation Exposures to Chemicals Chapter 11: Model Parameters There is a way to use an application of the well-mixed room model to measure the airflow rate through the room. When the release of a chemical ceases, the concentration will decrease according to Equation 11.1: Ct = C0 · e

–Q · t ——— V

(11.1)

Where C0: concentration at a time after generation of chemical ceases Ct: concentration at time t t: time between C0 and Ct Q: volumetric airflow through the room V: room volume Appropriate units should be chosen so that the quantity in the exponent is dimensionless. Concentration can be in any consistent units. This form of the well-mixed box model will model the decrease in concentrations if you know the volumetric airflow rate and the room volume. If concentration decay data are available or can be obtained experimentally, this equation can be rearranged and used to calculate the airflow rate. There are no historical ventilation data or concentration decay data for this room. However, you remember a trick of the trade from an old mentor, Figure 11.3: Carbon monoxide (CO) concentrations Dr. John Franke. You get ahold of a carbon monoxide during and after forklift is driven in the room. (CO) data logger and call on the services of a driver of a propane-fueled forklift. With the CO data logger recording concentration in the center of the room, the driver maneuvers the forklift around the room for 10 min and then leaves. You continue logging data for 50 min. The concentration pattern is illustrated in Figure 11.3. The slope of the decay curve is a function of Q/V, the air change rate in the room. Equation 11.1 can be rearranged algebraically into the following form: – ln

Ct —— C0

Q = — · t V

(11.2)

Equation 11.2 is the form of a linear equation, y = m ∙ x, where –ln(Ct/C0) is y, Q/V is m, and t is x. If a scatter plot is made using –ln(Ct/C0) and t, the slope of the line is Q/V. With the volume (V) of the room known, the airflow rate (Q) can be determined. This will be demonstrated below. In your current case, t0 for the concentration decay is the 10-min mark in the experiment, when the emission of CO in the room ceases and concentration decay begins. Figure 11.4 is a screenshot of the spreadsheet with the concentration decay data. The first few rows are shown, and the formula used in Column C is presented for illustrative purposes in Column D.

88

Copyright AIHA®. For personal use only. Do not distribute..

Modeling Occupational Inhalation Exposures to Chemicals Chapter 11: Model Parameters

Figure 11.4: Data used for the determination of the ventilation rate (showing first few rows). Figure 11.5: Scatter plot for the determination of ventilation rate.

Figure 11.5 is the scatter plot created with Columns A and C from the spreadsheet. The least-squares fit line to the data is drawn through the data points. The equation of the line is shown as well. The y intercept in the equation of the line is an artifact of the calculation—a result of the variation in the data from a “perfect” decay curve. The slope of the line, 0.0134, is Q/V, which is also the number of air changes per minute. The room volume is 120,000 ft3. Therefore, Q is 1,600 ft3/min, or about 0.8 ACH. This value is in the general range of the air change rates of naturally ventilated buildings reported in Table 2.1. This experiment was done in the middle of the room. It could be repeated with measurements taken at different places in the room. Then, the range of results could be considered when selecting the ventilation rate to use in modeling the potential concentrations created by the new process. The experiment could also be repeated with the ceiling windows open and in different weather conditions, both of which will likely affect the natural ventilation rate in the room. This is one example of using concentration decay data to estimate ventilation rates. The same approach can be undertaken using data from fixed location monitors that record data continuously or using tracer gases and detectors specifically brought in for the characterization of ventilation. More sophisticated methods are also available for more refined ventilation evaluations. The approach presented here is a simple but effective one that is within the reach of most industrial hygienists.

References 1. Hummel AA, Braun KO, Fehrenbacher MC. Evaporation of a liquid in a flowing airstream. American Industrial Hygiene Association Journal, 57(6):519–525, 1996. https://doi.org/10.1080/15428119691014729. 2. Arnold FC, Engel AJ. Evaporation of pure liquids from open surfaces. In: Modelling of Environmental Chemical Exposure and Risk, edited by Linders JBHJ. NATO ASI Series; Springer Netherlands: Dordrecht, 2001. pp. 61–71. https://doi.org/10.1007/978-94-010-0884-6_6. 3. Keil CB, Simmons CE, Anthony TR (Eds.). Mathematical Models for Estimating Occupational Exposure to Chemicals, 2nd edition. AIHA Press: Fairfax, VA, 2009. 4. Jones R, Lehr W, Simecek-Beatty D, Reynolds RM. ALOHA® (Areal Locations of Hazardous Atmospheres) 5.4.4: Technical Documentation. NOAA Technical Memorandum NOS OR&R 43. National Oceanic and Atmospheric Administration: Seattle, WA, 2013. p 96.

Copyright AIHA®. For personal use only. Do not distribute.

89

Modeling Occupational Inhalation Exposures to Chemicals Chapter 11: Model Parameters 5. López Lilao A, Sanfélix Forner V, Mallol Gasch G, Monfort Gimeno E. Particle size distribution: a key factor in estimating powder dustiness. Journal of Occupational and Environmental Hygiene, 14(12):975–985, 2017. https://doi.org/10.1080/15459624.2017.1358818. 6. Lennert A, Nielsen F, Breum NO. Evaluation of evaporation and concentration distribution models — a test chamber study. The Annals of Occupational Hygiene, 41(6):625–641, 1997. https://doi.org/10.1093/ annhyg/41.6.625. 7. Arnold SF, Shao Y, Ramachandran G. Evaluating well-mixed room and near-field–far-field model performance under highly controlled conditions. Journal of Occupational and Environmental Hygiene, 14(6):427–437, 2017. https://doi.org/10.1080/15459624.2017.1285492. 8. Simmons CE, Jones RM, Boelter FW. Factors influencing dust exposure: finishing activities in drywall construction. Journal of Occupational and Environmental Hygiene, 8(5):324–336, 2011. https://doi.org/10.10 80/15459624.2011.570239. 9. Lopez R, Lacey SE, Jones RM. Application of a two-zone model to estimate medical laser-generated particulate matter exposures. Journal of Occupational and Environmental Hygiene, 12(5):309–313, 2015. https://doi.org/10.1080/15459624.2014.989361. 10. American Conference of Governmental Industrial Hygienists. Industrial Ventilation: A Manual of Recommended Practice for Design, 30th edition. Signature Publications; ACGIH: Cincinnati, OH, 2019.

90

Copyright AIHA®. For personal use only. Do not distribute..

Modeling Occupational Inhalation Exposures to Chemicals Chapter 12: Lab Exposure and Monte Carlo Modeling

Chapter 12

Organic Chemistry Teaching Lab Exposure and a First Exploration of Monte Carlo Modeling Chris Keil, PhD, CIH

Scenario You are an industrial hygienist at a mid-sized state university. You are approached by a couple of faculty and staff members from the chemistry department. During a few of the laboratory sessions for organic chemistry, the odor of solvent is particularly noticeable. They ask if you could evaluate the exposures. Because it is a state university, budgets are tight. Before you take any samples, you think that you could make a good initial estimate of exposures using modeling. Although several lab sessions use solvents, one exercise stands out to you. It is the lab in which methylene chloride is used for an extraction experiment. The relatively low occupational exposure limit (OEL; 87 mg/m3) for methylene chloride, along with the quantity used, puts it at the top of your list for assessing the exposure. You read the lab manual and talk to the instructors and some students to get information on the process of the exercise. This case is based on previously published work.1

Information Gathering The NIOSH Pocket Guide2 provides basic physical properties for methylene chloride. Reading the lab manual and talking with those who run the lab reveals that there is a maximum of 24 students in each lab section, the lab session is scheduled to last 180 min, and each student is given 60 mL of methylene chloride (MeCl2) for this lab (Table 12.1). Table 12.1: Information on Methylene Chloride

Physical Properties

Release Characteristic Health Effects

Exposure Guidelines

Description Molecular weight Liquid density Vapor pressure Lower explosive limit Volume Time Target organs Symptoms PEL TWA-8 PEL STEL TLV TWA IDLH

Colorless with a chloroform-like odor 84.9 1.33 g/mL 350 mmHg 13% Each student is given 60 mL The lab period is 180 min long Eyes, skin, cardiovascular system, central nervous system Eye and skin irritation; lassitude, drowsiness, dizziness; numbness and tingling in limbs; nausea; potential occupational carcinogen 87 mg/m3 434 mg/m3 174 mg/m3 ≈ 8,000 mg/m3

IDLH, immediately dangerous to life or health; PEL, permissible exposure limit; STEL, short-term exposure limit; TLV, threshold limit value; TWA-8, 8-hr time-weighted average.

Copyright AIHA®. For personal use only. Do not distribute.

91

Modeling Occupational Inhalation Exposures to Chemicals Chapter 12: Lab Exposure and Monte Carlo Modeling Blueprints and measurements in the room determine that the room has a floor area of 102 m2 and a volume of 294 m3. There are three student workbenches. Four students work on each side of a bench. Other benches are around the room perimeter for equipment and reagents. There are four ceiling diffusers spread evenly throughout the room providing supply air. There are six lab hoods in the room. However, students do their work at their assigned bench space, not at the lab hoods. Figure 12.1 illustrates the room layout. There is no exhaust in the room other than the lab hoods, and the room is under negative pressure. The room ventilation rate is determined by measuring the face area of the lab hoods and the average velocity across the face of the hoods. The total airflow through the room is 97 m3/min. The fact that the evaporation of Figure 12.1: Room layout. MeCl2 occurs throughout the room at each student’s workstation as well as the geometry of the supply and exhaust suggest that this scenario could be modeled as a well-mixed space. Each student is given 60 mL of MeCl2. If all of it evaporates into the room over the 180 min scheduled for the lab session, the average emission rate from each student can be calculated. Using this information and the density of MeCl2, each student workstation emits the following: 60 mLMeCl2 1.33 g 1,000 mg 443 mgMeCl2 ————— · ———— · ————— · ————— = —————— student mL g 180 min student min If there are 24 students in the section, the total emission rate into the room is 10,640 mg/min.

Steady-State Well-Mixed Room Model The dispersed nature of the MeCl2 evaporation locations and the relative positions of the air supply and exhaust make the use of a well-mixed room model appropriate. Using the average emission rate (G) for a full classroom of 24 students, complete evaporation of the solvent calculated, and the measured ventilation rate (Q), a first-pass estimate steady-state concentration in the room can be calculated: G Css = — Q

(12.1)

10,640 mg min 110 mg = —————— · ———— = ————— min 97 m3 m3 The 3-hr average concentration estimate in the room is 110 mg/m3, which is higher than the permissible exposure limit (PEL) time-weighted average (TWA) of 87 mg/m3. If the MeCl2 exposure for the remainder of the day is zero, the TWA-8 exposure would be 41 mg/m3. Although this is below the PEL, there is further opportunity to clarify the exposure assessment.

92

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 12: Lab Exposure and Monte Carlo Modeling

Capturing Variability in Parameters: Best Case and Worst Case The estimated concentration assumes that the lab session takes the full scheduled 3 hr. The actual length of time the students spend doing the lab will affect both the estimated emission rate and the TWA exposure. It also assumes that all the liquid MeCl2 issued to the student evaporates. Information from the instructors confirms that the full 3 hr is typically used for this exercise. However, they report that rarely does a student use all of the MeCl2. The experiment is an extraction exercise, and much of the solvent is recovered and returned. Of the 60 mL that each student is issued, students return a minimum 20 mL and up to about 50 mL of the liquid. Thus, the actual evaporation of MeCl2 is not the full 60 mL but between 10 and 40 mL. The actual emission rate per student, then, is in the range of 74–296 mg/min. You also discover that the lab sections are sometimes the full 24 students but can also be as low as 12 students. If there are 12 students and their work practices are excellent, then each workstation emits 74 mg/min of MeCl2. This means the total emission rate would be 888 mg/min. If there is a full section of 24 students and their work practices are sloppy, then each creates an emission rate of 296 mg/min. In this case, the total emission rate would be 7,100 mg/min. You talk to the building engineers to find out what they think of your estimate for the room flow rate. They think that a typical airflow through the room is about 100 m3/min, so your measured value of 97 m3/min seems right to them. However, they also estimate that the flow rate can vary by about one-third. On any given day, Q might range from 67–133 m3/min. How can this information about variability in the generation rate and the ventilation rate be integrated into your exposure assessment? A best-case scenario regarding MeCl2 concentrations would be if the generation rate was at the low end of its possible range and the ventilation rate was at the upper end of its possible range. In this case—12 students with the best work practices and a high ventilation rate—the steady-state concentration would be the following: 888 mg min 6.7 mg ———— · ———— = ———— min 133 m3 m3 If there were a full class of 24 students, all with poor work practices and a low ventilation rate, the steady-state concentration would be the following: 7,100 mg min 106 mg ————— · ———— = ———— min 67 m3 m3 With the information you have on the variability of the emission rate and ventilation rate, the range of modeled steady-state concentrations is a best case of 6.7 mg/m3 to a worst case of 106 mg/m3. But is every value in that range equally likely? Would it be common for everyone in a low-enrollment class to be extra careful with the chemical, or would the opposite be true? We can use the low end, middle, and upper end of the ranges to get best, typical, and worst-case estimates. That may be enough to make a judgment about the acceptability of the exposure profile. We can also further refine our understanding of the exposure profile by using a computational approach called probabilistic or “Monte Carlo” modeling. This approach uses a distribution of values for each input parameter to calculate a distribution of model outputs.

Copyright AIHA®. For personal use only. Do not distribute.

93

Modeling Occupational Inhalation Exposures to Chemicals Chapter 12: Lab Exposure and Monte Carlo Modeling

Capturing Variability in Parameters: An Introduction to Monte Carlo Approaches Before continuing with the organic chemistry lab exposure assessment, we will take a short detour to introduce Monte Carlo modeling in principle. Then we will return to the organic chemistry lab to apply it. The pattern of variability in model inputs affects the pattern of concentrations that are modeled. By experience, we know this to be true even for common examples. Consider the rolling of a six-sided dice. Each time you roll a die, there is an equal probability of each side coming up. With each roll of one dice, there is a 16.7% chance of a result of a one, a 16.7% chance of a result of a two, etc. This is called a uniform distribution and can be represented in a histogram, as shown in Figure 12.2. If you roll two dice and add the results together, the result is going to be a value from 2 to 12. Yet we also know that each value from 2 to 12 is not equally likely. Of the 36 possible combinations of results from rolling dice A and dice B, many combinations sum to 7 and only one combination sums to 12.

Figure 12.2: Uniform distribution of results from rolling a fair six-sided dice.

Before we return to our organic chemistry lab scenario, let’s consider a steady-state well-mixed box model in which the generation rate ranges uniformly from 1–6 mg/min and the ventilation rate ranges from 1–6 m3/ min. You have a green six-sided dice to randomly determine a generation rate (G) and a quince-colored dice to randomly determine a ventilation rate (Q). You roll both dice and divide the result of the green dice (the generation rate, G) by the result of the quince-colored dice (the ventilation rate, Q) to calculate the concentration for that trial. Now you pick up the dice and do it again, 10,000 times. The distribution of the results from each dice will be uniform between 1 and 6. However, the distribution of the results of the calculation will not be uniformly distributed. Fortunately, we do not have to roll dice all those times. Modern computing allows us to generate random numbers very easily and then perform calculations with the numbers. Readily available productivity software can be used to make these sorts of calculations. Figure 12.3 shows a screenshot of the results using Excel to do 10,000 iterations of this example, where both G and Q are whole numbers uniformly distributed from 1 to 6 and C = G/Q. The inset in the figure shows the formulas in the first row that are copied 10,000 times with select-anddrag to perform the simulation. The example of dice rolling is used here to introduce the idea of a Monte Carlo approach. Model inputs for parameters such as generation rate will not often be limited to whole numbers. There are also a variety of distributions other than uniform that may describe the variability of the parameters. With the proper software and/or coding, all of this can be addressed. IHMOD 2.0 is a freely available program built on Excel that can do the calculations for most of the models in this book using either deterministic approaches (a set value for each model input) or a Monte Carlo approach. IHMOD 2.0 will be introduced more in Chapter 13. Another resource available for licensing is Task Exposure Assessment Simulator (TEAS) by Exposure Assessment Solutions, Inc. It is an outstanding tool for Monte Carlo exposure modeling and report generation and is introduced in Chapter 17.

94

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 12: Lab Exposure and Monte Carlo Modeling

Figure 12.3: Example modeling output using a Monte Carlo approach, where C = G/Q and G and Q are both uniformly distributed between 1 and 6.

Random Versus Uniform Monte Carlo approaches rely on randomly selecting a value from a distribution. Many people are familiar with randomness being expressed by events such as the flip of a coin—a 50/50 chance of a heads or tails—or with dice rolling, as we have used here. But just because a value selected from a distribution is done randomly does not mean that each value in that distribution is equally likely to be selected. For example, imagine a six-sided dice with one pip on one of the sides, two pips on three sides, and three pips on two sides. When you roll that dice, each side is just as likely to show up as the other side, but there are more two-pip sides than the other sides. Performing 10,000 rolls of that dice will produce a distribution or results that look like those in Figure 12.4. Figure 12.4: Results from random rolls of the pips dice.

Copyright AIHA®. For personal use only. Do not distribute.

95

Modeling Occupational Inhalation Exposures to Chemicals Chapter 12: Lab Exposure and Monte Carlo Modeling

Applying a Monte Carlo Approach to the Chemistry Lab Good computer software is extremely helpful for Monte Carlo modeling. The programs mentioned previously (IHMOD 2.0 and TEAS) are built for use by industrial hygienists for exposure modeling. For this case, we will use Excel for illustrative purposes because most people have access to the program and can experiment with the approach if they are so inclined. There are limitations when using Excel, particularly in the types of distribution that can easily be used in the spreadsheet. All these calculations with Excel can also be done in Google Sheets using the XLMiner Analysis ToolPak. In using Excel for illustrative purposes in this example, we will use distributions that are readily available in that software. Other distributions might be better suited based on the information available on the variability, but for this example, we will use uniform and normal distributions. Other software has capabilities for many different types of distributions. The number of students will be uniformly distributed between 12 and 24. The evaporative loss of MeCl2 per student will be normally distributed with a mean of 185 mg/min/student, which is the middle of the possible range (74–296 mg/min) determined earlier. A standard deviation of 30 was selected. This value was selected so that the mean plus or minus 3 times the standard deviation (holding 99% of the values) will be 95–275 mg/min. This range is similar to the range calculated previously, where careful students would result in a release rate of 74 mg/min and sloppy students would result in a release rate of 296 mg/min. The ventilation rate will be modeled as 67–133 m3/min, uniformly distributed. There will be 10,000 iterations of the calculations done. Figure 12.5 illustrates how this modeling looks in an Excel spreadsheet. Columns A, B, and D are columns of values generated by the spreadsheet according to the distributions described above. The steps to creating these arrays of values are described in the next section. Column C (overall emission rate) is the product of Columns A (number of students) and B (emission rate per student). Column E is the concentration calculated as the total emission rate, Column C, divided by the ventilation rate, Column D. The inset on the figure shows the formulas used in the cells. Doing some basic descriptive statistics in the spreadsheet reveals that given these distributions of model inputs, the resulting concentrations are log-normally distributed. The geometric mean is 33 mg/m3 with a geometric standard deviation of 1.4. The 95th percentile of the modeled concentration distribution is 57 mg/m3. This is below the PEL TWA-8 of 87 mg/m3.

Figure 12.5: Monte Carlo modeling the scenario in Excel.

96

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 12: Lab Exposure and Monte Carlo Modeling This information can be taken back to the chemistry faculty for a conversation about the exposure. That conversation should include the assumptions of the model and framework for understanding the results, including: • The room being well mixed and why this was assumed • Looking at the steady-state concentration as a maximum given other assumptions • How the input values, point estimates, and distributions were determined • Concentrations given in TWA-8 OELs are designed to be protective for exposures lasting 8 hr/day, 5 days/ week. With this information, the faculty stakeholders can participate in the decision-making process regarding the acceptability of exposure, the need for air sampling, and any changes in work practices that may be implemented.

Using Excel to Do the Monte Carlo Simulation of the Scenario Set up the columns that will do your repeat calculations for the Monte Carlo Analysis. There will be a column for the number of students, the emissions per student, the total emissions, the ventilation rate, and the concentration that is calculated. Figure 12.6 illustrates one way to set this up. The number of students (Column A) is uniformly distributed, and random values can be generated with the RANDBETWEEN function. This function will not produce distributions other than uniform.

Figure 12.6: Column headers for the example Monte Carlo analysis.

The spreadsheet has an add-on called “Analysis ToolPak” that is available in many installations of the program. On my PC version, I go to File  Options  Add-ins and select “Go…” next to the dropdown box near the bottom that says “Excel Add-ins.” Select “Analysis ToolPak” and click OK. The Data Analysis option should now be available in the Data menu, shown by the arrow in Figure 12.7 below.

Figure 12.7: Where to find the Data Analysis Add-in once loaded. Selecting Data Analysis will bring up a variety of data analysis tools, one of which is a random number generator shown in Figure 12.8. The random number generator allows you to create arrays of random numbers. To generate the emissions per student (Column B), indicate 1 variable with 10,000 random numbers, a normal distribution, a mean of 185, and a standard deviation of 30. It is important to plug in a different random seed (a number or vector used to initialize a random number generator) each time you generate an array of random

Copyright AIHA®. For personal use only. Do not distribute.

Figure 12.8: Data Analysis initial page.

97

Modeling Occupational Inhalation Exposures to Chemicals Chapter 12: Lab Exposure and Monte Carlo Modeling numbers—just key mash. Set the output range to the top of the column for the emission rate per student. This example is shown in Figure 12.9. The column for total emission is the product of the number of students and the emission rate per student. This can be coded in the row as A2*B2. The ventilation rate random number column can be obtained using the same method as the emission per student evaporation rate but selecting a uniform distribution between 67 and 133. Be sure to change the random seed. Figure 12.10 illustrates how this is set up. This uniform distribution could be modeled with RANDBETWEEN as well. The range of concentrations at steady state is obtained by dividing the total emission (Column C) by the ventilation rate (Column D). Descriptive statistics of the 10,000 concentrations can be determined for the minimum, maximum, mean, median, and 95th percentile. These values can be compared to the TWA-8 OEL to determine the acceptability of the exposure profile. Occasionally, the random number generator in Excel will produce a negative number when it physically does not make sense (e.g., a negative generation rate or negative ventilation rate). This especially can be the case if generating many random numbers with a normal distribution where the x̅ - 3S gets close to zero. If this occurs, you can remove the negative number from the array of random numbers manually. There are add-ins for Excel and other standalone programs that improve on the approach illustrated in this chapter. These example uses of Excel for Monte Carlo modeling were presented just as an introductory window into probabilistic solution approaches.

Figure 12.9: Random number generator set to produce 10,000 normally distributed values with a mean of 185 and a standard deviation of 30. The data will start at cell B2.

References 1. Keil C, Murphy R. An application of exposure modeling in exposure assessments for a university chemistry teaching laboratory. Journal of Occupational and Environmental Hygiene, 3(2):99–106, 2006. https://doi. org/10.1080/15459620500498109.

Figure 12.10: Random number generator set to produce 10,000 uniformly distributed values between 67 and 133. The data will start at cell D2.

2. National Institute for Occupational Safety and Health (NIOSH). NIOSH Pocket Guide to Chemical Hazards [DHHS (NIOSH) Publication Number 2005-149]. Department of Health and Human Services. September 2007.

98

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 13: Using IHMOD 2.0 for Calculations

Chapter 13

Filling a Container With a Mixture and Using IHMOD 2.0 for Modeling Calculations Tom Armstrong, PhD, CIH, FAIHA; Daniel Drolet, MSc; Chris Keil, PhD, CIH IHMOD 2.0 is a free suite of mathematical models used to estimate air concentrations of chemicals. It is currently available from the AIHA webpage at bit.ly/3QSjTbv. IHMOD 2.0 uses a Microsoft Excel spreadsheet to do the calculations for many of the models found in Mathematical Models for Estimating Occupational Exposure to Chemicals, 2nd edition1 and used in the cases in this book. The current updated version of IHMOD allows calculations to be done with a deterministic approach, a single value for the model inputs and a point estimate of the concentration, or with a Monte Carlo approach (as introduced in Chapter 12). In the latter approach, a distribution is provided for each of the model inputs and a distribution of concentrations is calculated. This chapter presents another container-filling scenario, but this time with a liquid mixture rather than a pure substance. The IHMOD 2.0 software is introduced as a tool for carrying some of the burden of doing the calculations. A full support file containing a similar scenario and many more details of the workings of the tool is available with IHMOD 2.0 at the website provided. The support file for IHMOD 2.0 contains a good review of many of the modeling approaches in this book.

Scenario You are asked to review a new use of 2-butoxyethanol (2BE) as an additive to consumer automobile windshield washing solution. 2BE is also known as ethylene glycol monobutyl ether (EGBE) or by the trade name Butyl CellosolveTM. Adding the windshield solution to a vehicle’s reservoir will displace vapors from the empty headspace of the reservoir. This may happen outdoors where 2BE vapors will quickly disperse, but the filling could also occur indoors in a residential garage. What is the generation rate and exposure from filling an automotive windshield wash reservoir with this material? Does the exposure present unacceptable risk?

Information About the Chemical and the Emission Rate 2BE is semi-volatile, with a molecular weight of 118 (g/mol) and a pure substance vapor pressure of 0.88 mmHg at 25ºC (298 K). The 8-hr time-weighted average (TWA-8) occupational exposure limits (OELs) are as follows: permissible exposure limit (PEL) = 240 mg/m3, threshold limit value (TLV) = 96 mg/m3, and recommended exposure limit (REL) = 24 mg/m3. Because this is a potential consumer exposure, we can also consider the reference concentration (RfC) for chronic inhalation exposure of 1.6 mg/m3 (refer to EPA).2 Reference concentrations are developed by the EPA and are an estimate of continuous chronic inhalation exposure without appreciable risk. The windshield washing solution will be sold in 4-L containers, with a proposed composition (by weight) of water > 70% and 2BE < 30%. For the purpose of this scenario, we will assume the entire 4 L is poured into the reservoir and that the filling takes about 0.5 min. The liquid filling, and therefore the headspace air displacement rate, is 4 L/0.5 min = 8 L/min. From previous chapters, we know that container filling will release vapors at a rate described by Equation 6.2 (applied here as Equation 13.1):

Copyright AIHA®. For personal use only. Do not distribute.

99

Modeling Occupational Inhalation Exposures to Chemicals Chapter 13: Using IHMOD 2.0 for Calculations MW Pv G = F · ——— · —— · 103 24.45 Patm

(13.1)

Where G: vapor release rate in mg/min F: fill rate in L/min MW: molecular weight of the chemical (g/mol) Pv: vapor pressure of the liquid Patm: atmospheric pressure above the liquid Pv and Patm must be in the same units The earlier container-filling cases were from pure substances. This liquid is a mixture. To determine the vapor pressure of a component of a mixture, Raoult’s law is applied. Raoult’s law states that the vapor pressure of the component of a liquid is the vapor pressure of the pure liquid multiplied by the mole fraction of the liquid in the mixture. It is important to remember that vapor pressure is highly dependent on temperature, so always check that the vapor pressures being used are for the temperature of the exposure scenario. Pa = xa · Pv,a

(13.2)

Where Pa: vapor pressure of the chemical in the mixture Pv,a: vapor pressure of the pure chemical xa: mole fraction of the pure chemical Table 13.1 illustrates the calculation of the mole fraction of the two components in the mixture. Table 13.1: Calculation of Mole Fractions of Mixture Components

Chemical Water 2BE

Weight Fraction 0.70 0.30

MW 18 118 Total moles

Moles (for 100 g of mixture) 3.89 0.25 4.14

Mole Fraction 0.94 0.06

2BE, 2-butoxyethanol; MW, molecular weight. Raoult’s law is reasonably accurate when the chemicals in the mixture are structurally similar and when the mole fraction of the chemical under consideration approaches 1. Dilute components in mixtures behave less than ideally. The vapor pressure they exert is not the same as the vapor pressure of the pure substance times the mole fraction. This can be accounted for using an activity coefficient. The effective vapor pressure of a dilute component of a mixture can be calculated using a modified form of Raoult’s law: Pa =

a

· xa · Pv,a

(13.3)

Where

a: activity coefficient for component a in the mixture

100

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 13: Using IHMOD 2.0 for Calculations The activity coefficient depends on the other chemicals in the mixture and their relative proportions. There are a variety of references for activity coefficients, but these references can sometimes be hard to access.3–5 There is an older but very useful tool that uses the UNIFAC6 approach for determining activity coefficients. The UNIFAC Calculator (Figure 13.1) was developed in 1996 by Bruce Choy and Danny D. Reible in the Departments of Chemical Engineering at the University of Sidney and Louisiana State University, respectively. The UNIFAC calculator is available at the AIHA tools website (bit.ly/3QSjTbv). To use the calculator, one enters the temperature of the mixture and the components of the mixture. The calculator comes with a few preloaded chemicals, as shown in Figure 13.2a. Other chemicals must be entered and built using the “Define” application of the program. UNIFAC then uses component units of the molecule, as shown in Figure 13.2b, to determine activity coefficients.

a

b

Figure 13.1: UNIFAC calculator opening screen (a) and blank input screen (b).

a

b

Figure 13.2: Preloaded chemicals (a) and building chemicals from UNIFAC units (b).

Copyright AIHA®. For personal use only. Do not distribute.

101

Modeling Occupational Inhalation Exposures to Chemicals Chapter 13: Using IHMOD 2.0 for Calculations Entering in the temperature for our scenario (298 K), the components in the mixture, and the mole fractions of each component calculated in Table 13.1 gives us the activity coefficients for each component, as shown in Figure 13.3. Note that the vapor pressure of water with a mole fraction close to 1 is not very much affected by being in a mixture. 2BE on the other hand expresses a vapor pressure more than 7.3 times higher than would be expected without considering activity factors. Equation 13.3 can now be used to get the vapor pressure for 2BE in the mixture:

2BE

· x2BE · Pv,2BE = P2BE

7.31∙0.06 · 0.88 mmHg = 0.39 mmHg

Figure 13.3: Results from UNIFAC Calculator for the scenario mixture.

This value can then be used in Equation 13.1 to calculate the vapor emission rate during the filling process: 8L mol 118 g 0.39 mmHg 103 mg 20 mg —— · ———— · ——— · —————— · ———— = ———— min 24.45 L mol 760 mmHg 1g min

Information About the Room For this product stewardship assessment, we will consider a scenario with low ventilation in order to look at the “reasonable worst case” concentration. The filling takes place in a small garage without mechanical ventilation. The garage is 4.0 m × 3.5 m with a 3-m ceiling, giving a volume of 42 m3. There is no mechanical ventilation. Based on some of the air change rates for unventilated spaces from Chapter 2, we will assume 0.5 air changes per hour, which is equivalent to 21 m3/hr or 0.35 m3/min. To model the higher concentration that will result close to the action of filling the reservoir, we will use a nearfield/far-field approach. The geometry of the near field will be a 1-m radius hemisphere above the open car hood. A half-hemisphere is chosen because although some air mixing may occur down into the engine compartment, most will likely be in an upward direction toward the breathing zone of the pourer. To use a near-field/far-field model, we also need to know the random airspeed in the near field. A study that included indoor random airspeed measurements suggested that motion in the near field creates its own random air motion.7 For calculating the interzonal airflow rate for this case, we will use the geometric mean random airspeed from the “stillest” room in the study. This value is 0.05 m/sec, which is equivalent to 3 m/min.

102

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 13: Using IHMOD 2.0 for Calculations

Figure 13.4: Using IHMOD 2.0 to calculate concentrations for the scenario.

Using IHMOD 2.0 Figure 13.4 illustrates using IHMOD 2.0 Two-Zone Model with a Constant Emission Rate to do the calculations for this exposure scenario. The concentration scale may need to be altered. This can be completed by clicking on the vertical axis and adjusting accordingly. The inputs to the model are as follows: G: 20 mg/min Q: 0.35 m3/min S: 3 m/min Near-field geometry: ½ hemisphere with a 1-m radius ½ hemisphere to account for vapors rising primarily upward and not down into the engine compartment Room volume: 42 m3 Calculate concentrations for 10 min Generation of the pollutant stops at 0.5 min The graph shows the pattern of concentration in the near field and far field over the course of time set for the length of the simulation. The results in the bottom left box report calculated concentrations. The first value reported for the near field and far field is the concentration at the point in time set in the box labeled “t” with the blue font. In this case, that value was set to 0.5 min—the time when the pouring for the scenario is over. This can be set to other points of time as well. The next value is the TWA over the next 15 min from the currently set point in time. The last value is the potential steady-state concentration if the pollutant release continued long enough.

Copyright AIHA®. For personal use only. Do not distribute.

103

Modeling Occupational Inhalation Exposures to Chemicals Chapter 13: Using IHMOD 2.0 for Calculations The exposure from this relatively brief filling operation is rather low. The exposure in the near field peaks at 2.0 mg/m3 and decreases rapidly. The long-term concentration in the garage will decrease slowly but is low—less than 0.25 mg/m3. These exposures are much lower than the OELs. The RfC is exceeded at the peak concentration, but RfCs are developed as a “daily exposure to the human population—including susceptible subgroups—that is likely to be without an appreciable risk of adverse health effects over a lifetime.”2 Risks for this one-time exposure scenario are low. If we have a service automotive technician that performs multiple refills throughout the workday in conditions such as these, then the RfC is likely to be exceeded during the workday. The results of the calculations can be viewed by clicking the “See results” button. Figure 13.5 illustrates the Monte Carlo capabilities of IHMOD 2.0. For simplicity, the inputs for the generation rate, ventilation rate, and random airspeed were entered as having a uniform distribution approximately ± 20% of the point estimates used in the deterministic solution. Other distribution shapes can be used depending on the information you have on the model inputs.

Figure 13.5: Using IHMOD 2.0 to do Monte Carlo concentration modeling. The dotted lines on the concentration plot show the percentiles of the modeled concentrations. The top dotted line is of particular note, as it shows the 95th percentile of the near-field concentrations when the pouring is completed. About 95% of the time, the concentrations will be below 2.5 mg/m3. With the assumptions used for this Monte Carlo approach, the 95th percentile value is still below the OELs. The 95th percentile peak concentration (2.5 mg/m3) at the end of generation does exceed the RfC, but as stated before, this value would only be applicable if an individual has repeated exposures throughout the workday and lifetime. The results of the Monte Carlo simulation support the initial evaluation that the exposure risk is low for this filling of the reservoir. If the TWA concentration is desired, the TWA concentration box can be clicked, as shown in Figure 13.6. This will show the 95th percentile (top dotted line) in the near field as a TWA concentration. As expected, these values are less than the OELs and right at the RfC value. The results of the Monte Carlo simulation can be viewed by clicking the “See results” button at the bottom right.

104

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 13: Using IHMOD 2.0 for Calculations

Figure 13.6: Viewing time-weighted average (TWA) concentration in IHMOD 2.0 Monte Carlo concentration modeling.

Conclusion IHMOD 2.0 serves as a very useful tool in calculating potential exposure either before or after the exposures occur. The addition of the Monte Carlo simulation allows exposures to be validated based on initial modeling over some range, such as ± 20%, for the various inputs. The tool can also be used for initial modeling of exposures. This simulation can provide greater assurances of potential exposures and their degrees of variance.

References 1. Keil CB, Simmons CE, Anthony TR (Eds.). Mathematical Models for Estimating Occupational Exposure to Chemicals, 2nd edition. AIHA Press: Fairfax, VA, 2009. 2. U.S. Environmental Protection Agency (EPA). Toxicological Review of Ethylene Glycol Monobutyl Ether (EGBE) (CAS No. 111-76-2) (p. 206). EPA/635/R-08/006F. EPA: Washington, DC, 2010. 3. Gmehling J, Onken U, Arlt W, Grenzheuser P, Weidlich, U, Kolbe B, Rarey J. Vapor-Liquid Equilibrium Data Collection. Frankfurt, Germany: Dechema Chemistry Data Series, 1991. 4. Landolt-Börnstein - Group IV Physical Chemistry; Kehiaian HV (Ed.) Binary Liquid Systems of Nonelectrolytes. Part 1 (Vol. 13A1). Springer Berlin Heidelberg: Berlin, Heidelberg, 2007. https://doi. org/10.1007/978-3-540-49315-0. 5. Landolt-Börnstein - Group IV Physical Chemistry; Kehiaian HV, Martienssen W (Eds.). Binary Liquid Systems of Nonelectrolytes. Part 2 (Vol. 13A2). Springer Berlin Heidelberg: Berlin, Heidelberg, 2008. https:// doi.org/10.1007/978-3-540-70745-5. 6. Fredenslund A. Vapor-Liquid Equilibria Using UNIFAC: A Group-Contribution Method; Elsevier, 2012. 7. Keil C, Zhao Y. Interzonal airflow rates for use in near-field far-field workplace concentration modeling. Journal of Occupational and Environmental Hygiene, 14(10):793–800, 2017. https://doi.org/10.1080/154596 24.2017.1334903. Copyright AIHA®. For personal use only. Do not distribute.

105

Modeling Occupational Inhalation Exposures to Chemicals Chapter 14: Oxygen Deficiency Hazard

Chapter 14

Oxygen Deficiency Hazard in a Storage Tank Elena Stefana, PhD; Filippo Marciano, PhD; Daniel Drolet, MSc; Thomas W. Armstrong, PhD, CIH, FAIHA Oxygen deficiency hazard (ODH) occurs when the indoor oxygen (O2) content drops to a level that may expose workers to the risk of asphyxiation.1 ODH is different from traditional chemical air contamination scenarios and typical inhalation exposures because of the following peculiarities that are described throughout the chapter:1 • There is no consensus on a unique safe exposure limit for the O2 content that can be used for judging the acceptability of exposures. • The O2 reduction and/or depletion can be created by other substances potentially present and/or released in the working environment, which are responsible for the displacement of the O2 available in the air.

Scenario A worker enters an above-ground stainless-steel storage tank. They enter through a manhole to perform metal inert gas (MIG) welding. The tank is a confined space because it (1) is large enough and so configured that an employee can bodily enter and perform assigned work, (2) has limited or restricted means for entry or exit, and (3) is not designed for continuous occupancy.2,3 To allow the worker to safely enter the tank and perform the required activity, the tank is first purged with nitrogen (N2). This will remove flammable or explosive vapors or gases and decrease the O2 concentration to a target value of about 2% by volume. Then, the tank will be ventilated with a forced outdoor airflow to increase the O2 level inside for safe occupancy. The nitrogen purging process is 3,600 sec (60 min). After the end of the purging, 400 sec elapse, and then ventilation with outside air starts and lasts 3,600 sec. Fifteen minutes (900 sec) later, the MIG welding process begins and lasts 1,500 sec (25 min). The MIG process uses argon (Ar) as the inert gas. The timeline of the scenario is schematically displayed in Figure 14.1.

Figure 14.1: Timeline of the scenario.

Copyright AIHA®. For personal use only. Do not distribute.

107

Modeling Occupational Inhalation Exposures to Chemicals Chapter 14: Oxygen Deficiency Hazard

Information Gathering Information About the Substance ODH is different than most other chemical exposure hazards in that it is the absence of a chemical, O2, that is the concern rather than the presence of a chemical. A concentration of < 18% at sea level (corresponding to a partial pressure of O2 of 136.8 mmHg) tends to be where symptoms and physiological effects begin. Appendix A of this chapter contains an overview of O2 deficiency health effects and causes. In the scenario under investigation, N2 is used for purging the tank, while Ar is involved in the MIG welding. The N2 purging flow rate is equal to 0.025 m3/sec, begins at t = 0 sec, and lasts for 3,600 sec; the welding flow rate equals 0.010 m3/sec, starts at t = 8,500 sec, and ends at t = 10,000 sec. Both N2 and Ar flows are not completely pure (there are some impurities, equal to 1% O2). The temperature and pressure are 20°C (293.15 K) and 101.325 kPa, respectively.

Information About the Room The storage tank where the worker performs the Figure 14.2: Storage tank. welding activity is a vertical cylinder with the radius equal to 1.3 m and a height of 5.65 m. Therefore, the total volume is equal to 30.0 m3. Initially, the in-tank air composition is the same as the outdoor air composition: CN2 = 78.1%, CO2 = 20.9%, and CAr = 1.0%. The air temperature is 20°C (293.15 K), and the air pressure is 101.325 kPa both inside and outside the tank. From t = 4,000 sec to t = 7,600 sec, the tank is ventilated with outdoor air by means of a forced supply airflow with a flow rate equal to 0.018 m3/s. Figure 14.2 schematically displays the tank under investigation.

Modeling Approach: Near-Field/Far-Field Model Occupational exposures to an O2-reduced atmosphere can be estimated by means of mathematical models predicting indoor O2 levels.4 A near-field/far-field (NF/FF) model represents a valuable tool for ODH assessments because it permits estimating the O2 levels near the release point(s) of asphyxiant gases.5 A NF/FF model is proposed by Stefana et al.4 to estimate the time profile of indoor O2 concentrations by volume and partial pressure in any working environment where inert gas releases can occur. The model can include conditions other than standard temperature and pressure. It can also consider multiple and simultaneous releases (if these can be assumed as point), mechanically supplied or exhausted ventilation airflows, and natural ventilation flows. The breathing zone of the exposed worker and the sources of inert gas releases are located in the NF, whereas forced and natural ventilation flows occur in the FF. In this scenario, the NF is considered a full open cylinder, the bottom of which starts 0.3 m above the floor of the tank. The cylinder has a radius of 0.5 m and a height of 1.0 m. This gives the NF a volume equal to 0.7854 m3 and a free surface area (FSA) of 4.712 m2. A schematic representation of the NF and FF model with the different flows during the welding activity is proposed in Figure 14.3.

108

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 14: Oxygen Deficiency Hazard This model is based on the ideal gas law, balances of mass of air, of moles of O2, and of internal energy in the NF and FF. The model calculates mass balances for discrete time steps. Conceptually, the amount of O2 in the NF at the end of a time step is as follows: • the amount of O2 in the NF from the previous time step; • plus O2 added to the NF from the welding gas release; • plus O2 carried into the NF from the FF by airflow between the fields; • minus O2 carried from the NF into the FF by airflow between the fields; • plus or minus O2 exchange between the fields due to pressure differences created by NF processes. Similarly, the amount of O2 in the FF at the end of a time step is as follows: • the amount of O2 in the FF from the previous time step; • plus O2 added to the FF by mechanical ventilation; • minus O2 removed from the FF by mechanical ventilation;

Figure 14.3: Near-field/far-field (NF/FF) model in the tank during the welding.

• minus O2 carried from the FF into the NF by airflow between the fields; • plus O2 carried into the FF from the NF by airflow between the fields; • plus or minus O2 exchange between the fields due to pressure differences created by NF processes; • plus or minus O2 exchange thanks to the natural ventilation flow between the FF and the outside environment due to pressure differences created by FF processes. The full time-step mass balance model that accounts for these conceptual steps is presented in full in Appendix B of this chapter. The complexity of the model is simplified for the user in ODHMOD, which is an Excel spreadsheet that performs the extensive calculations for the user and quickly provides answers to exposure assessment questions around ODHs. From the scenario description above, we can collect the information needed to use the ODHMOD spreadsheet. • The atmospheric pressure is 101.325 kPa both inside and outside the tank, and the temperature outside as well as inside the tank is 20ºC. • The composition of outdoor air is 78.1% N2, 20.9% O2, and 1% Ar, which is also the initial composition of the NF and the FF. • The volume of the tank is 30 m3. • The geometry of the NF is a full cylinder with an open top and bottom having a radius of 0.5 m and a height of 1 m. • The first inert gas release is the 3,600 sec purge using 99% N2 and 1% O2 gas at a rate of 0.025 m3/min. • The forced supply ventilation of outdoor air starts at 4,000 sec and lasts 3,600 sec. • The second inert gas release is the welding process, which starts at 8,500 sec and lasts 1,500 sec. The gas is a 99% Ar, 1% O2 gas mixture and is released in the NF at a rate of 0.010 m3/sec. Copyright AIHA®. For personal use only. Do not distribute.

109

Modeling Occupational Inhalation Exposures to Chemicals Chapter 14: Oxygen Deficiency Hazard All this information can be entered into the ODHMOD spreadsheet, as shown in Figure 14.4. The time step, t, must be set to 2 sec to have the spreadsheet solve for the 10,000 sec of the overall process.

Figure 14.4: Summary of the input values in the investigated scenario (from ODHMOD, bit.ly/3QSjTbv).

110

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 14: Oxygen Deficiency Hazard

Interpreting the Modeled Outputs (Concentration and Partial Pressure) The application of the NF/FF model permits obtaining the O2 concentration by volume and partial pressure over time in both the zones. Figure 14.5 shows the time trends of O2 levels in the NF and FF of the storage tank considering all the activities of the scenario (i.e., purging, ventilation, and welding), whereas Figure 14.6 focuses on the time interval when the worker performs the welding operation in the tank. Note that, for convenience, the time period in Figure 14.6 ranges between 0 and 1,500 sec starting at the beginning of the welding activity. Both these figures have two y-axes: the left axis reports the values of the O2 concentrations by volume, and the right axis shows the values of the O2 partial pressure (in mmHg) for the two zones.

Figure 14.5: O2 levels in the near field (NF) and far field (FF) in the entire scenario.

Copyright AIHA®. For personal use only. Do not distribute.

111

Modeling Occupational Inhalation Exposures to Chemicals Chapter 14: Oxygen Deficiency Hazard

Figure 14.6: O2 levels in the near field (NF) and far field (FF) during the welding inside the tank.

Figure 14.6 highlights that when the worker is in the tank, before the beginning of welding activity, the O2 content is acceptable: the concentration and partial pressure values in the NF are equal to 18.7% and 142.3 mmHg, respectively. During the welding, the O2 level continues to decrease until the end of the activity, when the O2 content in the proximity of the Ar release is equal to 11.1% in terms of concentration by volume and to 84.1 mmHg in terms of partial pressure. These are critical values from an ODH point of view: in accordance with Table 14.2 in Appendix A, the worker may experience severe physiological effects, such as loss of consciousness, possible damage to the heart, and/or poor muscular coordination. The time to reach the O2 concentration and partial pressure values of concern (refer to Table 14.2) in the NF is reported in Table 14.1. Note that these time values refer to the time period from the instant when the worker begins the welding activity in the tank. Table 14.1: O2 Values of Concern and Time to Reach Them From the Beginning of the Welding

O2 Concentration by Volume (%) 18 15 12

112

O2 Partial Pressure (mmHg) 137.0 114.0 91.2

Time to Reach the O2 Level (sec) ≈4 ≈ 510 ≈ 1,230

Time to Reach the O2 Level (min) 90%. 136

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 16: Furniture Stripping With MeCl2: Introducing the ART Tool The answers to the questions regarding local controls, surface contamination, fugitive emissions, the dispersion questions, and secondary sources will be the same as those for Activity 1 for our scenario. Table 16.3 summarizes the activity configuration inputs required for the ART model and how they differ (or not) for Activities 1 and 2. Review the narrative of the scenario and the discussion of inputs above to assess the source of the values. Table 16.3: ART Model Example Activity Configuration Inputs

Input Product type Process temperature Vapor pressure Liquid mole fraction Activity coefficient Source proximity Activity class Rate Level of contamination Primary localized controls Full enclosure Effective housekeeping General housekeeping Site Room size Ventilation rate Secondary emission source

Activity 1: Applying Liquid Room temperature 46,663 Pa Main component 1 Yes (within 1 m) Spreading of liquids 0.3–1.0 m2/hr Not applicable No localized controls No No Yes Indoors 100 m2 3 ACH No

Activity 2: Cleaning Liquid Room temperature 46,663 Pa Main component 1 Yes (within 1 m) Handling of contaminated objects 0.3–1.0 m2/hr > 90% of surface No localized controls No No Yes Indoors 100 m2 3 ACH No

ACH, air changes per hour.

Running the Art V1.5 Model and Interpreting the Results Once the ART inputs are entered, select “Finish & Run.” The model will “run,” and the results will be given. The default outputs are full-shift TWA-8, the 75th percentile of the exposure distribution, and interquartile confidence limits (25–75%). Figure 16.2 illustrates the default results of our model. To see the long-term average (over months/years rather than a shift), select “Long-Term” from the dropdown box that currently displays “Full Shift” (See Figure 16.2). If a different percentile of the distribution is needed, e.g., the 50th percentile, select “50th (Median)” from the dropdown menu labeled “Percentile.”

Copyright AIHA®. For personal use only. Do not distribute.

Figure 16.2: ART V1.5 default output for MeCl2 paint stripping exposure scenario.

137

Modeling Occupational Inhalation Exposures to Chemicals Chapter 16: Furniture Stripping With MeCl2: Introducing the ART Tool The confidence interval can also be changed and ranges from interquartile (25–75%) up to 95%. As the confidence level is increased, the width of the confidence interval on the exposure estimate will also increase. If a printed report containing all the inputs and the results is desired, it can be downloaded in either PDF or Excel format with the link located on the bottom left of the screen. Change the output selections and see if you can get the same answers to the questions below. Using the ART V1.5 model, estimate the long-term 50th percentile exposures of workers performing stripping and washing/rinsing. Answer: 770 mg/m3 MeCl2 Estimate the 90th percentile of the exposure distribution for the same workers. Answer: 1,800 mg/m3 Edit the activity configurations and see if you can answer the following questions using the ART V1.5 model: Estimate the 50th percentile full-shift TWA-8 if the room ventilation is increased to 10 ACH. Answer: 360 mg/m3 If increasing general ventilation was the only available exposure control, the ART mechanistic model estimates the 50th percentile exposure ventilation would be reduced from 640 to 360 mg/m3. Estimate the full-shift 50th percentile exposure if general ventilation were increased to 10 ACH and a new stripper were substituted that contains 40% MeCl2 instead of 50–90%. Answer: 170 mg/m3 As expected, ventilation and substitution result in a reduction in the 50th percentile of the exposure distribution from 650 mg/m3 to 170 mg/m3. The median concentration of MeCl2 for the initial configurations of 640 mg/m3 is far above its permissible exposure limit (PEL) TWA-8 (87 mg/m3), the PEL-short-term exposure limit (STEL) (434 mg/m3), and threshold limit value (TLV) (174 mg/m3). Based on these modeling results, it appears that increasing the general ventilation (more air changes per hour) helps some; substitution of another stripping solution combined with increased outside air reduces the concentration further, but not enough to keep the estimated exposures to acceptable levels. Additional controls and interventions are needed. Remember that these estimates are for the 50th percentile of the exposure distribution, meaning that 50% of the full-shift exposures will be higher than 640 mg/m3. In this discussion, we did not make use of the confidence limits provided, but they provide additional valuable information when interpreting the results.

138

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 16: Furniture Stripping With MeCl2: Introducing the ART Tool

Appendix A: More About ART Many evaluation/validation studies have been performed recently on ART V1.0 and V1.5. Specific studies will be discussed as needed in this chapter and where appropriate for applying ART V1.5. ART V1.5 is currently approved by the European Union (EU) to assess exposure to inhalable dust, vapors, and mists. ART cannot (for the time being) be used for the assessment of fumes, fibers, gases, and dust resulting from emissions during hot metallurgical processes due to lack of calibration data (https://www.advancedreachtool. com/). The U.S. EPA acknowledges ART as an occupational exposure assessment tool in the Exposure Factors Handbook.14 ART V2.0 and dART are reported to be in development. ART V2.0 has been reported to allow the use of a wider range of sampling data to refine exposure estimates, and dART is designed to assess dermal exposures.4,5 The ART model works by calculating an “ARTscore,” which is based on the determinants of exposure (refer to Table 16.2) and the use of “Modifying Factors” (MF). For example, a room with 3 ACH would have a different MF than a room with 10 ACH. The ARTscore is a combination of MF for each of the 9 categories of exposure determinants listed in Table 16.1. The ARTscore, which is unitless, then becomes part of a linear mixed-effects model (Equation 16.1). ln(Yijk) = Xijk = ln() + ln(ARTscore) + i + сij + ijk

16.1

Where Yijk: the exposure for the kth measurement within the jth company in the ith scenario Xijk: the natural log-transformed exposure ln : the intercept (natural logarithm of the slope) i: the random effect of the ith scenario сij: the random effect of the jth company in the ith scenario ijk: the residual error term The model uses 2,056 real-world personal sampling data sets to “calibrate” the model to predict exposures (in mg/m3).15,16 For a full explanation of the calibration process, see Schinkel et al.17 The model incorporates between-company and between-worker variance estimates into its exposure estimates. These variance estimates are taken from real-world studies and are included in each analysis.3,18,19 The result is that if several workers are performing the same task, in the same environment, in the same manner, under the same conditions, ART will provide the same exposure estimate for each worker. The result of the regression is a concentration (in mg/m3).

Recent Evaluation and Validation Studies of ART V1.5 Mechanistic Model Lee et al.20 evaluated the performance of ART V1.5 in 42 different exposure scenarios with real-world data and recommend the 50th percentile for estimating real-world measurements for liquids with VP > 10 Pa (~9.8 × 10–5 mmHg). The authors found that the 90th percentile confidence limit underestimated the observed data. For a worst-case estimate, Lee et al.20 recommend using the upper 90th percentile confidence limit of the 90th percentile of the exposure distributions. Landberg et al.11 applied ART V1.5 to exposure scenarios with real-world measured data in 7 industries, 12 companies, and 29 different exposure scenarios. Landberg et al.11 used the 50th and 95th percentiles of the exposure distribution from ART and compared those estimates to the geometric mean (GM) and 95th percentiles of the measured data from those exposure scenarios. The authors found that ART underestimated the exposures relative to the real-world data and worked least well for wood dust. LeBlanc et al.3 also found that ART V1.5 underestimated exposures (using the 50th percentile) by 15%; however, with Bayesian adjustment, the underestimation was corrected.

Copyright AIHA®. For personal use only. Do not distribute.

139

Modeling Occupational Inhalation Exposures to Chemicals Chapter 16: Furniture Stripping With MeCl2: Introducing the ART Tool

Appendix B: Bayesian Module The exposure estimates produced by the ART mechanistic model can be adjusted/refined using the Bayesian algorithms available in ART V1.5. In the ART Bayesian analysis, the mechanistic model result becomes the prior distribution and the actual personal sampling data become the likelihood distribution. With the prior and likelihood distributions, the posterior distribution can be calculated and the “adjusted” exposure estimates can be obtained by combining the modeling information with sampling data. For a more complete discussion of ART and its Bayesian module in ART V1.5, see McNally et al.4 For further discussion of Bayesian analysis and occupational hygiene data, see Hewett et al.21 The sampling data used in the ART Bayesian model must be “fully analogous” to the exposure scenario being modeled. McNally et al.4 investigated the use of partially analogous data, but ART V1.5 always assumes that the data used are fully analogous. ART V2.0 is anticipated to permit the use of partially analogous sampling data.4

Applying the ART V1.5 Bayesian Module to the Scenario To apply the Bayesian model in ART V1.5, fully analogous data need to be supplied to ART. The sampling data can include nondetect (ND) values, but the same restrictions apply, i.e., the data must be from an analogous exposure scenario to the one being modeled. If it is not analogous, ART V1.5 will automatically assume it is and run the analysis under that assumption. McCammon et al.22 studied furniture stripping operations in a variety of facilities. Two of the shops had operations with poor controls/ventilation (like the scenario being modeled here). Personal air samples (TWA-8s) were taken during stripping. The average full-shift exposures for these four workers were sampled as they performed stripping operations in four different companies. The TWA-8s were 124 ppm (430 mg/m3), 277 ppm (961 mg/m3), 122 ppm (423 mg/m3), and 366 ppm (1,270 mg/m3). To illustrate the use of the Bayesian model in ART V1.5, these samples will be uploaded to the Bayesian module of ART V1.5. Note: The sampling data need to be supplied to ART V1.5 in a specific format that is described on the “Upload Analogous Data” screen. By using these data, the assumption is made that the data from McCammon et al.22 are fully analogous to the Scenario Description in this chapter. To try this out with the scenario for this chapter, start with the initial configurations for Activities 1 and 2 that gave a full-shift median estimate of 640 mg/m3. Select “Proceed to Bayesian Model” and then “Upload own analogous data.” The format of the data must be in a comma-separated value (CSV) file with the elements illustrated by following the “see help” link on the “Upload Analogous Data” page. The help pop-up has a link to example data. You can download it, open it with Excel, and modify it with the data above. Be sure to save it again as a csv file. Figure 16.3 illustrates what the file should look like.

Figure 16.3: Format of the analogous data csv file required for ART.

140

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 16: Furniture Stripping With MeCl2: Introducing the ART Tool Choose and upload the file. Click on “Proceed to Bayesian Model.” Then, select “Run Bayesian Model.” The results of running the ART V1.5 Bayesian model are shown in Figure 16.4. The median has been selected for the percentile.

Figure 16.4: Results of ART V1.5 Bayesian Model for Furniture Stripping [18 min applying MeCl2; 81 min cleaning surface, and 81 min washing (unexposed time) and 3 ACH].

Copyright AIHA®. For personal use only. Do not distribute.

141

Modeling Occupational Inhalation Exposures to Chemicals Chapter 16: Furniture Stripping With MeCl2: Introducing the ART Tool

Appendix C: Activity 1 Configuration Screens

Figure 16.8: Liquid mole fraction screen.

Figure 16.5: Product type screen.

Figure 16.6: Process temperature screen.

Figure 16.7: Vapour pressure screen.

142

Figure 16.9: Activity coefficient screen.

Figure 16.10: Primary emission source proximity screen.

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 16: Furniture Stripping With MeCl2: Introducing the ART Tool

Figure 16.14: Surface contamination/fugitive emissions sources screen.

Figure 16.11: Activity class screen.

Figure 16.12: Spreading of liquid products screen.

Figure 16.15: Dispersion screen.

Figure 16.16: Secondary emission source screen.

Figure 16.13: Primary localized controls screen.

Copyright AIHA®. For personal use only. Do not distribute.

143

Modeling Occupational Inhalation Exposures to Chemicals Chapter 16: Furniture Stripping With MeCl2: Introducing the ART Tool

References 1. McCammon CS. Health hazard evaluation report: HETA-92-0360-2372. Investigation of Furniture Stripping Operation at Ackerman & Sons, Littleton, Colorado. NIOSH: Cincinnati, OH, 1993. 2. Fransman W, Van Tongeren M, Cherrie JW, Tischer M, Schneider T, Schinkel J, Kromhout H, Warren N, Goede H, Tielemans E. Advanced Reach Tool (ART): Development of the mechanistic model. The Annals of Occupational Hygiene, 55(9): 957–979, 2011. https://doi.org/10.1093/annhyg/mer083. 3. LeBlanc M, Allen JG, Herrick RF, Stewart JH. Comparison of the near field/far field model and the advanced reach tool (ART) model V1.5: Exposure estimates to benzene during parts washing with mineral spirits. International Journal of Hygiene and Environmental Health, 221(2):231–238, 2018. https://doi.org/10.1016/j. ijheh.2017.10.016. 4. McNally K, Warren N, Fransman W, Entink RK, Schinkel J, van Tongeren M, Cherrie JW, Kromhout H, Schneider T, Tielemans E. Advanced REACH tool: A Bayesian model for occupational exposure assessment. The Annals of Occupational Hygiene, 58(5):551–565, 2014. https://doi.org/10.1093/annhyg/ meu017. 5. McNally K, Gorce JP, Goede HA, Schinkel J, Warren N. Calibration of the dermal Advanced Reach Tool (dART) mechanistic model. Annals of Work Exposures and Health, 63(6):637–650, 2019. https://doi. org/10.1093/annweh/wxz027. 6. TNO report V9009. Development of a Mechanistic Model for the Advanced REACH Tool (ART) Version 1.5. Dutch Ministry of Social Affairs and Employment, Health and Safety Executive: TNO Quality of Life, Utrecht, Netherlands, 2013. 7. Esmen N, Corn M, Hammad Y, Whittier D, Kotsko N. Summary of measurements of employee exposure to airborne dust and fiber in sixteen facilities producing man-made mineral fibers. American Industrial Hygiene Association Journal, 40(2):108–117, 1979. https://doi.org/10.1080/15298667991429408. 8. Cherrie J, Krantz S, Schneider T, Öhberg I, Kampstrup O, Linander W. An experimental simulation of an early rock wool/slag wool production process. Annals of Occupational Hygiene, 31(4B):583–593, 1987. https://doi.org/10.1093/annhyg/31.4B.583. 9. ART User Guide. https://www.advancedreachtool.com/support.aspx (accessed July 19, 2021) p.4. 10. European Chemicals Agency (ECHA). Chapter R.14: Occupational exposure estimation V.2. In: Guidance on Information Requirements Chemical Safety Assessment. 2016. 11. Landberg HE, Axmon A, Westberg H, Tinnerberg H. A study of the validity of two exposure assessment tools: Stoffenmanager and the advanced REACH tool. Annals of Work Exposures and Health, 61(5):575– 588, 2017. https://doi.org/10.1093/annweh/wxx008. 12. National Institutes of Health. PubChem: Methylene Dichloride. https://pubchem.ncbi.nlm.nih.gov/compound/ dichloromethane (accessed July 27, 2021). 13. National Institute for Occupational Safety and Health (NIOSH). NIOSH Pocket Guide to Chemical Hazards [DHHS (NIOSH) Publication Number 2005-149]. Department of Health and Human Services Centers for Disease Control and Prevention. September 2007. 14. U.S. Environmental Protection Agency (EPA). Chapter 19―Building characteristics; Section 19.5.2 [Revised: 2018]. In: Exposure Factors Handbook, EPA/600/R-18/121F, EPA: Washington, D.C., 2011. 15. Schinkel J, Warren N, Fransman W, van Tongeren M, McDonnell P, Voogd E, Cherrie JW, Tischer M, Kromhout H, Tielemans E. Advanced REACH tool (ART): calibration of the mechanistic model. Journal of Environmental Monitoring, 13(5):1374–1382, 2011. https://doi.org/10.1039/C1EM00007A. 16. Tielemans E, Warren N, Fransman W, van Tongeren M, McNally K, Tischer M, Ritchie P, Kromhout H, Schinkel J, Schneider T, Cherrie JW. Advanced REACH tool (ART): overview of version 1.0 and research needs. Annals of Occupational Hygiene, 55(9):949–956, 2011. https://doi.org/10.1093/annhyg/mer094.

144

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 16: Furniture Stripping With MeCl2: Introducing the ART Tool 17. Schinkel J, Fransman W, McDonnell PE, Entink RK, Tielemans E, Kromhout H. Reliability of the advanced REACH tool (ART). Annals of Occupational Hygiene, 58(4):450–468, 2014. https://doi.org/10.1093/annhyg/ met081. 18. Kromhout H, Symanski E, Rappaport SM. A comprehensive evaluation of within-and between-worker components of occupational exposure to chemical agents. Annals of Occupational Hygiene, 37:253–270, 1993. https://doi.org/10.1093/annhyg/37.3.253. 19. Rappaport SM, Weaver M, Taylor D, Kupper L, Susi P. Application of mixed models to assess exposures monitored by construction workers during hot processes. Annals of Occupational Hygiene, 43(7):457–469, 1999. 20. Lee EG, Lamb J, Savic N, Basinas I, Gasic B, Jung C, Kashon ML, Kim J, Tischer M, Van Tongeren M, Vernex D, Harper M. Evaluation of exposure assessment tools under reach: Part II—higher tier tools. Annals of Work Exposures and Health, 63(2):230–241, 2019. https://doi.org/10.1093/annweh/wxy098. 21. Hewett P, Logan P, Mulhausen J, Ramachandran G, Banerjee S. Rating exposure control using bayesian decision analysis. Journal of Occupational and Environmental Hygiene, 3(10):568–581, 2006. https://doi. org/10.1080/15459620600914641. 22. McCammon CS, Glaser RA, Wells VE, Phipps FC, Halperin WE. Exposure of workers engaged in furniture stripping to methylene chloride as determined by environmental and biological monitoring. Applied Occupational and Environmental Hygiene, 6(5):371–379, 1991. https://doi.org/10.1080/104732 2X.1991.10387898.

Copyright AIHA®. For personal use only. Do not distribute.

145

Modeling Occupational Inhalation Exposures to Chemicals Chapter 17: Reconstruction of a Lab Spill

Chapter 17

Reconstruction of a Laboratory Spill Paul Hewett, PhD, CIH, CSP, FAIHA

Introduction At the 2017 AIHce EXP, Jeff Burton described a project where he was asked to estimate the employee exposures that occurred immediately after a spill of concentrated acetic acid. Two employees attempted to clean up the spill (Figure 17.1). Both experienced upper respiratory tract and pulmonary irritation, which required medical treatment. Burton’s objective was to predict the room (i.e., far-field) concentration of acetic acid vapor within the first 20 min of the spill using the one-box, constant emission well-mixed room (WMR) model.1 He estimated the generation rate for a spill of acetic acid using an equation that predicts the emission rate due to “evaporation from open surfaces.”2 Burton concluded that within a few minutes, the far-field concentrations probably rose to “hazardous levels.” The ACGIH short-term exposure limit (STEL) of 15 ppm was probably exceeded within 2 min, and concentrations approached the NIOSH immediately dangerous to life or health (IDLH) level of 50 ppm after 20 min. He observed that the concentrations near the spill (i.e., the near-field concentrations) were most likely greater. The objectives of this chapter are to (1) calculate the likely “near-field” concentrations immediately after the spill using the two-box, constant emission WMR model and (2) demonstrate the probabilistic approach to modeling where statistical distributions are assigned to the important model variables. Refer to Keil and Nicas3 and Nicas4 for more on the two-box, constant emission WMR model.

Figure 17.1: Slides from Burton (2017). The spill was assumed to be roughly 2 ft × 2 ft. The safety cabinet had a measured exhaust rate of roughly 1,000 cubic feet per minute (cfm). Most of the supply air came through the one ceiling register, with the remainder from underneath the hallway door. Copyright AIHA®. For personal use only. Do not distribute.

145

Modeling Occupational Inhalation Exposures to Chemicals Chapter 17: Reconstruction of a Lab Spill

Background Burton used the one-box, WMR transient equation, which predicts the average room concentration (i.e., the far-field, well-mixed concentration) at time t after the spill.1 I will use the two-box, WMR transient equations, which predict both the far- and near-field concentrations at time t.4 Figure 17.2 illustrates the two-box concept. Inside a room (i.e., the “far-field” box) having a ventilation rate Q (m3/min), there is a source emitting a vapor or gas (or a fine particulate) at a reasonably constant rate G (mg/min). This source is surrounded by a conceptual “near field,” or high concentration volume, which a worker may stand in or lean into. The near field must be assigned an approximate ventilation rate  (m3/min), representing the air that moves through the near field. We know that for a constant emission source and a relatively small room, the far- and near-field concentrations will rapidly approach the steady-state concentrations. For example, Figure 17.3 shows a set of predicted far- and near-field concentration curves for this scenario. (The calculation of the curves will be presented later.) Within 10 min, the far- and near-field concentrations approach the steady-state concentrations. At these concentrations, the mass per minute emitted by the source equals the mass removed from the room. Furthermore, the sum of the mass per minute entering the near field from the far field and the mass per minute being generated within the near field equals the mass leaving the near field.

Figure 17.2: Illustration of the two-box, near-field/farfield model. , near-field ventilation rate; G, constant generation rate; Q, room ventilation rate; VN, near-field volume.

Initial Modeling Approaches A Quick Calculation of Possible “Worst Case” Concentrations Before doing any complex calculations using software, we can calculate approximate “worst case” far- and near-field concentrations using the standard two-box, constant emission “steady-state” equations.4

Figure 17.3: Far-field (black) and near-field (grey) concentration curves calculated using the best estimates of the two-box model variables. The farand near-field 30-min averages are 17.3 and 74.0 ppm, respectively.

G CF,SS = — Q

(17.1)

G CN,SS = CF,SS + — 

(17.2)

146

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 17: Reconstruction of a Lab Spill Where CF,SS: steady-state concentration in the far field (mg/m3) CN,SS: steady-state concentration in the near field (mg/m3) G: pollutant emission rate (mg/min) Q: room ventilation rate (m3/min)  : near-field (interzonal) airflow rate (m3/min) Note that Ganser and Hewett5 provide a modified version of the steady-state equations for constant emission sources that turn on and off during the task. The critical variable in any modeling exercise is the emission rate, G. Burton assumed that the size of the spill was reasonably constant for the initial 20 or more minutes and used the Hummel Equation2 to calculate an approximate constant emission rate. The Hummel Equation was developed for estimating the solvent emission rates of fixed surface degreasing tanks. Keil and Nicas3 showed that with a simple modification, the Hummel Equation can be applied to circular spills. Using their approach (refer to Appendix A), a likely emission rate (G) of 1,340 mg/min can be calculated. When calculating near-field concentrations, the near-field ventilation rate () is a critical variable. Using a hemispherical near-field shape, a diameter of 1 m, and an air velocity (measured at the floor level by Burton) of 40 ft/min (fpm; 12.2 m/min), an approximate near-field ventilation rate () of 9.6 m3/min can be calculated (refer to Appendix B). Applying these values—as well as the measured room ventilation rate of 28.3 m3/min—to the above equations results in a far-field concentration at 47.3 mg/m3 (19.2 ppm) and a near-field concentration of 186.9 mg/m3 (76.0 ppm). The steady-state concentrations are the highest concentrations calculated using a “constant emission” WMR model and, as such, represent reasonable “worst case” concentrations. Because it takes a few minutes for the far-field concentrations to reach steady state, the steady-state concentrations slightly overestimate the 30min average far- and near-field concentrations of 17.3 ppm and 74.0 ppm (Figure 17.3). However, these values are clearly useful for the following conclusions: In the far field, the STEL concentration was probably exceeded, whereas in the near field, both the STEL and the IDLH concentrations were probably exceeded. At this point, one could conclude that both the far- and near-field concentrations of acetic acid resulting from a fairly large spill of acetic acid in a relatively small room would rapidly approach hazardous levels. However, let us determine what additional insights can be obtained from a more comprehensive modeling evaluation.

Accuracy of the Well-Mixed Room Models Before we start, let us consider two questions. First, how accurate are the WMR models? Nicas4 suggested that a WMR model with reasonable input values will predict concentrations that are often within 50–200% of the true concentration. Second, how accurate do we need a WMR model to be? Perry Logan (3M, personal conversation) observed that we are often interested in making green, yellow, and red light decisions. A green light decision is one where the model suggests that the actual exposure profile can be assigned a provisional AIHA exposure control rating6 of Category 0, 1, or 2. That is, the true 95th percentile exposure is likely to be less than 1%, 10%, or 50% of the occupational exposure limit (OEL). A red light decision is one where the model suggests a high Category 4 exposure profile, where the true 95th percentile is much greater than the OEL. In either case, high prediction accuracy is not strictly necessary, as it is unlikely that different input values will substantially change the predicted exposure profile and the resulting decision. A yellow light decision is a Category 3 or a low Category 4 decision. Here, it is more likely that different input values could result in a green light or a red light decision. In this situation, the model could be improved by identifying the critical variables and improving the estimates of each variable or collecting air concentration measurements. In summary, it is usually not necessary to agonize over the specific values assigned to the model variables. Reasonable, perhaps first-approximation values are usually all that are necessary. For example, for range finding and semi-quantitative exposure assessment using WMR models, we often use default values for the near-field volume and flow rate. However, if a WMR model for a commonly occurring unit operation is expected to be used frequently throughout a business entity, some level of model prediction validation should be considered.5

Copyright AIHA®. For personal use only. Do not distribute.

147

Modeling Occupational Inhalation Exposures to Chemicals Chapter 17: Reconstruction of a Lab Spill

Probabilistic Modeling In this example, I will use the systematic approach to WMR modeling depicted in Figure 17.4.

Start Select your modeling software. The WMR modeling equations1,4 can be programmed into a spreadsheet, but most modelers use the IH modeling spreadsheet IHMod2.xlsx (available from AIHA). It is capable of both deterministic and probabilistic modeling using either the one- or two-box WMR model. In this chapter, I will use the Task Exposure Assessment Simulator (TEAS). This is a commercially available program that I developed. (A student version will be available in 2023 from www.easinc.co.)

State the Problem There was a spill of a concentrated solution of acetic acid from a 750-mL bottle in a small hospital lab and storage room. Two employees attempted to clean up the spill using sponges, a bucket, rubber gloves, and cleaning materials. Both experienced upper respiratory tract and pulmonary irritation, which required medical treatment. The goal of this analysis is to (1) reconstruct the spill incident and predict far-field (i.e., general room) concentrations and near-field concentrations and (2) compare the predictions to the STEL and IDLH for acetic acid.

Figure 17.4: A systemic approach to well-mixed room modeling.

Define the Scenario During this step, assemble the needed information.

Chemical Information Table 17.1 presents information about the chemical, acetic acid, which was used as an 80–90% solution. 17.1: Information on Acetic Acid (CAS 64-19-7)

Physical Properties

Health Effects Exposure Guidelines

Molecular weight 60.1 Liquid density 1.05 g/mL Vapor pressure 11 mmHg (at 20°C) Lower explosive limit 4% = 98,000 mg/m3 Upper respiratory tract and eye irritation; pulmonary function effects OSHA PEL TWA-8 10 ppm ACGIH TLV TWA-8 10 ppm NIOSH IDLH 50 ppm (Based on acute inhalation toxicity data in humans. More than 50 ppm is intolerable to most persons.)

IDLH, immediately dangerous to life or health; PEL, permissible exposure limit; TLV, threshold limit value; TWA-8, 8-hr time-weighted average. 148

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 17: Reconstruction of a Lab Spill

Room and Ventilation Information The effective room volume, V (i.e., the actual volume minus the volume of objects, such as cabinets), was estimated to be 85 m3. The room ventilation rate, Q, was provided by a lab hood and measured to be 28.3 m3/min. There were no local controls.

Process Information The task was the attempted cleanup of a spill on the floor of concentrated acetic acid. The spill was estimated by the employees to be approximately 2 ft in diameter. No respiratory protection was worn by the cleanup workers.

Add One or More Tasks In this critical step, we must select a one- or two-box WMR model and identify all the variables that must be assigned either a single value (i.e., a constant) or a statistical distribution (e.g., a uniform, triangular, normal, or lognormal distribution). If all variables are assigned a constant, the WMR model is considered to be deterministic; that is, it produces a single estimate of concentration or exposure. A deterministic approach can generate low, medium, and high concentration estimates by manually changing the input variables. If one or more variables is assigned a statistical distribution, the WMR model is considered to be probabilistic; that is, it produces an estimate of the exposure profile. The distribution assigned to a variable represents our attempt to describe the true task-to-task or day-to-day variability for the variable. The distribution can also reflect our uncertainty in values assigned to the variable. I selected the two-box (aka two-zone) WMR model, which estimates both far-field and near-field concentrations. Next, we must choose between a “constant emission” WMR model or a “decreasing emission” WMR model. A constant emission model is appropriate for processes that emit a gas, vapor, or fine particulate (e.g., welding fumes) at a reasonably constant rate. We then must determine the generation rate (G; mg/min) for the process when it is emitting. The generation rate, G, is basically the mass emitted per unit time. Keil2 described several models for predicting the generation rate: e.g., the “complete evaporation” model, where a specific volume evaporates within a fixed time period; the “air displacement” model for container filling; and the “open surface” model for degreasing and cleaning tanks. Keil and Nicas3 showed how the open surface model can be used to estimate the emission rate from a circular spill. Their recommendation will be used and is described in Appendix B. A decreasing emission model is appropriate whenever the substance emission is due to evaporation (e.g., the evaporation of solvents from a spill or from freshly applied paint) or emission of a vapor or gas from an object (e.g., formaldehyde from manufactured wood products or ethylene oxide from a bag of sterilized surgical instruments). For decreasing emission WMR models, it is assumed that the generation rate is not constant but varies with time. For example, the generation rate for a paint solvent is greatest immediately after application of the paint and is assumed to decrease exponentially until the solvent has completely evaporated.7 Calculation of the time-dependent generation, G(t), requires knowledge of the mass of the substance available for evaporation and the emission rate constant.3,7 This scenario involves a spill, which will gradually decrease in size due to evaporation. However, because we are attempting to estimate short-term and peak exposures shortly after the spill, we will assume that the surface area of the spill did not change substantially during the initial 15–30 min. The variables for the two-box, constant emission WMR model, as used in the TEAS program, are listed in Table 17.2. The TEAS program allows the user to describe a job as either a single task or a series of different tasks. The duration and frequency of each task can also be modified to match the scenario. In this scenario, we are attempting to reconstruct a single event; therefore, the frequency was set to 1 and the duration was set to any period of interest, such as the first 30 min after the spill.

Copyright AIHA®. For personal use only. Do not distribute.

149

Modeling Occupational Inhalation Exposures to Chemicals Chapter 17: Reconstruction of a Lab Spill Table 17.2: Variables for the “Two-Box, Constant Emission” WMR Model

Variable Frequency Duration G Q V VN β tg t1 t2

Units min mg/min m3/min m3 m3 m3/min min min min

Statistical Model Constant Constant Triangular Constant Constant Constant Triangular Constant Constant Constant

Constant (or Mode) 1 30 1,330 28.3 85 0.26 9.6 30 0 0 or 30

Min

Max

810

1,900

7.1

11.9

Comments a b c

d e f g g

All variables must be assigned either a constant or a statistical distribution, such as uniform, triangular, normal, or lognormal distribution. a – In TEAS, the Frequency is used for repeating tasks. Here we have a single incident. b – Set the Duration of the task or the time of interest. c – Refer to Appendix A. d – Based on a 1-m diameter hemisphere on the floor surrounding the spill. (Refer to Appendix B.) e – Refer to Appendix B. f – The tg variable indicates the time that the “source” is emitting. In this scenario, it is assumed that the spill is emitting vapor at a constant rate during the 30 min of interest. g – In TEAS, the worker is assumed to enter the near field at time t1 and leave the near field at time t2. For far-field concentrations, set both to 0. The two critical variables for a two-box constant emission WMR model are the generation rate, G (mg/min), and near-field ventilation rate,  (m3/min). I describe how I modeled both variables in Appendices A and B. For each, I chose to use a triangular distribution. The triangular distribution—which consists of a mode (i.e., the most frequently occurring value), min, and max—is perhaps the most popular distribution used in modeling for describing the actual distribution for a variable and the variability and/or uncertainty in the estimates of the variable. Following Burton, I used the Hummel Equation2 for estimating the generation rate of a spill with a fixed surface area but added variability to several of the variables to reflect uncertainty. I also followed the recommendations of Keil and Nicas3 for applying the Hummel Equation to circular spills (refer to Appendix A). In the two-box model, the near-field volume (VN) represents a zone of high concentrations near the source. Cube, box, or cylinder shapes have been used in the literature; however, in recent years, the hemisphere seems to be the most popular near-field shape and will be used here. I set the diameter at 1 m, which seemed a reasonable width for encompassing both the spill and the upper torso of an employee kneeling nearby. To estimate the near-field flow rate  (m3/min), I used Burton’s measured air velocity at the floor of 40 fpm but added ± 25% to account for possible day-to-day variability and measurement error. This allowed me to model  using a triangular distribution. The details regarding the calculation of VN and  are in Appendix B. The room volume V (m3) was calculated, and the room ventilation rate Q (m3/min) was measured by Burton. The time of generation (tg) was set to 30 min. In TEAS, the t1 and t2 variables (Table 17.2) are used to indicate the location of the worker relative to the near field. The worker is assumed to enter the near field at time t1 and leave the near field at time t2. This allows flexibility in describing patterns of worker exposure. To calculate near-field concentrations, set t1 = 0 and t1 = duration. To calculate far-field concentrations, set both to zero. TEAS then assumes that the worker is in the far field for the duration of the task. 150

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 17: Reconstruction of a Lab Spill Comments: 1. We will ignore any evaporative cooling effects, which will slightly reduce the emission rate. 2. A spill can also be treated as a decreasing emissions scenario, but we would need to know the mass of acetic acid that was spilled and a scenario-specific emission rate constant for acetic acid. 3. We are not trying to predict an “exposure profile.” This was a one-time exposure scenario. We are attempting to describe the range of possible exposures experienced by the first responders to this spill.

Simulate Random Exposures Test the Model With TEAS, you can simulate the concentration curve for a single task or a series of tasks. First, test that the WMR model is predicting what you think it should by generating several single-task concentration curves. Four are shown in Figure 17.5. Charts A and B in the figure show examples of the far-field concentration curve. Charts C and D show examples of the near-field concentration curve. In each instance, TEAS generated random values for the G and  variables. The remaining variables were assigned the fixed values in Table 17.2. The nearand far-field concentration curves—which represent the concentration at each time, t—were calculated using Equations 15.3–15.6 from Chapter 15. For each chart in Figure 17.5, the black and blue curves represent the concentration curve (i.e., the concentration at time t) and task average curve (i.e., the average concentration for the duration of the task). For reference,

Figure 17.5: Example concentration curves. Charts A and B represent far-field concentrations. Charts C and D represent near-field concentrations. IDLH, immediately dangerous to life or health; STEL, short-term exposure limit.

Copyright AIHA®. For personal use only. Do not distribute.

151

Modeling Occupational Inhalation Exposures to Chemicals Chapter 17: Reconstruction of a Lab Spill the STEL and IDLH values are also shown. The curves show that the far- and near-field concentrations tended to exceed the STEL. The near-field concentrations tended to exceed the IDLH. Both the far- and near-field instantaneous concentrations approached a steady-state concentration within 10–15 min. In conclusion, the predicted near- and far-field concentration curves meet our expectations.

Simulate a Distribution of Concentrations Using TEAS, I generated 10,000 simulated far-field and near-field concentration curves. TEAS calculated the 30min average exposure for each curve. The resulting 10,000 thirty-minute average concentrations were displayed as a histogram. (Log-probability and cumulative probability plots are also available.) Figures 17.6 and 17.7 show the histograms of the far- and near-field predicted 30-min concentrations. Most of the random far- and near-field averages exceeded the STEL and IDLH, respectively.

Figure 17.6: Predicted far-field exposure profile for the 30-min average concentrations after the spill. IDLH, immediately dangerous to life or health; STEL, short-term exposure limit.

152

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 17: Reconstruction of a Lab Spill

Figure 17.7: Predicted near-field exposure profile for the 30-min average concentrations after the spill. IDLH, immediately dangerous to life or health; STEL, short-term exposure limit.

Evaluate the Results Table 17.3 contains the nonparametric statistics for near- and far-field simulations. The predicted far- and nearfield 95th percentile exposures exceed the STEL and IDLH, respectively. From Table 17.4, we note that 78% of the predicted task average (i.e., 30 min) far-field concentrations exceeded the STEL. For the near field, 100% of predicted task average concentrations exceeded the STEL and 98% exceeded the IDLH. Comments: 1. It is clear from the near-field concentration curves that virtually any 15-period time-weighted average (TWA) would have exceeded the STEL and most likely the IDLH. 2. The values plotted in Figures 17.6 and 17.7 represent the “average” concentrations within either the far field or the near field. One of the assumptions of the WMR models is that the vapor is “well mixed” within both the far and near fields. 3. Uniform concentrations in the far and near fields will not happen in reality, but experience using the models has shown that the concentrations are often relatively uniform. However, in this scenario, relatively uniform concentrations should not be expected. This is because the room had a single air register and a single exhaust point, which resulted in non-random, directional air movement through the room. 4. Remember, we are aiming not at a perfect model and perfect predictions. The goal in this modeling exercise is to generate reasonable estimates of exposure for determining whether the employees were exposed to high levels of acetic acid.

Copyright AIHA®. For personal use only. Do not distribute.

153

Modeling Occupational Inhalation Exposures to Chemicals Chapter 17: Reconstruction of a Lab Spill Table 17.3: Simulation Statistics

Nonparametric Statistics Minimum 5th percentile 50th percentile (median) 95th percentile Maximum

Far Field (ppm) 11 13 17 22 25

Near Field (ppm) 40 53 75 99 125

Sample size (n) = 10,000 simulations. Immediately dangerous to life or health (IDLH) = 50 ppm; short-term exposure limit (STEL) = 15 ppm. Table 17.4: Fraction of Simulated Concentrations Exceeding the STEL and IDLH

Nonparametric Statistics Fraction > STEL Fraction > IDLH

Far Field 78% 0%

Near Field 100% 98%

Sample size (n) = 10,000 simulations. Immediately dangerous to life or health (IDLH) = 15 ppm; short-term exposure limit (STEL) = 50 ppm.

Summary of Modeling Results The results of applying the two-box, WMR model strongly suggest that both far-field and near-field concentrations would rapidly increase. Within minutes, the STEL would likely be exceeded in the far field and the IDLH would almost certainly be exceeded within the near field. Figure 17.5 suggests that concentrations within the near-field approach and exceed the IDLH level within minutes and reach an equilibrium concentration within 10 min. The far- and near-field histograms in Figures 17.6 and 17.7 show the range of possible far- and near-field 30-min average concentrations. The variability in the far-field predicted values reflects the uncertainty in the generation rate calculation for a circular spill (Appendix A). Regardless of the uncertainty in the estimates for G, the predicted concentrations strongly suggest that the far-field levels exceeded the STEL and would represent an eye and upper airways irritation hazard. The variability in the near-field predicted values reflects the uncertainty in both the generation rate and the estimates of the near-field flow rate,  (Appendix B). Regardless of the uncertainty in these estimates, the predicted near-field concentrations suggest that concentrations exceeded the IDLH, resulting in a health hazard for any employee bending over or near the spill. Based on WMR developed for this scenario, this is clearly a “red light” simulation. It is unlikely that reasonable changes to the input variables will substantially alter the predicted instantaneous and average concentrations.

Discussion Burton concluded that it is possible the far-field acetic acid concentrations reached hazardous levels within a few minutes. Furthermore, he observed that concentrations near the spill were probably greater. As a result of this incident and Burton’s estimates of the likely exposures, protocols were established for handling future spill incidents. Burton observed that similar WMR modeling approaches could be used to predict the concentrations from similar incidents and help guide the development of emergency response guidelines. Burton used the one-box, WMR transient equation, which predicts the average room concentration (i.e., the far-field, well-mixed concentration) at time t after the spill.1 I used the two-box, WMR transient equations, which predict both the far- and near-field concentrations at time t.4 Figure 17.3 shows that the concentration in the near

154

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 17: Reconstruction of a Lab Spill field will rapidly approach equilibrium (under the assumption that the emission rate for the spill was reasonably constant for the initial 15–30 min). If the room is fairly small, the far-field concentration will also rapidly approach equilibrium. When this is the case, the much simpler steady-state equations for the two-box model4 can provide reasonable predictions of far- and near-field worst-case exposures. The first objective of this chapter was to estimate the near-field concentrations. Considering the predicted nearfield concentrations, it is no surprise that the employees experienced health effects while cleaning up the spill. As Burton pointed out, even the far-field, general room concentrations approached IDLH levels. The second objective was to demonstrate probabilistic modeling. The resulting histograms of predicted concentrations are usually interpreted as an estimate of the true exposure profile (for the job or task). From the random concentrations, we can determine the probable minimum and maximum values, the 5th and 95th percentiles, and estimates of the most likely or typical values, such as the median, mean, and geometric mean. In this scenario, histograms in Figures 17.6 and 17.7 represent the possible 30-min average concentrations for a single incident.

Assumptions of WMR Models Versus Reality The instantaneous and 30-min TWA concentrations predicted by the two-box WMR model are accurate, provided the actual conditions match the assumptions built into the model. As discussed by Nicas,4 the WMR predictions are reasonably accurate provided the input values to the WMR are reasonable. One of the WMR assumptions is that the room is well mixed, resulting in a relatively uniform concentration throughout the room. In this scenario, we know that it is unlikely that both the room and the near-field volume were well mixed. There was one supply air inlet and one exhaust point. This most likely resulted in a channeling effect that prevented uniform mixing within the far field. The two-box model also assumes that the air within the hypothetical near field is well mixed. Furthermore, the model assumes that the air moving into and out of the near field is not directional. In other words, the air direction is random at all locations on the interface surface of the near field. In this scenario—based on Burton’s description and air velocity measurements—there appeared to be a directionality to the air movement, where relatively clean air entered on one side of the near field, picked up vapor from the spill, and exited at the opposite side in the direction of the exhaust hood. My view is to imagine a near field that is not centered on the spill but shifted in the direction of the room exhaust (i.e., the safety cabinet). In summary, we do not expect the WMR model to routinely provide highly accurate estimates of exposure. However, we expect that the estimates will be useful for making green light and red light decisions. In this case, Burton demonstrated that the WMR predictions could be useful for showing that the far-field concentrations experienced by the employees engaged in the cleanup of an acetic acid spill most likely approached and exceeded the acetic acid STEL and approached the IDLH. He also demonstrated the utility of WMR modeling for predicting appropriate emergency response actions. In this chapter, we showed that the near-field concentrations almost certainly exceeded the STEL and more than likely exceeded the IDLH.

Adding Variability or Uncertainty to Additional Model Variables With deterministic modeling, a single value is assigned to each variable. With probabilistic modeling, a simple range (i.e., minimum and maximum) can be assigned to any or all variables. I prefer to use a triangular distribution where the minimum and maximum can be specified, as well as mode (i.e., most likely value). The statistical distribution—whether it is uniform, triangular, normal, or lognormal—represents my attempt to describe the natural day-to-day or task-to-task variability for that variable. I might make the statistical distribution wider to partially account for my uncertainty.

Advantages of a Probabilistic Approach to Modeling As was demonstrated, probabilistic modeling allows you to assign a range or distribution to the critical variables and produces a histogram (i.e., distribution) of probable exposures. The histogram can be interpreted as an approximate exposure profile or, as in this case, a set of possible concentrations for a single event. Probabilistic modeling can also be used to explore “what if” scenarios. For example, we could use WMR modeling to predict concentrations for spills of other solvents for different size spills, larger or smaller rooms, different ventilation rates, and so on.

Copyright AIHA®. For personal use only. Do not distribute.

155

Modeling Occupational Inhalation Exposures to Chemicals Chapter 17: Reconstruction of a Lab Spill Variables for one “what if” scenario are shown in Table 17.5. I evaluated the effect of different room sizes and a wide range of ventilation rates (expressed in air changes per hour). The resulting far- and near-field histograms are in Figure 17.8. The predicted far-field concentrations tended to approach and exceed the IDLH, suggesting that in rooms with smaller volumes and/or less ventilation, even the far-field levels may exceed the IDLH. However, the predicted near-field 30-min TWA concentrations are virtually identical to those predicted earlier. This is because the near-field concentrations are strongly influenced by the near-field ventilation rate. The size of the room and the room ventilation rate have less of an effect on near-field exposures. Table 17.5: Variables for a “What If” Simulation

Variable Frequency Duration G Q V VN β tg t1 t2

Units min mg/min ACH m3 m3 m3/min min min min

Statistical Model Constant Constant Triangular Uniform Uniform Constant Triangular Constant Constant Constant

Constant (or Mode) 1 30 1,330

0.26 9.6 30 0 0 or 30

Min

Max

Comments

810 5 50

1,900 20 200

a

7.1

11.9

a – In this simulation, the Q unit was ACH, or air changes per hour. For each simulated 30-min task, a random ACH was selected. The ventilation rate (in m3/min) was calculated from the random ACH value and the room volume.

Figure 17.8: Far-field (light) and near-field (dark) 30-min concentrations for the “what if” scenario. IDLH, immediately dangerous to life or health; STEL, short-term exposure limit. 156

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 17: Reconstruction of a Lab Spill

An Alternative Approach Using a Decreasing Emissions Scenario In principle, the decreasing emission two-box WMR model would have been more appropriate for a scenario involving evaporation. Decreasing emission models calculate the constantly changing generation rate G(t) using the mass M (mg) that was applied (or spilled) on a surface and the emission rate constant  (per min). Although M can be easily calculated from available information on the volume spilled and the density of acetic acid,  is usually experimentally determined and will vary with the quantity applied (or spilled), the substrate (or spill surface), and the environmental conditions (e.g., air and surface temperatures, air velocity, humidity). For example,  will vary considerably for spills on linoleum, concrete, carpet, and wood.

Conclusion Applying WMR models can be used to estimate past exposures, as was shown here, as well as predict future exposures. A probabilistic approach, where reasonable ranges are assigned to the critical variables, provides estimates of the task or job exposure profile and allows us to explore the likely effects on exposures for different exposure scenarios.

Acknowledgments I am extremely grateful to Jeff Burton for allowing me to use his 2017 AIHce presentation as the basis for this chapter.

Copyright AIHA®. For personal use only. Do not distribute.

157

Modeling Occupational Inhalation Exposures to Chemicals Chapter 17: Reconstruction of a Lab Spill

Appendix A: Estimating the Generation Rate The Hummel and Fahrenbacher “Open Surface” model (hereafter referred to as the Hummel Equation) was designed for estimating the steady evaporation rate from an open tank (e.g., degreasing tank), open container (e.g., an open jar), or a fixed surface area spill2,3: 0.25 1 1 0.166 · MW · PV · ——— + —— · A MW 29 G = ———————————————————————— · T0.05 0.833

√

S 1,000 mg ————— · fs · ————— L · Patm g

(17.3)

Where G: generation rate (mg/min) MW: molecular weight A: surface area (m2) L: length of the container in the direction of the air movement (m) S: airspeed across the length dimension (m/s) PV: substance vapor pressure (Pa) Patm: atmospheric pressure (Pa) fs: substance volume fraction in the liquid (unitless) T: ambient temperature (K) The Hummel Equation is an empirical model originally applied to open surface degreasing tanks. For circular spills, Keil and Nicas3 recommend substituting the “average chord length” (spill diameter · π/4). The Hummel Equation can be used to estimate the “initial” generation rate for a spill, under the assumption that the spill surface area will not change greatly for several minutes. Therefore, when using this equation, we are assuming that for a short time the surface area will be reasonably constant and therefore the generation rate G will also be reasonably constant. Table 17.6 contains estimates of the variables in the Hummel Equation. Using the best estimate of each variable, the generation rate (G) is 1,340 mg/min. Because the actual spill diameter, air velocity, and fraction of acetic acid in the solution might have varied from the best estimate, I assigned a triangular or uniform distribution to several of the variables. For the spill diameter (D) and the airspeed (S), I devised a triangular distribution by setting a reasonable mode, minimum value, and maximum value. I used a uniform distribution to model the fraction fs of acetic acid in the bottle, where the minimum and maximum were taken from a typical safety data sheet for acetic acid. Using TEAS, I generated 50,000 sets of the variables and calculated a random G using each set. The histogram of the estimates of G is shown in Figure 17.9. The histogram was well described by a triangular distribution: G ~ Triangular (mode = 1,330; minimum = 810; maximum = 1,900) mg/min.

158

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 17: Reconstruction of a Lab Spill Table 17.6: Constants and Statistical Distributions Assigned to the Hummel Equation Variables

Variable Process Temperature Spill Diameter (D)

Units °C

Statistical Model Constant

Mode (or Constant) 21

Min

Max

feet

Triangular

2

1.5

2.5

0.46

0.76

Constant

0.61 760

Triangular

101,324 40

30

50

m/min 60.1

9.1

15.2

Constant

12.2 60.1

mmHg

Constant

11

Uniform

1,467 pa 0.85

0.8

0.9

Pressure (P)

m mmHg

Air Velocity (s)

pa feet/min

Molecular Weight (MW) Vapor Pressure (VP)

pa Fraction Available (fs)

Figure 17.9: Histogram of 50,000 estimates of the generation rate.

Copyright AIHA®. For personal use only. Do not distribute.

159

Modeling Occupational Inhalation Exposures to Chemicals Chapter 17: Reconstruction of a Lab Spill

Appendix B: Estimating the Near-Field Volume and Flow Rate () Experience has shown that the shape and volume of the near field are not as important as the flow rate of air () moving through the near field. This is because almost all near fields are relatively small compared to the minute volume of air moving through the near field. Therefore, the concentration within the near field will rapidly approach an equilibrium concentration for a constantly emitting source (such as we have here) and rapidly approach the far-field concentration whenever the source stops emitting. In this scenario, I chose a hemisphere for the shape of the near field. Other options are a box, cube, or cylinder. Assuming an approximate diameter of 1 m leads to a near-field volume (VN) of 0.26 m3: 2 VN = — · π · 3

D — 2

3

2 = — · π · 3

1 — 2

3

= 0.26 m3

The near-field flow rate, , is traditionally calculated using Equation 17.4:4 1  = — · FSA · s 2

(17.4)

Where FSA: the “free surface area”; the unblocked surfaces of the assumed near-field volume (m2) s: the average speed of the air moving into and out of the FSA For a hemisphere where the base is blocked, the near-field flow rate is calculated from Equation 17.5: 1  = — · (2 · π · r2) · s 2

(17.5)

The airspeed across the floor was measured at 40 fpm (12.2 m/min). Assuming a radius of 0.5 m leads to a  of 9.6 m3/min. To construct a probabilistic WMR model, I assumed that the single airspeed measurement had roughly a ± 25% error. This error represents both temporal variability in the floor level air velocity and instrument error. In effect, I modeled the floor level airspeed using a triangular distribution, where the most likely value is 40 fpm (12.2 m/min) and the minimum and maximum are 30 fpm (9.1 m/min) and 50 fpm (15.2 m/min). Using the values in Table 17.7 and the above equation resulted in a triangular distribution model for :  ~ Triangular (mode = 9.6; minimum = 7.1; maximum = 11.9) m3/min

160

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 17: Reconstruction of a Lab Spill Table 17.7: Statistical Models Used for Estimating the Near-Field Flow Rate β (Using a Hemisphere Near Field)

Variable Radius Air Velocity (s)

Units m fpm

Statistical Model Constant Triangular

m/min

Mode (or Constant) 0.5 40

Min

Max

30

50

12.2

9.1

15.2

Comments: 1. Estimation of the near-field ventilation rate () is always a critical step in two-box, WMR modeling. An estimate of the random airspeed moving into and out of the near field is needed. For range finding or as a first approximation, the work by Baldwin and Maynard8 is frequently cited source for typical air velocity measurements. 2. Ganser and Hewett5 describe a method for estimating the “effective near-field flow rate” using near- and far-field measurements.

References 1. Reinke PH, Keil CB. Chapter 4: Well-mixed box model. In: Mathematical Models for Estimating Occupational Exposure to Chemicals, 2nd edition. AIHA: Fairfax, VA, 2009. 2. Keil CB. Chapter 3: Modeling pollutant generation rates. In: Mathematical Models for Estimating Occupational Exposure to Chemicals, 2nd edition. AIHA: Fairfax, VA, 2009. 3. Keil CB, Nicas M. Predicting room vapor concentrations due to spills of organic solvents. AIHA Journal, 64(4):445–454, 2003. https://doi.org/10.1080/15428110308984838. 4. Nicas M. Chapter 6: The near field/far field (two-box) model with a constant contaminant emission rate. In: Mathematical Models for Estimating Occupational Exposure to Chemicals, 2nd edition. AIHA: Fairfax, VA, 2009. 5. Ganser GH, Hewett P. Models for nearly every occasion: Part II - Two box models. Journal of Occupational and Environmental Hygiene, 14(1):58–71, 2017. https://doi.org/10.1080/15459624.2016.1213393. 6. Jahn SD, Bullock WH, Ignacio JS (Eds.). A Strategy for Assessing and Managing Occupational Exposures, 4th edition. AIHA: Fairfax, VA, 2015. 7. Reinke PH, Jayjock M, Nicas M. Chapter 5: Well-mixed rooms with changing conditions. In: Mathematical Models for Estimating Occupational Exposure to Chemicals, 2nd edition. AIHA: Fairfax, VA, 2009. 8. Baldwin P, Maynard A. A survey of wind speeds in indoor workplaces. Annals of Occupational Hygiene, 42(5):303–313, 1998. https://doi.org/10.1016/S0003-4878(98)00031-3.

Copyright AIHA®. For personal use only. Do not distribute.

161

Modeling Occupational Inhalation Exposures to Chemicals Chapter 18: The Limitations of Mathematical Models

Chapter 18

Understanding the Limitations of Mathematical Models and Proactive Communication of Modeling Results Kang Chen, PhD, CIE, CHMM Since the publication of the first edition of Mathematical Models for Estimating Occupational Exposure to Chemicals in 2000, the role of mathematical modeling in industrial hygiene has continued to be enhanced and expanded.1 Mathematical exposure modeling has been adopted as a primary exposure and risk assessment tool given its ubiquitous presence in a range of industries in the past decades.2 The accelerated pace of the development of computational capability has simplified numerical simulations.3,4 Today, it can be relatively easy to reach impressive accuracy using simple models thanks to high-performance computers.

Knowing the Limitations of Mathematical Models “All models are wrong, but some are useful” is a famous quote often attributed to the British statistician George E.P. Box. The idea of this quote is that no single model will be perfect or one-size-fits-all, meaning that it will never fully represent the real-world object or process. Furthermore, there are multiple opportunities, even for industrial hygiene modeling professionals, to become confused and consequently misunderstand or misapply the principles of model selection and utilization. Having said that, even if a model cannot exactly describe reality, it can be fairly useful for risk assessors to make preliminary decisions if the modeling result is close enough to reality. This leads to the recognition of two key limitations of any model. The first limitation is the scope of the model’s applicability. The second limitation is the set of underlying assumptions of both the model and the inputs selected for the model. These two limitations are the foundations of incompleteness in representing modeling results. It is often of great importance that such limitations be understood, interpreted properly, and communicated thoroughly by industrial hygiene professionals to the concerned stakeholders who may have a totally different level of expertise in industrial hygiene and mathematical modeling. The following are some examples of potential pitfalls when modeling if the model’s scope of applicability and underlying assumptions are not well understood and considered. • The well-mixed room model is widely recognized as the easiest approach to modeling indoor chemical contaminant concentrations. This model, however, is a useful approach only when the room air is indeed well mixed or if the contaminant is released at multiple locations throughout a room.5 From the author’s perspective, the use of mixing factors should be discouraged because mixing factors defy the law of conservation of mass and cannot be adequately quantitative in most modeling applications. For instance, the impact from airflow short circuits might not be explained well by using a mixing factor; thus, variations or stratifications in pollutant concentration may exist in the room that would not be considered at all in the model. • Another similar limitation of a well-mixed room model assumption is the application of carbon dioxide (CO2) monitoring as an indicator of ventilation effectiveness. In relatively small and multi-occupant spaces like laboratories, air ventilation rates can be calculated using the steady-state concentration of CO2. This approach can enable better awareness and support informed public and organizational decision making on reducing the exposure to airborne contaminants.6 However, in large spaces without proper design of ventilation, CO2 concentrations are an ineffective indicator of the ventilation in specific areas because the indoor air cannot be assumed to be fully mixed and the data acquired may be highly variable and less representative.7

Copyright AIHA®. For personal use only. Do not distribute.

163

Modeling Occupational Inhalation Exposures to Chemicals Chapter 18: The Limitations of Mathematical Models • Understanding the volume of a room is an important step in modeling concentrations. For a relatively complex indoor environment, when doing the geometrical calculation of room volume, the room might be divided into smaller volumes and/or include the calculation of the volume displaced by all solid objects in the room. This step is indispensable before room volume parameters can be incorporated into various models. In many cases of a real chemical laboratory environment, the displaced volume could be large due to the placement of larger instruments or equipment. However, this decrease in volume is often neglected or simplified with incorrect assumptions by modelers, which can lead to underestimation of actual concentrations of airborne contaminants. Consequently, this can result in overexposure to hazardous materials due to inadequate protective actions being taken based on the modeling results. It is imperative for industrial hygiene modeling professionals to acquire accurate data on room volume. This is true for simple models such as the well-mixed room model or complex modeling such as computational fluid dynamics. • The quality of the make-up air for rooms can also be overlooked. In many models, the concentration of contaminants in the make-up air is usually assumed to be zero. This might not be true when the outdoor ambient environment is highly polluted, together with high potential of outdoor airflow infiltration or reentry of exhausted air into the laboratory environment. • Turbulent eddy diffusion models often focus on concentration gradients near a source. Thus, general ventilation pollutant removal mechanisms in the vicinity of the emission source (e.g., about 1 m) could be negligible in reducing exposure intensity.1 However, a mathematical limitation of the model is that the concentration value becomes infinite when the distance from the source approaches zero, but an infinite concentration is impossible in reality. Thus, the turbulent diffusion models may not be appropriate for estimating concentrations at the positions very close to the emission source (e.g., less than 10 cm).1 • The near-field/far-field (NF/FF) model is a useful model for approximating a chemical concentration gradient in a room when a worker is using a hazardous chemical close to their breathing zone. An important assumption of the NF/FF model is that the contaminant is relatively well mixed in both fields. However, in the presence of a cooling fan close to the worker (which is common in hot work environments), the result can be inaccurate when using the standard approach to calculating the interzonal airflow rate () using random air velocity. The random air velocity is not sufficient to describe air mixing between fields in the presence of forceful, mechanical air movement close to the emission source. Similar cases include exposure scenarios that involve local exhaust ventilation (LEV; with the return of filtered air to the workspace), partial recirculation of the filtered general exhaust ventilation (GEV), and combinations of GEV and LEV.8 If the modeler does not account for this limitation, the NF concentration will be overestimated. In extreme cases where the fields are very well mixed, the NF/FF model can become less accurate than the well-mixed room model! These examples illustrate the importance of fully understanding the model assumptions and the scope of a model’s applicability. These are just a few examples of why it is highly recommended to review the modeling results. Make sure they are logical and reasonable according to the current scientific axioms and formulas before interpreting and communicating the results to stakeholders and the general public.

Proactive Communication of Modeling Results It is usually not wise to simply give the raw data derived from empirical equations and numerical simulations directly to the corporate leadership team members or the general public. These audiences usually have no familiarity with industrial hygiene and mathematical modeling. Therefore, modeling results must be interpreted and communicated proactively with extreme caution and professional judgment from an industrial hygiene modeling expert. A first step can be to identify the invested parties and potentially interested recipients of the industrial hygiene mathematical modeling results. Potential candidates might be identified through multiple approaches, including • Brainstorming among the management team (especially the risk communication team) members • Talking with colleagues in other functions • Reading reports from an internal public relations department

164

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 18: The Limitations of Mathematical Models • Reviewing media coverage • Contacting environmental health and safety (EHS) consulting companies • Interacting with related local government agencies • Recognizing other important community members Figures 18.1 and 18.2 are adapted from a previous AIHA publication (The Occupational Environment: Its Evaluation, Control, and Management, 3rd edition).9 Figure 18.1 exemplifies some questions that are valuable in identifying and prioritizing stakeholder groups for those who have fewer concepts of key stakeholders. Figure 18.2 renders a relatively comprehensive list of typical potential stakeholder groups in terms of environment, health, and safety.10 Again, these are just guidance questions and not all-inclusive for identification of key stakeholders. Thus, readers should check with their own situations and identify key stakeholders at their discretion. The strategy or intention of communication to people in various groups could be fairly different, as detailed in Figures 18.1 and 18.2. Questions to Help Identify Key Parties of Interest • Which groups have been involved in this issue or mentioned in retrospective references? • Which groups are likely to be affected directly or to think they are affected directly by the company’s action? • Which groups are likely to have great concern if they are not consulted or alerted to the issue? • Which groups would be helpful for you to consult with because they might have important information, ideas, or opinions? • Which groups should you involve to ensure that the company has communicated with a balanced range of opinions on the issue? • Which groups have responsibility relevant to the company’s action (e.g., firefighters, legal counsels, business practice officers)? • Which groups may not especially want input, but do need to know what your company is doing (e.g., peer sites or subsidiary companies)? Figure 18.1: Example questions to help identify key parties of interest. Potential Groups of Key Participants • Group 1: Corporate senior leadership members—The President, CEO, CFO, COO, or CTO and other functional leaders (Administrative, HR, Engineering Directors) or board directors. • Group 2: Employees—The on-site workforce and their family members, including retirees. • Group 3: Neighboring plants and the general public—Local facilities and people from other businesses and communities. • Group 4: Government supervision agencies—Local or state EPA, CDC, FEMA, and legislative committees. • Group 5: Educational and academic organizations—Industrial research institutions, colleges, or universities. Figure 18.2: Example list of potential key participants. When reporting exposure modeling results to the first group listed in Figure 18.2, corporate senior leadership members, it is important to prepare a detailed presentation including the case background, input information, assumptions, limitations, improvement opportunities, and (critically for this group) a return-on-investment analysis. Such an overall description, not just the raw data, will assist the management team in making the Copyright AIHA®. For personal use only. Do not distribute.

165

Modeling Occupational Inhalation Exposures to Chemicals Chapter 18: The Limitations of Mathematical Models decision to (conditionally) accept or reject the risk and then prioritize the limited human and technical resources available for improvement projects. Furthermore, it might be helpful to demonstrate the impact of making a poor decision regarding risk acceptance on the company’s reputation. Thus, the corporate leaders could be persuaded to take actions and address potential risks for their employees or the community close to their facility. The second and third groups include the on-site workforce, neighboring plants, and the general public. The methodology of communicating modeling results with people in these groups should be straightforward, easily understood, and open to two-way communication and questions. First and foremost, it is mandatory that current industrial hygiene ethical codes and guiding principles be strictly followed when interpreting the modeling results because all the affected groups have the right to know the calculated level of risk. Additionally, the industrial hygiene modeling professionals want the people who are potentially impacted by a chemical exposure to be adequately protected.10 The risk profile derived from modeling results should never be misrepresented or tampered with in a way in which personal health and safety might be compromised. The author recommends that the source of the industrial hygiene modeling tool or the algorithm used be described briefly at the very beginning of communication in order to give a good impression of the reliability of the whole modeling process. For example, the effectiveness of the tool is either verified in peer-reviewed literature or research projects, or the approach is advocated by professional organizations like AIHA, BOHS, ASHRAE, etc. Second, two-way communication is of great importance when industrial hygiene modeling results are delivered to internal and external parties of interest, who may have their own concerns or even challenges to the results. Do not forget that employee concerns may be very different from the concerns of the general community. Additionally, different groups of employees may have different concerns based on their workplace culture. It is important to consider what method of two-way communication would be most successful (townhall meetings, an online forum, supervisor-led discussions, etc.). Industrial hygienists must have patience and tolerance, allowing for opportunities to further answer the questions from different people that make up the employee and community groups. Industrial hygienists may want to conduct and communicate small-scale field, experimental validation of the mathematical modeling results, which is highly effective in increasing people’s comprehension of the modeling process and building up their trust with the models. During the communication, it is really the icing on the cake to show what protective actions are already in place or need to be implemented to mitigate the risk. This approach is beneficial in convincing people that the risk is acceptable or at least could be reduced to a level as low as reasonably practicable (ALARP). In this way, people are encouraged to voluntarily support risk control measures taken by the EHS authorities. Last but not least, mathematical modeling results should be delivered in a way that facilitates better understanding of the model outputs and potential impacts. Characterization with proper illustrations can usually help laypeople acquire important information quickly and much more readily than presenting them with complicated mathematical equations that they may not understand or have interest in. A recently developed modeling freeware (Expostats) for Bayesian decision analysis (BDA) has drawn great attention from the chemical industrial hygiene community.11–13 BDA has been used in the interpretation of chemical exposure measurements, as it renders quantitative integration of professional judgment14–17 together with data gained from numerical simulations or physical measurements, which permits us to calculate decision probabilities with small sample sizes.18,19 Integrated with the functions in other tools (e.g., IHSTAT and IH Data analyst), BDA’s assessment outputs are presented through a variety of illustrations with different aims, including the exposure distribution plot, overexposure risk plot, the comparative exposure band plot, and a risk gauge with a needle indicating the probability value within the corresponding risk category. Refer to Figures 18.3–18.5 for a glance at some useful graphs interpreting the modeling results.11 Just like the pictogram with its signal word under the Globally Harmonized System of Classification and Labelling of Chemicals (GHS) framework, the graphical outputs from Expostats can facilitate an easy understanding of the level of pertinent exposure risk as well as the effectiveness of an intervention or improvement for both internal employees and external communities. When communicating with the fourth group, government supervision agencies, the raw data of industrial hygiene modeling might be more important than for any other audience. This is mainly because EHS regulatory authorities rely on huge quantities of industrial hygiene modeling outputs (in addition to exposure measurements) to compile the overall supervision and emergency response system at state or country level. Pertinent staff from EHS regulatory authorities likely have adequate knowledge of industrial hygiene modeling approaches and can further investigate the collected data. The raw data from companies in various industries could be

166

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 18: The Limitations of Mathematical Models

Figure 18.3: The exposure distribution plots to show the occupational exposure limit (OEL) exceedance proportion.

Figure 18.4: The risk gauge indicating the probability value falling into the corresponding risk category.

Figure 18.5: The comparative exposure band plot (left) and overexposure risk plot (right). OEL, occupational exposure limit.

Copyright AIHA®. For personal use only. Do not distribute.

167

Modeling Occupational Inhalation Exposures to Chemicals Chapter 18: The Limitations of Mathematical Models uploaded onto the centralized risk control platform at the local EPA/CDC, and then regulatory authorities can prioritize the potential risks and reinforce on-site inspection or audit as appropriate based on the modeling results. Furthermore, analysis of the industrial hygiene mathematical modeling results guides the drafting and implementation of regional categorized emergency response plans (ERPs). Industrial research institutions and universities are the typical members of the fifth communication group. Considering the expertise of people in this group, the industrial hygiene mathematical modeling methods, assumptions, outputs, and limitations could be summarized and characterized in a more scientific way during the communication. For research institutions, modeling inputs, methodologies, and results might be detailed with a full description of initial parameters, boundary conditions, simplifications, and assumptions. In this way, experts in industrial research institutions can validate (or even improve) the modeling process through proper design and implementation of field experiments. As a next step, collaboration could lead to optimized solutions of exposure risk control, or process improvement could be figured out based on the modeling outputs and mitigation of the known limitations of a model. Communication of the modeling results with professors or researchers at colleges and universities follows the same process. Additionally, it is beneficial to provide some illustrations that can be used as vivid case studies for tutors to show in their lectures with students. As a result, students’ interest, trust, and understanding of industrial hygiene mathematical modeling will be aroused and established at the very beginning of their lessons on industrial hygiene. By this means, the value chain of industrial hygiene mathematical modeling is extended, and the value is maximized through effective communication of the modeling outputs with the professionals in this group. This is also the fundamental purpose of this chapter and the whole book. During communication with members of each group, record their concerns or questions completely for further improvement of the robustness of the models, consideration of the limitations of the approach, and continuous improvement of your communication processes. For some sensitive cases, especially with the blue-collar workers in the workshop, the communication records need to be anonymous because there might be conflicts of interest with other participants. Additionally, once an improvement project is completed, the updated modeling results can be communicated and compared with the previous data to demonstrate the effectiveness of an engineering control in occupational exposure risk mitigation.

Conclusions Knowing how to proactively communicate the modeling results with different parties of interest is equally important as developing a good mathematical model. Readers may consider making a separate standard operating procedure (SOP) of proactive communication of modeling results or combining the SOP with the existing risk communication plan (RCP) to manage the whole process of communication well. Here is where the industrial hygiene expert interactions are more needed. Industrial hygienists know how to illustrate the results successfully to not only the industrial hygiene practitioners but also the general public who are concerned with their personal exposure to hazardous chemicals. There are numerous tricks and traps that users of exposure models face because of the existence of two key limitations: the scope of a model’s applicability and the underlying assumptions used in the approach. If users are not careful, imperfect or incorrect decisions can be made when implementing a model, as was illustrated for the WMB model, turbulent eddy diffusion model, and the NF/FF model. Thorough understanding of the scope of a model’s applicability and the assumptions is a critical and fundamental step for beginners who want to use modeling tools with success. Assessing the limitations of a model is also an important prerequisite for proper and successful communication of the modeling results with key groups in the next step. Previous sections guide readers in identifying potential groups of interest and further explain specific tactics for communicating modeling results with such groups. Although the strategies of communication with different groups are not exactly the same, the purposes are fairly similar—namely, enhancing peoples’ understanding of the models, building up their comfort with the models, and finally expanding the opportunity to apply models. Just like awareness changes in safety culture, so it will take time for the advancement and prosperity of mathematical modeling in the industrial hygiene field. This may be especially true in the traditional occupational health community, where there has historically been a heavy focus on sampling, testing, occupational health surveillance, and medical intervention. 168

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 18: The Limitations of Mathematical Models Industrial hygiene mathematical modeling is of extreme importance to the science and practice of industrial hygiene and is indispensable in implementing an effective exposure risk management program. With robust modeling results and proactive communication, pertinent countermeasures can be proposed for engineering solutions to control and reduce hazardous material exposure in various working conditions, even with a very limited number of exposure measurements. Furthermore, industrial hygiene mathematical modeling can support retrospective exposure estimation, e.g., for the purpose of legal prosecution or an epidemiology study. Industrial hygiene mathematical models are also extremely useful as tools for performing a quick and preliminary exposure scan that can provide early and timely alerts about potential overexposure risks, in which case further investigations are warranted. Although industrial hygiene mathematical models vary in their complexity and applicability, most of them are time efficient and cost effective. Hopefully, this chapter serves as a useful introductory reference for industrial hygiene modeling beginners and shines new light onto the application, interpretation, and communication of industrial hygiene mathematical modeling outputs in an occupational exposure evaluation.

References 1. Keil CB, Simmons CE, Anthony TR (Eds.). Mathematical Models for Estimating Occupational Exposure to Chemicals, 1st edition. AIHA: Falls Church, VA, 2009. ISBN: 978-1-935082-10-1. 2. Chen K, Martin LF. Proper selection and application of mathematical models for estimating occupational exposure to chemicals. The Chemist, 92(1):95–107, 2021. 3. Shiling Z, Paul P, Liangbo H, Parmesh V, Richard L, Larry B. CFD modeling of flammable refrigerant leaks inside machine rooms: Emergency ventilation rates for different size chillers. Science and Technology of the Built Environment, 24(8):878–885,2018. https://doi.org/10.1080/23744731.2018.1448676. 4. Stern F, Wilson RV, Coleman HW, Paterson EG. Comprehensive approach to verification and validation of CFD simulations—Part 1: Methodology and procedures. Journal of Fluids Engineering, 123(4):793–802, 2001. https://doi.org/10.1115/1.1412235. 5. Heinsohn RJ. “General Ventilation Well-Mixed Mode.” In: Industrial Ventilation: Engineering Principles. John Wiley & Sons, Inc., New York, 1991, ch. 5. ISBN: 978-0-471-63703-5. 6. The Royal Academy of Engineering. Infection Resilient Environments: Buildings that keep us healthy and safe: Initial Report. RAEng: London, 2021. 7. SAGE EMG. Role of Ventilation in Controlling SARS-CoV-2 Transmission, 2020. https://assets.publishing. service.gov.uk/government/uploads/system/uploads/attachment_data/file/928720/S0789_EMG_Role_of_ Ventilation_in_Controlling_SARS-CoV-2_Transmission.pdf (accessed July 6, 2021). 8. Ganser GH, Hewett P. Models for nearly every occasion: Part II - Two box models. Journal of Occupational and Environmental Hygiene. 14(1):58–71, 2017. https://doi.org/10.1080/15459624.2016.1213393. 9. Anna DH (Ed.). The Occupational Environment: Its Evaluation, Control, and Management, 3rd edition. AIHA: Falls Church, VA, 2011. 10. Hance BJ, Chess C, Sandman P. Industry Risk Communication Manual: Improving Dialogue with Communities. Lewis Publishers: Boca Raton, FL, 1990. 11. Jérôme L, Lawrence J, Peter K, Hugh D, France L, Frédéric C, Gautier M, Tracy K. Expostats: A Bayesian toolkit to aid the interpretation of occupational exposure measurements. Annals of Work Exposures and Health. 63(3):267–279, 2019. https://doi.org/10.1093/annweh/wxy100. 12. Gurumurthy R. Progress in Bayesian statistical applications in exposure assessment. Annals of Work Exposures and Health. 63(3):259–262, 2019. https://doi.org/10.1093/annweh/wxz007. 13. Eun GL, Diana MC. Adoption of exposure assessment tools to assist in providing respiratory protection recommendations. Annals of Work Exposures and Health. 64(5):547–557, 2020. https://doi.org/10.1093/ annweh/wxaa023.

Copyright AIHA®. For personal use only. Do not distribute.

169

Modeling Occupational Inhalation Exposures to Chemicals Chapter 18: The Limitations of Mathematical Models 14. Susana GH, Mariscala MA, Javier GR, Dale OR. Working conditions, psychological/physical symptoms and occupational accidents. Bayesian network models. Safety Science, 50(9):1760–1774, 2012. https://doi. org/10.1016/j.ssci.2012.04.005. 15. Ramachandran G, Vincent JH. A Bayesian approach to retrospective exposure assessment. Applied Occupational and Environmental Hygiene, 14(8):547–557, 1999. https://doi. org/10.1080/104732299302549. 16. Sottas PE, Lavoué J, Bruzzi R, David V, Nicole C, Droz PO. An empirical hierarchical Bayesian unification of occupational exposure assessment methods. Statistics in Medicine, 28(1):75–93, 2009. https://doi. org/10.1002/sim.3411 . 17. Banerjee S, Ramachandran G, Vadali M, Sahmel J. Bayesian hierarchical framework for occupational hygiene decision making. The Annals of Occupational Hygiene, 58(9):1079–1093, 2014. https://doi. org/10.1093/annhyg/meu060. 18. Jones RM, Burstyn I. Bayesian analysis of occupational exposure data with conjugate priors. Annals of Work Exposures and Health, 61(5):504–514, 2017. https://doi.org/10.1093/annweh/wxx032. 19. Paul H, Perry L, John M, Gurumurthy R, Sudipto B. Rating exposure control using Bayesian decision analysis. Journal of Occupational and Environmental Hygiene, 3(10):568–581, 2006. https://doi. org/10.1080/15459620600914641.

170

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 19: Feeding Models

Chapter 19

Feeding Models Using Simple and Some Not-So-Simple Means Mike Jayjock, PhD, CIH

Introduction After reading the other chapters of this book, it should be fairly obvious that physical-chemical models applied to predict the airborne concentration of toxicants can be extremely useful. Indeed, it is safe to say that the value of these models has done much to drive the professional careers of the authors. Over the last few decades, software tools that run on personal computers1,2 have rendered the calculation burden on the modeler to be essentially nil. Given this situation, one might be tempted to think that most (if not all) practicing industrial hygienists would have by now naturally come to use modeling as a primary (if not dominant) aspect of their practice. The reality is that most do not. Indeed, the vast majority of industrial hygienists continue to cling to specific air monitoring as their primary operational paradigm. Therefore, models may be useful, but they are clearly being underutilized by the profession. For years, the author and other colleagues have asked: What changes are needed to promote modeling as a widespread tool? We believe that some modest growth has occurred and further believe that this growth is primarily a result of courses presented under the auspices of AIHA. This book aims to promote and accelerate that trend by providing specific examples and tools to aid in the use of models. This specific chapter hopefully provides some understandable background and immediately useful advice.

Models Are Beasts That Need to Be Fed Most of the models we use or recommend are based on basic first-principles that are quite simple and true at face value. Specifically, they calculate the airborne concentration of toxicant in a relevant volume of air by accounting for the amount going in and the amount coming out of that volume over time. Given that information, one can calculate the average concentration of the toxicant in that volume at any point in time and over any period of time as a time-weighted average. It is that simple. However, actually modeling this reality in a meaningful context can get remarkably complex—or at least fill the industrial hygienist with uncertainty. As a general rule, most of the model input data are relatively easy to acquire by simple measurements. The most difficult input appears to be the generation rate of the toxicant. As an illustrative case, consider modeling the potential chemical exposure to an individual working with a volatile chemical on a desktop. To date, perhaps the best model available to apply to this scenario is the two-zone nearfield/far-field model.3 Given this model and the broad description of modeling provided previously, the modeler needs to come up with the following information or data in order to feed it.

Estimate of the Breathing Zone Volume Models require a quantitative description of the relevant volume of air that is immediately in front of the worker’s mouth and nose. This is the so-called breathing zone (BZ). Assuming all the air in the room is well mixed means this BZ is the volume of the entire room. As discussed elsewhere and consistent with most industrial hygienists’ understanding, such rooms are really quite rare. Especially when dealing with a relatively small (point) emission source, we know that the concentration in the immediate vicinity of the released chemical (by evaporation or entrainment) is typically much higher than the concentration in the room away from the source. This adjacent or proximate space is known as the near field

Copyright AIHA®. For personal use only. Do not distribute.

171

Modeling Occupational Inhalation Exposures to Chemicals Chapter 19: Feeding Models (NF), and it is perhaps best described as the smallest geometric volume that includes both the source and the BZ. In this case, the two-zone model assumes that BZ = NF volume because the assumption is made that there is good mixing within the NF. In our example, one might consider a BZ/NF volume to be a hemisphere over the source on the table with a radius of slightly more than an arm’s length.

Rate of Material Going Into the Volume Also required in modeling is a quantitative description of the amount of chemical being released to that volume per unit of time. This is the source, or generation rate term (G), and it is typically the most difficult quantity to accurately determine or estimate. In this scenario, the source rate (G) is the evaporation or possible physical entrainment of the chemical that the individual is working with. For simplicity, the evaporation rate is often assumed to be constant. However, depending on how the chemical is used, the function of time may be more complicated. As mentioned in earlier chapters, there are examples of evaporative rate models8,9 and assumptions based on drying time that could be used. It has been my experience that these models for estimating evaporation rate provide relatively uncertain predictions compared to properly conducted experiments. Later in this chapter, chambers will be presented as a relatively sophisticated way to determine G. We will also go over some fairly simple measurements that can significantly lower the uncertainty associated with this critical modeling parameter.

Rate of Material Coming Out of the NF or BZ Volume In most cases, once the chemical is in the NF or BZ volume, the only way for it to come out is by removal via ventilation. In the case of large airborne particles, it is necessary to use both ventilation and settling as removal mechanisms, in which the contaminated air is being moved out of the air volume. There are rare instances where a chemical is so reactive that it transforms to degradants in a relatively short time frame, thereby effectively removing a significant portion of it from the BZ/NF. These situations are quite rare, however. Thus, we are left with the following need for our example of chemical vapors: a quantitative description of the ventilation rate, which is the BZ/NF volume that is exchanged with fresh(er) air within the room. Descriptions for measuring or estimating this ventilation rate as used in the two-zone model are presented elsewhere in the text. Basically, the idea is to use the NF geometry, the ambient air movement within the room, and the general ventilation rate of mixing fresh air to the room to estimate this factor. The reason for laying out the above requirements is to show or confirm that although the concept of the model is simple, its execution is not effortless. To any industrial hygienist new to modeling, this aspect is almost certainly daunting. In fact, it can be so daunting that industrial hygienists often go back to directly monitoring the scenario by air sampling because it is a straightforward, familiar, and comfortable exercise. What we hope to convey in this book and chapter is that the “light” (knowledge and insight) provided by modeling is definitely worth the “candle” of effort to make it happen. We hope to do this by providing real and practical approaches that industrial hygienists will actually try.

Trading Conservatism for Data One of the most compelling and rewarding aspects of exposure modeling is that it allows industrial hygienists to overestimate the quantitative level of exposure when they have very little data. These overestimates, although technically inaccurate, can still be valuable if they show that the overestimated exposure is still below concentration levels of concern. This is a very cost-effective way of doing exposure assessment. For example, in the previous assessment, we can easily overestimate G by assuming that an unusually large portion (or even all) of what is typically handled becomes airborne. This would certainly overestimate G. We could assume that the individual works with the chemical 8 hr/day and the ventilation rate and air movement in the room are relatively low. All this allows for the confident overestimation of the exposure potential. If this purposeful “stacking the deck” to render an overestimate of the true exposure still results in a predicted concentration that is below the allowable exposure limit, then this was a very valuable exercise. Indeed, in this case, relatively little effort allowed for this documented conclusion of relative safety vis-à-vis an accepted exposure limit. The reality is, of course, that the actual worker exposure potential was most likely significantly less. We used conservatism.

172

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 19: Feeding Models If the purposely overestimated exposure did not show relative safety compared to an exposure limit, then we could trade conservatism for data by doing more work. Doing so will cost more, but it will most certainly increase our knowledge confidence and lower the estimated exposure. The remainder of the chapter will concentrate on two areas; namely, the development of real tools and the information gained from surprises during modeling.

Chamber Case Studies Chambers are remarkable tools. Indeed, they could be considered the ultimate modeling tool. That statement may sound outrageous or even contradictory, but it is not. We can use chambers to provide us with real-time insight into the actual time course of emission or generation. This is something we will develop later in the chapter; for now, just keep in mind that chambers are critically important to model development. Chambers can be incredibly sophisticated and expensive scientific entities, but they really do not have to be. If you have access to a room with a high (>12 ft) ceiling and a relatively large volume, building your own chamber can be relatively simple and inexpensive. Starting with 8-ft long pine 2 × 4s, industrial hygienists can readily build a room into an 8 ft cube with 7 studs on each wall and a ceiling with 16-inch centers. Some building skills may be needed to frame and hang a door inside and perhaps a glass window for observation, but contractors can easily be hired to do this work. To simulate a real room in a residence, apply painted wallboard with spackled joints to the interior. If you want to apply a much more inert material surface, put thin sheets of stainless steel or other metal over the wallboard and door interior, sealing the wall seams with silicon sealant. The next step is supplying controlled ventilation to the little room chamber. Cut an 8- to 10-inch hole in the back of the chamber and attach an external duct with an adjustable gate before a relatively small radial fan of a ¼ to ½ horsepower (117 V AC). Arrange the fan to blow out away from the chamber. This will put the chamber under mild vacuum and contain the air and generated contaminant within the chamber and its exhaust. If there is concern about what is coming out of the chamber, duct the outflow outdoors or to a filter. The ventilation rate is simply the cross-sectional area of the exhaust duct times the measured air velocity through the duct. In a later section, we will show another relatively simple method for measuring ventilation rate (Q) of a walled-in chamber. Another “quickie” chamber would be an exterior room in a residential building. Apply negative ventilation as described previously and you have a test chamber. For not much output in time or money, you now have a functioning chamber capable of determining the generation rate of any operation conducted within the chamber. This may not be critical for evaporative sources such as those from spills. In fact, these evaporation rates can be determined without such a chamber. However, the chamber will be invaluable for determining the emission rate of chemicals or particles during any cleaning, spraying, rolling, cutting, sanding, or machining operations that can be conducted within the chamber. It can also be used to measure the off-gassing of building materials such as carpeting. One advantage of a “walk-in” chamber is that you can have a person performing tasks as they are being monitored. It is, of course, important that this person wears the appropriate personal protective equipment during these tests. Say we used the chamber to estimate the generation rate of a small source (e.g., remember the desktop chemical example above?). The first task would be to define the dimensions of the NF (remember the hemisphere?). Then, measure the air movement at a few points on that imaginary hemisphere using a recording thermo-anemometer during the operations with the chemical in the chamber. This will most likely be quite variable, but distribution data can be used as well as an average or lower bound limit for conservatism. In a more sophisticated analysis, you could use the distributional data in a Monte Carlo analysis, as was done by Shade and Jayjock.4 You will, of course, need to measure the BZ concentration of whatever you are testing at steady state. Given the following model and inputs: CNF,SS =

G

G

— + —

Q



Copyright AIHA®. For personal use only. Do not distribute.

(19.1)

173

Modeling Occupational Inhalation Exposures to Chemicals Chapter 19: Feeding Models Input data: CNF,SS: BZ/NF concentration in the near-field at steady state (mg/m3). This can be measured on the person doing the work in the chamber. Q: ventilation rate [measured at the fan duct = duct area (m2) × velocity (m/min) = m3/min]. : calculated from anemometer data for airspeed at the NF/FF interface and geometry of the hemisphere BZ/NF. Solve for G in the steady-state two-zone model equation (Eq. 19.1) and you have it! You can also normalize this generation rate (G) by dividing by the volume of the chemical handled. The assumption is that G is proportional to the amount handled. One question comes up if you are using this technique: How long does it take to reach steady state? The math to answer this question is pretty straightforward. The faster you turn over the air in the chamber via ventilation from the fan, the more quickly steady state is established. Mt

1/2

= M0 · e–k

(19.2)

· t1/2

Where: Mt1/2: number of original air molecules remaining at time t1/2 , one half-life M0: number of original air molecules at time t = 0

t1/2 : half-life, time elapsed to remove half of the original molecules from the system k: the mixing air changes per hour (1 per hour) Because the ratio Mt1/2/M0 is 0.5 by definition 0.5 = e–kt

(19.3)

Take the natural log of both sides of the equation and rearrange: ln(0.5) 0.693 t1/2 = ——— = ——— k k

(19.4)

At 4 mixing air changes per hour (k = 4), the half-life of the average air molecule in the chamber is 0.693 ÷ 4, or 0.173 hr. After 5 half-lives [or 0.866 hr (52 min)] elapse, only about 3% of the original air in the chamber remains and steady state has essentially been attained. Thus, conducting the operation under interest but delaying the start of measuring the air concentration for 1 hr (60 min) after initiating the release process seems reasonable. After 1 hr, sampling only needs to proceed for as long as needed to attain enough sample for analytical requirements. What did all this get you? Well, you have a value or distribution for G in this type of operation that can be used to estimate the NF or BZ concentration anywhere in scenarios with different values of Q or . All the above detail was presented not to prescribe a specific chamber configuration but more so to show you some possibilities. Next, we are going to recount a lesson learned by the author during the construction and “shakedown” of an 8-ft room cube chamber put together some years ago.

174

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 19: Feeding Models

The “Leaky” Chamber We constructed the 8-ft cubic room as described above with a door and small window. The inside was covered with 5/8-inch gypsum or standard “wallboard.” All joints were taped and spackled or otherwise sealed. We decided to measure the ventilation rate but not by measuring the exhaust rate as described above. Instead, we chose to measure ventilation by taking the decay rate of a tracer gas sprayed into the chamber as a single bolus and then measuring its fall-off concentration with time. The procedure and calculation are as follows: 1. Find a relatively nontoxic tracer gas whose concentration can be measured in real time. 2. Release a single bolus of this gas into the chamber and mix it briefly but thoroughly with a fan. 3. Take real-time measurements of C at time (t); that is, C,t pairs until the original C has fallen to less than 25% of the original value. Assuming the chamber is well mixed, C can be sampled anywhere in the chamber. 4. In Microsoft Excel, run a linear regression of the ln(C) versus t in hours for a series of data pairs. The slope of the regression equation is k, or air changes per hour. Ventilation (Q) = (k)(Room Volume)

19.5

For our first test, we wanted to see how “leaky” the room was; that is, we put in the tracer gas (we used CO2), closed the door, and measured the concentration decay over a few hours with no ventilation. The results really surprised us. The tracer decayed rapidly, indicating a short half-life and high ventilation rate of about 5 air changes per hour! This was for a closed room inside of a building. We went inside the room (which had no lights), closed the door, and looked for light leaks coming through openings. Any opening we saw, we sealed or covered and repeated the test with the same results! We hypothesized that the unpainted wallboard must be absorbing the CO2 and giving a false read for the general ventilation. To test this, we painted the wallboard with two coats of sealer, let it dry, and reran our test. This time we calculated a reasonably expected Q < 0.1/hr. Clearly, the unpainted wallboard was absorbing the CO2. We later found out that fresh wallboard typically contains a residual amount of calcium oxide (CaO), which is quite reactive with CO2, readily forming calcium carbonate (CaCO3) and creating a CO2 sink. Thus, the mystery was solved, but it definitely points to the possibility of surface or “sink” effects in inhalation modeling. Indeed, even ostensibly inert surfaces such as glass and metal can be at least temporary sinks by adsorbing the chemical, saturating it, and then releasing it. Even accounting for sinks, one can still be surprised—as the next case shows.

The Disappearing Biocide We set up a relatively small (6-L) glass chamber to test the emission rate and ultimate concentration of biocide from treated wood used indoors.5 The modeling was set up using very sophisticated software capable of continuously solving simultaneous differential equations. We were told that the biocide was very stable and not liable to degradation; thus, we did not consider degradation in our first series of models of the results. This caused problems. Our first-principle model did not include degradation and could not reasonably reproduce or explain the experimental data. When we put a degradation term into the model, we were able to size the effect of degradation and match the experimental output. Putting degradation into the model caused a level of controversy with the people who actually developed the biocide, who had advised us that the biocide was not liable to degradation. This led to a series of tests that showed definitively that degradation was occurring primarily as a result of: 1. Its relatively large surface area-to-volume plating onto the chambers glass surface. 2. The extended residence time of the biocide on this surface. 3. Fresh ambient air constantly moving over the surface.

Copyright AIHA®. For personal use only. Do not distribute.

175

Modeling Occupational Inhalation Exposures to Chemicals Chapter 19: Feeding Models The degradation products were analytically identified as the products of oxidation, most likely from oxygen and tropospheric ozone. Here, both parties in this controversy were correct. The biocide did not significantly degrade in the wood because it was not as exposed to the above conditions; however, it most likely does degrade after it adsorbs onto or into most environmental surfaces. This level of understanding was essentially provided by the modeling process.

Simple Stuff With Big Payouts The final part of this chapter is simply designed to show you how some simple experiments can really help feed your models. Perhaps the simplest way to get an evaporation rate is to put a small amount into a shallow watch glass on an open microscale. Try to have the air go over the surface at a normal speed for indoor environs. Published data indicated a range of median velocity of 4.2–12.4 cm/sec in an unoccupied residential room with the HVAC turned off.6 Measure or record the weight loss over time. If it is a mixture, you might be able to analyze the concentration at the beginning and after a certain period of time expires. For material that is applied to a surface, simply do a mass balance and assume it all dries. Weighing paint cans, paint brushes, etc. before and after will go a long way in quantifying what was applied and what was released. One last trick or bit of cleverness was given to me by a colleague many years ago. The colleague was Dr. John Franke, who told me of a simple way to get the ventilation rate in a plant that uses propane-powered forklift trucks. John would simply have the forklift operators take a break or he would do testing at lunch or after or between shifts. The real-time C,t data pairs of carbon monoxide decay gave him just the data he needed to calculate k, or the air change rate per hour, using the tracer gas methodology discussed above.

Postscript As mentioned above, the generation rate of pollutants into the workplace air represents perhaps the largest challenge to and source of uncertainty in modeling. A complete chapter decided to this topic is available in the premiere AIHA text on Mathematical Modeling.7

References 1. IH MOD. https://aiha-assets.sfo2.digitaloceanspaces.com/AIHA/resources/Public-Resources/IHMOD_2_0. xlsm, American Industrial Hygiene Association, 2022. 2. Task Exposure Assessment Simulator (TEAS). Exposure Assessment Solutions, Inc., 2019. 3. Nicas M. Estimating exposure intensity in an imperfectly mixed room. American Industrial Hygiene Association Journal, 57(6):542–550, 1996. https://doi.org/10.1080/15428119691014756. 4. Shade WD, Jayjock MA. Monte Carlo uncertainty analysis of a diffusion model for the assessment of halogen gas exposure during dosing of brominators. American Industrial Hygiene Association Journal, 58(6):419–424, 1997. https://doi.org/10.1080/15428119791012658. 5. Jayjock MA, Doshi DR, Nungesser EH, Shade WD. Development and evaluation of a source/sink model of indoor air concentrations from isothiazolone treated wood used indoors. American Industrial Hygiene Association Journal, 56(6):546–557, 1995. https://doi.org/10.1080/15428119591016773. 6. Matthews TG, Thompson CV, Wilson DL, Hawthorne AR, Mage DT. Air velocities inside domestic environments: An important parameter in the study of indoor air quality and climate. Environmental International, 15(1–6):545–550, 1989. https://doi.org/10.1016/0160-4120(89)90074-3. 7. Keil CB, Simmons CE, Anthony TR (Eds.). Mathematical Models for Estimating Occupational Exposure to Chemicals, 2nd edition. AIHA Press: Fairfax, VA, 2009.

176

Copyright AIHA®. For personal use only. Do not distribute.

Modeling Occupational Inhalation Exposures to Chemicals Chapter 19: Feeding Models 8. Choudhary R, Klauda JB. The simultaneous mass and energy evaporation (SM2E) model. Journal of Occupational Environmental Hygiene, 13(4):243–253, 2016. https://doi.org/10.1080/15459624.2015.11011 23. 9. Hummel AA, Braun KO, Fehrenbacher MC. Evaporation of a liquid in a flowing airstream. American Industrial Hygiene Association Journal, 57(6):519–525, 1996. https://doi.org/10.1080/15428119691014729.

Copyright AIHA®. For personal use only. Do not distribute.

177

HAZARD RECOGNITION AND EVALUATION

A Case-Based Introduction to Modeling Occupational Inhalation Exposures to Chemicals Edited by Chris Keil, PhD, CIH This collection of case studies is written to provide an application-based “on ramp” to the mathematical modeling of inhalation exposures to chemicals. Seeing the models as integral tools for everyday exposure assessment questions will make these resources more readily accessible for industrial hygienists and other OEHS professionals as they improve the quality of their exposure judgments.

STOCK NUMBER: AEAB23-826