This article was downloaded by: [Gazi University]

On: 05 January 2015, At: 15:22

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer
House, 37-41 Mortimer Street, London W1T 3JH, UK

Journal of Occupational and Environmental Hygiene

Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/uoeh20

A Comparison Study Using a Mathematical Model and

Actual Exposure Monitoring for Estimating Solvent
Exposures During the Disassembly of Metal Parts
a a
John W. Spencer & Marc J. Plisko
a
Environmental Profiles, Inc. , Baltimore , Maryland
Published online: 31 Oct 2007.

To cite this article: John W. Spencer & Marc J. Plisko (2007) A Comparison Study Using a Mathematical Model and Actual
Exposure Monitoring for Estimating Solvent Exposures During the Disassembly of Metal Parts, Journal of Occupational and
Environmental Hygiene, 4:4, 253-259, DOI: 10.1080/15459620701205253

To link to this article: http://dx.doi.org/10.1080/15459620701205253

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of
the Content. Any opinions and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied
upon and should be independently verified with primary sources of information. Taylor and Francis shall
not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other
liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or
arising out of the use of the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
 Journal of Occupational and Environmental Hygiene, 4: 253–259
ISSN: 1545-9624 print / 1545-9632 online
Copyright c 2007 JOEH, LLC
DOI: 10.1080/15459620701205253
A Comparison Study Using a Mathematical Model and
Actual Exposure Monitoring for Estimating Solvent
Exposures During the Disassembly of Metal Parts
John W. Spencer and Marc J. Plisko
Environmental Profiles, Inc., Baltimore, Maryland
The objective of this research was to compare the airborne
solvent concentrations measured during the disassembly of
solvent-coated metal parts with concentrations predicted by a
mathematical model. The study involved three test simulations
where cyclohexane, used as a penetrating solvent, was squirted
Downloaded by [Gazi University] at 15:22 05 January 2015
onto a gate valve while the valve was subsequently disas-
sembled. Three test simulations were performed to evaluate
the effect of varying the speed of random air movement in
the work area. For statistical considerations, six replicate
solvent application trials were conducted for each simulation.
Area and personal air samples were collected during the
trials performed under each test simulation. Cyclohexane was
applied to the valve at a consistent rate to obtain, to the greatest
extent possible, a constant generation rate of solvent vapors.
The Near Field-Far Field (NF-FF) model, applied using a
constant solvent generation rate, was selected to predict the
solvent vapor concentrations, and Monte Carlo analysis was
used to quantify uncertainty in the input parameters of the
model. Solvent concentration predictions obtained from the
modeling process were within a multiplicative factor of 0.1 to
1.5 of the arithmetic mean of the actual air sample results for all
three NF and FF conditions in each simulation. Application of
the NF-FF model under the conditions described suggests there
is a reasonable degree of reliability in forecasting airborne
contaminant levels in the workplace environment. Given the
limited resources faced today by many industrial hygienists,
exposure modeling can serve as a valuable tool for generating
the information needed to make informed decisions about
employee exposure.
Keywords exposure assessment, mathematical modeling, solvent
exposure estimation
Address correspondence to: John W. Spencer, Environmental
Profiles, Inc., 813 Frederick Road, Baltimore, MD 21228; e-mail:
jspencer@episervices.com.
INTRODUCTION
T he science of industrial hygiene involves the assessment
and management of occupational exposures. Tradition-
ally, air monitoring has been the primary means by which
Journal of Occupational and Environmental Hygiene
workplace exposures have been measured. However, more
recently, mathematical modeling has gained acceptance in the
industrial hygiene community. Mathematical modeling may
be more useful than traditional air monitoring in instances
where operations associated with certain exposures have been
discontinued, historical air monitoring data are not sufficient
or lacking entirely, when prospectively assessing the need
for exposure controls, or when limited resources prevent
the industrial hygienist from monitoring all operations at all
times.
Mathematical models have been successfully used in varied
disciplines, including environmental science, engineering, and
public health for a wide range of applications, which includes
risk assessments, product failure analysis, and community
noise exposure assessments. Regulatory agencies have long
relied on a wide variety of models for decision-making
purposes. Models are used to assess the fate of pollutants in a
variety of media, their impacts on human and ecological health,
and the costs and benefits of abatement/remediation activities.
The U.S. Department of Housing and Urban Development
(HUD) and the U.S. Department of Transportation (DOT)
have used different noise models in their planning and design
activities, whereas the U.S. Environmental Protection Agency
(EPA) has relied on models to support its decisions and
policies.(1–3)
The EPA currently uses a wide range of models to make
informed, science-based decisions that are important for hu-
man and ecological health and include atmospheric and indoor
air models, chemical equilibrium models, economic models,
exposure models, leaching and runoff models, multimedia
models, risk assessment models, ground water and surface
water models, and toxicokinetic models.(4) With this increased
focus and reliance on modeling, the EPA has established
the Center for Exposure Assessment Modeling to provide
the exposure assessment community with predictive exposure
assessment techniques for aquatic, terrestrial, and multimedia
pathways.
In the context of industrial hygiene and retrospective
epidemiology, as well as legal matters, the need often arises
April 2007 253
 to estimate occupational exposures to airborne chemicals
associated with certain job tasks. If sufficient information
on chemical use and physical and environmental factors
are available, then mathematical models can be used as a
means of estimating exposures.(5) However, the accuracy of
a model is an important consideration even when sufficient
input data are available. Verification of a model’s accuracy
is an important step to reduce the uncertainty associated
with a particular model’s outcome and to refine the exposure
assessment process.
This work was performed as the direct result of an earlier
modeling study performed by the authors.(6) In that study, the
Near Field-Far Field (NF-FF) model, also known as the two-
zone model, was used to estimate the benzene exposure of
a mechanic using a benzene-containing penetrating solvent.
Monte Carlo simulation was used in conjunction with the
model to predict an output distribution rather than a fixed
theoretical point estimate.
Three trials were performed where three batches of penetrat-
ing solvent were spiked with 1%, 7%, and 30% concentrations
by weight of benzene for use in a work practice simulation.
Downloaded by [Gazi University] at 15:22 05 January 2015
Two-hour work practice simulation trials were conducted with
each solvent preparation, which involved using the solvent to
loosen rusted nuts and bolts from pipe flanges. Air samples
were collected from the worker’s breathing zone and the results
were compared with the estimates generated by the model. The
model was found to underestimate the solvent air concentration
at lower generation rates and overestimate the concentration at
higher generation rates. Based on this data, the study concluded
that additional work simulation trials and air sampling data
were needed to better determine the accuracy of the predicted
model outcome.
The authors also recently collaborated on an effort that
demonstrated the utility of using the NF-FF model for es-
timating benzene exposure at a parts washer.(7) This model
presumed an exponentially decreasing benzene emission rate,
since the benzene evaporated from the parts washer solvent
as a function of time, and used Monte Carlo analysis to
assign distributions for respective model inputs. The general
concordance between the measured and predicted benzene
concentrations were within a multiplicative range of 1/2 to
2 and suggested that the model was adequate for making
retrospective exposure estimates using a similar apparatus.
The model predictions and work practice trials reported
herein were conceived to further the work previously per-
formed by the authors. The objective of this research was to
evaluate the accuracy and usefulness of a mathematical model
by comparing airborne solvent concentrations measured during
three controlled simulated parts disassembly processes with
concentrations predicted by the mathematical model. This was
achieved by comparing measured air concentrations generated
under well-controlled environmental conditions with modeled
concentrations that used the same environmental parameters
that were recorded during the controlled work simulation.
The physical conditions associated with the work practice
trials were carefully controlled such that input parameters for
254 Journal of Occupational and Environmental Hygiene
the model would be known with greater certainty. Greater
certainty in the true values of model input parameters will
lead to more precise (narrower) distributions of the predic-
tive outcomes when using stochastic probability analysis.
The work practice trials were repetitively performed, thus
generating sufficient data to facilitate a robust statistical
analysis.(8)
METHODS
The Predictive Model
The NF-FF model with a constant generation rate was
selected to predict the solvent vapor concentrations under
three different physical conditions and was selected because
it accounts for spatial variability in exposure intensity.(5) This
representation is based on the concept that the closer a person
is to a source of generation, the greater the exposure, and vice
versa.(9) The NF-FF model, also known as a two-box model,
is analogous to a box-inside-of-a-box. The inner box, or NF, is
a conceptual fixed space centered on the contaminant source
and also contains the worker’s breathing zone. The outer box
(FF) comprises the remainder of the room. The model is based
on the concept of a contaminant mass being emitted over time
into the NF where it is then carried to the FF by means of
air exchange between the NF and the FF. The exchange of air
between the NF and FF is also considered a generation source
for the NF, since air in the FF containing the contaminant
source is being exchanged with the NF. The contaminant is
eventually removed from the room by the airflow from the FF
to the outside.(7) This construct also presumes that the air in
each zone is well mixed and that air exchange occurs between
each zone. The concept of the model as explained above is also
described by the following mass balance equations for the NF
and FF, respectively:(5)
VNF · dCNF = G · dt + β · CFF · dt − β · CNF · dt (1)
VFF · dCFF = β · CNF · dt − [β + Q] · CFF · dt (2)
where
VNF and VFF = the NF and FF volumes, respectively (m3 )
CNF and CFF = the NF and FF concentrations, respectively
(mg/m3 )
G = constant mass emission rate (mg/min)
β = interzonal airflow rate (m3 /min) between the NF and FF
Q = room air supply rate (m3 /min)
dt = an infinitesimal time interval (min)
The NF zone was represented by a hemisphere with radius
(r) = 1 m centered on the contaminant source and with its base
on the workbench. The basis for the size of the NF zone was
that the NF encompassed not only the contaminant source but
also the worker’s breathing zone. For each simulation VNF was
2.1 m3 . VFF consisted of the remainder of the room, less the
VNF , which in this case was 111 m3 .
April 2007
 The interzonal airflow rate (β) was estimated as the product
of one-half the available free surface area (FSA) of the NF and
the random air speed (s) m/min in the vicinity of the NF zone
boundary, and is represented below by Eq. 3.(10,11)
1
β= (FSA · s) (3)
2 (60 minutes) = 1283.3 mg/min.
In this situation, the base of the hemispherical NF zone was
situated on the workbench, thus prohibiting air exchange at
this surface. As a result, the base was not included in the FSA
calculation. The resulting FSA of the NF zone, based on (r) =
1 m, was 6.28 m2 .
Three simulations were performed where all variables were
held constant except for β, which was varied by changing
the intensity of random air movement (s) for each simulation.
This construct was intended to evaluate the effect of changing
a single parameter on the output of the model such that the
predictions could be compared with the actual air monitoring
results obtained from the work practice simulation. Addition-
ally, conditions where β was greater than Q were also evaluated
to demonstrate the principle that when β is much greater than
Downloaded by [Gazi University] at 15:22 05 January 2015
Q, the CNF is approximately equal to CFF , an indication that
the room is essentially well mixed.(5)
Values of air movement intensity were selected to repre-
sent conditions of (a) an enclosed area or shop with little
ventilation (3.3 m/min), (b) the same area or shop with open
doors/windows and/or people walking in the vicinity (23
m/min), and (c) a well-ventilated or semi-outdoor work envi-
ronment (60.8 m/min). Each of the three preceding conditions
were established based on the authors’ personal experience in
performing workplace ventilation studies. Whereas workplace
conditions are diverse and vary considerably even within the
same industry, the authors considered the parameters included
in this research to be representative of the work environments
described above.
For Simulation 1, (β) was 1/2 × (6.28 m2 ) × (3.3 m/min) =
10.34 m3 /min. With a given VNF of 2.1 m3 , (β) for Simulation 1
represented approximately 4.9 air changes per minute in the NF
zone. For Simulation 2, (β) was 1/2 × (6.28 m2 ) × (22.8 m/min)
= 71.6 m3 /min, representing approximately 34.1 air changes
per minute in the NF zone. For Simulation 3, (β) was 1/2 ×
(6.28 m2 ) × (60.8 m/min) = 190.9 m3 /min, and represented
approximately 90.9 air changes per minute in the NF zone.
Because Eq. 1 includes the CFF as a contributing source to
the overall CNF , the volumetric supply/exhaust rate of the room
air (Q) is also included in the calculation. The authors selected
an overall rate of 4.3 air changes per hour, resulting in a Q of
8.1 m3 /min.
Generation rate (G) was estimated based on 100 mL of
reagent-grade cyclohexane being consistently applied to the
valve during trial times of 60 min each. Constant and steady
application of solvent over the duration of each trial prevented
pooling of the solvent on the workbench. Situations involving
pooling created by spills of solvents are more aptly described
by model constructs that incorporate an exponentially decreas-
ing contaminant emission rate.(12) The emission rate decreases
Journal of Occupational and Environmental Hygiene
as the surface area available for mass transfer decreases and
evaporation cools the liquid.(12) With a reported specific gravity
of 0.77, the approximate quantity of cyclohexane applied was
77,000 mg. Owing to a vapor pressure of 78 mmHg at 20◦ C, it
was determined that all of the solvent would evaporate during
the trial time, resulting in a generation rate (G) of (77,000 mg)/
Each input parameter for the model has now been described.
Concentration predictions representing the mean cyclohex-
ane concentration above the workbench and at the worker’s
breathing zone (NF) and the remainder of the room (FF)
were calculated for each simulation. The predictions were
numerically computed from the modeled concentration time
series derived from Eqs. 1 and 2 for the NF and FF, respectively.
An example of the model output using Simulation 1 data is
shown in Figure 1.
Monte Carlo Simulation
Uncertainty in modeling, whether in industrial hygiene or
otherwise, may be quantified by means of stochastic analysis,
an example of which is Monte Carlo simulation. This concept
is based on a repeated random sampling from the distribution of
values for each of the parameters in a generic equation to derive
an estimate of the distribution of exposures in a population.(13)
An algorithm of the NF-FF model was run using Microsoft
Excel in conjunction with Crystal Ball 2000 (Decisioneering,
Inc., Denver, Colo.) to perform the Monte Carlo simulation.
Predictions were made based on a range of possible parameter
value to analyze the effect of varying the inputs on the outputs
of the modeled system.
For the situation described herein, uncertainty in air con-
centration predictions lay with the solvent generation rate (G)
and the random air movement (s) around the NF boundary.
Based on evaporation trials performed prior to the actual
simulations, G was assigned a normal distribution with a mean
of 1283.3 mg/min and a standard deviation of 60.3 mg/min.
The range of evaporation measurements, based on six trials,
was 1103 mg/min to 1463 mg/min. The evaporation rate and
thus (G) was determined by applying cyclohexane to the valve
FIGURE 1. Modeled NF and FF time series concentrations for
Simulation 1
April 2007 255
 being used for the simulation at a rate of 5 mL/min. under
the study conditions. The rate of application was selected
to prevent pooling of the liquid as it was applied. Repeated
evaporation trials were conducted and the average observed
time was selected as the model generation rate.
For the NF, the greatest uncertainty was due to the random
air movement (s). Air velocity measurements of the fan con-
ducted prior to the simulations were used to define distributions
of air velocity for each given simulation. Each simulation was
found to approximate a normal distribution of measured air
velocity. For Simulation 1, (s) was 3.2 m/min with a standard
deviation of 2.2. The range, based on 79 measurements, was 0
to 8.8 m/min. For Simulation 2, (s) was assigned a mean of 22.8
m/min with a standard deviation of 5.5. The range, based on 170
measurements, was 8.5 to 38 m/min. Simulation 3 was assigned
a mean (s) of 60.8 m/min with a standard deviation of 5.3. The
range, based on 49 measurements, was 51.7 to 73 m/min. The
observed ranges and corresponding coefficient of variations
(CVs) for each simulation reflected the performance results of
fan speed measurements, indicating greater variability at lower
fan settings than at the higher fan speed settings.
Downloaded by [Gazi University] at 15:22 05 January 2015
The Monte Carlo simulation predicted the 60-min mean air
concentrations and frequency distributions on the means for
each selected value of β. An example of the model output
provided by Crystal Ball for the NF prediction in Simulation
1 is shown in Figure 2.
Job Task Simulation
The study involved three test simulations, with each sim-
ulation consisting of six trials. Cyclohexane, representing a
penetrating solvent, was squirted from a Nalgene laboratory
wash bottle onto a 5.08 cm, Class 125 Iron Body Gate Valve
during disassembly of the valve. The wash bottle was selected
to simulate the use of a squirt can as opposed to a spray
bottle, which would have resulted in a mist application. Three
simulations were selected to evaluate the effect of varying
a single parameter (air movement [s]) during each of the
simulations, as well as for comparison with the predictions
using the NF-FF construct previously described.
For statistical considerations, six replicate solvent appli-
cation trials were conducted for each of the three simulations.
FIGURE 2. Distribution of NF concentration predictions for Sim-
ulation 1
256 Journal of Occupational and Environmental Hygiene
Each trial lasted 60 min and was of sufficient duration to ensure
that all of the solvent was used and evaporated during the trial.
The solvent was applied to the valve consistently to obtain,
to the extent possible, a constant generation rate of solvent
vapors. One hundred mL of reagent-grade cyclohexane were
used for each solvent application trial. The material safety
data sheet (MSDS) provided by the manufacturer indicated
a specific gravity of 0.77 at 20◦ C, resulting in 77,000 mg of
solvent used per trial. The center of the worker’s breathing
zone was approximately 1 m from the contaminant source.
The work was conducted on a bench in the center of
a rectangular room measuring approximately 6.33 m long,
5.33 m wide, and 3.35 m high, thus containing a volume of
approximately 113 m3 . A schematic of the room is shown in
Figure 3. This room size was selected to represent, based on the
authors’ professional experience, a small garage, workshop, or
engineering space aboard a ship. The room was constructed of
nominal 2 in. × 4 in. (5.08 cm × 10.16 cm) wood framing
and 6-mL (0.0254 mm) thick, fire-resistive polyethylene
sheeting and was ventilated to facilitate air exchange with
the surrounding environment. Ventilation was established by
means of a filtered exhaust device placed at the south side of
the room and a curtained/flapped entryway with a single-stage
airlock located at the north end of the room. The design was
similar to the concept of the negative-pressure enclosure used
FIGURE 3. Schematic showing the general layout of the testing
area used for the job task simulation
April 2007
 for decontamination of potentially hazardous environments.
Air movement around and through the breathing zone and con-
taminant source (conceptual near field) was established using a
40.64 cm diameter variable speed pedestal fan pointed upward
at approximately 45◦ and away from the center of the room.
Three values of air movement intensity were selected to
represent the modeled conditions and included 3.3 m/min for
Simulation 1, 23 m/min for Simulation 2, and 60.8 m/min
for Simulation 3. Smoke tubes were used to visually observe
the air movement around the room, including around the
workbench surface and the contaminant source, to verify that
it was nondirectional, since the model construct is based on
random air movement between the NF and FF zones. Air
velocity was measured using a TSI Incorporated Veloci-Check
hot wire anemometer (TSI Inc., St. Paul, Minn.). Accounting
for the nondirectional nature of the air movement, the probe of
the anemometer was held at right angles in three dimensions
to determine the average air velocity at a given location in
the room. An established air exchange rate of 8.1 m3 /min,
equivalent to 4.3 room air changes per hour, nominally, was
measured prior to the study following a method described by
Downloaded by [Gazi University] at 15:22 05 January 2015
American Society for Testing and Materials (ASTM) Method
E741-00 “Standard Test Method for Determining Air Change
in a Single Zone by Means of a Tracer Gas Dilution.”
During each trial, a personal breathing zone air sample was
obtained to measure the mean solvent concentration near the
contaminant source, and two area air samples were collected to
measure the solvent concentration away from the contaminant
source but within the room. Whereas the location of the
personal air samples was limited to the worker’s breathing
zone, the area air sample (conceptual FF) locations were
selected in a random manner. A grid was laid out in three
dimensions within the room, with each grid being equal to
1 m3 . A computerized random number generator was then used
to position two air sampling devices. Grids were automatically
disqualified if they were occupied by either the workbench, the
location of the worker’s breathing zone, or other obstructions.
The personal breathing zone air samples and the area
air samples were obtained over the full 60-min duration
of each trial. MSA Escort Elf low-flow sampling pumps
(MSA, Pittsburgh, Pa.) were used for sample collection, and
were calibrated before and after each trial using a BIOS
Dry-Cal DC-Lite primary gas flow calibrator (BIOS Int.,
Butler, N.J.). The air samples were collected and analyzed in
accordance with NIOSH Method 1500 (Hydrocarbons). The
sample analysis was performed by an independent, American
Industrial Hygiene Association-accredited laboratory.
RESULTS
P oint concentration estimates obtained from the modeling
process were within a multiplicative range of 0.1 to 1.5 of
the arithmetic mean of the actual air sample results for all three
NF and FF conditions. Furthermore, the model predictions
underestimated the observed mean air sample concentrations
in the NF for Simulation 3 and in all simulations for the FF
Journal of Occupational and Environmental Hygiene
conditions. Statistical analysis of the air sample data, including
calculations of the mean, standard deviation, normal and
lognormal distribution tests, and confidence interval evaluation
was performed using LogNorm2, Statistics for Exposure
Assessment (In Tech Software Corp., Tulsa, Okla.), a PC-based
statistical analysis program written specifically for the field of
industrial hygiene.
For Simulation 1, the predicted NF air concentration
was 243 mg/m3 . The mean measured NF air concentration,
calculated from six Simulation 1 trials, was 235 mg/m3 . The
predicted FF air concentration was 121 mg/m3 , and the mean
measured FF air concentration was 153 mg/m3 .
For Simulation 2, the predicted NF air concentration
was 139 mg/m3 . The mean measured NF air concentration,
calculated from six Simulation 2 trials, was 137 mg/m3 .
The predicted FF air concentration was 121 mg/m3 , and the
measured mean FF air concentration was 153 ppm.
For Simulation 3, the predicted NF air concentration was
127 mg/m3 and the mean measured air sample concentration
was 177 mg/m3 . The predicted FF air concentration was
120 mg/m3 and the mean measured FF air concentration was
180 mg/m3 .
Monte Carlo simulation was performed over the course of
105 iterations and included the previously described solvent
generation rate and airflow distributions. The mean of the
output distribution for Simulation 1, NF, was 245 mg/m3 ,
and 95% of the distribution fell between 185 mg/m3 and
328 mg/m3 . For Simulation 1, FF, the mean of the output
distribution was 121 mg/m3 and 95% of the distribution fell
between 98 mg/m3 and 143 mg/m3 .
For Simulation 2, NF, the mean of the output distribution
was 139 mg/m3 and 95% of the distribution fell between 116
and 176 mg/m3 . For the FF prediction, the mean of the output
distribution was 121 mg/m3 , and 95% of the distribution fell
between 101 mg/m3 and 154 mg/m3 .
Finally, for Simulation 3, the mean of the output distribution
for the NF concentration prediction was 127 mg/m3 with 95%
of the distribution falling between 106 mg/m3 and 152 mg/m3 .
For the FF prediction, the mean was 121 mg/m3 and the central
95% of the output data fell between 101 mg/m3 and 144 mg/m3 .
Statistical analysis of the data suggested that the replicate
measurements from the work practice trials were approxi-
mately normally distributed. Based on a t-distribution with 5
degrees of freedom, the predicted range of NF concentrations
fell within the 95% confidence interval (CI) of the mean of
the actual measurements in Simulation 1 and Simulation 2.
Looking at the t-test results for the FF data, the predicted range
of FF concentrations fell within the 95% confidence interval
of the mean of the actual measurements in Simulation 1 and
Simulation 2 as well. For Simulation 3, the model predictions
underestimated the measured concentrations for both the NF
and FF condition.
Figures 4 and 5 compare the predicted concentrations with
the measured concentrations for the NF and FF, respectively.
Error bars represent the 95% CI on the mean of the predicted
and measured concentrations.
April 2007 257

FIGURE 4. Comparison of predicted NF air concentrations to
measured NF air concentrations
DISCUSSION
T he comparison of predicted air concentrations to average
measured concentrations showed agreement to within a
multiplicative factor of 1.5. Furthermore, two of six predictions
were within the 95% confidence interval of their respective
observed results. These results raise two questions concerning
the findings. First, did the model construct represent the
Downloaded by [Gazi University] at 15:22 05 January 2015
conditions under study and, second, were there limitations
or presumptions inherent in the construct that, if accounted
for, may have contributed to closer agreement of predicted vs.
actual air concentrations?
The NF-FF constant generation rate model adequately
accounted for the physical conditions established for the work
practice simulation. Variables including VNF , VFF , and Q, were
distinct, fixed quantities whose values did not change over
the duration of the study. Uncertainty in the values of actual
parameters such as G and β contributed the greatest variability
to both the modeling predictions and to the actual work practice
simulation trials.
Solvent generation rates are not hard constants and are
influenced by a variety of factors including but not limited
to ambient temperature, atmospheric pressure, air movement
above the solvent, the quantity of solvent, and the shape of its
container. Therefore, the values of G used in this study will not
necessarily be applicable in all modeling situations. Because
the authors identified no published values of generation rates
for cyclohexane, G was estimated based on observations
conducted prior to the study.
FIGURE 5. Comparison of predicted FF air concentrations to
measured FF air concentrations
258 Journal of Occupational and Environmental Hygiene
The interzonal air exchange rate β was estimated based
on the FSA of the near field and the random air speed (s)
around the NF boundary. Because the theory of the NF-FF
model is based on random air exchange between the NF
and FF zones, the attainment of nondirectional air movement
within the test chamber was required. This was a challenging
endeavor in which considerable time was devoted to ensure that
the air movement was behaving in a nonpredictable manner,
indicating that a degree of randomness had been attained.
Measurements of the magnitude of air movement revealed that
air movement was not constant but changing throughout the
study process. This variability was considered as a component
of the estimated value for the model input of β.
Limitations, or factors contributing to disagreement be-
tween the predicted and measured concentrations, were con-
sidered. Whereas the design of the study incorporated a
constant generation rate G, the true rate of solvent generation
during the work practice simulation trials was most likely
not entirely constant. Small pools of liquid settling on the
workbench, changes in the magnitude of air movement through
the NF zone, and imprecision in the constant application
of the solvent were contributing factors serving to influence
a less than constant rate of vapor generation. In addition,
worker movement in the NF zone induced additional air mix-
ing/movement not accounted for in the model construct. The
effects of convective airflow caused by the worker were also
considered and may have contributed additional uncertainty to
the simulation process.
The relatively small size of the test chamber enabled the
authors to evaluate how the effect of changing the rate of
air movement would effect the airborne concentrations in
the NF zone, but it also demonstrated confirmation of the
principle that where β is much greater than Q, then CNF is
approximately equal to CFF , an indication that the room is
essentially well mixed. For Simulation 1, where β < Q, the
ratio of CNF :CFF was 1.5, suggesting fairly well-mixed room air
but still exhibiting a significant difference in the measurements.
Similar examination of the results from Simulations 2 and 3,
where β  Q, showed ratios of 0.9 and 1.0, respectively,
indicating that air movement around the room was sufficient
to create a well-mixed space.
The general agreement between the predicted and measured
cyclohexane concentrations suggests that this construct suf-
ficiently describes the cyclohexane emission and dispersion
around the valve and within the room. Although this evalu-
ation did not constitute a validation of this model, the data
provided valuable insight on how the model can predict room
concentrations under a range of specified parameters.
CONCLUSIONS
R isk analysis is a fundamental principle by which the
assessment and management of occupational exposures is
performed. The findings presented herein can be used by the in-
dustrial hygienist to better define the strengths and weaknesses
April 2007
 of using the NF-FF construct as a tool to prospectively or
retrospectively evaluate occupational exposure.
The output distribution of predicted air concentrations
generated using the NF-FF model with Monte Carlo analysis
under the conditions described suggests there is a reasonable
degree of reliability in estimating airborne solvent vapor levels
in the workplace environment. Modeling for purposes of risk
assessment frequently involves constructs that overestimate
the likely exposure, thus creating a worst-case situation from
which to make decisions for exposure control. There was
general agreement between the predicted and measured results
which were within a multiplicative range of 0.1 to 1.5. Refine-
ment of input parameters and stochastic analysis, however, will
prove beneficial for evaluating a range of exposure possibilities
under similar situations.
ACKNOWLEDGMENTS
T he authors wish to thank Mark Nicas, School of Public
Health, University of California, Berkeley, for his assis-
tance in reviewing this manuscript and for providing additional
Downloaded by [Gazi University] at 15:22 05 January 2015
technical insight.
REFERENCES
1. U.S. Department of Housing and Urban Development (HUD): The
Noise Guidebook. Washington, D.C.: HUD, 1985.
Journal of Occupational and Environmental Hygiene
2. “EPA Guidelines on Exposure Assessment.” Federal Register
57(104):22888–22938 (1992).
3. “Guidelines for Carcinogenic Risk Assessment,” Federal Register
70(66):17765–17817 (2005).
4. U.S. Environmental Protection Agency (USEPA): Draft Guidance on
the Development, Evaluation and Application of Regulatory Environmen-
tal Models. Washington, D.C.: USEPA, 2003.
5. American Industrial Hygiene Association (AIHA): Mathematical
Models for Estimating Occupational Exposure to Chemicals. Fairfax, Va.:
AIHA Press, 2000.
6. Plisko, M., and J.W. Spencer: “A Retrospective Evaluation of Oc-
cupational Exposure to Benzene Using Mathematical Modeling and
Work-Practice Simulation.” Roundtable presentation at AIHCE 2005 in
Anaheim, Calif. (available on request).
7. Nicas, M., M.J. Plisko, and J.W. Spencer: Estimating benzene exposure
at a solvent parts washer. J. Occup. Env. Hyg. 3:284–291 (2006).
8. National Institute for Occupational Safety and Health (NIOSH):
Occupational Exposure Sampling Strategy Manual. DHHS (NIOSH)
Publication No. 77-173. U.S. Department of Health, Education and
Welfare, January 1978.
9. Cherrie, J.W.: The effect of room size and general ventilation on the
relationship between near and far-field concentrations. Appl. Occup.
Environ. Hyg. 14:539–546 (1999).
10. Nicas, M: Estimating exposure intensity in an imperfectly mixed room.
Am. Ind. Hyg. Assoc. J. 57:542–550 (1996).
11. Hemeon, W.C.: Plant and Process Ventilation, Second Edition.
New York: Industrial Press, Inc, 1963.
12. Keil, C.B., and M. Nicas: Predicting room vapor concentrations due to
spills of organic solvents. Am. Ind. Hyg. Assoc. J. 64:445–454 (2003).
13. U.S. Environmental Protection Agency (USEPA): Dermal Exposure
Assessment: Principles and Applications. Washington, D.C.: USEPA,
1992.
April 2007 259