An Employee Thermal Comfort Model for Semiconductor Manufacturing

An Employee Thermal Comfort Model for Semiconductor Manufacturing

 Robbie Walls, Corporate Ergonomist Intel Corporation

Abstract
Utilizing new technology, next generation wafer fabrication facilities will be able to reduce airflow and filtering requirements translating to a major reduction in capital and expense costs. In addition, relaxed microcontamination requirements will also allow for use of more cost-effective gowning systems.

While these changes in the clean room environment are known, the impact on the employee is not fully understood. This project investigates how to quantitatively determine the effects of decreasing clean room airflow on employee thermal comfort and ascertain which of the current clean room gowning systems are perceived as “cooler” and more thermally comfortable.

The overall objective of this project is to use existing guidelines and methods to quantify thermal comfort in the clean room environment. Using methodology outlined in the International Organization of Standardization guideline, ISO 7730 - Determination of the PMV and PPD Indices and Specification of the Conditions for Thermal Comfort, a model of the 200mm clean room environment is developed to determine the thermal sensation and degree of discomfort of employees.

Using statistical analysis, the thermal comfort model is compared with employee thermal comfort survey results to validate its predictability. Based on this analysis, the thermal comfort model was proven valid thus substantiating that the results obtained represent the current 200mm fab environment.

The model is then used to determine what combination of environmental variables in future factory designs will allow a decrease in airflow while maintaining or improving the current clean room thermal comfort environment without impacting manufacturing processes.

Introduction
300mm semiconductor technologies will have less demanding air system requirements in the fab manufacturing environment. Process equipment utilizing the Front Opening Unified Pod (FOUP) and tool mini-environment concept will allow for a relaxed class clean room environment outside of the equipment. As a result, next generation fabs will be able to reduce airflow and filtering requirements translating to a major reduction in capital and expense costs for new factories. In addition, relaxed microcontamination requirements will also allow for use of more cost-effective gowning systems.

To ensure that the impact to the clean room employee is fully understood, Intel Corporate Environmental, Health and Safety was asked to quantitatively determine the effects of decreasing clean room airflow on employee thermal comfort and ascertain which of the current clean room gowning systems are perceived as “cooler” and more thermally comfortable.

Literature Search
A literature search was performed to find any research, guidelines, or standards applicable to the thermal evaluation of clean room working environments. It was found that ISO 7730, Moderate Thermal Environments - Determination of the PMV and PPD Indices and Specification of the Conditions for Thermal Comfort, is the most current and comprehensive standard dealing with thermal comfort. It should be noted that this standard was developed in an office environment and as such may not fully comprehend the differences in a fab manufacturing environment. ISO7730 details an algorithm based on empirical data that can be used to determine what combination(s) of indoor space environment and personal factors will produce thermal environment conditions acceptable to 80% or more of the occupants in that environment. The algorithm takes six variables into account: four of which relate to the working environment and two of which deal with the individual (personal variables).

Environmental Variables
The four environmental variables include:
• Air temperature (Ta)
• Mean radiant temperature (Tr)
• Relative humidity (RH)
• Air velocity (var)
Since a clean room is a controlled environment, the current environmental variables and the proposed changes to these variables (reduced air velocity) are already known.

Personal Variables
The two personal variables include:
• Clothing insulation (Icl)
• Physical activity level (metabolic rate) (M)

Measurement of Clean Room Environmental Variables
Table 1 summarizes the current environmental variable levels currently used in 200mm wafer factories and the proposed 50% airflow reduction for 300mm factories.

Table 1. Environmental Variable Levels

Environmental
Variable

Current
Level

Proposed
Level

Air Temperature

72 oF

?? oF

Mean Radiant Temp.

72 oF

?? oF

Relative Humidity

40%

??%

Air Velocity

80 ft./min.

40 ft./min.

The insulation value of the clean room ensemble and the physical activity level of the employee (metabolic rate) are the two unknown variables that must be determined before the ISO 7730 algorithm can be utilized.

Determining the Clothing Insulation Value
To determine a realistic insulating value, the entire “typical” ensemble of the clean room employee had to be taken into account. An ensemble consists of the gowning system (Gore-TexTM or polyester), consumables (hairnet, gloves, shoe covers, etc.) and the personal clothing worn underneath.

The most widely used and accepted method for obtaining clothing insulation values for ensembles in moderate environments is the heated mannequin technique to determine resistance of an entire ensemble to dry heat loss. In a partnership with W.L. Gore, manufacturer of the Gore-TexTM suit, Intel contracted with the Institute for Environmental Research at Kansas State University to develop a testing design and procedure to obtain the insulation values needed.

Testing Design and Setup
To more accurately reflect real conditions in the clean room, six different combinations of ensembles were tested using a full-size heated mannequin testing apparatus in a controlled environmental chamber. Figure 1 shows an example of the testing setup.

 Figure 1. Heated Mannequin Testing Apparatus

Testing Results
Table 2 below shows the results of the testing. Each value is an average of three repeated testing trials on the same ensemble.

Table 2. Clothing Insulation Value Results

Ens
No.


Personal
Clothing

Gowning
System

Hood/
Helmet System

Clothing
Ins. -Icl
(clo)

1

Summer

Polyester

Polyester

1.17

Summer

Gore-
TexTM

Gore-
TexTM

1.15

 3

Summer

Gore-
TexTM

Polyester

1.10

Summer

Gore-
TexTM

Dryden
Shield

1.08

5  

Winter

Polyester

Polyester

1.25

6

Winter

 Gore-
TexTM

Gore-
TexTM

1.21

7  

Winter

Gore-
TexTM

Polyester

1.19

8

Winter

Gore-
TexTM

Dryden
Shield

1.18

From the results, it can be seen that there is virtually no difference in the insulating values of the ensembles. From a thermal comfort standpoint, there is statistically no difference between a Gore-TexTM and a polyester clean room suit meaning no one ensemble is noticeably “cooler” than another.

Determining Physical Activity Level
Physical activity is expressed in terms of the rate of energy production of the body or metabolic rate. To determine a physical activity level for each job in the clean room would be impractical and costly. As a result, it was decided to select three representative jobs that characterize working in the clean room environment as shown in Table 3.

Table 3. Representative Clean Room Jobs

 

Job

Task Description

Wafer
Inspection

Microscope inspection, mostly sedentary, some walking.

Deposition

Equipment operation, walking, manual material handling.

Maintenance

Maintenance and PM tasks, high peak exertions in awkward postures for short duration.

 A third party consultant was contracted to design and implement a limited heart rate monitoring study to collect heart rate data for use in determining the average metabolic rate while performing each job.

Testing Setup and Procedure
Each subject was outfitted with a small heart rate monitoring device to record data continuously throughout the shift. Prior to actual data collection in the clean room, each subject was required to complete a short questionnaire and perform “step test” to obtain baseline heart rate data.

Once in the clean room, each subject was asked to maintain an activity log detailing their tasks throughout their shift. The heart rate data and activity log were then used to calculate average energy expenditure for each subject over an entire 12-hour shift.

In addition, each subject completed a subjective thermal comfort self-survey twice during their shift to help validate the thermal comfort algorithm. The surveys were completed one hour after the start of shift and thirty minutes prior to the end of shift.

Study Results
Table 4 shows the results of the heart rate monitoring study.

Table 4. Physical Activity Level Testing Results

 

Job

Sample
Size

Avg.
Metabolic
Rate (met)

Standard
Deviation

Wafer Inspect

13

 2.6

0.84

Deposition

12

2.6

0.82

Maintenance

15

2.6

0.44

Overall

40

2.6

0.69

 Establishing a Baseline of Current Clean Room Thermal Comfort
With the clothing insulation and physical activity levels determined, a baseline of the current clean room thermal comfort environment was established using the thermal comfort model detailed in ISO 7730. Indices for the model are explained below.

Predicted Mean Vote (PMV) Index
The thermal sensation for the body as a whole can be predicted by calculating the Predicted Mean Vote (PMV) index. The PMV represents a prediction of the mean vote of a large population of people on the following 7-point thermal sensation scale:

Figure 2. PMV: 7-Point Thermal Sensation Scale

PMV is derived from the physics of the human thermoregulatory system combined with an empirical fit to thermal sensation. PMV establishes a thermal strain based on steady-state heat transfer between the body and the environment and assigns a comfort vote to that amount of strain. The PMV is given by the following simplified equation.

PMV = (0.303e-0.036M + 0.028) * ((M-W) - H - Ec - Cres - E res)**
Equation 1. Predicted Mean Vote

Where:
M = Metabolic rate. Rate of transformation of chemical energy into heat and mechanical work by aerobic and anaerobic activities within the body.
W = External work, equal to zero for most activities.
H = Dry heat loss. Heat loss from the body surface through convection, radiation and conduction.
Ec = Evaporative heat exchange at the skin, when the person experiences a sensation of thermal neutrality.
Cres = Respiratory convective heat exchange.
Eres = Respiratory evaporative heat exchange.
Note: All units in W/m2 unless denoted otherwise.

Predicted Percentage Dissatisfied (PPD) Index
The PPD index predicts the percentage of a group of people likely to feel too warm or cool (i.e. voting hot (+3), warm (+2), cool (-2) or cold (-3) on the 7-point thermal sensation scale). After the PMV value has been determined, the PPD for that PMV value be can calculated using the following equation:

PPD = 100 - 95 x e -(0.03353 x PMV4 + 0.2179 x PMV2)
Equation 2. Predicted Percentage Dissatisfied

It should be noted that as the PMV deviates away from zero in either the positive or negative direction, the PPD increases.

PMV and PPD Results from Thermal Comfort Model
Because the ISO 7730 thermal comfort model is a long and mathematically intensive algorithm, the analysis process was performed utilizing the ASHRAE Thermal Comfort Program, a software package developed to determine both the PMV and PPD indices quickly and easily. Table 5 summarizes the PMV and PPD indices for each clean room ensemble detailed earlier with all environmental and personal variables at current 200mm factory levels (Ta, Tr = 72oF, RH = 40%, Var = 80 ft./min., M = 2.6 mets).

Table 5. Current Clean Room PMV and PPD Index Results

 Ens.
No.

 Predicted
Mean
Vote
(PMV)

 % Pop.
Dissatisfied
(PPD)

 Intrinsic
Clothing
Ins. (Icl)

 Avg.
Metabolic
Rate (M)

 1

 1.47

 49

 1.17

 2.6

 2

 1.45

 48

 1.15

 2.6

 3

 1.42

 47

 1.10

 2.6

4

 1.41

 46

 1.08

 2.6

  5

 1.52

 52

 1.25

 2.6

6

 1.49

 50

 1.21

 2.6

  7

 1.48

 50

 1.19 

 2.6

 8

 1.47

  49

 1.18

 2.6

As can be seen from the table, the PMV scores range from 1.41 to 1.52 which equates between “slightly warm” to “warm” on the 7-point thermal comfort index. The PPD index ranges from 46% to 52%.

PMV and PPD Results from Employee Self-Survey
As mentioned earlier, each participant in the heart rate monitoring study was asked to complete a subjective thermal comfort self-survey twice during their respective shift. The survey data was collected to compare it with the output from the thermal comfort model to confirm its validity in predicting clean room thermal comfort environments. The results are shown in Table 6. Only the thermal comfort results from the Gore-Tex/Dryden ensembles (#4 and #8) were compared since these were the ensembles worn by subjects during the heart rate monitoring study

Table 6. PMV and PPD Comparison (Model vs. Survey Results).

 

Ensemble

Predicted
Mean Vote
(PMV)

% Population
Dissatisfied
(PPD)

Summer/Gore-Tex
Suit and Dryden

1.41

 46

Winter/Gore-Tex Suit
and Dryden

1.47

49

Survey Results (Begin
of Shift)

1.35

43

Survey Results (End
of Shift)

1.44

48

Statistical Analysis of PMV and PPD Results
To prove the validity of the thermal comfort model, statistical hypothesis testing using a t-statistic was performed to compare the survey PMV results (1.35, 1.44) to the model PMV results (1.41 and 1.47). The hypothesis for the Summer/Winter Gore-Tex Suit and Dryden ensemble (PMV=1.41, 1.47) is:

Ho: μ = 1.41 (or 1.47)
H1: μ ≠ 1.41 (or 1.47)
α = 0.05; .
t α/2 =±1960(two-tailed test)

Critical region: t < -1.96 and t > 1.96, where

 Equation 3. t-Test Statistic Example Calculation

Analysis Results
The statistical analysis is summarized in Table 7.

Table 7. Statistical Analysis Results

 

Survey
(begin/shift)

Survey
(end/shift)

Survey
(begin/shift)

Survey
(end/shift)

Test Mean = Value

Hypoth.
Value

1.41

1.41

1.47

1.47

Actual
Estimate

1.35294

1.44118

1.35294

1.44118

t-Test Results

Test
Statistic

-0.311

0.1533

 -0.6381

-0.1418

Decision

Accept
Ho

Accept
Ho

Accept
Ho

Accept
Ho

 From the statistical analysis, it can be seen that none of the test statistics fall within the critical region defined. Therefore the null hypothesis should be accepted and it should be concluded that the thermal comfort model is valid and the PMV results obtained from the model represent the current clean room thermal comfort environment. Since the PPD is a direct calculation using the PMV, these results should also be considered valid.

Predicting Impact on Employee Thermal Comfort
The second objective was to determine the employee impact of altering environmental conditions in the clean room environment. In particular, determining the impact of lowering the airflow from 80ft./min. to 40 ft./min.

300mm Fab Modeling Constraints
As shown earlier, the PMV and PPD indices in the current 200mm fabs exceed the ISO guidelines for acceptable employee thermal comfort. Given this, the proposed temperature and humidity levels must at least maintain the current PMV and PPD indices.

Temperature Constraints
Holding the other variables constant and lowering the clean room airflow increases the PMV and PPD indices. Other variables including temperature have to be reduced to counteract this effect. However, to maintain process stability, the temperature cannot fall below 69oF.

Relative Humidity Constraints
If the temperature is lowered to counteract the effects of reduced airflow, the “absolute” humidity level will be impacted. Varying the “absolute” humidity also has a detrimental impact on manufacturing processes. The lower the temperature, the higher the relative humidity level required to maintain the same “absolute” humidity as shown in graph 1 below.

"Absolute" Humidity Comparison

 Graph 1. Relative Humidity vs. Temperature

Model Prediction Results
Using the constraints provided, several combinations of temperature and humidity were evaluated using the thermal comfort model. A temperature of 69 oF and relative humidity level of 44% were found to be optimal. A comparison of the proposed and current variable levels are shown below in Table 7.

Table 7. 200mm vs. 300mm PMV and PPD Comparison

 

Current 200mm Environment

Proposed 300mm
Environment

Ens.
No.

PMV

 PPD

PMV

PPD

1

1.47

49

1.34

42

2

1.45

48

1.33

42

3

1.42

47

1.30

40

4

1.41

46

 1.28

39

5

1.52

52

1.39

45

6

1.49

50

1.36

44

7

1.48

50

 1.35

43

8

1.47

49

1.35

43

From the data, it can be seen that there is a small improvement in the overall PMV and PPD indices with the proposed variable levels for 300mm factories (~7% across all ensembles). The proposed levels do not accommodate 80% of the clean room employee population per the recommendations of the ISO 7730 thermal comfort standard. However, the results of thermal comfort surveys taken during the heart rate monitoring study show that most employees find the current environment acceptable. While there are some complaints that the clean room is too warm, there are also complaints of it being too cool as well.

Predicted Cost Savings
See Table 8 for predicted cost savings percentages from the reduction in clean room airflow and temperature.

Table 8. Cost Savings Summary

 

Cost Category

% Savings

Savings
Type

Initial
Construction

10.3%

One-Time

Operations

26.3%

Annual

Gowning

6.6%

One-Time

 

 It should be noted that these potential cost savings are for one factory only and that the operational cost savings are annual savings.

Summary
In summary, a thermal comfort model was developed and a current baseline established. The model was statistically validated using actual thermal comfort data collected from employees. Potential employee impact of altering the clean room environment was predicted. The results were used to justify reduction in clean room airflow for future wafer factories.

References
1. ASHRAE (1981). Thermal Environmental Conditions for Human Occupancy (ASHRAE Standard 55-1981). Atlanta, GA: American Society of Heating, Refrigerating and Air Conditioning Engineers
2. ISO (1994). Moderate Thermal Environments - Determination of the PMV and PPD Indices and Specification of the Conditions for Thermal Comfort (ISO Standard 7730-1994). Geneva, Switzerland: International Organization for Standardization.
3. Salvendy, Gavriel (1987). Handbook of Human Factors. New York: John Wiley & Sons Publishing.
4. Walpole, Ronald E. and Myers, Raymond H. (1989). Probability and Statistics for Engineers and Scientists. New York: Macmillan Publishing Company.

Acknowledgements
The author would like to acknowledge the following people that made this project a success.
• Nora Sylvestre, W. L. Gore and Associates, Inc.
• Elizabeth McCullough, Ph.D., Institute for Environmental Research, Kansas State University
• Joseph Selan, Ph.D., Advanced Ergonomics, Inc.
• Fab 12 Occupational Health Staff, Intel Corporation
• Doug Grant, Intel Corporation