13JR EAST Technical Review-No.33

Special edition paper

Passengers have long said that the thermal environment of commuter train cabins is too hot or cold. They continue to submit such opinions despite the measures we have taken, such as improving air conditioning capacity.

The cabin thermal environment of commuter trains is considered from the development stage, and the their temperature distribution is uniform when empty. However, when investigating the thermal environment in operating trains with passengers on them, we confirmed that irregularity in temperature occurs at times such when they are crowded. In this way, uncomfortable situations that we could not imagine when studying empty trains occurred when there were passengers on the train, leading to such opinions.

We therefore need to study under the condition that there are passengers present in order to make a more comfortable cabin thermal environment. In that, we may be able to apply simulation technologies that have seen remarkable advancement in recent years to reproduce cabin thermal environments assuming trains in commercial operation and predict psychological response of people under such environments (thermal sensation, comfort, etc.). We believe reproduce uncomfortable situations in cabins of operating trains in this manner will allow us to consider measures to improve on those at the design stage, thereby contributing to creation of comfortable cabin thermal environments.

Up to now, ordinary computational fluid dynamics (CFD) has been utilized as a method of reproducing thermal environments by simulation,1) and human thermal models have been developed2) taking into account thermoregulatory functions, perspiration, and the like in order to predict psychological response from physiological response. Studies have been done in the past to couple those and predict thermal comfort under environments that are not uniform.3) However, there have not been many case examples of reproducing by CFD thermal environments with many people, such as in cabins of operating trains, and evaluating psychological response.

In this study, we thus attempted to reproduce the thermal environment in cabins of operating trains when coolers are operating by simulation using CFD and human thermal models. The objective was to use this as a technique to improve

Introduction1uncomfortable conditions that occur in cabins of operating trains and make a comfortable cabin thermal environment and to predict psychological response under that environment.

Cabin Thermal Environment in Commuter Trains2

2.1 Cooling Method for Commuter TrainsThe cabin cooling method studied is that which is ordinarily used in commuter trains as shown in Fig. 1. Cool air produced by air conditioners mounted on the train roof flows through air conditioning ducts in the ceiling and is blown into the cabin by two rows of outlets on the ceiling. It is then drawn in through return inlets at the center of the ceiling and circulated. Cross flow fans are also mounted on the ceiling to improve the cooling effect and serve as circulators.4)

Cooing control is done by detecting cabin temperature, humidity, outside temperature, occupancy rate, door opening/

Modeling and Evaluation of Thermal Environment in Railway Vehicles by Simulation

•Keywords: Air conditioning, Simulation, Computational Fluid Dynamics (CFD), Human thermal model, Thermal sensation prediction

In this study, simulation of the thermal environment in the cabin of a train in commercial operation using CFD analysis and human thermal models was conducted to achieve a comfortable thermal environment in the design phase. Psychological response under the environment was predicted as well, and the precision of reproduction and predictability were verified. As a result, it was confirmed that the thermal environment could be simulated and that thermal sensation under such an environment could be predicted with practical accuracy.

*Advanced Railway System Development Center, Research and Development Center of JR East Group

Shinichi Hasegawa*Ryohei Shimamune*Nobuaki Hayashi*

Air conditionerAir conditioning ductFlow of air

Door

Window

(a) Schematic of lengthwise cross section of cabin

Air conditioning outlet

Air conditioning outlet

(b) Schematic of ceiling part of cabin (viewed from beneath)

Cross flow fan Return inlet

Air conditioning outlet

Air conditioning duct

Crossflow fan

Swinging motion

(c) Schematic of short direction cross section of cabin

Flow of air

Fig. 1 Commuter Train Air Conditioning System

14 JR EAST Technical Review-No.33

Special edition paper

3.2 Method of Measuring Thermal EnvironmentTemperature, humidity, wind speed, and wall temperature in the cabin, temperature and humidity directly under outlets, and outlet air flow volume/direction were measured as the thermal environment. Temperature in the cabin was measured at multiple locations at heights of 0.15 m, 1.1 m, and 1.7 m, humidity was measured at multiple points at a height of 1.1 m, wind speed was measured at multiple points at a height of 1.1 m, and wall temperature was measured at multiple points of differing materials.

3.3 Method of Measuring Psychological ResponseThe psychological response of people to the thermal environment was measured by having test subjects remain in the cabin for 21 minutes positioned at the locations shown in Fig. 3 with the temperature controlled to be 25 ºC and then taking a

closing information, and the like and automatically controlling to maintain the cabin temperature at the set temperature called the reference temperature.

2.2 Thermal Environment of Trains in Operation In order to identify the actual conditions on trains in operation, we rode a train in the greater Tokyo area during the morning commute in October and measured the thermal environment. At the time, outside temperature was 22 to 25 ºC.

Fig. 2 shows the changes over time for vertical temperature distribution, horizontal (lengthwise) temperature distribution, and occupancy rate measurement results in a certain section. The measurement points were heights of 0.15 m, 0.8 m, and 1.5 m for vertical temperature distribution and height of 1.0 m at the center and ends of the car for horizontal temperature distribution.

From Fig. 2, we can see that temperature difference between measurement points was low when occupancy rate was lower than about 80%. But as the occupancy rate becomes greater, temperature differences of up to 5 ºC were seen in the vertical direction and up to 4 ºC in the horizontal direction, confirming that there was irregularity in temperature in the cabin. As uniform distribution was maintained when the occupancy rate was low, air from air conditioning being prevented from uniformly reaching all parts of the car due to the presence of passengers could be a factor behind irregularity in temperature.

Such irregularity in temperature in the cabin is believed to create localized uncomfortable environments and lead to the aforementioned opinions by passengers, so improvements should be made.

Obtaining Data for Verifying Simulation Accuracy3

In order to verify accuracy of reproducing the temperature environment by simulation and accuracy of psychological response prediction, we conducted measurements on an active duty car to obtain measurement data to compare. Only cooling conditions were measured for air conditioning.

3.1 Measurement ConditionsMeasurements on an active duty car were conducted in September 2014. Multiple test subjects boarded a commuter train car parked outside a depot to simulate the environment of an operating train. The subjects were given sufficient explanation on the purpose and details of the measurement, and their informed consent was obtained.

There were 43 test subjects without bias towards sex or age, and they were placed standing at the locations shown in Fig. 3 to create an environment equivalent to that with an occupancy rate of 75%. Cooling was automatically controlled to keep cabin temperature at 25 ºC. Measurements were taken under the two conditions of cross flow fans not operating and of them operating by swinging in cycles of 13.4 sec. to send air into the cabin at air flow of 9 m3/min. In order to reduce the affect of sunlight, windows were covered by curtains and the like, preventing intrusion of sunlight.

(a) Vertical distribution

020406080100120140160180200

2021222324252627282930

0 5 10 15 20 25 30

Occ

upan

cy ra

te (%

)

Tem

pera

ture

(ºC

)

Time (minutes)

(b) Horizontal distribution

Occ

upan

cy ra

te (%

)

Tem

pera

ture

(ºC

)

Time (minutes)

Height: 0.15 mHeight: 0.8 mHeight: 1.5 mOccupancy rate

020406080100120140160180200

2021222324252627282930

0 5 10 15 20 25 30

Center of carEnd of carOccupancy rate

(a) Placement of test subjects

SeatDoor

Test subject

(b) Photo at time of measurements

Fig. 2 Thermal Environment Measurement Results for Cabin of Operating Train

Fig. 3 Placement of Test Subjects at Time of Measurements

15JR EAST Technical Review-No.33

Special edition paper

questionnaire survey. The item for which response was obtained was thermal sensation, and a nine-level scale was used to be compatible with Thermal Sensation Vote (TSV) used by Fukai et al.5) Specifically, the responses were hot: 4, slightly hot: 3, warm: 2, slightly warm: 1, neutral: 0, slightly cool: -1, cool: -2, slightly cold: -3, and cold: -4.

Conducting Simulation4Simulation was done by placing human thermal models within a commuter train model and conducting coupled analysis of CFD and human thermal models to reproduce a thermal environment simulating the cabin of a commuter train with passengers. Psychological response under that reproduced environment was then predicted.

4.1 Calculation Accuracy TargetsFirst we defined the targets for calculation accuracy by simulation. In this study, we predict thermal sensation in the nine levels described in 3.3 from the results of reproducing the thermal environment by simulation, aiming for error of within one level taking into account practicality use.

We then found accuracy of reproducing the thermal environment needed to meet this target. In this study we used a prediction formula that utilized Standard new Effective Temperature (SET*) as the method of predicting thermal sensation from the thermal environment. SET* is a type of perceived temperature presented in 1971 by the American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE). It is defined as the equivalent air temperature of an isothermal environment at 50% rh in which a subject, wearing clothing standardized for the activity concerned, has the same heat stress and thermoregulatory strain as in the actual environment, and it has the units ºC 6). It can be calculated from temperature, humidity, radiant temperature, wind speed, amount of clothing, and metabolic level. And studies have been conducted to solve a prediction regression formula of TSV utilizing SET* like that below.5)

TSV = a × SET* − b ····· (1)Here a and b are coefficients, and they are in calculation to

solve a regression formula from TSV and SET* measurement data. In the research by Fukui et al., coefficient a of formula (1) is about 0.3 to 0.5, and SET* error needs to be within 2.0 ºC to make thermal sensation error within one level. For that reason, error in temperature needs to be within 2.0 ºC and wind speed within 0.3 m/s for the cabin thermal environment. In this study, we set targets for difference between measured and calculated values in the simulated thermal environment to be within the aforementioned range.

4.2 Car Body ModelWe built a car body thermal model to reproduce the thermal environment. The car body model creates in 3D-CAD the cabin of the commuter train used for measurements on an active duty car, and it reproduces equipment such as seats, luggage racks, hand straps, hanging ads, and handrails that affect the flow field in the cabin (Fig. 4 (a)). In order to shorten calculation time,

luggage racks, hand straps, hanging ads, and handrails were only reproduced in the area where test subjects were present in the measurements on an active duty car (Fig. 4 (b)).

The method of analysis by CFD was a finite volume method commonly used, and the turbulence model was the k- ε model commonly used as a turbulent flow field model.

Input conditions of wind speed and direction from outlets, temperature, humidity directly under the outlets, cabin wall temperature, and conditions of whether or not cross flow fans are operating were given in a time-series manner. From those, cabin wind speed, temperature, and humidity were found in a time-series manner.

4.3 Human Thermal ModelsAs shown in Fig. 5, human thermal models were placed at the locations where test subjects were present (Fig. 3 and 4) in the measurements on an active duty car, and the thermal environment at the time of active duty car measurements was reproduced by conducting coupled analysis with CFD.

The human thermal models used are the JOint System Thermoregulation Model (JOS)7) developed by Tanabe et al.

(b) Overall view of cabin

(a) Reproduced cabin equipment

Area with test subjects present

HangingadvertisementLuggage rack

Longitudinal seating

Heater (only shape reproduced, no heating)

Handrail

Hand strap

Air conditioning outletCross flow fan

Luggage rack shape

Fig. 4 Car Body Model Created in 3D-CAD

Fig. 5 Placement of Human Thermal Models

16 JR EAST Technical Review-No.33

Special edition paper

The thermal environment surrounding the model (temperature, humidity, wind speed, etc.) obtained from CFD analysis and test subject information (sex, age, body fat rate, amount of clothing, calorific value, etc.) at active duty car measurements were added to reproduce heat generation, perspiration, and the like of human bodies. Body fat rate for ordinary adults was assumed to be 20% for males and 30% for females, amount of clothing 0.5 clo as the test subjects were instructed to wear short-sleeved shirts and long pants, and calorific value 116 W for males and 65 W for females found from the metabolic rate when standing.

Comparison of Measurement and Simulation Results5

Results of measurements in an active duty car and of CFD - human thermal model coupled calculation were compared, and we evaluated the accuracy of reproducing the thermal environment by simulation and the accuracy of predicting psychological response.

5.1 Cabin Wind Speed Distribution Prediction Accuracy Fig. 6 shows the cabin wind speed distribution obtained by CFD analysis under the condition that cross flow fans are operating. From this, we can confirm flow from outlets and cross flow fans.

For cabin wind speed with cross flow fans operating, we compared values measured and value calculated. Fig. 7 shows the results of comparing wind speed values at different measurement points. We were able to reproduce wind speed average values within an error range of the target accuracy of 0.3 m/s. Also, wind speed fluctuates cyclically due to cross flow fan swinging motion, and the range of wind speed fluctuation in the areas directly under cross flow fans (points 2, 3, and 4) is large. We are able to reproduce this phenomenon even in calculations.

Similar reproduction accuracy could also be gained even under the condition that cross flow fans are stopped.

5.2 Cabin Temperature Distribution Prediction AccuracyFig. 8 shows the cabin temperature distribution obtained by CFD analysis under the condition that cross flow fans are operating. Change over time of cabin temperature measured and calculated values showed similar tendencies at all locations and heights, so we show the results at 1.1 m height for point 2 (see Fig. 7) as a comparative example in Fig. 9. From calculations, we are able to reproduce change in temperature in fluctuating measurements within an error range of the target accuracy of 2.0 ºC. Calculated values fluctuating in a short cycle is the result of the flow field and temperature field both being measured in a non-stationary manner. Change in temperature could also be reproduced at other locations and heights within an error range of 2.0 ºC as in Fig. 9. Similar reproduction accuracy could also be gained even under the condition that cross flow fans are stopped.

5.3 Cabin Humidity Distribution Prediction AccuracyFig. 10 shows the cabin humidity distribution obtained by CFD analysis under the condition that cross flow fans are operating. From this, we can see that locations where cool air is introduced to the cabin from outlets have higher humidity.

0.0 2.01.0

m/sec

(b) Calculation results (cross section AA)

(c) Calculation results (cross section BB)

SeatDoor

Test subjectCross flow fan

↑A

↑A

B→

B→(a) Schematic of cabin

C→

C→

(d) Calculation results (cross section CC)

20

22

24

26

28

30

0 500 1000 1500

Tem

pera

ture

(ºC

)

Time (seconds)

Measured value Calculated value

Fig. 6 Example of Wind Speed Distribution Calculation Results

Point 1 Point 2 Point 3 Point 4 Point 5(a) Wind speed measurement point

(b) Comparison of wind speed values (at 1.1 m height)

00.10.20.30.40.50.6

Point 1 Point 2 Point 3 Point 4 Point 5

Win

d sp

eed

(m/s

ec)

Measured value Calculated value

*Values in graph are average wind speeds, and error bar shows range of wind speed fluctuation.

Seat Door

Test subjectCross flow fan

Fig. 7 Comparison of Measured and Calculated Values (Wind Speed with Cross Flow Fan Operating)

15 3020

oC

(a) Calculation results (cross section AA)

(b) Calculation results (cross section BB)

(c) Calculation results (cross section CC)

25

*See Fig. 6 (a) regarding cross section names

Fig. 8 Example of Temperature Distribution Calculation Results

Fig. 9 Change Over Time of Measured and Calculated Values (Temperature with Cross Flow Fan Operating)

(Point 2, at 1.1 m height)

17JR EAST Technical Review-No.33

Special edition paper

Also, Fig. 11 shows change over time of cabin humidity measured and calculated values at 1.1 m height for point 2 in Fig. 7. From the calculations, we see that fluctuating change in humidity over time is reproduced with error of about 10% RH.

Similar reproduction accuracy could also be gained with error of about 20% RH even under the condition that cross flow fans are stopped. That error is small enough not to affect prediction of thermal sensation.

5.4 Psychological Response Prediction AccuracyIn this research, we worked to predict psychological response to thermal sensation. The method of predicting thermal sensation was to use the thermal environment values (see 5.1 to 5.3) calculated by coupled calculation of CFD and human thermal models to find SET* and substituting that in formula (1). For coefficients a and b in formula (1), the results of tests on subjects performed in a similar commuter train (a=0.5379, b=12.74)8)

were used. And when calculating SET*, radiation temperature was set at the same as air temperature, and metabolic rate was 70 W/m2 and clo value 0.5.

In order to confirm that thermal sensation prediction could be done by individual location in the cabin, we set seven groups of four people lengthwise in the cabin (Fig. 12) and evaluated by average thermal sensation in the individual groups. Calculation values used for prediction of thermal sensation were obtained at a height of 1.1 m at the center of the four-person groups (Fig. 12).

Fig. 13 shows the change over time of thermal sensation (TSV) measured and calculated values under the condition that cross flow fans are operating. From calculation, we were able to reproduce change trends for thermal sensation at each location. Prediction error for thermal sensation was mainly due to individual differences and wind speed value calculation error.

Fig. 14 shows a comparison of measured and calculated values for the averages of each group under the conditions that cross flow fans are operating or stopped. From this, we found that prediction error of thermal sensation using with simulation can be obtained with accuracy of within about one of the nine levels.

From the above, we reproduced by simulation a cabin thermal environment of operating trains with passengers

present, confirming that thermal sensation felt by people in that environment can be predicted at practical accuracy.

0 10050

%RH

(a) Calculation results (cross section AA)

(b) Calculation results (cross section BB)

(c) Calculation results (cross section CC)

*See Fig. 6 (a) regarding cross section names

40

45

50

55

60

65

70

0 500 1000 1500

Tem

pera

ture

(ºC

)

Time (seconds)

Measured value Calculated value

Group 1Group 2

Group 3Group 4

Group 5Group 6

Group 7

Location where thermal environment values are obtained

Seat Door

Test subject

-4-3-2-101234

0 500 1000 1500TSV

Measured valueCalculated value

Measured valueCalculated value

-4-3-2-101234

0 500 1000 1500TSV

-4-3-2-101234

0 500 1000 1500TSV

-4-3-2-101234

0 500 1000 1500TSV

-4-3-2-101234

0 500 1000 1500TSV

-4-3-2-101234

0 500 1000 1500TSV

-4-3-2-101234

0 500 1000 1500TSV

-4-3-2-101234

0 500 1000 1500TSV

Hot

Cold

Warm

Cool

(a) Group 1 (c) Group 3

(e) Group 5 (g) Group 7

(b) Group 2 (d) Group 4

(f) Group 6 (h) Overall average

Neutral

Measured valueCalculated value

Measured valueCalculated value

Measured valueCalculated value

Measured valueCalculated value

Measured valueCalculated value

Measured valueCalculated value

Time (seconds)Time (seconds)

Time (seconds) Time (seconds)

Time (seconds)

Time (seconds)

Time (seconds)

Time (seconds)

Fig. 13 Change Over Time of Measured and Calculated Values (Thermal Sensation with Cross Flow Fan Operating)

Fig. 10 Example of Humidity Distribution Calculation Results

Fig. 12 Group Settings for Test Subjects

Fig. 11 Change Over Time of Measured and Calculated Values (Humidity with Cross Flow Fan Operating)

(Point 2, at 1.1 m height)

18 JR EAST Technical Review-No.33

Special edition paper

Reference:1) Akihiro Fujita, Junichi Kanemaru, Hiroshi Nakagawa, Yoshiichi

Ozeki, “A Practical Method of Numerical Simulation to Predict Thermal Environment Conditions of Car Cabin [in Japanese]”, HONDA R&D Technical Review Vol. 11, No.2 (1999): 89-98.

2) Adolf P. Gagge, Jan A. Stolwijk, Yasunobu Nishi, “An effective temperature scale based on a simple model of physiological regulatory response”, ASHRAE Trans. Vol. 77, No.1 (1971): 247-262

3) Yang Ling , Shinsuke Kato, Hideaki Nagano, Zhu Sheng wei, “Comparisons of Thermal Adaptive Effect Of Positional Adjustment of a Seated Human Body Between Fanger's and Sakoi's Thermal Model [in Japanese]”, Architectural Institute of Japan Journal of Environmental Engineering Vol. 73, No.630 (2008): 979-984

4) Kazuhiko Shiraishi, Osamu Sakai, “Air conditioning system for advancement of amenity in the train [in Japanese]”, Mitsubishi Denki Giho Vol. 81, No.10 (2007): 681-684

5) Kazuo Fukai, Hiromu Ito, Shigeru Gotoh, Satoru Akui, Junji Saito, “Experimental Study on Correlation between Standard New Effective Temperature (SET*) and Japanese Thermal Sensation: Part 2 - Comparison of Thermal Sensation in Winter and Summer Seasons [in Japanese]”, Transactions of the Society of Heating, Air-Conditioning and Sanitary Engineers of Japan No.51, (1993): 139-147

6) ASHRAE Handbook Fundamentals, ASHRAE (1989): 137) Xu Li, Takahiro Sato, Kazuaki Ogawa, Shin-ichi Tabe, “Numerical

Comfort Simulator for Thermal Environment (Part 22): Development of Human Thermoregulation Model JOS with the Vascular System Using AVA [in Japanese]” Summaries of technical papers of Annual Meeting Architectural Institute of Japan D-2, (2002): 361-362

8) Nobuaki Hayashi, Ryohei Shimamune, Shinichi Hasegawa, Hiroharu Endoh, Tetsuyuki Ohe, “Modeling and evaluation of thermal environment in railway vehicle cabin [in Japanese]”, Papers Submitted to the 22nd Transportation and Logistics Conference, The Japan Society of Mechanical Engineers, (2013): 47-50

Conclusion6In this study, we reproduced by simulation using CFD and human thermal models the thermal environment of an operating train when coolers are operating and predicted psychological response under that environment. This is used as a method to consider improvement of unpleasant situations in operating train and creation of a comfortable cabin thermal environment.

First, we conducted measurements in an active duty car to obtain measurement data for verifying accuracy of reproducing the thermal environment and psychological response by simulation and input some of the measurement data in that simulation.

As a result, we were able to reproduce wind speed, temperature, and humidity changing over time in the thermal environment of an operating train. Furthermore, accuracy of reproducing could be kept within the accuracy required for calculations to predict thermal sensation. And while some localized error occurred, thermal sensation predicted from the reproduced thermal environment values prediction error within about one of the nine levels could be obtained for the car as a whole.

From the above findings, we were able to confirm that the thermal environment in operating trains with passengers present could be reproduced by simulation and that thermal sensation felt by people in that environment can be predicted at practical level of accuracy. In the future, we will study using simulation technology to make improvements to conditions that make passengers feel uncomfortable, such as irregularity in temperature.

Acknowledgment We would like to express out thanks to Akira Kondo of Toyota Technical Development Corporation for his cooperation in conducting this research.

Fig. 14 Comparison of Average Thermal Sensation by Measured and Calculated Values

-2

-1

0

1

-2 -1 0 1TS

V c

alcu

late

d va

lue

TSV measured value

With cross flow fan operatingWith cross flow fan stopped