University of Southern California Dissertations and Theses

Developing a data-driven model of overall thermal sensation based on the use of human physiological information in a built environment

(USC Thesis Other)

Developing a data-driven model of overall thermal sensation based on the use of human physiological information in a built environment

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content
1

Developing a Data-Driven Model of Overall Thermal Sensation based on the Use
of Human Physiological Information in a Built Environment

By

Qi Wang

Presented to the
FACULTY OF THE
SCHOOL OF ARCHITECTURE
UNIVERSITY OF SOUTHERN CALIFORNIA
In partial fulfillment of the
Requirements of degree
MASTER OF BUILDING SCIENCE

August 2017

2

COMMITTEE

Joon-Ho Choi
Assistant Professor
USC School of Architecture
joonhoch@usc.edu
(213) 740-4576

Marc Schiler
Professor
USC School of Architecture
marcs@usc.edu
(213)740-4591

Douglas Noble
Director of Graduate Building Science
Chair of the Ph.D. Program in Architecture
dnoble@usc.edu
(213)4465416

3

To my mom and dad, I couldn't finish this without you.

I love you forever.

4

ACKNOWLEDGEMENTS

First and foremost, I want to express my deep gratitude to my committee chair, Professor
Joon-Ho Choi. He is so creative and positive. He gave me great advice, the advice he
gave me can benefit my whole life. Besides, he provided device support for the whole
project, the experiment can’t be completed without his help.

Second, I would also like to acknowledge to Professor Marc Schiler and Douglas Noble.
They are so nice and patient, they gave me lots of encouragement during the whole
project, their advice also help me improve the writing habits and skills.

Last but not the least, I would like to thank all the MBS students, we are like a family. I
have a great time with you and it is so wonderful to meet you in my life.

5

ABSTRACT

Traditional air-conditioning methods maintain temperatures in a whole room at a constant level,
and much work has been done to assess and improve the thermal comfort and sensations of
people in a building environment. This study endeavors to identify the potential of using some
local body area for predicting thermal stress. A total of 20 human subjects were tested in the
University of Southern California’s climate chamber to determine their physiological parameters
and subjective perceptions of environment. Ambient temperature was documented during the
tests, while the human subjects were exposed to a warm, cool, or neutral environment. Based on
these tests, correlation analysis and algorithms are applied to identify the relative thermally
sensitive skin areas, their contribution rate to the overall thermal sensation, and potential skin area
combinations that have high correlation with thermal sensation. Besides, the study also identifies
the different impacts of local thermal sensation and local skin temperature while predicting the
overall thermal sensation. When only the baseline attributes (environment temperature, Body
Mass Index [BMI], gender) are considered, the estimation can be 94.5%. If baseline attributes are
combined with the temperature of one local spot on the skin, the estimation accuracy can be
around 97%; on the other hand, if baseline attributes are combined with one local thermal
sensation, the estimation accuracy can be around 98%.
With the correlation analysis and the application of the data-driven model, a method that uses
some local body area to predict the overall thermal sensation can be developed.
Keywords: Thermal sensation, Thermal comfort, Thermal environment, Skin temperature

6

TABLE OF CONTENTS

ACKNOWLEDGEMENTS ...................................................................................................... 4
ABSTRACT .............................................................................................................................. 5
TABLE OF CONTENTS .......................................................................................................... 6
1. Introduction ....................................................................................................................... 9
1.1. Energy consumption in the building industry ............................................................ 9
1.2. Indoor environment quality ..................................................................................... 10
1.3. Basics of thermal comfort ........................................................................................ 11
1.3.1. Factors in human comfort ................................................................................ 12
1.3.2. PMV and PPD model ...................................................................................... 13
1.4. Hypothesis statement ............................................................................................... 13
1.4.1. Hypothesis ....................................................................................................... 13
1.4.2. Goal ................................................................................................................. 14
1.4.3. Objectives ........................................................................................................ 14
1.5. Study statement ....................................................................................................... 14
1.6. Important terminology ............................................................................................. 14
1.7. Summary.................................................................................................................. 16
2. Background research ....................................................................................................... 17
2.1. Physiological basics of thermal sensation ............................................................... 17
2.1.1. Skin structure ................................................................................................... 17
2.2. The physiological balance and thermal regulation model ....................................... 19
2.3. Evaluation model of thermal sensation ................................................................... 21
2.4. Study of thermal sensation in different thermal conditions ..................................... 24
2.4.1. Study of thermal sensation in uniform conditions ........................................... 24
2.4.2. Study of thermal sensation in non-uniform conditions .................................... 25
2.5. Thermal comfort control method ............................................................................. 27
2.6. Conclusions for background research ...................................................................... 28
3. Project methodologies ..................................................................................................... 29
3.1. Workflow overview ................................................................................................. 29
3.2. Experimental chamber ............................................................................................. 30
3.3. Equipment for the experiment ................................................................................. 31
3.3.1. Heating and cooling system ............................................................................. 31
3.3.2. Data acquisition tool ........................................................................................ 34
3.3.3. Data acquisition software ................................................................................ 37
3.4. Human subject tests ................................................................................................. 38
3.4.1. Volunteer recruitment ...................................................................................... 39
3.4.2. Experiment procedure ..................................................................................... 40
3.4.3. Survey Questionnaire ...................................................................................... 42
3.5. Statistical analyses ................................................................................................... 42

7

3.5.1. Data analysis and mining tools ........................................................................ 42
3.5.2. Exploring data with graphics ........................................................................... 43
3.5.3. Statistical methods ........................................................................................... 43
3.5.4. Data-driven method ......................................................................................... 43
3.6. Summary.................................................................................................................. 44
4. Human subject experiment data and results .................................................................... 45
4.1. Overall thermal sensation analysis .......................................................................... 45
4.1.1. Overall thermal sensation and environmental temperature ............................. 45
4.1.2. Overall thermal sensation and local thermal sensation.................................... 45
4.1.3. Overall thermal sensation and local skin temperature ..................................... 46
4.2. Local thermal sensation analysis ............................................................................. 48
4.2.1. Local thermal sensation and environmental temperature ................................ 48
4.2.2. Local thermal sensation and local skin temperature ........................................ 50
4.3. Demographic factor analysis ................................................................................... 51
4.3.1. BMI and overall thermal sensation .................................................................. 52
4.3.2. Gender and overall thermal sensation ............................................................. 53
4.4. Conclusions ............................................................................................................. 54
5. Correlation analysis and thermal sensation estimation models ....................................... 56
5.1. Overall thermal sensation and local thermal sensation............................................ 56
5.1.1. Correlation analysis of the overall thermal sensation and local thermal
sensation…. ..................................................................................................................... 56
5.1.2. Correlation analysis of the overall thermal sensation and local thermal sensation
by gender ………………………………………………………………………………..58
5.1.3. Correlation analysis of the overall thermal sensation and local thermal sensation
by BMI… ........................................................................................................................ 59
5.1.4. Estimation of the overall thermal sensation by the local thermal sensation .... 60
5.1.5. Summary.......................................................................................................... 65
5.2. Overall thermal sensation and local skin temperature ............................................. 65
5.2.1. Correlation analysis of the overall thermal sensation and local skin
temperature…… .............................................................................................................. 65
5.2.2. Correlation analysis of the overall thermal sensation and local skin temperature by
gender….. ........................................................................................................................ 67
5.2.3. Correlation analysis of overall thermal sensation and local skin temperature by
BMI……. ........................................................................................................................ 68
5.2.4. Estimation of the overall thermal sensation by skin temperature .................... 69
5.2.5. Estimation accuracy with consideration for temperature change .................... 72
5.2.6. Summary.......................................................................................................... 72
5.3. Accuracy estimation based on other algorithms ...................................................... 73
6. Conclusions ..................................................................................................................... 77
6.1. Local thermal sensation and overall thermal sensation ........................................... 77

8

6.2. Local skin temperature and overall thermal sensation ............................................ 78
6.3. Potential use of the research findings ...................................................................... 78
7. Future work ..................................................................................................................... 80
References ............................................................................................................................... 81
Appendix A: CITI Program Certification ................................................................................ 86
Appendix B: Stepwise analysis of the local thermal sensations and overall thermal sensation by
gender ...................................................................................................................................... 87
B-1 Stepwise analysis of male subjects ............................................................................... 87
B-2 Stepwise analysis of female subjects ............................................................................ 88
Appendix C: Stepwise analysis of local thermal sensations and overall thermal sensation by
BMI.. ....................................................................................................................................... 90
C-1 Stepwise analysis by low BMI ..................................................................................... 90
C-2 Stepwise analysis by high BMI .................................................................................... 91
Appendix D: Stepwise analysis of the local skin temperature and overall thermal sensation by
gender ...................................................................................................................................... 92
D-1 Stepwise analysis of male subjects ............................................................................... 92
D-2 Stepwise analysis of female subjects............................................................................ 93
Appendix E: Stepwise analysis of the local skin temperature and overall thermal sensation by
BMI. ........................................................................................................................................ 94
E-1 Stepwise analysis by low BMI ..................................................................................... 94
E-2 Stepwise analysis by high BMI .................................................................................... 95
Appendix F: Prediction model (decision tree) of using forehead sensation and baseline attributes
to predict the overall thermal sensation ................................................................................... 96
Appendix G: Prediction model (decision tree) of using arm, back sensation and baseline attributes
to predict the overall thermal sensation ................................................................................... 99

9

1. Introduction
In recent years, the building industry has been facing two challenges: the growing energy
consumption and the increasing need for indoor environment comfort (Michal Veselý 2015).
Scientists from different countries have worked diligently to increase building energy efficiency
and improve the indoor environment quality (IEQ). People spend most of their time indoors, it
has been demonstrated that IEQ can significantly influence human mood and working
productivity (Fisk 2000). There is a significant energy consumption in an indoor environment,
and the buildings are designed to make more than 80% of the residents feel comfortable with the
indoor thermal condition (ASHRAE Std 55 2010). But research from Huizenga shows that in
only 11% of the 215 surveyed buildings in different countries, 80% or more of the occupants are
satisfied with their thermal comfort in building (C.Huizenga 2006).
Considering the large amount of energy used in an air-conditioned building and people’s extreme
dissatisfaction with the environmental quality in the interiors, a method that uses local body areas
to predict the overall thermal sensation may give the designer practical and accurate way to
evaluate the thermal environment of the building.
1.1. Energy consumption in the building industry
Many of today’s building designers take IEQ issues into consideration. At the same time,
buildings use 20% to 40% of the total energy use in developed countries (Luis Pe´rez-Lombard
2007). From the U.S. Energy Information Administration (EIA), American residential and
commercial buildings consumed approximately 40% of the total country’s energy use (Figure
1.1). Heating, ventilation and air conditioning [HVAC] occupies a large portion of the total
world’s energy consumption. It occupies approximately 50% of the building energy consumption
and nearly 20% of the total energy use in USA (Luis Pe´rez-Lombard 2007).

10

Figure 1.1: Primary energy use in American commercial and residential buildings in 2010 (U.S.
Energy Information Administratioin 2010)
Traditional air-conditioning methods such as “Forced Air” ventilation method fully ventilate the
room without considering the resting and active area of the occupant, and it is likely that
traditional air-conditioning methods waste a lot of energy because of the redundant and extra air
treatment. Research has also found that traditional HV AC systems use a large portion of energy
due to its uniform control of the indoor with a temperature range of 2 ℃ and there could be a
10% energy reduction when extending 1 ℃ of the HV AC set point during cool or warm climate
(Hoyt 2014). EIA also conducted a survey of commercial building energy usage in 2012 （Figure
1.2 ）Space heating accounts for the largest portion of total building energy use, and energy use in
HVAC accounts more than 40% (U.S. Energy Information Administratioin 2016).

Figure 1.2:Result of the 2012 survey on the energy consumption of commercial buildings (U.S.
Energy Information Administratioin 2016)
1.2. Indoor environment quality
With the steady improvement of living standards, people continue to become more concerned
about their health and comfort indoors. From the Environmental Protection Agency (EPA), the
average of American people only spends 7% of their daily life outdoors. Indoor Environmental
Quality (IEQ) is the condition of the building which can influence the work efficiency,
productivity or mood of the residents (Centers for Disease Control and Prevention 2013). As
shown in Figure 1.3, IEQ can be determined by many factors, including indoor environment
conditions such as temperature, lighting, humidity, glare, and air quality(ASHRAE 2016). An
acceptable IEQ may consider ventilation, humidity and light at the same time.

11

Figure 1.3: Influential factors of IEQ
1.3. Basics of thermal comfort
Thermal comfort is a subjective feeling concerning the thermal environment: people may feel
satisfied or dissatisfied with a specific environment (ASHRAE Std 55 2010). Thermal comfort is
one of the significant factors that highly correlate with the satisfaction of IEQ.

Figure 1.4: Influencing factors of indoor environment quality (A. Wagnera 2007)
Figure 1.4 shows the result of a survey conducted in 17 office buildings in Germany. The factors
which can influence the IEQ are weighted by their correlation coefficient with the employee’s
working efficiency and productivity in office building (A. Wagnera 2007). The upper-right level
shows parameters with low satisfaction and high importance. We can conclude that the room
temperature is one of the parameters that has high weighting factor, but majority of the indoor
people are not satisfied with.

12

1.3.1. Factors in human comfort
There are six significant influencing factors for thermal comfort:
 Metabolic rate (met)
Metabolic is a kind of physiological function which can help maintain the basic activities of
people. Metabolic rate is the intensity of the metabolic (ASHRAE Std 55 2010). The ASHRAE
Standard also provides some common values for different activities: 0.7 met for activity of
sleeping, 1.0 met for rest like sitting and 2.0 met for slight activity like walking (ASHRAE Std 55
2010).
 Clothing insulation (CLO)
Clothing insulation can have a significant impact on residents’ thermal comfort and overall
thermal sensation because it affects the heat gain and heat loss of the human body. (Wikipedia
2017).
 Air temperature
Air temperature, as referred in this thesis, is dry-bulb temperature (DBT). DBT is the temperature
of air without considering the effect of humidity or radiation (ASHRAE Std 55 2010). To
evaluate the temperature of an indoor place, the spatial average temperature considers the
temperature of the ankle, waist, and head levels.
In this thesis, DBT was investigated as the factor influencing local skin temperature and thermal
sensation.
 Mean radiant temperature [MRT]
MRT is the temperature considering the radiation of surrounding objects’ effect to human body
(International Organization for Standardization 1998). MRT can be an index of IAQ which can
express the influence of the surface temperature on the occupants in the building.
 Airspeed
In this study, the experiment room was equipped with a specified HVAC system to supply air. The
air speed for heating and cooling in the experiment room was controlled to less than 0.2 m/s,
based on the ASHRAE Standard (ASHRAE Std 55 2010).

13

 Relative humidity
Relative humidity (RH) is the ratio of the real-time humidity and the maximum humidity in a
specific temperature (Wikipedia 2017). From the previous study, the optimum comfortable
workplace for relative humidity is typically in the range between 40–60% (Shaharon 2012).
1.3.2. PMV and PPD model
Thermal comfort is strongly related to indoor satisfaction and work efficiency (W. Fisk 2000).
There are multiple attempts to predict indoor comfort. One of the most popular models is
Predicted Mean Vote Model (PMV model), which was proposed by P. O. Fanger (P. O. Fanger
1970). This basic model was later developed into the ASHRAE Standard 55-Thermal
Environmental Conditions for indoor thermal comfort design.
The PMV model is represented by a seven-point sensation scale index (Table 1.1) and it is based
on the internal heat gain and heat loss of the human body. Further, 0 (zero) is the ideal value,
which stands for neither warm or cool; +3 means very hot, -3 means cold, the ideal PMV value is
between –0.5 and +0.5 (ASHRAE Std 55 2013).
Table 1.1: Seven-point thermal sensation index
Sensation Cold Cool Slightly
cool
Neutral Slightly
warm
Warm Hot
Index -3 -2 -1 0 +1 +2 +3

Although this comfort model based on heat gain and heat loss is a practical way to predict
thermal sensation, Fanger proposed Predicted Percentage of Dissatisfied (PPD), which is based
on PMV (P. O. Fanger 1970).
1.4. Hypothesis statement
1.4.1. Hypothesis
Skin temperature and local thermal sensations in different body areas can be an index for the
sensation in the whole body; an air-conditioning method that uses some local body area can help
to predict the overall thermal sensation.

14

1.4.2. Goal
The goal of the thesis is to acquire an intense understanding of thermal sensation by performing
human subject tests and data analysis. The correlation between overall thermal sensation and
local thermal sensation, overall thermal sensation, and local skin temperature are analyzed.
Besides, this thesis also figures out the inherent factors (gender, Body Mass Index [BMI]) that
influence overall thermal sensation.
1.4.3. Objectives
 Identify the correlation between the thermal sensation in the whole body and local thermal
sensation
 Find the correlation between skin temperature and local thermal sensation
 Detect which body area relatively correlates with the overall thermal sensation
 Identify the inherent factors such as gender and BMI that can influence thermal sensation
1.5. Study statement
Based on human subject tests, correlation between the local skin temperature and overall thermal
sensation as well as correlation between the local thermal sensation and overall thermal sensation
will be studied. At the same time, accuracy estimation to predict the overall thermal sensation is
also evaluated. The potential use of some local skin locations to predict the overall thermal
sensation can be achieved based on the method of correlation analysis, step-wise analysis, and the
application of data-driven algorithm.
1.6. Important terminology
 Thermal sensation
Thermal sensation is kind of cognitive psychology which is related to the thermal condition, and
it is an index which can evaluate indoor environment quality. We cannot use a direct way to
measure the thermal sensation of residents, even different people with different gender, race or
BMI may have different thermal sensations in the same environment.
 BMI

15

BMI is a body index value which related to the height and weight of an individual and it can
evaluate the obese level (Wikipedia 2017).
 Personal Comfort System (PCS)
The PCS refers to systems which can be controlled for personal use individually, the working
condition of PCS can be different between different people, PCS usually operates without
supplying outside air. (E. A. Hui Zhang 2015).
 Data-driven approach
The data-driven approach is a kind of process that applies data analysis method to a specific data
set. The analysis can include correlation analysis and stepwise regression.
 Statistical classification
Classification problem is a typical problem in machine learning. In the classification problem, the
input data has specific characteristics or features. The target of machine learning method is to
classify the input based on these characteristics or features (Wikipedia 2016).
 Decision tree
Decision tree is a kind of algorithm tool which use tree-like model to express the correlation
between attributes and outcome. (Wikipedia 2013).
 Stepwise regression
Stepwise regression is a data analyzing method which can identify the significant influence
factors (Minitab 2016). In the process of stepwise regression analysis, important factors which
have high Pearson R value are usually in the former step, factors that are in low correlation with
the target are in the latter step.
 Cross validation
Cross validation is a practical method to acquire and examine the prediction accuracy, with the
application of the cross validation, over fitting problem can also be minimized. (Edson Duarte
2017).

16

 Artificial neuron network (ANN)
ANN is a data model which has similar structure with the neural network in biology (techopedia
2015) . It typically consists of multiple layer and several nodes in each of the layers and the
attributes are in the front of the ANN model.
1.7. Summary
This chapter introduces contemporary problems in the building industry including the energy
consumption and indoor environment quality. Some basics of thermal sensation are also
introduced. To have a better understanding of thermal sensation and air-conditioning strategies, a
background study of human thermal regulation and personal comfort system are documented in
chapter 2.

17

2. Background research
2.1. Physiological basics of thermal sensation
When there is hot or cold stimulation in some parts of the human body, the thermal regulation
response of the whole body will occur. For the overall thermal sensation, if there exists a cold
stimulation (such as from ice water, or cool air) to a part of the body, sweating phenomenon of the
whole body could be reduced. When a person is in the overall cold sensation, if there is hot
stimulation (such as warm air from a heater or hot water on a part of the body), the sensation of
cold might be alleviated.
Early in the 1880s, investigators described that there exist spots on the skin associated with
thermal perceptions under different environment condition (N. Vastani 2009). By building a
device that allowed him to pass warm and cold stimulations, Blix found warm and cold spots
related to the thermal sensation. By doing experiment, he also concluded that warm and cold
spots could not only be activated by warm and cold stimuli (Blix 1883). Also by performing
experiments, Goldscheider used different tools for stimulus of cold, heat and pain, he found that
the number of warm spots are less than the number of cold spots (N. Vastani 2009).
2.1.1. Skin structure
Skin is the largest organ of the human body. The total surface area of skin can be 1.8 m² for
adults. Skin has many functions: it can be a physical fence to the environment, which can allow
or limit the inward and outward passage of water (N Pan 2006). As Figure 2.1 shows there are
three structural layers to the skin, the layer of epidermis, the layer of dermis, and the layer of
hypodermis (The University of Waikato 2008).

18

Figure 2.1: Cross-sectional view of skin (the University of Waikato 2008)
The thickness of the epidermis is relatively smaller, as it is between 0.075~0.15 mm (N Pan
2006). From Figure 2.1 we can see that the outer layer of the epidermis is the stratum layer. The
subcutaneous layer is the deepest in the skin.
Thermoreceptors are specific neurons that are sensitive to the changes of the skin temperature.
The thermoreceptors can detect temperature changes. . When detecting the change of temperature,
thermoreceptors will also transmit information about the environmental temperature to the brain
(Wikipedia 2017).
The researcher Hensel carried forward a study on thermal physiology in man that includes the
investigation of overall thermal sensation and overall thermal comfort (H 1982). His study also
included the number of cold and warm sensation spots from previous study (
Table 2.1).
Table 2.1: Number of cold and warm spots per cm² in the human skin (H 1982)
Body parts Cold spots Warm spots
Forehead 5.8–8
Nose 8 1
Chest 9–10.2 0.3
Abdomen 8–12.5
Back 7.8
Upper arm 5–6.5
Fore arm 6–7.5 0.3–0.4
Back of hand 7.4 0.5
Palm of hand 1–5 0.4

Thermal receptors in the skin can send signals to the brain and the hypothalamus. When the brain
receives the signal, a sensation of cold or warmth will be produced and subjective human
behavior will appear. At the same time, when the hypothalamus receives the signal from brain, a
physiological reaction such as shiver or sweating will also appear (JA 1981).

19

2.2. The physiological balance and thermal regulation model
The thermal difference between humans and inanimate objects is that the temperature of the
human body and heat release from their bodies is not completely dependent on the environment
temperature. This is because humans have a thermoregulatory system. The function of the
thermoregulatory system is to maintain the core body temperature of the human within a
relatively narrow and stable temperature range (Wikipedia 2017).
The thermal regulation model has been investigated and developed for many years. Before the
1960s, the thermal regulation model did not consider the physiological mechanisms of the human
body, and the model can only be thought of as the temperature distribution model of the human
body. In 1963, Crosbie first introduced the human thermal regulation function to the study of the
thermal regulation model and this model is a one-node model (R.J. Crosbie 1961). In 1971,
Stolwijk first proposed the positive and negative feedback of the thermal regulation system (J.
Stolwijk 1966). In 1977, Werner first proposed to study the thermal regulation model by
mathematical system analysis (Werner 1980). In 1978, Wissler proposed a model with
consideration of thermal transfer of the blood vessels (E. Wissler 1985). These studies made the
regulation model more accurate. Below are some representative thermal regulation models.
 Stolwijk model
The total number of nodes in Stowijk’s model is 25, including 24 tissue nodes and one node of
blood. Stolwijk formulated energy balance equations for each node. The considered parameters
included blood convection, metabolic rate, evaporation off of the skin and radiation (J. Stolwijk
1966).
 Wissler model
In Wissler model, the human body is divided into 15 cylinders, this model also considers
physiological control functions (for example, control of the skin’s blood flow, thermal exchange
by blood). This model can reflect the thermal regulation of human body more accurate (E. H.
Wissler 1985).
 University of California, Berkeley multi-node model
To improve the model’s ability of predicting thermal comfort, the UC Berkeley multi-node
comfort model divides the body into sixteen segments. A thermal manikin which could measure
the heat flux in each body segments was also made to check whether the multi-node model can
predict the thermal comfort accurately. Besides, this multi-node model also considered the

20

different blood temperatures throughout the human body. This can make the model more precise
(Charlie Huizenga 2006, Charlie Huizenga 2006).

Figure 2.2: The 16 body segments of the Berkeley model (Charlie Huizenga 2006)
When compared to two-node thermoregulation models, multi-node thermoregulation models can
simulate the human body’s thermal condition more precisely. A new model that is called the
Universal Thermal Climate Index [UTCI]-Fiala was proposed in 2002. This model contains 12
compartments and 187 nodes. It also considers the behavioral factor of the clothing insulation
with outdoor environment including the effect of wind (Dusan Fiala 2011). Further, in 2002,
based on the Stolwijk model, a radiation model including human body thermal radiation,
radiation of the solar, and CFD model wasproposed by researcher Tanabe and he also discussed
the shape of the human body as one of a factor to predict the accurate boundary conditions of the
thermoregulation model. (Shin-ichi Tanabe 2002)
Thermoregulatory mechanisms can play a significant role in maintaining the thermal temperature
during light activities like sitting in a chair or lying on a bed, but some researchers studied
thermal regulation during exercise. By analyzing heat transfer between the body skin and the
indoor environment that occurs through the physical phenomenon such as conduction, convection
and radiation, the researcher Chin Leong Lim recommended that strenuous exercise should be
conducted during the cool hours of the day like early morning or late afternoon(Chin Leong Lim
2008). Researcher Daniël Wendt also studied the thermal regulation during exercise. He found
that although the human body has mechanisms such as shivering to generate internal heat or
sweating to prevent a rise of the body’s core temperature, environment factors like a hot or
humidcan also affect the human thermal comfort and thermoregulation (Daniël Wendt 2007).

21

2.3. Evaluation model of thermal sensation
The traditional thermal sensation model is based on the thermal equilibrium of heat input and
output. Researchers tried to combine four environmental factors (temperature, humidity, airspeed,
and mean radiant temperature) with thermal sensation. In 1970, P. O. Fanger proposed the
equilibrium of heat input and output of the human body (P. O. Fanger 1970). This basic model
later developed into the standard of designing indoor environment in the ASHRAE Standard and
International Organization for Standardization [ISO].
In 1989, Wyon proposed the concept of “Equivalent Homogeneous Temperature (EHT)” to
evaluate the non-uniform environment in a car based on the combination of thermal manikin and
five years of vehicle environment research. (Wyon 1989).
In 1992, Ingersoll proposed a method to predict local body PMV based on Fanger’s PMV model.
This method calculated the overall thermal sensation based on the local thermal sensation and
area ration of each location area of the human body (Ingersoll 1992). The PMV model can be
used to predict the overall thermal sensation in uniform condition, but it is not precise in
predicting the local thermal sensation by using PMV. Besides, calculating the overall thermal
sensation based on the area ration of local skin is also inaccurate.
Hagino and other researchers studied the effects of environmental factors like temperature,
airflow, and solar radiation on passenger comfort in vehicles with air conditioning equipment. A
method was devised to predict the thermal comfort level of passengers in the vehicle. This study
also found that local thermal sensation can be affected by the airflow and solar radiation (Hagino
1992).
In 1993, Matsunaga described a method to measure non-uniform thermal environments by using
the thermal manikin which the skin surface temperature can be controlled by the researcher. An
“equivalent temperature (t eq)” which is based on the built thermal manikin was proposed and
discussed based on the thermal performance of the thermal manikin. This model did not simulate
the sensation in a local part of the body separately. Instead, by combining the temperature of the
head, abdomen, and feet with weighting factors of 0.1, 0.7, and 0.2. An equivalent temperature of
the human body was calculated (Matsunaga 1993). Although this method can be used to estimate
overall thermal sensation, it is still not accurately in predicting the local thermal sensation.
Brown and Jones presented a new model considering the transient passenger thermal comfort.
This model used environmental variables such as air temperature, air velocity, and relative
humidity that can vary in different time or space. This model also read the clothing layers and
original physiological state of the body as input factors. The model first calculated the skin
temperature and humidity and then based on these data to establish the correlation between
thermal comfort and thermal sensation (Brown 1997). But this model can only predict the overall

22

thermal sensation and comfort in non-uniform environment without concerning itself with the
local thermal sensation and comfort.
In 1998, Fiala and other researchers collected mass data from the previous research, and a new
physiological model was developed by using regression analysis (KJ Lomas 2003). Fiala used the
human adjustment model to calculate physiological data such as skin temperature, change rate of
skin temperature, and core skin temperature. The research found that the overall thermal sensation
had a linear regression correlation with the mean skin temperature. This model is widely used for
predicting the overall thermal sensation but not including the prediction of local thermal
sensation.

Figure 2.3: Mean skin temperature difference and overall sensation (KJ Lomas 2003)
In 2003, with the development of computational fluid dynamics, the study conducted by Nilsson,
and he constructed a new thermal manikin which is numerical. Based on the data from the skin
temperature of the manikin as well as subjective results such as thermal sensation from several
hundred experiments. He found that human beings can react differently to the local heat transfer
in different parts of their bodies (Nilsson 2006).

23

Figure 2.4: MANIKIN3 built with PROSTAR (Nilsson 2006)
Nilsson model also displayed the comfort temperature range of each body location during
summer and winter (Figure 2.5 Figure 2.6). This model can be widely used in a car and building
environment.

Figure 2.5: Comfort zone for light summer clothing ensembles (Nilsson 2006)

24

Figure 2.6: Comfort zone for enhanced winter clothing ensembles (Nilsson 2006)

2.4. Study of thermal sensation in different thermal conditions
2.4.1. Study of thermal sensation in uniform conditions
People mostly spend time in uniform conditions. The study of thermal sensation in uniform
conditions can help mechanical designer to optimize the set point of the HV AC system. P. O.
Fanger and Olesen studied the skin temperature distribution of a man resting in a comfort
condition. They found although a human is in an environment with a uniform temperature, the
temperature distribution of the whole body is not uniform (B.W. Olesen 1973). The temperature
of the breathing zone is higher than other spots of the skin. The researcher Edward Arens and
others in the Center for the Built Environment (CBE) studied the thermal sensation and comfort
in different uniform environments including slightly warm, warm, cool, and cold environments.
The result from their study showed that the thermal sensation of the breathing zone tended to be
warmer than the sensations of other body parts like head, neck, and face in cool and cold
environments (Figure 2.7). Besides, the thermal sensation of feet was also studied (Edward Arens,
Partial- and whole-body thermal sensation and comfort 2006). Although the research team studied
the local thermal sensation of human body under different overall thermal sensation, the total
number of human subject tests in their study was relative small.

25

Figure 2.7: Local and overall thermal sensation and comfort in uniform/cold environment
(Edward Arens 2006)
Some researchers studied the correlation between local skin temperature/sensation and the overall
thermal sensation. Humphreys measured temperatures of 2000 fingertip and collected the overall
thermal sensation data in different thermal conditions. He proposed that a combination of hand
temperature and environment temperature can predict the overall thermal sensation well (M.A.
Humphreys 1999). Researcher Danni Wang studied finger temperature and the combination of air
and finger temperatures to predict the overall thermal sensation. He found that the skin
temperature range of fingers was narrow when compared to that of other local skin temperatures
in cool conditions. This indicated that the skin temperature of the finger is a significant sensitive
indicator of the body’s thermal sensation and comfort in cool conditions (Danni Wang 2007).

Researchers also use machine learning approaches to predict thermal demands based on skin
temperatures. Researcher Changzhi found that by using the support vector machine (SVM)
model, thermal demands can be well predicted based on the local skin temperatures (Changzhi
Dai 2017). Although the thermal demands can be well predicted in this model, other machine
learning methods such as decision tree or Artificial Neuro Network are not used to predict the
thermal demand or sensation.
2.4.2. Study of thermal sensation in non-uniform conditions
Although researchers have worked considerably to study the thermal sensation in a uniform
environment, local body sensations in non-uniform environments is hard to define (for example,
how to predict and quantify human sensation response to the non-uniform environment). Several
researchers have conducted numerous studies concerning thermal comfort and sensation in non-
uniform environments.

26

Huizenga’s study early related thermal sensation and comfort to local skin temperature and core
temperature based on the high correlation between thermal sensation and heat transfer of local
skin area (C. Huizenga 2004).
Researchers from CBE conducted more than two hundred tests in which local body parts were
separately cooled or warmed and then back to the uniform thermal condition. The study examined
how the thermal sensation and comfort perceived differently between local parts and overall body
(Edward Arens 2006). To provide the constant air to the local body area, the experiment used air
sleeves. However, this might make local thermal sensation inaccurate because the air sleeves
contact with the skin.
In 2015, Qihong Deng in China investigated overall and local thermal sensation and comfort in a
cool environment condition and studied the personalized heating effect of thermal sensation and
thermal comfort (Qihong Deng 2017). The result of this investigation (Figure 2.8) showed that
the overall thermal sensation and comfort votes were neutral and the human subjects felt
comfortable respectively in a low environment temperature of 16 °C with local personal heating
system, this indicated that local thermal strategy can help residents get thermal comfort.

Figure 2.8: Diagram result of the non-uniform environment research
Quan Jin proposed a local thermal sensation model which is based on the skin temperature and
the overall thermal sensation model by using the weighting factor of each local skin area. The
model he developed can reflect a human body’s thermal sensation in different non-uniform
thermal environments. He found that in a constant non-uniform thermal environment, the
weighting factor to predict the thermal sensation is relatively constant for each of the local body
parts. In the environment of cooling conditions, the local cooling of the trunk can affect other
local body part differently (Quan Jin 2012).

27

2.5. Thermal comfort control method
The current air-conditioning systems in buildings control the temperature or humidity based on
general and traditional assumption of the thermal comfort, and occupants usually adjust the set
points based on their subjective thermal sensation.
Over the past few decades, many researchers proposed control algorithms and compared the
energy use performance of the air-conditioning system. By comparing three novel control
algorithms, the researcher Siddharth Goyal found that one controlling algorithm that had
feedback control was more accurate for “occupancy-based zone–climate” control (Siddharth
Goyal 2013). Further, Yoshifumi proposed a kind of new HVAC controlling systems which can
receive requests data from occupants. The system can collect the preference of the occupants and
has a controlling logic based on the operating strategies such as energy-saving HVAC strategy
and occupants’ thermal satisfaction with their thermal condition (Yoshifumi Murakamia 2007).
The Personal Comfort System [PCS] is a kind of system that can control thermal and air quality
based on personal real-time sensation. One can traditionally use a desk fan or radiant heater
beside the feet as a PCS (E. A. Hui Zhang 2015).
Individual thermal sensation may be different due to their race, gender, and BMI. Even human
subjects in the same environment may be different. Instead of finding the optimal set point of
HVAC for everybody, a PCS system can have the possibility of letting most of the people feel
comfortable in the same building environment. Figure 2.9 shows that in a PCS system, the
satisfaction rate can be above 80%, which can content the ASHRAE Standard target.

Figure 2.9: Comfort rate in different PCS systems and No-PCS system (E. A. Hui Zhang 2015)

28

2.6. Conclusions for background research
Although studies have already explained the physiological basics of thermal sensation, existing
studies related to the local thermal sensation and the overall thermal sensation are still very
limited. There were studies on the difference in thermal sensation between male and female, but
other physiological factors such as BMI were not included in them. The correlation of the local
thermal sensation and overall thermal sensation, the local skin temperature and overall thermal
sensation are also analyzed by developing different thermal sensation model and identifying the
physiological and psychological respond in different thermal condition; however, there are few
studies that consider overall thermal sensation by combining two or three local spots, which can
potentially predict the overall thermal sensation with high accuracy.

29

3. Project methodologies
A method to investigate how local skin temperature and local thermal sensation correlates to
human thermal response is proposed. This chapter documents the overall methodology and
workflow used to pursue this goal. The overall condition of the climate chamber, experiment
procedure, and basic data analysis method are also documented in this chapter.
3.1. Workflow overview
The workflow of the proposed method can be divided into three steps: chamber arrangement, data
acquisition, and data analysis (Figure 3.1).

Figure 3.1: Overall workflow of proposed method

30

1. Step of chamber arrangement
These steps are prerequisites for the human subject test in step 2. A heater and cooler are installed
for getting a specific temperature of the climate chamber. Sensors (air temperature, humidity,
radiant temperature, CO₂) connected to the computer in the chamber help monitor the real-time
heating or cooling condition. A skin temperature sensor is used for acquiring physiological data
during the human subject test.
2. Step of data acquisition
Step 2 gathers the data source for the study. During the human-subject test, environmental data
(humidity, temperature) and physiological data (skin temperature) are recorded and transferred to
a computer by using the LabVIEW software (see Figure 3.11). Besides, human subjects are also
required to submit their answers to a questionnaire concerning the thermal sensation and thermal
comfort.
3. Step of data analysis
The data collected in step 2 are compiled and analyzed by using various methods (independence
test, correlation analysis, comparison analysis, and stepwise analysis) for comparing the overall
and local thermal sensations and the overall sensation and local skin temperatures. Data-driven
algorithms are applied to achieve the potential combination of body locations that will accuracy
predict the overall thermal sensation.
3.2. Experimental chamber
This study involved conducting an experiment with a human subject at the University of Southern
California’s (USC) experiment chamber (Figure 3.2).

31

Figure 3.2: Environmental chamber
In the basement of Watt Hall, the chamber next to the undergraduate studio is 4.0 m long, 2.8 m
wide, and 2.4 m high (Figure 3.3).

Figure 3.3: Floor plan of environmental chamber at USC (Unit: meter)
Although there exists an air-conditioning system in Watt Hall, the environmental chamber has a
separate system for the human subject test. The chamber is a concrete-enclosed room. In order to
reduce the thermal transfer from the room near the chamber, the wall surface of the chamber was
covered with foamed plastic. Besides, because the environmental chamber is in the basement, the
relative humidity can remain at a constant level.
3.3. Equipment for the experiment
To investigate thermal sensation, thermal comfort, and local skin temperature in different
environmental conditions, the environmental chamber was equipped with heaters, coolers,
sensors to detect the environment condition, and sensors to read the subject’s psychological data.
3.3.1. Heating and cooling system
The temperature of the environmental chamber was regulated by two air conditioners and four
portable heaters.
For the air-conditioning system, one window AC unit was connected to a wiring board (Figure
3.4). The compressor and fan were also controlled separately by the computer.

32

Figure 3.4: Window AC unit
Another air conditioning unit in the chamber is a portable air conditioner (Figure 3.5).

Figure 3.5: Portable AC
For the heating system, the heater (Figure 3.6) was plugged to a socket. As with the wire board,
the socket could also be controlled by the computer.

33

Figure 3.6: Portable heater
The conditioned air was supplied through flexible ducts and connected to four diffusers (
Figure 3.7) that were fixed symmetrically on both sides of the desk.

Figure 3.7: Diffuser
Specifications for the HV AC systems are summarized in Table 3.1.
Table 3.1: Specification for the HVAC system
Device Model Specification
Window AC unit Frigidaire FAA062P7A

Capacity: 6,000 BTU

34

Portable AC Honeywell HL10CESWW Capacity: 10,000 BTU
Portable AC Lasko 754200 Maximum 1500 Watt

Figure 3.8 shows the heater and AC connection in the chamber. In order to distribute the
conditioned air uniformly, four diffusers were placed on both sides of the desk.

Figure 3.8: Heater and AC connection

3.3.2. Data acquisition tool
To monitor the real-time environmental conditions of the chamber, four air temperature sensors,
one mean radiant temperature sensor, one humidity sensor, and one CO 2 sensor were used. All
these sensors were installed on a tripod (Figure 3.9), and the four air temperature sensors were
placed at the four different levels (0.1 m, 0.6 m, 1.1 m, and 1.6 m) above the ground floor. The
other sensors were arranged in the level of 1.1 m, which is similar with the working level.

35

Figure 3.9: Tripod with environmental sensors
Surface temperature sensors that have a rapid response rate were used to measure the skin
temperature during the test. The sensors (Figure 3.10) were in contact with the skin surface
directly for getting the precise data. The physiological data was transmitted to the computer by
the data acquisition system (DAQ). The DAQ board was inserted in two wrist bags so that they
could be carried conveniently.

Figure 3.10: Wrist bag of the sensor DAQ

36

Specifications for the data acquisition devices are summarized in Table 3.2.

37

Table 3.2: Specification for Data Acquisition Devices
Device Model Specification
Temperature sensor
LM35DT

Range: –55° C to 150 °C
Resolution: 0.01 °C
Accuracy: ± 0.5 °C
Skin temperature sensor STS-BTA
Range: –25° C to 125 °C
Resolution: 0.03 °C
Accuracy: ± 0.2 °C

CO₂ sensor Telarire6004
Range: 0 to 2000 ppm
Accuracy: ± 40 ppm
Radiant temperature sensor OS-542
Resolution: 0.1° C
Accuracy: ± 2° C

Humidity sensor HIH-4000-003
Range: 0 to 100%
Resolution:0.5%
Accuracy: ± 3.5%
Data acquisition board1 Sensor DAQ
Resolution: 13 bit
Sampling rate: 10 KS/s
Data acquisition board1 NI-DAQ 60008
Resolution: 12 bit
Sampling rate: 10 KS/s
3.3.3. Data acquisition software
One of the challenge for the project was to identify a program that could efficiently collect
environmental data in the chamber. The Laboratory Virtual Instrument Engineering Workbench
(LabVIEW) software was used in this project to achieve the purpose.
LabVIEW is an experiment platform which is based on visual programming (Wikipedia 2017). It
is an ideal tool for many measurements or control systems. One of the significant advantages is
that by using LabVIEW, scientists can acquire and analyze the experiment data conveniently.
Figure 3.11 and Figure 3.12 present the functions and setting buttons and logic of LabVIEW. The
interface was designed and developed based on the previous research provided by Professor Joon-
ho Choi.

38

Figure 3.11: Front panel of the LabVIEW (Source: Choi’s Human-Building Integration Lab)

Figure 3.12: Logic diagram of the designed program in LabVIEW (Source: Choi’s Human-
Building Integration Lab)
3.4. Human subject tests
The environmental data and physiological data used in the data analysis comes from the human
subject test. To reduce the unpredictable factors and make the test data accurate, the test
procedure is carefully designed, details of the procedure are shown in chapter 3.4.

39

3.4.1. Volunteer recruitment
A total of 23 volunteers (12 males, and 11 females) participated the experiment; they were all
college students from the University of Southern California. All subjects selected were in good
health and had no background information about the thesis topic. After the test, each volunteer
received a reward of 20 dollars.
Three results were excluded from the whole data set due to false data recording (one female and
two males). The skin temperature and environment record are abnormal (above 50 ºC) in the false
data recording.
The individual subjects’ data are shown in Table 3.3.
Table 3.3: Individual subjects’ data
ID Gender Age BMI ID Gender Age BMI
A1 male 20 21.5 A11 female 20 18.4
A2 male 21 19.7 A12 female 21 23.2
A3 male 23 24.2 A13 female 22 19.7
A4 male 23 23.5 A14 female 22 21.5
A5 male 24 20.7 A15 female 24 19.1
A6 male 25 22.3 A16 female 25 22.6
A7 male 22 21.1 A16 female 23 18.5
A8 male 24 19.1 A18 female 26 20.1
A9 male 26 24.3 A19 female 27 22.6
A10 male 30 20.4 A20 female 28 22.4

Demographic information about these 20 subjects is summarized in Table 3.4.
Table 3.4: Demographic information about subjects
Age BMI
Avg. 23.8 21.3
St. Dev 2.668 1.822

40

Range
20–25 26–30 Subtotal
Underweight
Normal
weight
Overweight
Subtotal
≤ 18.5 ≤ 24.9 ≥ 30
Number
Female 8 2 10 2 8 - -
Male 7 3 10 - 10 - -
Total 15 5 20 2 18 - -

To qualify for performing human subject tests, an authorization was required for this thesis. As a
student investigator, fundamental training on historical development of human subject protection,
ethical issues, and current regulatory and guidance information was taken. The student
investigator then passed the training session and was deemed eligible to conduct the experiment.
The course completion report and grade are shown in Appendix A.

3.4.2. Experiment procedure
The experiment carried on for around 100 minutes, which consists of 90 minutes in the
environmental chamber and 10 minutes’ rest outside the chamber (Figure 3.13).

Figure 3.13: Experiment procedure diagram
At the beginning of the experiment (10 minutes), the subjects stayed out of the chamber (25 ℃
constantly). During this step, personal data such as age and BMI were collected. Then, the test
schedule but not the goal of the test was introduced to them. Further, some basic concepts such as
thermal sensation and thermal comfort were also explained.

41

After ten minutes outside the chamber, the subjects were asked to stay not outside the chamber
but in the chamber. Skin temperature sensors were set up with the help of the experiment’s
assistants. In total, seven sensors were attached to different body locations with medical tape
(Figure 3.14).

Figure 3.14: Surveyed local body spot
Based on thermoregulation models of thermal sensation and physiological response (J. K. Choi
1997), the seven local spots were forehead, neck (at the back), back (upper back), chest, arm
(upper arm), wrist, and belly.
After all the sensors were installed, the subjects were asked to rest and calm down in the rest time
of step 2 (Figure 3.15).

Figure 3.15: Photograph of a subject wearing sensors in a calm condition

42

In step three, the temperature of the chamber was increased from 20 ℃ to 30 ℃ in about 70
minutes. The humidity of the chamber was controlled within 40–50%. Environmental and
physiological data of the subjects was collected and transferred to a computer with LabVIEW
every ten seconds. The subjects could engage in some simple activities such as reading a book or
listening to soothing music.
Thermal sensation and comfort data from different parts of the subjects’ body and from their
whole body were documented every five minutes.
3.4.3. Survey Questionnaire
Subjects’ local and whole body thermal sensations and comfort were surveyed by the
questionnaire. The survey of thermal sensation and thermal comfort in the questionnaire is shown
in Table 3.5.
Table 3.5: Thermal sensation and comfort survey
1. What is your overall level of thermal comfort?
Very unsatisfied Unsatisfied
Slightly
unsatisfied
Neutral
Slightly
satisfied
Satisfied Very satisfied
□ □ □ □ □ □ □
2. What is your overall thermal sensation?
Cold Cool
Slightly cool
Neutral
Slightly warm
Warm Hot
□ □ □ □ □ □ □

3.5. Statistical analyses
The data collected from the human subject tests was presented by graphics (such as boxplot,
interval plot) and analyzed by using various statistical methods (such as the 2-sample t-test,
correlation analysis, stepwise regression) in Minitab and a data-driven model (such as J48) with
the use of a data mining software called Weka.
3.5.1. Data analysis and mining tools
 Minitab

43

Minitab is a software for data analysis (Wikipedia 2017). Functions such as exploring data with
graphs, conducting statistical analyses, assessing quality can be applied in the Minitab software.
This thesis use Minitab software to deal with the environment data and physiological data.
 Weka
Weka is a software for the application of learning algorithm which is based on java (The
University of Waikato n.d.). In this thesis, algorithms like decision tree, Artificial Neural
Network, Sequential Minimal Optimization are applied in Weka software.
3.5.2. Exploring data with graphics
 Boxplot
Boxplots can be useful while comparing different datasets, it can reveal the feature of the dataset
like maximum value, minimum value (Minitab 2016).
 Interval plot
This thesis uses interval plot to compare the sensation vote or skin temperature distribution of
different body locations under neutral condition.
3.5.3. Statistical methods
 T-test
T-test is used to test the difference between two samples. The output of T-test can be used to value
the varying degrees of the two samples (Wikipedia 2017).
 Correlation analysis
Correlation refers to the relationship between two factors or variables. A weak relation indicates
that there is no significant relationship between the two factors and that they are hardly related.
Correlation analysis refers to the study process of the correlation.
3.5.4. Data-driven method
 Decision tree (J48)
A decision tree is a tree-like decision method which is usually used for classification problems. It
usually has several independent variables and outcomes (Korting 2006). This thesis use Decision
tree method to predict the overall thermal sensation.

44

J48 is a Java program which can applicate the decision tree algorithm (Wikipedia 2016).
3.6. Summary
This chapter states detailed information about the workflow of the research, chamber
arrangement, data acquisition tool, software development, and the procedure of the human subject
tests. As a result, it will be easier to comprehend and analyze the results in the next chapters upon
being familiar with the methodology.

45

4. Human subject experiment data and results
4.1. Overall thermal sensation analysis
4.1.1. Overall thermal sensation and environmental temperature
It is obvious that as the environment’s temperature goes up, the overall thermal sensation level
also goes up from –3 (very cool) to +3 (very warm) (Figure 4.1). The Pearson R value is 0.882
with a significant p-value (p < 0.05); this indicates that the overall thermal sensation is in high
correlation with the environment’s temperature.

Figure 4.1: Box plot of indoor temperature by the overall thermal sensation
4.1.2. Overall thermal sensation and local thermal sensation
It has been proved that the thermal sensation for local body parts varies greatly under different
environmental temperatures or different overall thermal sensation (Edward Arens 2006).
Additionally, thermal sensation of the entire body is related to the local thermal sensation (E. A.
Hui Zhang 2010).
When the overall thermal sensation is neutral, the sensation level of different body parts is around
zero (Figure 4.2).
The local thermal sensation of the wrist is close to neutral (p < 0.05). The sensations of forehead,
back, chest, and belly are warmer than neutral (p < 0.05), while the sensation of the arm (p <
0.05) is cooler than other locations on the body. This phenomenon indicates that in the condition

46

of neutral overall thermal sensation, the sensations in the breathing zone and head region are
more likely warm; the sensation in the trunk is more likely cool.

Figure 4.2: Interval plot of sensation votes under neutral condition (Sample size: 20)
4.1.3. Overall thermal sensation and local skin temperature
When the local thermal sensation is neutral, the skin temperature of different locations on the
body varies (Figure 4.3). The temperature of the head region and the breathing zone is higher than
that of the trunk, which is similar to the local thermal sensation vote under the neutral condition.

47

Figure 4.3: Interval plot of local skin temperature distribution under a neutral condition (Sample
size: 20)

48

4.2. Local thermal sensation analysis
4.2.1. Local thermal sensation and environmental temperature

49

Figure 4.4: Interval plots of the indoor temperature by the local thermal sensation
Similar to the correlation between the environment and the overall thermal sensation, the local
thermal sensation went up as the indoor temperature rose. Although the environment temperature
for the wrist at the sensation level of –3 (cold) was a little bit higher than the sensation level of –2
(cool), the Pearson R of wrist sensation and environment temperature was 0.747 with a significant
p-value (p < 0.05), while the Pearson R of arm sensation and environment temperature was 0.806
with a significant p-value (p < 0.05) (Table 4.1). It is safe to say that the local thermal sensation
of the wrist and arm correlated positively with the environment’s temperature.
Table 4.1: Pearson correlation between the environment’s temperature and the local thermal
sensation
Forehead Neck Back Chest Belly Wrist Arm
Pearson R 0.758 0.768 0.742 0.763 0.688 0.747 0.806
p-value 0.000* 0.000* 0.000* 0.000* 0.000* 0.000* 0.000*

50

4.2.2. Local thermal sensation and local skin temperature

51

Figure 4.5: Interval plots of the local skin temperature by the local thermal sensation
The temperature of the trunk was lower than that of the head region and breathing zone when
they were in the same local thermal sensation. The neutral skin temperature of the forehead was
around 35 ℃, which was higher than that of the other body locations; the neutral skin temperature
of the back, chest, and belly was around 34 ℃. The trunk area, like the arm and wrist, had the
lowest neutral skin temperature, which was around 32 ℃.
Table 4.2: Pearson correlation between the local skin temperature and the local thermal
sensation
Forehead Neck Back Chest Belly Wrist Arm
Pearson R 0.492 0.620 0.595 0.661 0.565 0.744 0.647
p-value 0.000* 0.000* 0.000* 0.000* 0.000* 0.000* 0.000*
Table 4.2 shows the Pearson correlation between the local skin temperature and the local thermal
sensation. The wrist has the most correlation value (Pearson R = 0.744), while the forehead has
the least correlation value (Pearson R = 0.492) when compared with other local areas on the body.
While comparing the local skin temperature and the environment temperature, the environment
temperature is found to be more correlated with the local thermal sensation.
4.3. Demographic factor analysis
Based on Fanger’s PMV model, individual thermal sensation is affected by environmental factors
such as air temperature, radiant temperature, air speed, relative humidity and individual factors
such as clothing insulation and metabolism rate (P. O. Fanger 1970). However, other
physiological factors (such as gender and BMI) are not considered.

52

4.3.1. BMI and overall thermal sensation
BMI, which is highly correlated with the ordinary metabolic rate and the fat content of the human
body, can be a potential factor that influences thermal sensation. Countries that have different
races or economic levels may have different BMI standards. According to the standard set by the
World Health Organization (WHO), “the BMI can be divided into four categories, which are
underweight (≤ 18.5 kg/m
2
), normal weight (18.5 kg/m
2
~24.9 kg/m
2
), overweight (25 kg/m
2
~29.9
kg/m
2
), and obese (≥ 30 kg/m
2
).” (World Health Organization 2006).
In total, 20 subjects who were no underweight and obese subjects participated the experiment.
The subjects were divided into two groups: six subjects were in the low BMI group (< 20 kg/m
2
)
and 14 subjects were in the high BMI group (≥ 20 kg/m
2
).

Figure 4.6: Interval plot of the indoor temperature by the overall thermal sensation and BMI
(Sample size: 20)
The indoor temperatures of both high- and low-BMI groups increased consistently when the
overall thermal sensations change from –3 (cold) to 3 (hot). In the sensation level of –1 (slightly
cool) and +1 (slightly warm), the environment’s temperature experienced by the high BMI and
low BMI groups were similar. The biggest temperature difference occurred when the overall
thermal sensation level was neutral (0) (show in Table 4.3).

53

Table 4.3: Two-sample t-test of the indoor temperature by the overall thermal sensation and BMI
Overall thermal sensation level
–3 –2 –1 0 1 2 3
Mean
Low 21.13 ℃ 21.38 ℃ 22.38 ℃ 23.90 ℃ 26.23 ℃ 28.37 ℃ 29.34 ℃
High 20.07 ℃ 21.06 ℃ 22.42 ℃ 25.19 ℃ 26.36 ℃ 28.30 ℃ 29.84 ℃
∆ (L. - H.) 1.06 ℃ 0.32 ℃ 0.04 ℃ 1.29 ℃ 0.13 ℃ 0.07 ℃ 0.5 ℃
p-value 0.000* 0.000* 0.000* 0.000* 0.000* 0.000* 0.000*
∆ (L. - H.) is the absolute value of (Low-High) at the same overall thermal sensation level.
4.3.2. Gender and overall thermal sensation
Gender is one of the significant factors that can influence thermal sensation in similar thermal
environments (JoonHo Choi 2010). In recent years, researchers also investigated the gender
differences in thermal comfort (W. Pasut 2015). The researcher Karjalainen (Karjalainen 2012)
found that females are more sensitive than males when the environment condition is cool.
Other researchers also investigated sensation of male and female in different environment
conditions. It has been proved that there is only small overall thermal sensation difference
between the genders in neutral or slightly warm environment conditions, but females tend to feel
cooler when the environmental temperature is relative cool (KC 2002).

54

Figure 4.7: Interval plot of the indoor temperature by the overall thermal sensation and gender
(Sample size: 20)
In total, 20 human subjects (10 male and 10 female subjects) were tested in the experimental
chamber. As shown in Figure 4.7, the thermal sensation level of both male and female subjects
increased continuously. The thermal sensation of male subjects increased from –2 (cool) to +3
(hot), while that of female subjects increased from –3 (very cool) to +3 (hot). There was no male
subject who felt cold during the test; this indicated that in very cool conditions, females tend to
feel cooler than males. When male and female subjects had the same thermal sensation level
(except +3), the mean environment temperature for male subjects was lower than that of female
subjects (Table 4.4). When the overall thermal sensation was hot (+3), the environment’s
temperature for male subjects were higher than that of female subjects. This indicted that in very
warm conditions, males tend to feel warmer than females.
Besides, the difference of the environmental temperature between the male and female subjects in
cool thermal sensation is larger than the warm thermal sensation (Table 4.4). When the overall
thermal sensation was slightly cool, the temperature difference between male and female subjects
was 1.81℃, which is the largest difference in all thermal sensation levels. When both male and
female subjects were experiencing neutral thermal sensation, the environment’s temperature for
male subjects was a little bit higher than that of female subjects with a temperature difference of
0.68 ℃.
Table 4.4: Two-sample t-test of the indoor temperature by the overall thermal sensation and
gender
Overall thermal sensation level
–3 –2 –1 0 1 2 3
Mean.
Male - 20.22℃ 21.26℃ 24.51℃ 25.75℃ 28.06℃ 29.77℃
Female 20.66℃ 21.36℃ 23.07℃ 25.19℃ 27.29℃ 28.71℃ 29.20℃
∆ (M.-F.) - 1.14℃ 1.81℃ 0.68℃ 1.54℃ 0.65℃ 0.57℃
p-value - 0.000* 0.000* 0.000* 0.000* 0.000* 0.000*
∆ (M.-F.) is the absolute value of (Male-Female) at the same overall thermal sensation level.
4.4. Conclusions
Results from the human subject tests showed that the local thermal sensation is correlated with
the environment’s temperature and the local skin temperature. In general, the index of the local
thermal sensation goes up as the environment’s temperature rises; skin temperature varies

55

differently under a neutral overall thermal sensation. Specifically, the temperature of the head
region and breathing zone is higher than that of the trunk area. At the same time, the overall
thermal sensation is correlated with the environment’s temperature and local thermal sensation.
Similar with the local thermal sensation, the index of the overall thermal sensation goes up when
the environment temperature rises. Besides, with different genders and BMIs, there exists a
sensation difference even in the same environmental temperature. The correlation between the
overall thermal sensation and the local thermal sensation, the overall thermal sensation and the
local skin temperature while considering the physiological factors (gender, BMI) will be further
analyzed in the next chapter.

56

5. Correlation analysis and thermal sensation estimation models
5.1. Overall thermal sensation and local thermal sensation
5.1.1. Correlation analysis of the overall thermal sensation and local thermal sensation
We can describe the strength of the correlation by using the guide from Evans. As shown in Table
5.1, for absolute values of R, “The range 0–0.19 indicates that there is a very weak correlation
between two factors; 0.2–0.39 indicates a weak correlation; 0.4–0.59 signifies a moderate
correlation; 0.6–0.79 suggests a high relationship; and 0.8–1 indicates a very strong correlation”
(Evans 1996).
Table 5.1: Evaluation of Pearson R (Evans 1996)
Pearson R 0–0.19 0.2–0.39 0.4–0.59 0.6–0.79 0.8–1
Strength Very weak Weak Moderate Strong Very strong

Table 5.2 shows the Pearson correlation between the overall thermal sensation and local thermal
sensation. Each local thermal sensation had a very strong correlation (Pearson R > 0.8) with the
overall thermal sensation (p < 0.05). The thermal sensation of the arm was particularly strongly
correlated with the overall thermal sensation with the reference to the Pearson R value of 0.918;
the forehead sensation also had a strong correlation with the Pearson R value of 0.871. The
thermal sensation of the belly showed a relatively low correlation with the overall thermal
sensation (Pearson R = 0.82, p < 0.05).
Table 5.2: Pearson correlation between the overall thermal sensation and local thermal sensation
Forehead Neck Back Chest Belly Wrist Arm
Pearson R 0.871 0.869 0.842 0.865 0.820 0.85 0.918
p-value 0.000* 0.000* 0.000* 0.000* 0.000* 0.000* 0.000*

High correlation between the single body thermal sensation and overall thermal sensation
indicated that there exists a potential use of some local body area for predicting the overall
thermal sensation.

57

Table 5.3: Stepwise analysis of the local thermal sensation and the overall thermal sensation
Step 1 Step 2 Step 3 Step 4

Coef. P Coef. P Coef. P Coef. P
Arm .968 .000 .684 .000 .599 .000 .528 .000
Forehead .417 .000 .309 .000 .3061 .000
Chest .217 .000 .1697 .000
Wrist .1271 .000
R-sq 84.3% 86.8% 87.4% 87.7%
∆ R-sq - 2.5% 0.6% 0.3%
Step 5 Step 6
Coef. P Coef. P

Arm .455 .000 .445 .000
Forehead .235 .000 .216 .000
Chest .104 .000 .059 .000
Wrist .159 .000 .150 .000
Neck .183 .000 .179 .000
Back .093 .000
R-sq 88.0% 88.1%
∆ R-sq 0.3% 0.1%

Table 5.3 shows the stepwise analysis of the local thermal sensation and overall thermal
sensation. As shown in Table 5.2, thermal sensation of the forehead had a strong correlation with
overall thermal sensation. The first step of the stepwise analysis focused on the arm with an R-sq
value of 84.3%. Forehead sensation combined with that of the arm had the R-sq value of 86.8%;

58

the ∆ R-sq value was 2.5% in total. With the combination of the thermal sensation of seven local
bodies, the R-sq value could up to 88.1%. By combining the arm and the forehead, the R-sq value
could be 84.3%, which is close to 88.1%.
5.1.2. Correlation analysis of the overall thermal sensation and local thermal sensation by
gender
Table 5.4: Correlation analysis of the overall thermal sensation and local thermal sensation by
gender
Forehead Neck Back Chest Belly Wrist Arm
Male
Pearson R 0.834 0.879 0.806 0.857 0.817 0.846 0.902
p-value 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Female
Pearson R 0.871 0.838 0.840 0.842 0.781 0.822 0.914
p-value 0.000 0.000 0.000 0.000 0.000 0.000 0.000

As shown in Table 5.4, female subjects showed a higher correlation than male subjects with the
reference to these body locations: forehead, back, and arm. The local body thermal sensation of
the belly in the female group showed a relatively low Pearson R value, which means that the
sensation of the belly was not highly correlated with the overall thermal sensation. Although the
Pearson R value was different in the male and female group, the difference in the value was not
significant.
Table 5.5: R-sq summary of the stepwise analysis of the overall thermal sensation and local
thermal sensation by gender
Rank Overall Male Female
Local
spot and
R-sq
1 Arm 84.3%
Arm
81.30%
Arm
83.45%
2 Forehead 86.8%
Chest
85.2%
Forehead
86.12%
3 Chest 87.4%
Neck
86.02%
Wrist
87.45%
4 Wrist 87.7%
Forehead
86.45%
Back
88.06%
5 Neck 88.0%
Belly
86.77%
Belly
88.45%

59

6 Back 88.1%
Back
86.90%
Neck
89.21%
7 - -
Wrist
86.93
Chest
89.25%

As illustrated in Table 5.5, the arm was most correlated with the overall thermal sensation in both
the male and female group. The priority of the forehead in the male group was lower than that of
the female group. By combining several body thermal sensations, the maximum R-sq value of the
male subjects was 86.93%; on the other hand, the maximum R-sq value of female subjects was
89.25%.
For more details, stepwise analysis of the local thermal sensations and overall thermal sensation
by gender is shown in Appendix B.
5.1.3. Correlation analysis of the overall thermal sensation and local thermal sensation by BMI
Table 5.6: Correlation analysis of the overall thermal sensation and local thermal sensation by
BMI
Forehead Neck Back Chest Belly Wrist Arm
Low BMI
Pearson R 0.906 0.863 0.832 0.869 0.854 0.892 0.937
p-value 0.000 0.000 0.000 0.000 0.000 0.000 0.000
High BMI
Pearson R 0.853 0.874 0.848 0.866 0.816 0.832 0.909
p-value 0.000 0.000 0.000 0.000 0.000 0.000 0.000
As shown in Table 5.6, all body locations in the low BMI and high BMI group showed a very
strong correlation with the overall thermal sensation. The low BMI group showed a higher
correlation than the high BMI group with the reference to most body locations (forehead, chest,
belly, wrist, arm). The Pearson R difference was not significant while considering the local
thermal sensation of the neck, back, and chest. It is worth noting that there existed a relatively
high correlation difference of the forehead between the low BMI (Pearson R = 0.906) and the
high BMI (Pearson R=0.853) group.
Table 5.7: R-sq summary of the stepwise analysis of the overall thermal sensation and local
thermal sensation by BMI
Rank Overall Low BMI High BMI
1 Arm 84.3%
Arm
87.71%
Arm
82.71%

60

Local
spot and
R-sq
2 Forehead 86.8%
Wrist
89.81%
Back
86.58%
3 Chest 87.4%
Forehead
90.84%
Neck
88.19%
4 Wrist 87.7%
Back
91.88%
Chest
88.48%
5 Neck 88.0%
Neck
92.22%
Belly
88.67%
6 Back 88.1%
Chest
92.23%
Forehead
88.83%
7 - -
-
-
Wrist
88.94%

Table 5.7 summarizes the R-sq values of the stepwise analysis of the overall thermal sensation
and local thermal sensation by BMI. The arm was the strongest correlated spot in both the low
BMI group and high BMI group. The table could assert that the sensation in the arm had a high
relevance to the overall thermal sensation, regardless of physiological status (gender and BMI).
Further, it was also worth noting that the low BMI group had a higher R-sq value than the high
BMI group and overall group.
For more details, stepwise analysis of the local thermal sensations and overall thermal sensation
by BMI is shown in Appendix C.
5.1.4. Estimation of the overall thermal sensation by the local thermal sensation
Correlation and stepwise analysis proved that the local thermal sensations were highly correlated
with the overall thermal sensation in the environment temperature range of 20~30 ℃. Although
physiological status (gender and BMI) can influence the Pearson R value, the order of local
sensation steps, and the final R-sq value in stepwise analysis, the Pearson R and final R-sq value
were still high (Pearson R > 0.8; R-sq > 80%), which means that local thermal sensations were
highly correlated with the overall thermal sensation regardless of physiological status.
To estimate the overall thermal sensation, this thesis takes the environment’s temperature and the
subject’s physiological status (gender and BMI) as the baseline attributes and combines the local
thermal sensation in different spots as the changeable attributes.
The data-driven algorithm decision tree (J48) is used to estimate the accuracy (Correctly
Classified Instances) of the prediction result and uses the ten-fold cross validation method for the
test option. For the final process of analysis, the overall thermal sensation was set as a nominal
attribute.

61

For the test option, the ten-fold cross validation method was used. this method divided the initial
data into ten sets of size n/10, and then trains the other nine datasets and tests on one of the
datasets, by repeating this procedure for ten times, the final average accuracy of the prediction
can be much precise. (Wikipedia 2017).
 Estimation of the overall thermal sensation by the local thermal sensation in one spot
Table 5.8: Accuracy of using baseline attributes (environment’s temperature, BMI, gender) or
combining baseline attributes with a single local thermal sensation factor to estimate the overall
thermal sensation
Forehead Neck Back Chest Belly Wrist Arm
Accuracy (gender, BMI
environment temperature)
98.58% 98.02% 98.44% 97.19% 98.05% 97.73% 98.05%
Accuracy (gender, BMI)
72.93% 68.71% 73.65% 71.46% 69.28% 74.13% 75.77%

Table 5.8 shows the accuracy of using non-local or single local thermal sensation to estimate the
overall thermal sensation. When only the baseline attributes (environment’s temperature, BMI,
gender) were used to estimate the overall thermal sensation, the estimation accuracy was 94.5%.
With the consideration of the local thermal sensation, the accuracy could be above 97%. This
indicated that by adding an extra attribute of local thermal sensation, the accuracy of predicting
the overall thermal sensation can be improved. While only considering one local spot, the
forehead, the estimation accuracy was 98.58%, which is the most accurate figure when compared
with other local spots.
The accuracy is at a relatively low level while not considering the environment temperature in the
baseline attributes. The arm had maximum accuracy (75.77%), and the neck had minimum
accuracy (68.71%). Although the forehead had the highest accuracy when the environment’s
temperature, the BMI and gender in baseline attributes were considered, the accuracy was not the
highest without considering the environment’s temperature.

62

Figure 5.1: Prediction model of using forehead sensation and baseline attributes to predict the
overall thermal sensation
As an example, Figure 5.1 shows a part of the prediction model (decision tree) of using forehead
sensation and baseline attributes (gender, BMI) to predict the overall thermal sensation. The
whole decision tree is shown in Appendix F.
Based on the stepwise analysis, while considering the local thermal sensation in a specific order,
the correlation value (R-sq) can be increased. It is possible to predict the overall thermal sensation
more precisely by combining two or three local spots.
 Estimation of the overall thermal sensation by combining two spots of the local thermal
sensation
Table 5.9: Accuracy of combining two thermal sensation factors with gender, BMI, environmental
temperature to estimate the overall thermal sensation
Neck Back Chest Belly Wrist Arm
Forehead 98.93% 99.08% 99.16% 98.96% 98.89% 99.40%
Neck 99.03% 98.81% 98.41% 98.58% 99.08%
Back 98.88% 98.80% 98.99% 99.29%
Chest 98.43% 98.72% 99.27%

63

Belly 98.43% 99.15%
Wrist 98.62%

By adding one more local thermal sensation attribute, the overall estimation accuracy improved
from the minimum 97.19% to 98.43%. Maximum estimation (99.4%) was for the combination of
arm and forehead, which were the two most correlated (Pearson R) factors of the overall thermal
sensation and the first two factors in the stepwise analysis.
Table 5.10: Accuracy of combining two thermal sensation factors with gender and BMI to
estimate the overall thermal sensation
Neck Back Chest Belly Wrist Arm
Forehead 81.44% 83.13% 83.64% 80.94% 85.78% 86.96%
Neck 83.57% 80.54% 78.24% 82.84% 82.43%
Back 81.13% 81.13% 85.32% 87.23%
Chest 77.89% 84.98% 85.80%
Belly 79.43% 84.16%
Wrist 85.54%

Because of the high accuracy of back, the combination of arm and back had maximum accuracy
(87.23%). The combination of arm and forehead also had a relatively high accuracy (86.96%).

64

Figure 5.2 Prediction model of using arm, back sensation and baseline attributes to predict the
overall thermal sensation
Figure 5.2 shows a part of the prediction model, of using arm, back sensation and baseline
attributes (gender, BMI) to predict the overall thermal sensation. The tree size of the decision tree
is 115, and has 58 leaves. Details of the decision tree is shown in Appendix G.

 Estimation of the overall thermal sensation by combining three spots of the local thermal
sensation
The maximum estimation accuracy (99.58%) existed while combining local thermal sensations in
the forehead, chest, and arm with baseline attributes of the environment’s temperature, gender,
and BMI. This combination also constitutes the first three steps in the stepwise analysis. When
the environment’s temperature was not considered in the baseline attributes, the combination of
arm, chest, and forehead also had maximum estimation accuracy (90.19%).

65

Figure 5.3 Prediction model of using chest, arm, forehead and baseline attributes to predict the
overall thermal sensation
Figure 5.3 shows a part of the prediction model, of using chest, arm, forehead sensation and
baseline attributes (gender, BMI) to predict the overall thermal sensation. The tree size of the
decision tree is 117, and has 59 leaves. Details of the decision tree is shown in Appendix H.
5.1.5. Summary
The results from the correlation analysis of the local thermal sensation and overall thermal
sensation showed that all the local thermal sensations have a strong correlation level with the
overall thermal sensation. The stepwise analysis showed the exact combination order of local
thermal sensations (arm, forehead, and chest), which were most correlated with the overall
thermal sensation. Besides, two attributes (arm, forehead or back, arm) and three attributes (arm,
forehead, chest) that had the most estimation accuracy were also calculated by using the data-
driven model J48.
5.2. Overall thermal sensation and local skin temperature
5.2.1. Correlation analysis of the overall thermal sensation and local skin temperature
Table 5.11: Pearson correlation between the overall thermal sensation and local skin temperature
Forehead Neck Back Chest Belly Wrist Arm
Pearson R 0.570 0.714 0.727 0.750 0.719 0.786 0.721

66

p-value 0.000* 0.000* 0.000* 0.000* 0.000* 0.000* 0.000*

As shown in Table 5.11, every local skin temperature had a strong correlation with the overall
thermal sensation except the forehead (p < 0.05). The Pearson R of the forehead was only 0.570
(p < 0.05), which indicates a moderate correlation with the overall thermal sensation. Compared
with the local thermal sensation, the local skin temperature was less correlated with the overall
thermal sensation.
Table 5.12: Stepwise analysis of the local skin temperature and overall thermal sensation
Step 1 Step 2 Step 3 Step 4
Coef. P Coef. P Coef. P Coef. P
Belly 0.789 .000 0.334 .000 0.217 .000 0.528 .000
Wrist 0.916 .000 0.503 .000 0.3061 .000
Arm 0.683 .000 0.1697 .000
Back 0.1271 .000
R-sq 32.47% 56.83% 64.13% 70.42%
∆ R-sq - 24.36% 7.3% 6.29%
Step 5 Step 6 Step 7
Coef. P Coef. P Coef. P

Belly 0.107 .000 0.103 .000 0.103 .000
Wrist 0.387 .000 0.367 .000 0.363 .000
Arm 0.214 .000 0.158 .000 0.155 .000
Back 0.483 .000 0.462 .000 0.450 .000
Neck 0.329 .000 0.317 .000 0.313 .000
Forehead 0.113 .000 0.108 .000

67

Chest 0.028 .000
R-sq 74.77% 74.96% 74.97%
∆ R-sq 4.35% 0.19% 0.01%

The first step involved the forehead, which had the lowest Pearson R value with the overall local
thermal sensation. By combining the temperature of the back, the R-sq value could be raised up to
56.83%. The maximum R-sq value existed when all local skin temperatures (back, chest, belly,
wrist, arm, neck) were combined in order.
5.2.2. Correlation analysis of the overall thermal sensation and local skin temperature by gender
Table 5.13: Correlation analysis of the overall thermal sensation and local skin temperature by
gender
Forehead Neck Back Chest Belly Wrist Arm
Male
Pearson R 0.439 0.794 0.704 0.743 0.729 0.755 0.728
p-value 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Female
Pearson R 0.751 0.798 0.751 0.796 0.855 0.807 0.763
p-value 0.000 0.000 0.000 0.000 0.000 0.000 0.000
As shown in Table 5.13, by comparing the Pearson R, the female group had a larger Pearson R
than the male group in all body locations. It is worth noting that the Pearson R of the forehead
(0.439) in the male group was significantly low when compared to other local skin spots. This
indicated that predicting the overall thermal sensation based on forehead skin temperature in the
male group is unreliable.
Table 5.14: R-sq summary of the stepwise regression analysis of overall thermal sensation and
local skin temperature by gender

Ra
nk
Overall Male Female
Local
spot &
R-sq
1 Belly 32.47%
Back
49.52%
Forehead
56.35%
2 Wrist 56.83%
Belly
69.06%
Chest
69.30%
3 Arm 64.13%
Wrist
74.10%
Belly
78.99%

68

4 Back 70.42%
Neck
75.61%
Back
80.5%
5 Neck 74.77%
Chest
76.44%
Neck
81.28%
6 Forehead 74.96%
Arm
76.71%
Arm
81.56%
7 Chest 74.97%
Forehead
76.83%
-
-

Table 5.14 is the R-sq summary of stepwise analysis of overall thermal sensation and local skin
temperature by gender. Although back was not the highest correlated local spot with overall
thermal sensation in male group, forehead was not the highest correlated local spot with overall
thermal sensation in female group; they had the priority in stepwise analysis compared with
another skin spot. By combining six skin spot (forehead, chest, belly, back, neck and arm), R-sq
value in female group could go up to 81.56 which is higher than male group.
For more details, stepwise analysis of the local skin temperatures and overall thermal sensation
by gender is shown in Appendix D.
5.2.3. Correlation analysis of overall thermal sensation and local skin temperature by BMI
Table 5.15: Correlation analysis of the overall thermal sensation and local skin temperature by
BMI
Forehead Neck Back Chest Belly Wrist Arm
Low
BMI
Pearson R 0.431 0.600 0.626 0.555 0.756 0.839 0.590
P-value 0.000 0.000 0.000 0.000 0.000 0.000 0.000
High
BMI
Pearson R 0.723 0.756 0.778 0.834 0.702 0.792 0.785
P-value 0.000 0.000 0.000 0.000 0.000 0.000 0.000

As shown in Table 5.15, in the high BMI group, the local skin temperature was more correlated
with the overall thermal sensation in general (local spots includes forehead, neck, back, chest, and
arm). In the low BMI group, the Pearson R value of the forehead was lower than the Pearson R
value of 0.431 (p-value: 0.000). In local spots of forehead and chest, the Pearson R difference was
more than 0.2.
Table 5.16: R-sq summary of the stepwise analysis of the overall thermal sensation and local skin
temperature by BMI

69

Rank Overall Low BMI High BMI
Local
spot and
R-sq
1 Belly 32.47%
Belly
57.06%
Forehead
52.33%
2 Wrist 56.83%
Wrist
70.62%
Back
64.82%
3 Arm 64.13%
Arm
74.69%
Chest
74.18%
4 Back 70.42%
Back
75.03%
Belly
77.12%
5 Neck 74.77%
Neck
75.22%
Wrist
78.67%
6 Forehead 74.96%
Forehead
75.39%
Arm
79.14%
7 Chest 74.97%
Chest
75.45%
Neck
79.16%

As illustrated in Table 5.16, in the high BMI group, the forehead is focus of the first step in the
stepwise analysis due to its high correlation (Pearson R) with the overall thermal sensation.
For more details, stepwise analysis of the local skin temperatures and overall thermal sensation
by BMI is shown in Appendix E.
5.2.4. Estimation of the overall thermal sensation by skin temperature
From the result of correlation and stepwise analysis, the thesis can come to the conclusion that the
local skin temperature is highly correlated with the overall thermal sensation. The potential use of
some local skin spot to predict the overall thermal sensation may be possible.
The data-driven algorithm decision tree (J48) was used to estimate the accuracy (Correctly
Classified Instances). The ten-fold cross validation method was used for the test option.
 Estimation of the overall thermal sensation by the local skin temperature of one spot
Table 5.17: Accuracy of using baseline attributes (environment’s temperature, BMI, gender) or by
combining baseline attributes with the skin temperature of a single spot to estimate the overall
thermal sensation
Forehead Neck Back Chest Belly Wrist Arm
Accuracy (gender, BMI
environment’s temperature)
97.53% 96.79% 97.05% 96.98% 96.44% 97.64% 97.41%

70

Accuracy (gender, BMI)
86.49% 87.54% 88.27% 89.93% 88.05% 89.52% 91.58%

As can be seen in Table 5.17, the accuracy of predicting the overall thermal sensation is 94.5%
when any local skin temperature is not considered. By using the skin temperature of one spot, the
estimation accuracy improved. The wrist, which had a high correlation with the overall thermal
sensation (Pearson R 0.786), had high estimation accuracy when compared with the local skin
temperature of other spots.
The accuracy went down when the environment’s temperature in baseline attributes was not
considered. This indicated that the environment’s temperature is an important indicator to predict
the overall thermal sensation. By adding the factor of the environment’s temperature to the
baseline attributes, the accuracy of using the skin temperature of one spot can increase by more
than 5%.

Figure 5.4: Prediction model of using forehead temperature and baseline attributes to predict the
overall thermal sensation
As an example, Figure 5.4 shows part of the prediction model (decision tree) by using the
forehead temperature and baseline attributes (BMI, gender) to predict the overall thermal
sensation.

71

Based on the stepwise analysis, when the local skin temperature in a specific order is considered,
the correlation value (R-sq) can be increased. It is possible to predict the overall thermal sensation
more precisely by combining two or three local spots.
 Estimation of the overall thermal sensation by the local skin temperature of two spots
Table 5.18: Accuracy of combining the local skin temperature of two spots with gender, BMI, and
the environment’s temperature to estimate the overall thermal sensation
Neck Back Chest Belly Wrist Arm
Forehead 97.56% 97.89% 98.17% 97.99% 98.09% 97.83%
Neck 97.06% 97.33% 97.64% 97.77% 97.42%
Back 97.72% 97.60% 97.80% 97.68%
Chest 97.56% 98.05% 97.85%
Belly 97.63% 97.47%
Wrist 98.09%
As shown in Table 5.18, although the forehead had low correlation with the overall thermal
sensation (Pearson R = 0.570), by combining it with the chest, the accuracy of estimation was
the highest when compared with other combinations.
Table 5.19: Accuracy of combining the local skin temperature of two spots with gender and BMI
to estimate the overall thermal sensation
Neck Back Chest Belly Wrist Arm
Forehead 96.34% 95.68% 96.55% 95.89% 96.52% 96.73%
Neck 96.64% 96.36% 96.47% 96.46% 96.46%
Back 96.29% 96.62% 96.91% 96.56%
Chest 97.00% 97.18% 96.`80%
Belly 95.15% 96.95%
Wrist 97.22%

In Table 5.19, it is apparent that the combination of arm and wrist, which had a high Pearson R
value, also had the highest predicting accuracy (97.22%). The overall prediction accuracy when

72

two local skin temperatures, BMI, and gender are considered is higher than the accuracy
achieved while combining two local skin temperatures with the BMI and gender.
 Estimation of the overall thermal sensation by the local skin temperature of three spots
The maximum estimation accuracy (98.20%) occurred when the local skin spots of the forehead,
chest, and wrist were combined with baseline attributes (environment’s temperature, gender, and
BMI). The combination of the belly, wrist, and arm, which was used in the first three steps of the
stepwise analysis, had the estimation accuracy of 97.59%. While considering the baseline
attributes without the environment’s temperature, the combination of arm, chest, and wrist also
had maximum accuracy (97.98%).
5.2.5. Estimation accuracy with consideration for temperature change
The change of skin temperature was another factor that can influence the estimation accuracy. By
combining the baseline attributes (environment’s temperature, BMI, gender), skin temperature,
and temperature change in three skin spots (forehead, chest, and wrist), new estimation accuracy
was updated in Table 5.20.
Table 5.20: Estimation accuracy of the overall thermal sensation with consideration for
temperature change
0 seconds 30 seconds 40 seconds 50 seconds 60 seconds 120 seconds
Accuracy 98.20% 97.91% 97.61% 97.98% 97.88% 97.63%

The data in Table 5.20 shows that when considering the temperature change in skin, the
estimation of the overall thermal sensation decreased. The result also indicated that to estimate
the overall thermal sensation, real-time skin temperature without considering the temperature
change was enough to estimate accuracy.
5.2.6. Summary
The results from the correlation analysis of the local skin temperature and the overall thermal
sensation indicate that the skin temperature has high correlation with the overall thermal
sensation and that there exists a potential use of the skin temperature of some local spots to
predict the overall thermal sensation. When compared with the process of using thermal sensation
(maximum: 99.58%) to predict the overall thermal sensation, local skin temperature had relatively
low estimation accuracy (maximum 98.20%) when the environment’s temperature was
considered in the baseline attributes. Without considering the environment’s temperature in the

73

baseline attributes, maximum accuracy by using the skin temperature is higher than using the
local thermal sensation.
5.3. Accuracy estimation based on other algorithms
This thesis used the data-driven algorithm decision tree (J48) method to estimate the accuracy of
predicting the overall thermal sensation. Other algorithms such as multilayer perceptron (MLP)
and Sequential Minimal Optimization (SMO) are also used to evaluate the estimation accuracy.
To compare the estimation accuracy between decision tree (J48)) and different data-driven
algorithms, thermal sensation (arm, forehead, chest) and skin temperature (chest, forehead, wrist)
are used to predict the overall thermal sensation.
 Multilayer perception [MLP]
This thesis also uses MLP method to test the accuracy in the prediction of the overall thermal
sensation.
Figure 5.5 is the multilayer perceptron analysis of the overall thermal sensation based on
forehead, chest, and arm sensation. The lowest relative absolute error of the model was 25.23%
(Table 5.21), which had three nodes in the first layer and one node in the second layer. When
compared with the accuracy estimation result from the decision tree (J48), the absolute error was
higher than the error from J48.

Figure 5.5: Multilayer perceptron analysis of the overall thermal sensation based on forehead,
chest, and arm sensation

74

Table 5.21: Relative absolute error of different multilayer perceptron models based on the local
thermal sensation
First layer Second layer Third layer Relative absolute error
1 2 0 26.89%
1 3 0 29.49%
2 1 0 28.39%
2 2 0 27.11%
2 3 0 27.13%
2 4 0 27.02%
3 1 0 25.23%
3 2 0 25.82%
3 3 0 25.73%
1 2 3 27.51%
2 2 2 27.31%
3 2 1 28.32%

Figure 5.6 illustrates the multilayer perceptron analysis of the overall thermal sensation based on
the temperature of forehead, chest, and wrist. The lowest relative absolute error of the estimation
was 37.98% (Table 5.22). When compared with the accuracy (98.2%) from the decision tree
(J48), the accuracy of multilayer perceptron was relative low.

75

Figure 5.6: Multilayer perceptron analysis of the overall thermal sensation based on local
temperatures of forehead, chest, and wrist
Table 5.22: Relative absolute error of different multilayer perceptron models based on the local
skin temperature
First layer Second
layer
Third layer Relative absolute error
1 2 0 42.65%
1 3 0 44.08%
2 1 0 39.57%
2 2 0 39.47%
2 3 0 39.09%
2 4 0 38.43%
3 1 0 38.72%
3 2 0 38.91%
3 3 0 39.02%
1 2 3 42.44%
2 2 2 39.83%
3 2 1 37.98%
 Sequential Minimal Optimization [SMO]
SMO is a new developed algorithm which can also deal with classification problems (Wikipedia
2016). In this thesis, SMO algorithm is used to predict the overall thermal sensation based on
local thermal sensations or skin temperatures.
Figure 5.7 shows the result from the data-driven model SMO. The correctly classified instance of
the estimation accuracy was 79.79%; the accuracy was lower than the result (99.58%) from the
decision tree (J48)

76

Figure 5.7: Confusion matrix of predicting the overall thermal sensation based on the sensations
in the arm, forehead, and chest
Figure 5.8 shows the result of predicting the overall thermal sensation based on local skin
temperature (chest, forehead, wrist). The accuracy of the estimation was 59.02%, which was
lower than the estimation result (98.20%) from the decision tree (J48).

Figure 5.8: Confusion matrix of predicting the overall thermal sensation based on skin
temperatures of the chest, forehead, and wrist

77

6. Conclusions
The study aimed to enhance the researcher’s understanding about the potential use of local body
skin areas to estimate thermal sensations in a built environment. To study the relationship
between the overall and local thermal sensations, the overall thermal sensation and individual
local skin temperatures, human subject experiments were conducted in the environmental
chamber of the Human-Building Integration Lab at the University of Southern California, which
was equipped with multiple environmental control systems and sensory devices.
Further, 20 volunteers participated in the human subject experiment. Each experiment lasted
about 100 minutes on an average. During the experiment, the participant was required to answer
environmental satisfaction surveys in computer-based or paper-based questionnaires while
experiencing various levels of ambient thermal conditions. At the same time, the ambient
environmental conditions and local body skin temperatures were also monitored and documented
automatically during the experiment.
The relationship between the local thermal sensation/local skin temperature and the overall
thermal sensation was statistically analyzed with consideration for an individual’s physiological
factors (e.g., gender, BMI).
6.1. Local thermal sensation and overall thermal sensation
Based on the correlation analysis, local thermal sensations are highly correlated with the overall
thermal sensation (Pearson R > 0.8). By conducting stepwise regression analysis, the first step
(arm) can have a high R-sq value of 84.3%.
When the baseline attributes are gender, BMI, and the environment’s temperature, the local
thermal sensation in one spot that has the highest accuracy (Correctly Classified Instances) comes
from the consideration of baseline attributes and forehead (98.58%). If we consider two local
spots, a combination of the forehead and arm can predict the overall thermal sensation better than
predicting it for combinations of other local spots with an accuracy of 99.4%. When considering
three spots, the combination of arm, forehead, and wrist can have the highest accuracy (99.58%)
in predicting the overall thermal sensation.
When the baseline attributes are gender and BMI, the sensation of arm has the maximum
accuracy (75.77%). If the local thermal sensations in two spots are combined, the combination of
arm and back can have an accuracy of 87.23%. If three local spots are combined, the combination
of arm, forehead, and chest can have an accuracy of 90.19%.

78

6.2. Local skin temperature and overall thermal sensation
By calculating and comparing the correlation value (Pearson R) of the local skin temperature and
local thermal sensation, the Pearson R value of the local skin temperature of each spot is lower
than the local thermal sensation.
If the baseline attributes are the environment’s temperature, gender, and BMI, the temperature of
the wrist can predict overall thermal sensation better than any other singer thermal sensation with
an accuracy of 97.64%. While considering two sensation spots, the combination of forehead and
chest can have the maximum accuracy (98.17%). While considering local thermal sensations in
three spots, the sensation of chest, forehead, and wrist can have an accuracy of 98.20%.
If the baseline attributes are gender and BMI and only one local skin temperature is used to
predict the overall thermal sensation, the arm has the maximum accuracy (91.58%). If the local
skin temperature of two spots is combined, the combination of arm and wrist can have the highest
accuracy of 97.22%. If three local spots are considered, the temperature of arm, chest, and belly
has the maximum accuracy (97.98%).
6.3. Potential use of the research findings
As discussed and investigated in the research, based on the use of the data on local skin
temperatures and sensations, we could predict a user’s overall thermal sensation in real time.
Current technologies such as a thermal image processing technique and smart watch could help
integrate a building mechanical control system with the data-driven thermal sensation model
developed in this research. When the temperature image processing system or smart watch
identifies skin temperatures of the selected local body areas (Figure 6.1), the overall thermal
sensation can be estimated based on the human physiological features. Then, a control system can
be actuated for optimizing the thermal condition by using the HV AC system (Figure 6.2).

79

Figure 6.1: Potential use of skin spots

Figure 6.2: Conceptual strategy to adjust the HVAC system based on the local skin temperature
and physiological features

80

7. Future work
 Enlarge the sample size
The research was conducted based on 23 human subjects, but the data collected from three
subjects should be abandoned for data analysis due to the corrupted method of sensing data. All
the subjects were students and were recruited as volunteers from USC (aged between 22 and
28). Among the participants, 12 were Chinese, three belonged to other Asian ethnicities, and five
were Caucasian. Besides, the diversity of the BMI is also very simple, as only two subjects were
underweight and there was no overweight person among all the subjects. To increase the
precision and applicability, more human subjects of greater diversity need to be tested.
 Improvement of the chamber design
The temperature of the chamber was regulated by two air conditioners: one window AC that
could be controlled by a computer and one portable air conditioner. The portable air conditioner
could not be controlled by the computer. To make the temperature control of the chamber more
accurate, a new wiring board for additional control needs to be developed.
 Another index to evaluate thermal stress
We traditionally use the seven-point PMV model to evaluate thermal stress in different
conditions. A new evaluation method, such as the five-point model or three-point model, can be
used to gain a new perspective of the thermal stress. Besides, worker productivity or efficiency
(with references to qualities such as attention and memory) can also be used to evaluate thermal
stress.

81

References

A. Wagnera, E. Gossauera, C. Moosmanna, Th. Groppa, R. Leonhartb. "Thermal comfort and
workplace occupant satisfaction—Results of field studies in German low energy office
buildings." Energy and Buildings, 7 2007.
ASHRAE Std 55. Thermal environment conditions for human occupancy. 2015.
Thermal environment conditions for human occupancy. 2013.
ASHRAE Std 55. "Thermal environment conditions for human occupationcy." 2010.
ASHRAE. “The Standards For Ventilation And Indoor Air Quality. ” 2016.
B.W. Olesen, P.O. Fanger. "The skin temperature distribution for resting man in comfort." Arch
Sci Physiol, 1973.
Blix, Magus. "Experimentela bidrag till lösning af frågan om hudnervernas specifika energi."
Uppsala LäkFör. Förh, 1883: 427-440.
Brown, J. "A New Transient Passenger Thermal Comfort Model." SAE Technical Paper, 1997.
C. Huizenga, H. Zhang, E. Arens, D. Wang. "Skin and core temperature response to partial- and
whole-body heating and cooling." Thermal biology, 2004.
C.Huizenga, S.Abbaszadeh, L.Zagreus, E.Arens. Air Quality and Thermal Comfort in Office
Buildings: Result of a Large Indoor Environmental Quality Survey. 2006.
http://escholarship.org/uc/item/7897g2f8.
centers for disease control and prevention. INDOOR ENVIRONMENTAL QUALITY. 5 2013.
https://www.cdc.gov/niosh/topics/indoorenv/.
Centers for Disease Control and Prevention. INDOOR ENVIRONMENTAL QUALITY. 5 2013.
https://www.cdc.gov/niosh/topics/indoorenv/.
Changzhi Dai, Hui Zhang, Edward Arens, Zhiwei Lian. "Machine learning approaches to predict
thermal demands using skin temperatures: Steady-state conditions." Building and
Environment, 3 2017.
Charlie Huizenga, Zhang Hui, Thomas Duan, Edward Arens. "AN IMPROVED MULTINODE
MODEL OF HUMAN PHYSIOLOGY AND." 2006.
Chin Leong Lim, Chris Byrne, Jason KW Lee. "Human Thermoregulation and Measurement of
Body Temperature in Exercise and Clinical Setting." Ann Acad Med Singapore, 4 2008.
CMU. Cross Validation. 2 7, 1997. https://www.cs.cmu.edu/~schneide/tut5/node42.html.
Daniël Wendt, Luc J.C. van Loon, Wouter D. Marken Lichtenbelt. "Thermoregulation during
Exercise in the Heat." Sports Medicine, 8 2007.
Danni Wang, Hui Zhang, Edward Arens, Charlie Huizenga. "Observations of upper-extremity
skin temperature and corresponding overall-body thermal sensations and comfort."
Building and Environment, 12 2007.
Dusan Fiala, George Havenith, Peter Bröde, Bernhard Kampmann, Gerd Jendritzky. "UTCI-Fiala
multi-node model of human heat transfer and temperature regulation." 4 19, 2011.
Edson Duarte, Jacques Wainer. "Empirical comparison of cross-validation and internal metrics for
tuning SVM hyperparameters." Pattern Recognition Letters, 3 2017: 6-11.

82

Edward Arens, Hui Zhang, Charlie Huizenga. "Partial- and whole-body thermal sensation and
comfort." Journal of Thermal Biology, 1 2006: 53-59.
"Partial- and whole-body thermal sensation and comfort—Part II: Non-uniform environmental
conditions." Journal of Thermal Biology, 1 2006.
Energy Information Administratioin. "International Energy Outlook." 2014.
Evans, James D. Straightforward statistics for the behavioral sciences. Pacific Grove, 1996.
Fanger, P. O. Thermal comfort: Analysis and applications in environmental engineering.
McGraw-Hill, 1970.
Fanger, P.O. Thermal comfort. Analysis and applications in environmental engineering. McGraw-
Hill: 1972, 1972.
Fisk, WJ. "Health and productivity gains from better indoor environments and their relationship
with building energy efficiency." Annual Review of Energy and the Environment, 2000.
FiskJ.William. “REVIEW OF HEALTH AND PRODUCTIVITY GAINS FROM BETTER. ”
Proceedings of Healthy Buildings, 2000.
Gaganjot Kaur, Amit Chhabra. "Improved J48 Classification Algorithm for the Prediction of
Diabetes." International Journal of Computer Applications , 7 2014.
H, Hensel. Thermoreception and Temperature Regulation. 1982.
Hagino, M. "Development of a Method for Predicting Comfortable Airflow in the Passenger
Compartment." 9 1, 1992.
Hoyt, Tyler, Edward Arens, and Hui Zhang. "Extending Air Temperature Setpoints: Simulated
Energy Savings and Design Considerations for New and Retrofit Buildings." Building
and Environment, 9 10, 2014: 89-96.
Hui Zhang, Edward Arens, Yongchao Zhai. "A review of the corrective power of personal
comfort systems in non-neutral ambient environments." Building and Environment, 9
2015: 15-41.
Hui Zhang, Edward Arensa, Charlie Huizenga, Taeyoung Han. "Thermal sensation and comfort
models for non-uniform and transient environments, part III: Whole-body sensation and
comfort." Building and Environment, 2 2010: 399-410.
Ingersoll, J.G, T.G. Kalman. "Automobile passenger compartment thermal comfort model-Part 2:
Human Thermal Comfort Calculation." SAE Technical Paper Series, 1992.
International Organization for Standardization. "ISO 7726." 1998.
J. K. Choi, K. Miki, S. Sagawa, K. Shiraki. "Evaluation of mean skin temperature formulas by
infrared thermography." International Journal of Biometeorology, 11 1997.
J. Stolwijk, J.D. Hardy. "Temperature regulation in man-a theoretical study." Pflugers Arch.,
1966.
JA, Boulant. "Hypothalamic mechanisms in thermoregulation." 12 1981.
Jiangjiang Wang, Zhiqiang (John) Zhai, Youyin Jing, Chunfa Zhang. "Influence analysis of
building types and climate zones on energetic, economic." Applied Energy, 3 15, 2011.
JoonHo Choi, Azizan Aziz, Vivian Loftness. "Investigation on the impacts of different genders
and ages on satisfaction with." Building and Environment, 1 4, 2010.

83

Karjalainen, S. "Thermal comfort and gender: a literature review." Indoor air, 2012.
KC, Parsons. "The effects of gender, acclimation state, the opportunity to adjust clothing and
physical disability on requirements for thermal comfort. ." Energy and Buildings, 2002.
KJ Lomas, D Fiala, M Stohrer. "First principles modeling of thermal sensation responses in
steady-state and transient conditions." 2003.
Korting, TS. "C4. 5 algorithm and multivariate decision trees." 2006.
Luis Pe´rez-Lombard, Jose´ Ortiz, Christine Pout. "A review on buildings energy consumption
information." Energy and Buildings, 2007.
M.A. Humphreys, K.J. McCartney, J.F. Nicol, I.A. Raja. "An analysis of some observations of the
finger temperature and thermal comfort of office workers." Proceedings of Indoor Air,
1999.
Matsunaga, K., Sudo, F., Tanabe, S., and Madsen, T. "Evaluation and Measurement of Thermal
Comfort in the Vehicles with a New Thermal Manikin." SAE Technical Paper 931958,
1993.
Michal Veselý, WimZeiler. "Personal heating: effectiveness and energy use." Building Research
& Information, 1 30, 2015: 346-354.
Minitab. 2-sample t test. 2017. http://support.minitab.com/en-us/minitab/17/topic-library/basic-
statistics-and-graphs/hypothesis-tests/tests-of-means/why-use-2-sample-t/.
Basics of correlation and covariance. 2016. http://support.minitab.com/en-us/minitab/17/topic-
library/modeling-statistics/regression-and-correlation/correlation-and-covariance/basics-
of-correlation-and-covariance/.
"Basics of stepwise regression." 2016. http://support.minitab.com/en-us/minitab/17/topic-
library/modeling-statistics/regression-and-correlation/basics/basics-of-stepwise-
regression/.
Interpret all statistics and graphs for Descriptive Statistics. 2016. http://support.minitab.com/en-
us/minitab-express/1/help-and-how-to/basic-statistics/summary-statistics/descriptive-
statistics/interpret-the-results/all-statistics-and-graphs/#boxplot.
Interval plot. 2016. http://support.minitab.com/en-us/minitab/17/topic-library/basic-statistics-and-
graphs/graphs/graphs-that-compare-groups/interval-plots/interval-plot/.
N Pan, P Gibson. Thermal and moisture transport in fibrous materials. 2006.
Nilsson, Håkan O. "Thermal comfort evaluation with virtual manikin methods." Building and
Environment, 4 2006.
Qihong Deng, Runhuai Wang, Yuguo Li. "Human thermal sensation and comfort in a non-
uniform environment with personalized heating." Science of the total environment, 2017.
Quan JinLi, Lin Duanmu, Haiwen Shu, Yuming Sun, Qianru DingXiangli. “Predictive model of
local and overall thermal sensations for non-uniform environments. ” Building and
Environment, 2012.
R.J. Crosbie, J.D. Hardy, E. Fessenden. "Electrical analog simulation of temperature regulation in
man." Bio Med. Electron, 1961: 234-252.

84

Shaharon, M.N, Jalaludin, J. "Thermal Comfort Assessment-A Study Toward Workers’
Satisfaction in a Low Energy Office Building." American Journal of Applied Sciences,
2012.
Shin-ichi Tanabe, Yoshiichi Ozeki, Toru Takabayashi. "Numerical Comfort Simulator for Thermal
Environment (Part 1) A Comprehensive Combined Analysis Method with Multi-Node
Thermoregulation Model, Radiation Model and CFD for Evaluation of Thermal
Comfort." SAE International , 1 2002.
Siddharth GoyalA. Ingley, Prabir BarooahHerbert. “Occupancy-based zone-climate control for
energy-efficient buildings: Complexity vs. performance. ” Applied Energy, 2013.
Std, ASHRAE 62. n.d.
techopedia. Artificial Neural Network (ANN). n.d.
https://www.techopedia.com/definition/5967/artificial-neural-network-ann.
Artificial Neural Network (ANN). 2015. https://www.techopedia.com/definition/5967/artificial-
neural-network-ann.
the University of Waikato. 7 29, 2008.
http://legacy.sciencelearn.org.nz/layout/set/lightbox/Contexts/You-Me-and-UV/Sci-
Media/Images/Looking-at-skin.
The University of Waikato. Cross section of skin. July 2008.
https://www.sciencelearn.org.nz/images/1612-cross-section-of-skin.
Weka 3: Data Mining Software in Java. n.d. http://www.cs.waikato.ac.nz/ml/weka/.
Tina R. Patil, Mrs. S. S. Sherekar. "Performance Analysis of Naive Bayes and J48." International
Journal Of Computer Science And Applications , 2013.
U.S. Energy Information Administratioin. 2012 Commercial Buildings Energy Consumption
Survey: Energy Usage Summary. 3 18, 2016.
https://www.eia.gov/consumption/commercial/reports/2012/energyusage/.
Vastani, N. Mechanisms of thermal sensitivity in rodent primary afferent neurons innervating the
skin. 2009.
Vastani, Nisha. Mechanisms of thermal sensitivity in rodent primary afferent neurons innervating
the skin. University College London, 2009.
W. Pasut, H. Zhang, E. Arens, Y . Zhai. "Energy-efficient comfort with a heated/cooled chair:
results from human subject tests." Building and environment, 2015.
Werner. "A contribution to the topography of temperature regulation in man." European Journal
of Applied Physiology and Occupational Physiology, 10 1980.
Wikipedia. Artificial neural network. 2 27, 2017.
https://en.wikipedia.org/wiki/Artificial_neural_network.
Body mass index. 2 14, 2017. https://en.wikipedia.org/wiki/Body_mass_index.
C4.5 algorithm. 6 27, 2016. https://en.wikipedia.org/wiki/C4.5_algorithm.
Clothing insulation. 1 2017. https://en.wikipedia.org/wiki/Clothing_insulation.
Cross-validation (statistics). 3 2017. https://en.wikipedia.org/wiki/Cross-validation_(statistics).
"Decision tree." 10 2013. https://en.wikipedia.org/wiki/Decision_tree.

85

LabVIEW. 3 31, 2017. https://en.wikipedia.org/wiki/LabVIEW.
Minitab. 1 30, 2017. https://en.wikipedia.org/wiki/Minitab#cite_note-2.
Multilayer perceptron. 2 1, 2017. https://en.wikipedia.org/wiki/Multilayer_perceptron.
Relative humidity. 3 2017. https://en.wikipedia.org/wiki/Relative_humidity.
Sequential minimal optimization. 11 10, 2016.
https://en.wikipedia.org/wiki/Sequential_minimal_optimization.
Statistical classification. 11 18, 2016. https://en.wikipedia.org/wiki/Statistical_classification.
Statistical classification. 11 2016. https://en.wikipedia.org/wiki/Statistical_classification.
Student's t-test. 3 2017. https://en.wikipedia.org/wiki/Student%27s_t-test.
Thermoreceptor. 4 12, 2017. https://en.wikipedia.org/wiki/Thermoreceptor.
Thermoregulation. 4 10, 2017. https://en.wikipedia.org/wiki/Thermoregulation.
Wissler, E.H. "Mathematical simulation of human thermal behaviour using whole body models."
Heat Transfer in Medicine and Biology, 1985: 325-373.
Wissler, Eugene H. "A NEW HUMAN THERMAL MODEL." 1985.
World Health Organization. "Global Database on Body Mass Index." 2006.
Wyon, D. "Standard Procedures for Assessing Vehicle Climate with a Thermal Manikin." SAE
Technical Paper Series, 1989.
Yoshifumi MurakamiaTeranoa, Kana Mizutanib, Masayuki Haradab, Satoru KunobMasaaki.
“Field experiments on energy consumption and thermal comfort in the office
environment controlled by occupants ’ requirements from PC terminal. ” Building and
Environment, 2007.
Zhang, Hui, Arens, Edward, Zhai, Yongchao. "A review of the corrective power of personal
comfort systems in non-neutral ambient environments." 2015.

86

Appendix A: CITI Program Certification

87

Appendix B: Stepwise analysis of the local thermal sensations and overall
thermal sensation by gender
B-1 Stepwise analysis of male subjects

Step 1 Step 2 Step 3 Step 4

Coef. P Coef. P Coef. P Coef. P
Arm 1.16 .000 .742 .000 .557 .000 .316 .000
Chest .461 .000 .259 .000 .1451 .000
Neck .373 .000 .1367 .000
Forehead .5311 .000
R-sq 81.30% 85.20% 86.02% 86.45%
∆ R-sq - 3.9% 0.82% 0.43%
Step 5 Step 6 Step 7
Coef. P Coef. P Coef. P

Arm .226 .000 .476 .000 .510 .000
Chest .113 .000 .201 .000 .207 .000
Neck .162 .000 .229 .000 .223 .000
Forehead .447 .000 .217 .000 .220 .000
Belly .194 .000 .147 .000 .169 .000
Back –.103 .000 –.106 .000
Wrist –.052 .000
R-sq 86.77% 86.9% 86.93%

88

∆ R-sq 0.32% 0.13% 0.03%

B-2 Stepwise analysis of female subjects

Step 1 Step 2 Step 3 Step 4

Coef. P Coef. P Coef. P Coef. P
Arm 0.97 .000 .676 .000 .513 .000 .488 .000
Forehead .433 .000 .391 .000 .324 .000
Wrist .229 .000 .171 .000
Back .182 .000
R-sq 83.45% 86.12% 87.45% 88.06%
∆ R-sq - 2.67% 1.33% 0.61%
Step 5 Step 6 Step 7
Coef. P Coef. P Coef. P

Arm .484 .000 .395 .000 .379 .000
Forehead .402 .000 .290 .000 .294 .000
Wrist .181 .000 .237 .000 .232 .000
Back .269 .000 .277 .000 .266 .000
Belly –.186 .000 –.312 .000 –.365 .000
Neck .265 .000 .260 .000
Chest 0.90 .000
R-sq 88.45% 89.21% 89.25%

89

∆ R-sq 0.39% 0.76% 0.04%

90

Appendix C: Stepwise analysis of local thermal sensations and overall
thermal sensation by BMI
C-1 Stepwise analysis by low BMI

Step 1 Step 2 Step 3 Step 4

Coef. P Coef. P Coef. P Coef. P
Arm .989 .000 .708 .000 .520 .000 .660 .000
Wrist .287 .000 .235 .000 .288 .000
Forehead .339 .000 .411 .000
Back –.327 .000
R-sq 87.71% 89.81% 90.84% 91.88%
∆ R-sq - 2.1% 1.03% 1.04%
Step 5 Step 6
Coef. P Coef. P

Arm .705 .000 .699 .000
Wrist .296 .000 .289 .000
Forehead .540 .000 .540 .000
Back –.295 .000 –.312 .000
Neck –.230 .000 –.242 .000
Chest .049 .000
R-sq 92.22% 92.23%
∆ R-sq 0.34% 0.01%

91

C-2 Stepwise analysis by high BMI

Step 1 Step 2 Step 3 Step 4

Coef. P Coef. P Coef. P Coef. P
Forehead .969 .000 .686 .000 .512 .000 .481 .000
Back .403 .000 .314 .000 .245 .000
Chest .308 .000 .262 .000
Belly .157 .000
R-sq 82.71% 86.58% 88.19% 88.48%
∆ R-sq - 3.87% 1.61% 0.29%
Step 5 Step 6 Step 7
Coef. P Coef. P Coef. P

Forehead .486 .000 .467 .000 .406 .000
Back .272 .000 .252 .000 .243 .000
Chest .277 .000 .245 .000 .278 .000
Belly .243 .000 .224 .000 .207 .000
Wrist –.138 .000 –.153 .000 –.188 .000
Arm .124 .000 .127 .000
Neck .094 .000
R-sq 88.67% 88.83% 88.94%
∆ R-sq 0.19% 0.16% 0.11%

92

Appendix D: Stepwise analysis of the local skin temperature and overall
thermal sensation by gender
D-1 Stepwise analysis of male subjects
Step 1 Step 2 Step 3 Step 4

Coef. P Coef. P Coef. P Coef. P
Back 0.873 0.000 0.566 0.000 0.397 0.000 0.332 0.000
Belly 0.940 0.000 0.690 0.000 0.557 0.000
0.000
Wrist 0.267 0.000 0.175 0.000
Neck 0.000
R-sq 49.52% 69.06% 74.10% 75.61%
∆ R-sq - 19.54% 5.04% 1.51%
Step 5 Step 6 Step 7
Coef. P Coef. P Coef. P

Back 0.271 0.000 0.267 0.000 0.253 0.000
Belly 0.506 0.000 0.478 0.000 0.467 0.000
Wrist 0.143 0.000 0.122 0.000 0.118 0.000
Neck 0.248 0.000 0.229 0.000 0.231 0.000
Chest 0.205 0.000 0.173 0.000 0.163 0.000
Arm 0.115 0.000 0.123 0.000
Forehead 0.043 0.000
R-sq 76.44% 76.71% 76.83%
∆ R-sq 0.83% 0.27% 0.12%

93

D-2 Stepwise analysis of female subjects
Step 1 Step 2 Step 3 Step 4

Coef. P Coef. P Coef. P Coef. P
Forehead 1.245% 0.000 0.597% 0.000 0.448% 0.00
0
0.363% 0.00
0
Chest 0.937% 0.000 0.242% 0.00
0
–
0.083%
0.00
0
Belly 1.311% 0.00
0
1.368% 0.00
0
Back 0.444 0.00
0
R-sq 56.35% 69.3% 78.99% 80.5%
∆ R-sq - 12.95% 9.69% 1.51%
Step 5 Step 6
Coef. P Coef. P

Forehead 0.297% 0.000 0.213% 0.000
Chest –
0.132%
0.000 –
0.276%
0.000
Belly 1.217% 0.000 1.294% 0.000
0.000

Back 0.405% 0.000 0.324% 0.000
Neck 0.292% 0.000 0.279% 0.000
Arm 0.230% 0.000
R-sq 81.28% 81.56%
∆ R-sq 0.78% 0.28%

94

Appendix E: Stepwise analysis of the local skin temperature and overall
thermal sensation by BMI
E-1 Stepwise analysis by low BMI
Step 1 Step 2 Step 3 Step 4
Coef. P Coef. P Coef. P Coef. P
Belly 1.900 .000 0.278 .000 –
0.290
.000 –
0.230
.000
Wrist 0.763 .000 0.830 .000 0.788 .000
Arm 0.433 .000 0.324 .000
Back 0.150 .000
R-sq 57.06% 70.62% 74.69% 75.03%
∆ R-sq - 13.56% 4.07% 0.34%
Step 5 Step 6 Step 7
Coef. P Coef. P Coef. P

Belly –0.306 .000 –0.318 .000 –
0.261
.000
Wrist 0.786 .000 0.776 .000 0.766 .000
Arm 0.302 .000 0.301 .000 0.322 .000
Back 0.138 .000 0.128 .000 0.145 .000
Neck 0.121 .000 0.114 .000 0.115 .000
Forehead 0.046 .000 0.048 .000
Chest –
0.072
.000
R-sq 75.22% 75.39% 75.45%
∆ R-sq 0.19% 0.17% 0.06%

95

E-2 Stepwise analysis by high BMI

Step 1 Step 2 Step 3 Step 4

Coef. P Coef. P Coef. P Coef. P
Forehead 1.310 0.000 0.572 0.000 0.263 0.000 0.192 0.000
Back 0.822 0.000 0.383 0.000 0.427 0.000
Chest 0.859 0.000 0.600 0.000
Belly 0.506 0.000
R-sq 52.33% 64.82% 74.18% 77.12%
∆ R-sq - 12.49% 9.36% 2.94%
Step 5 Step 6 Step 7
Coef. P Coef. P Coef. P

Forehead 0.149 0.000 0.090 0.000 0.087 0.000
Back 0.388 0.000 0.407 0.000 0.417 0.000
Chest 0.436 0.000 0.326 0.000 0.326 0.000
Belly 0.388 0.000 0.352 0.000 0.370 0.000
Wrist 0.211 0.000 0.186 0.000 0.197 0.000
Arm 0.194 0.000 0.205 0.000
Neck –
0.046
0.000
R-sq 78.67% 79.14% 79.16%
∆ R-sq 1.55% 0.47% 0.02%

96

Appendix F: Prediction model (decision tree) of using forehead sensation and
baseline attributes to predict the overall thermal sensation

forehead sensation <= 1
| forehead sensation <= 0
| | forehead sensation <= -1
| | | BMI <= 22.3
| | | | BMI <= 19.1
| | | | | BMI <= 18.4: -1 (56.0)
| | | | | BMI > 18.4: -3 (95.0)
| | | | BMI > 19.1
| | | | | forehead sensation <= -2
| | | | | | BMI <= 20.7: -3 (38.0)
| | | | | | BMI > 20.7: -2 (48.0/21.0)
| | | | | forehead sensation > -2: -2 (274.0/27.0)
| | | BMI > 22.3
| | | | BMI <= 22.6
| | | | | forehead sensation <= -2: -3 (15.0)
| | | | | forehead sensation > -2: 0 (470.0/230.0)
| | | | BMI > 22.6
| | | | | Gender <= 1: -2 (23.0)
| | | | | Gender > 1: -1 (120.0/30.0)
| | forehead sensation > -1
| | | BMI <= 23.2
| | | | BMI <= 22.6
| | | | | BMI <= 18.4: 0 (79.0)
| | | | | BMI > 18.4
| | | | | | Gender <= 1
| | | | | | | BMI <= 19.1: -1 (60.0/30.0)
| | | | | | | BMI > 19.1
| | | | | | | | BMI <= 19.7: 0 (60.0)
| | | | | | | | BMI > 19.7
| | | | | | | | | BMI <= 20.7: -1 (36.0)
| | | | | | | | | BMI > 20.7
| | | | | | | | | | BMI <= 21.5: 1 (120.0/60.0)
| | | | | | | | | | BMI > 21.5: 0 (61.0/13.0)
| | | | | | Gender > 1
| | | | | | | BMI <= 19.7
| | | | | | | | BMI <= 18.5: -1 (181.0/30.0)

97

| | | | | | | | BMI > 18.5
| | | | | | | | | BMI <= 19.1: 1 (149.0/89.0)
| | | | | | | | | BMI > 19.1: -2 (122.0/60.0)
| | | | | | | BMI > 19.7: -1 (523.0/173.0)
| | | | BMI > 22.6: 0 (162.0/30.0)
| | | BMI > 23.2: -1 (125.0)
| forehead sensation > 0
| | BMI <= 21.5
| | | BMI <= 20.1
| | | | BMI <= 19.1
| | | | | BMI <= 18.5: 1 (149.0/30.0)
| | | | | BMI > 18.5: 2 (317.0/164.0)
| | | | BMI > 19.1: 1 (180.0/30.0)
| | | BMI > 20.1
| | | | BMI <= 21.1: 0 (749.0/352.0)
| | | | BMI > 21.1
| | | | | Gender <= 1: 2 (180.0/60.0)
| | | | | Gender > 1: 0 (84.0)
| | BMI > 21.5
| | | Gender <= 1: 1 (679.0/237.0)
| | | Gender > 1
| | | | BMI <= 22.6: 2 (236.0/116.0)
| | | | BMI > 22.6: 1 (73.0/21.0)
forehead sensation > 1
| forehead sensation <= 2
| | BMI <= 22.3
| | | Gender <= 1
| | | | BMI <= 19.1: 3 (57.0)
| | | | BMI > 19.1: 2 (426.0/60.0)
| | | Gender > 1: 2 (540.0/15.0)
| | BMI > 22.3
| | | BMI <= 23.5: 3 (201.0/72.0)
| | | BMI > 23.5: 2 (247.0/67.0)
| forehead sensation > 2
| | Gender <= 1: 3 (618.0/30.0)
| | Gender > 1
| | | BMI <= 18.5: 3 (133.0)
| | | BMI > 18.5
| | | | BMI <= 21.5
| | | | | BMI <= 19.7: 2 (20.0)

98

| | | | | BMI > 19.7: 3 (23.0)
| | | | BMI > 21.5: 2 (29.0)

99

Appendix G: Prediction model (decision tree) of using arm, back sensation
and baseline attributes to predict the overall thermal sensation

upper arm sensation <= 0
| upper arm sensation <= -1
| | upper arm sensation <= -2
| | | BMI <= 19.1
| | | | Gender <= 1: -1 (30.0)
| | | | Gender > 1: -3 (67.0)
| | | BMI > 19.1
| | | | upper arm sensation <= -3
| | | | | back sensation <= -2: -3 (15.0)
| | | | | back sensation > -2: -1 (69.0)
| | | | upper arm sensation > -3
| | | | | back sensation <= -1
| | | | | | BMI <= 21.5
| | | | | | | back sensation <= -2
| | | | | | | | BMI <= 19.7: -2 (30.0)
| | | | | | | | BMI > 19.7: -3 (38.0)
| | | | | | | back sensation > -2: -2 (221.0/21.0)
| | | | | | BMI > 21.5: -2 (260.0)
| | | | | back sensation > -1: -1 (144.0/30.0)
| | upper arm sensation > -2
| | | back sensation <= -2
| | | | Gender <= 1: -2 (44.0)
| | | | Gender > 1: -1 (90.0/30.0)
| | | back sensation > -2
| | | | BMI <= 20.1
| | | | | Gender <= 1: 0 (43.0)
| | | | | Gender > 1
| | | | | | back sensation <= -1
| | | | | | | BMI <= 19.7: -1 (30.0)
| | | | | | | BMI > 19.7: -2 (38.0)
| | | | | | back sensation > -1: -1 (335.0/57.0)
| | | | BMI > 20.1
| | | | | back sensation <= -1
| | | | | | BMI <= 22.6: -1 (219.0)
| | | | | | BMI > 22.6: 0 (102.0/30.0)
| | | | | back sensation > -1

100

| | | | | | Gender <= 1
| | | | | | | BMI <= 21.1: 0 (30.0)
| | | | | | | BMI > 21.1: -1 (149.0/54.0)
| | | | | | Gender > 1: 0 (354.0)
| upper arm sensation > -1
| | BMI <= 23.2
| | | BMI <= 19.1
| | | | BMI <= 18.4: 0 (79.0)
| | | | BMI > 18.4
| | | | | back sensation <= 0: 1 (270.0/60.0)
| | | | | back sensation > 0: 0 (30.0)
| | | BMI > 19.1
| | | | BMI <= 22.3
| | | | | BMI <= 20.7: 0 (249.0/30.0)
| | | | | BMI > 20.7
| | | | | | Gender <= 1
| | | | | | | BMI <= 21.1
| | | | | | | | back sensation <= 0: -1 (30.0)
| | | | | | | | back sensation > 0: 0 (300.0/92.0)
| | | | | | | BMI > 21.1: 0 (78.0)
| | | | | | Gender > 1: 1 (7.0)
| | | | BMI > 22.3
| | | | | back sensation <= -1: 0 (77.0)
| | | | | back sensation > -1: 1 (118.0/30.0)
| | BMI > 23.2
| | | back sensation <= 0
| | | | BMI <= 24.2: -1 (60.0)
| | | | BMI > 24.2: 0 (41.0/10.0)
| | | back sensation > 0
| | | | BMI <= 23.5: 1 (122.0/30.0)
| | | | BMI > 23.5: 2 (33.0/9.0)
upper arm sensation > 0
| upper arm sensation <= 1
| | back sensation <= 1
| | | BMI <= 21.1
| | | | BMI <= 20.7
| | | | | back sensation <= 0
| | | | | | BMI <= 19.7: 0 (31.0/1.0)
| | | | | | BMI > 19.7: 1 (60.0)
| | | | | back sensation > 0

101

| | | | | | Gender <= 1: 2 (50.0/10.0)
| | | | | | Gender > 1: 1 (209.0/30.0)
| | | | BMI > 20.7: 0 (90.0/30.0)
| | | BMI > 21.1
| | | | Gender <= 1: 1 (478.0/56.0)
| | | | Gender > 1
| | | | | BMI <= 21.5: 2 (55.0/8.0)
| | | | | BMI > 21.5: 1 (135.0/23.0)
| | back sensation > 1
| | | Gender <= 1
| | | | BMI <= 19.7: 1 (60.0)
| | | | BMI > 19.7
| | | | | BMI <= 22.6: 2 (131.0)
| | | | | BMI > 22.6: 1 (42.0)
| | | Gender > 1: 2 (244.0/30.0)
| upper arm sensation > 1
| | back sensation <= 2
| | | BMI <= 21.5
| | | | Gender <= 1
| | | | | back sensation <= 1: 2 (273.0)
| | | | | back sensation > 1
| | | | | | BMI <= 19.1: 3 (87.0)
| | | | | | BMI > 19.1
| | | | | | | BMI <= 21.1: 2 (172.0/30.0)
| | | | | | | BMI > 21.1: 3 (54.0)
| | | | Gender > 1: 2 (386.0)
| | | BMI > 21.5
| | | | Gender <= 1: 2 (570.0/162.0)
| | | | Gender > 1
| | | | | back sensation <= 1: 2 (118.0/58.0)
| | | | | back sensation > 1: 3 (39.0)
| | back sensation > 2
| | | Gender <= 1: 3 (499.0)
| | | Gender > 1
| | | | BMI <= 18.5: 3 (133.0)
| | | | BMI > 18.5
| | | | | BMI <= 19.7: 2 (17.0)
| | | | | BMI > 19.7: 3 (23.0)

Abstract (if available)

Abstract Traditional air-conditioning methods maintain temperatures in a whole room at a constant level, and much work has been done to assess and improve the thermal comfort and sensations of people in a building environment. This study endeavors to identify the potential of using some local body area for predicting thermal stress. A total of 20 human subjects were tested in the University of Southern California’s climate chamber to determine their physiological parameters and subjective perceptions of environment. Ambient temperature was documented during the tests, while the human subjects were exposed to a warm, cool, or neutral environment. Based on these tests, correlation analysis and algorithms are applied to identify the relative thermally sensitive skin areas, their contribution rate to the overall thermal sensation, and potential skin area combinations that have high correlation with thermal sensation. Besides, the study also identifies the different impacts of local thermal sensation and local skin temperature while predicting the overall thermal sensation. When only the baseline attributes (environment temperature, Body Mass Index [BMI], gender) are considered, the estimation can be 94.5%. If baseline attributes are combined with the temperature of one local spot on the skin, the estimation accuracy can be around 97%

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Enhancing thermal comfort: air temperature control based on human facial skin temperature

PDF

Exploration for the prediction of thermal comfort & sensation with application of building HVAC automation

PDF

Human-building integration: Investigation of human eye pupil sizes as a measure of visual sensation in the workstation environment

PDF

Developing environmental controls using a data-driven approach for enhancing environmental comfort and energy performance

PDF

Building energy performance estimation approach: facade visual information-driven benchmark performance model

PDF

Enhancing thermal comfort: data-driven approach to control air temperature based on facial skin temperature

PDF

Energy use intensity estimation method based on building façade features by using regression models

PDF

Multi-occupancy environmental control for smart connected communities

PDF

Visualizing thermal data in a building information model

PDF

IEQ, sleep quality, and IoT: meta-analysis on improving IEQ and sleep quality using IoT

PDF

Natural ventilation in tall buildings: development of design guidelines based on climate and building height

PDF

Using bio-signals with smart windows

PDF

Exploring participatory sensing and the Internet of things to evaluate temperature setpoint policy and potential of overheating/overcooling of spaces on the USC campus

PDF

Tiny house in the desert: a study in indoor comfort using moveable insulation and thermal storage

PDF

Daylighting in Riyadh and Los Angeles: comparison of cultural factors in potential market penetration

PDF

Human-environmental interaction: potential use of pupil size for office lighting controls

PDF

Exterior glare simulation: understanding solar convergence from concave facades using heat maps

PDF

Human–building integration: machine learning–based and occupant eye pupil size–driven lighting control as an applicable visual comfort tool in the office environment

PDF

Double skin façade in hot arid climates: computer simulations to find optimized energy and thermal performance of double skin façades

PDF

Environmentally responsive buildings: multi-objective optimization workflow for daylight and thermal quality

Asset Metadata

Creator Wang, Qi (author)

Core Title Developing a data-driven model of overall thermal sensation based on the use of human physiological information in a built environment

School School of Architecture

Degree Master of Building Science

Degree Program Building Science

Publication Date 06/02/2017

Defense Date 04/28/2017

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag OAI-PMH Harvest,skin temperature,thermal comfort,thermal environment,thermal sensation

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Choi, Joon-Ho (committee chair), Noble, Douglas (committee member), Schiler, Marc (committee member)

Creator Email qiwang1201@163.com,wang375@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c40-377653

Unique identifier UC11255842

Identifier etd-WangQi-5362.pdf (filename),usctheses-c40-377653 (legacy record id)

Legacy Identifier etd-WangQi-5362.pdf

Dmrecord 377653

Document Type Thesis

Rights Wang, Qi

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA