Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Enhancing thermal comfort: air temperature control based on human facial skin temperature
(USC Thesis Other)
Enhancing thermal comfort: air temperature control based on human facial skin temperature
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
Enhancing Thermal Comfort: Air Temperature Control Based on Human Facial Skin Temperature By Bo Yi A Thesis Document SCHOOL OF ARCHITECTURE UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree MASTER OF BUILDING SCIENCE May 2015 Committee Chair: Joon-Ho Choi joonhoch@usc.edu Phone: (213)740-4576 Third Member: Douglas Noble dnoble@usc.edu Phone: (213)740-4589 Bo Yi yib@usc.edu Phone:213-610-6662 Second Member: Kyle Konis kkonis@usc.edu Web: performance-and-form.corn 2 Abstract Thermal comfort is defined as condition of mind that indicates human thermal sensation, such as cool or warm, with the thermal envirornnent. Achieving thermal comfort for occupants in buildings is an important purpose of heating, ventilation, and air conditioning (HV AC) systems. Air temperature is one of the significant parameters that affect thermal comfort, in addition to mean radiant temperature, air speed, relative humidity, metabolic rate, and clothing insulation. However, since physiological conditions and satisfaction standards vary considerably from person to person, each individual normally has a different acceptable range of air temperature in a certain thermal condition. It is possible to create an automatic system by utilizing physiological signals to help control and optimize buildings' indoor air temperature to help maintain thermal neutrality m a one-occupant room. Skin temperatures can signal appropriate thermoregulatory behavior in response to specific conditions, smce skin temperature changes with different thermal sensations. The main goal of this study is to better understand the relationship between facial skin temperature and human thermal sensation in order to achieve the appropriate conditioned air temperature dynamic settings, thus helping to achieve thermal comfort by reading facial temperature signal patterns. A series of human subject experiments were conducted in an envirornnental chamber at USC. Facial skin temperatures were tested continuously by six electronic thermocouples at six locations and sensation surveys were provided by subjects. During the tests, the air temperature was regulated between 20°C and 30°C, while the other five thermal parameters were kept constant. A thermal sensation diagnostic model and control logic with decision tree was built by using data mining and machine 3 learning technology. Hypothesis Statement: It is possible to create an automatic system by utilizing physiological signals to help control and optimize a building's indoor thermal conditions in a one-occupant room. 4 TABLE OF CONTENTS ABSTRACT ................................................................................................................... 3 Chapter 1 Introduction ................................................................................................... 7 1.1. Fundemantals of thermal comfort ...................................................................... 8 1.1.1. Significance of thermal comfort ............................................................ 8 1.2. Factors influencing thermal comfort ................................................................ 9 1.2.1. Metabolic rate ...................................................................................... 10 1.2.2. Clothing insulation ............................................................................... 11 1.2.3. Mean radiant temperature .................................................................... 12 1.2.4. Air temperature .................................................................................... 13 1.2.5. Air speed .............................................................................................. 14 1.2.6. Relative humidity ................................................................................. 14 1.3. Background of PMV/PPD model. ................................................................. 15 1.4. Background of current manual control strategies .......................................... 20 1.5. Hypothesis statement ........................................................................... 21 1.6. Potential of utilizing facial skin temperature to be integrated with a thermal control system ................................................................................. 22 1.7. Explanation ofterms ....................................................................................... 26 1. 7.1. Physiological signals ............................................................................ 26 1. 7.2. Region of interest ................................................................................. 26 1. 7.3. Building automation system ................................................................ 27 1. 7.4. Infrared ................................................................................................. 28 1.7.5. Core temperature .................................................................................. 29 1. 7.6. Thermoregulation ................................................................................. 29 1. 7. 7. Two sample tailed-test ......................................................................... 30 1.8. Research objectives ........................................................................................ .31 1.9. Project scope .................................................................................................. .34 Chapter 2 State of Art of Thermal Control Technology and Facial Skin Study ........... 35 2.1. Current thermal control technology ...................................................... 35 2.2. The role skin temperature plays in human thermoregulation ....................... .43 2.3. Review of studies on human thermal sensation ............................................. .45 Chapter 3 Project Methodology ................................................................................... 49 3 .1. Experiment resources ...................................................................................... 50 3.1.1. Environmental chamber. ...................................................................... 50 3.1.2. Heating and cooling system .................................................. 51 3.1.3. Data acquisition tools ........................................................................... 54 3.2. Experiment procedure ................................................................................... 61 Chapter 4 Human Subjects Experiment Data and Results ........................................... 64 5 Chapter 5 Interaction of Facial Temperature, Ambient Temperature, and Thermal Sensation ............................................................................... 73 5.1. Facial skin sensing location selection for thermal sensation estimation ......... 73 5.2. Discussion of ambient temperatures at each thermal sensation ...................... 86 5.3. Facial temperature based control model development.. .................................. 88 Chapter 6 Conclusion ................................................................................................. 100 Chapter 7 Future Work ............................................................................................... 106 References .................................................................................................................. 108 6 Chapter 1 Introduction 1.1. Fundamentals of thermal comfort 'Thermal comfort is the condition of mind that expresses satisfaction with the thermal environment and is assessed by subjective evaluation' (ASHRAE-55, 2013). According to the ASHRAE standard, there are six primary factors affecting thermal comfort, which can be categorized into two groups: personal factors, due to characters of occupants; and environmental factors, due to ambient conditions. Personal factors include clothing insulation and metabolic rate; environmental factors include air temperature, relative humidity, air speed, and mean radiant temperature (ASHRAE-55, 2013). Those factors can affect thermal comfort since they help determine heat gain and loss of the human body. Because thermal comfort is a condition of mind, it indicates how people feel about the thermal condition of the ambient environment. Achieving thermal comfort of the occupants of buildings is one of the main tasks of heating, ventilation, and air conditioning (HV AC) systems. Satisfaction with the thermal environment is significant because it influences productivity and health. After receiving 34,000 survey responses to thermal comfort questions in 215 buildings, researchers from UC Berkeley found that office workers who were satisfied with their thermal environment were more productive (Huizenga et al, 2006). Thermal discomfort has also been discovered to result to sick building syndrome symptoms (Myhren & Hohnberg, 2008). Thermal discomfort is significantly related to individual physiological and psychological mechanisms. Discomfort is linked to thermal stress, which can affect work performance and individual health (Wyon, 7 1996). According to Huizenga's research, in 215 buildings in the USA, Canada, and Finland, only in 11% of surveyed buildings did 80% or more occupants claim that they were satisfied with the building's thermal conditions (Huizenga et al, 2006). 1.1.1. Significance of Thermal Comfort There is evidence that the increased availability of individual control has positive effects far beyond just thermal comfort (de Dear & Brager, 2002). Hawkes has found that energy consumption was reduced if people had access to control their thermal envirornnent, since energy consumption was more closely aligned to needs rather than maintaining solid standards based on externally-imposed standards (Hawkes, 1982). Researchers found fewer symptoms of Sick Building Syndrome (SBS) and greater productivity were achieved because the perceived level of individual control increased (de Dear & Brager, 2002). The importance of individual control should not be underestimated, but needs to be studied further so its effects on comfort, health, productivity, and energy consumption can be fully understood. Thus, individual thermal control can be better applied to buildings. SBS is the situation in which health problems occur as a result of interior envirornnental air quality. It is not a specific illness, but it has negative effects linked to a building's indoor conditions on occupants. Frequently, problems result when a building is operated or maintained in a status that is different to its original design or prescribed operating procedures. Sometimes poor building design or occupant activities cause indoor air quality problems (EPA, 1991). 8 Keeping the majority of people thermally comfortable is a requirement of ANSI/ ASHRAE Standard 55 - Thermal Envirornnental Conditions for Human Occupancy (2013). As thermal comfort is a condition of mind, it is not only based on envirornnental thermal conditions, but also depends on human physiology and psychology. Therefore, ASHRAE's thermal comfort zone is suitable for the majority of people, but not for 100% of people because people have different physiological and psychological conditions and status. Since current systems that use the solid standard calculated by a comfortable combination of air temperature, relative humidity, mean radiant temperature, air speed, metabolic rate, and clothing level, aimed to make acceptable thermal conditions for the majority of people, those systems can hardly meet best preference of an individual but try to compromise and make it acceptable to a larger number of people instead. A system based on using human physiological signals is needed to improve thermal comfort for individual users. Due to the goal of HV AC systems, the whole system is human centered; therefore, a system with human physiological signals as feedback signals may be valuable to improve thermal comfort. Researchers from Japan conducted a study about automobile temperature control using neural technology; they created an automobile thermal control system by predicting skin temperature (UEDA et al, 1997). The details of their study will be discussed in Chapter 2. 1.2. Factors influencing thermal comfort As described previously, there are six primary factors that directly affect thermal comfort: metabolic rate, clothing level, air dry bulb temperature, mean radiant temperature, air speed, and relative humidity. In this thesis, the research mainly studies the air temperature factor while other factors were kept constant. The 9 background knowledge of those six factors is described in the following sections. 1.2.1. Metabolic rate Metabolic rates can reflect people's activity levels and are also affected by environmental conditions (Stanton, 2015). In ASHRAE Standard 55, metabolic rate is defined as the rate of chemical energy transforming into heat and mechanical work by metabolic activities within an organism. It is usually described in the form of unit area of the total body surface. Generally, the met unit is used to describe metabolic rate. The definition of met unit is defined as: 1 met~ 58.2 W/m 2 (18.4 Btu/h·ft2), which is equal to the energy produced per unit surface area of an average person seated at rest. The surface area of an average person is 1.8 m 2 (19 ft 2 ) (ASHRAE-55, 2013). ASHRAE Standard 55 provides a table of met rates for some general types of activities. For example, the value for sleeping is 0. 7 met, the value for typing is 1.0, and the value of packing is 2.1. For activity which is not continuous, the ASHRAE Standard recommends a time-weighted average metabolic rate if people are performing activities that change over a period of one hour or less. If the period is longer than one hour, it must be considered as different metabolic rates (ASHRAE-55, 2013). According to the ASHRAE 'Handbook of Fundamentals', predicting metabolic rates is very complex. 10 When the activity level has the metabolic rate of 2 met or above, the accuracy is not high. Therefore, the ASHRAE standard is not accurate to estimate metabolic rate for activities higher than 2 met. Besides ASHRAE, the 'Compendium of Physical Activities' is used to record physical activities in the medical industry. In the 'Compendium of Physical Activities', met is defined as a different method, that is the ratio of the metabolic rate of the activity to a resting metabolic rate (Ainsworth et al, 2000). Because the calculation method is different from the met that ASHRAE uses, these met values from the 'Compendium of Physical Activities' cannot be applied directly in a predictive mean vote (PMV) model, but it creates an innovative way of quantifying physical activity level. Food and beverage habits can affect metabolic rate, thereby indirectly affecting thermal preferences. These effects can be different based on the food and beverage intake (Szokolay, 2004). Body shape is another parameter will affect the thermal comfort. Radiation depends on the surface area. A tall and thin man with a large ratio of surface to volume can more easily have heat dissipation, and can withstand higher temperatures than a person with a round body shape (Szokolay & Brisbin, 2004). 1.2.2. Clothing insulation The quantity of heat insulation people wear has a great influence on thermal comfort, because it affects heat loss as well as heat balance (Havenith, 1999). The insulation layer reduces heat loss, and can help keep a person warm or possibly lead to overheating. In general, thicker clothing has an increased 11 insulating ability. Depending on the material type of the garment, air flow and relative humidity can decrease the capacity of the insulation material (McCullough, Eckels & Harms 2009). One clo is equal to 0.155 m 2 ·K/W (0.88 °F·ft2·h/Btu). This, typically, stands for clothing insulation with trousers, a long sleeved shirt, and a jacket. Clothing insulation values for other general clothes were given by ASHRAE (ASHRAE-55, 2013). 1.2.3. Mean radiant temperature Mean radiant temperature (MRT) is defined as the uniform temperature of an imaginary shell, from which the human body radiation heat transfer is equal to the radiation heat transfer in the actual non-uniform shell. MRT is a concept derived from the fact that the radiation net exchange between two objects can be approximately proportional to their temperature by emission and absorption of heat capacity. This applies around the object in the body of the area weighted mean temperature. This is the temperature difference with respect to effectiveness. This is valid under the condition that the absolute temperatures of the objects in question are obviously larger than the temperature differences, which allows linearization of the Stefan-Boltzmann Law in the linked temperature range (ISO 7726, 1998). Thermal comfort is related to the impact of both the air dry bulb temperature and the mean radiant temperature. The mean radiant temperature is expressed as the surface temperature of the room and is significantly related to enclosure performances. Keeping a balance between the operative temperature and 12 the mean radiant temperature can make spaces more comfortable. This is done with effective design of the building's interior, use of high temperature radiant cooling, and low temperature radiant heating (Mcintyre & Griffiths, 1972). 1.2.4. Air tern perature Air temperature is also known as dry bulb temperature. The dry bulb temperature (DBT) is the temperature of air measured by a thermometer which is freely exposed to the indoor air but is insulated from moisture and radiation. Dry bulb temperature is the true thermodynamic temperature which reflects the heat energy of air and is usually considered as the temperature of air. It indicates the amount of heat of the air and it is directly proportional to the mean kinetic energy of the air molecules. Unlike wet bulb temperature, dry bulb temperature is not related to humidity. In construction, it is an important parameter when designing a building for a certain climate. Nall has stated that dry bulb temperature is one of the most important climate variables for human comfort and building energy efficiency (Nall, 2004). Dry bulb temperature is an important characteristic and is shown on the horizontal axis of the Psychrornetric chart. In this thesis, the dry bulb temperature will be investigated as the factor influencing human facial skin temperature and thermal sensation, while the other five parameters that affect thermal comfort are constant. 13 1.2.5. Air speed Air speed is defined as the rate of air movement at a point, regardless of movement direction. According to ASHRAE Standard 55, air speed is the average speed of the air to the body part, which is exposed in the indoor air, at a certain time and location. However, in some spaces, the airflow velocity and direction varies at different locations. For example, in a conditioned room with floor grilles, the air velocity could be 0.3m/s on the occupant's feet but the velocity be close to zero on the occupant's face. It results in different heat loss rates from different body parts. Therefore, the designer should decide the average appropriately, especially air velocity incidents on exposed body parts, which have greater cooling effects and potential for local discomfort (ASHRAE-55, 2013). 1.2.6. Relative humidity Relative humidity is defined as the ratio of the quality of moisture vapor in the air to the quality of moisture vapor that the dry air could hold at a certain temperature and pressure. Unlike the human body, which has receptors in the skin that can feel heat and cold, relative humidity (RH) is detected indirectly by the human body. Sweating is an important heat loss mechanism based on moisture evaporation from the skin. However, at high relative humidity, air contains nearly the maximum moisture vapor that it can hold. Consequently, wet evaporation and heat loss rate is decreased (Balaras et al, 2007). On the other hand, very dry envirornnents (RH<20-30%) also make people uncomfortable because they result in a too high water evaporation rate from the skin and people lose heat too fast. The recommended relative 14 humidity of a conditioned space is from 30% to 60% (Wolkoff & Kjaorgaard, 2007). 1.3. Background of PMVIPPD model In order to establish a standard to achieve thermal comfort, Fanger developed the Predicted Mean Vote/Predicted Percentage of Dissatisfaction (PMV IPPD) model by utilizing heat balance equations and experience based research about skin temperature to determine thermal comfort. According to the PMV model, subjects were asked to complete thermal comfort surveys regarding their thermal sensation on a seven point scale from cold (-3) to hot (+3). Fanger's equations are used to calculate PMV in a large number of human samples for a particular thermal condition, which is a combination of dry bulb air temperature, mean radiant temperature (MRT), relative humidity, air speed, metabolic rate, and clothing insulation (Fanger, 1972). Zero value stands for thermal neutrality, which is the ideal value. The ASHRAE comfort zone is defmed by the combinations of the six parameters when the PMV is within recommended limits (-0.5<PMV<+0.5), i.e. where the occupants' metabolic rates are in the range 1.0 met to 1.3 met and clothing provides between 0.5 clo and 1.0 clo of thermal insulation (ASHRAE-55, 2013). The PMV was developed as the index to forecast the mean thermal sensation vote for occupants based on a standard scale by taking into account the four envirornnental factors and two personal factors. It was derived from taking measurements from more than 300 people. The PMV model is a flexible tool that can be utilized in different indoor envirornnents with different HVAC systems, clothing levels, and activity levels. The PMV model is represented by a 7-point thermal sensation scale shown in Table I: 15 Table I. Seven point thermal sensation scale Hot Warm Slightly Neutral Slightly cool cold warm cool +3 +2 +l 0 -1 -2 -3 Zero is the ideal value which stands for thermal neutrality; thermal comfort is described by AHRAE as the combination of votes on the six factors that are within -0.5 and 0.5. Although predicting the thermal sensation of an individual is a significant task in determining what conditions are comfortable, it is more applicable to consider whether or not occupants will be satisfied. Fanger developed another equation linked to the PMV, called the Predicted Percentage of Dissatisfaction (PPD) to indicate human thermal satisfaction. This was based on research that surveyed subjects in an envirornnental chamber where the indoor thermal conditions could be precisely controlled (Fanger, 1972). ASHRAE's standard set the requirement of thermal conditions based on PMV; it requires 80% of occupants to be satisfied (Figures I. I. and 1.2.). 16 30 25 20 15 10 5 ~~~~~~~~~~~~~~~--~~~~~~L-~~~~--+-0 10 12 14 16 18 20 22 24 26 28 30 32 34 36 Drybulb Temperature [°C) Figure 1.1. Psychometric chart of thermal comfo1t for PMV method 100 90 80 70 I - - - l - I 30 20 10 qo 12 14 16 1a 20 22 24 26 2s 30 32 34 36 Drybulb Temperature (°C) Figure 1. 2. Temperature-humidity chait of theimal comfort for Pl\!IV method 'ii ,, fi - ~ .S!? .Q (ii a: ~ E ~ J: 17 Predicted percentage of dissatisfaction represents the percentage of occupants who are dissatisfied with the indoor thermal conditions. The maximum of PPD is 100%, which indicates no one is satisfied with the ambient thermal condition. The AHRAE-55 standard indicates that the PPD of buildings should be less than 10%. As PPD is a function of PMV, when the PMV is closer to zero, the PPD is lower. The value of PPD can be defined according to the following equation: (1) Currently, the PMV model is the most popular thermal formula in industry. However, based on the PMV model, the ASHRAE-55 standard established requirements for indoor thermal conditions and built thermal comfort zones in which at least 80% of people are satisfied (ASHRAE-55, 2013). Based on this standard, there may be up to 20% of occupants who are not satisfied with the condition, even though it meets the requirement. In fact, since thermal comfort is a condition of mind (ASHRAE-55, 2013), it is impossible for a special thermal condition to exist that satisfies 100% of occupants. In addition, the PMV model does not consider an occupant's physiological conditions, such as gender, age, body mass, fat rate, or health condition, that are proven important factors for individual thermal sensation (Cena & de Dear, 2001). However, in a single occupant space, there is a good chance to enhance the thermal comfort for one person based on that person's physiological signals. Therefore, a new method of control, using human physiological signals such as skin temperature, has the potential to improve thermal 18 conditions, especially in one-occupant spaces, like a private office. Barlow and Fiala suggested that future service engineers and architects should have a better understanding of thermal adaptation and occupants' thermal comfort, focusing more on the indoor climate and 'human aspects' in order to include adaptive models in their building design work (Barlow & Fiala, 2007). Besides the PMV /PPD model, the adaptive model is another type of model used when discussing thermal comfort and is based on the idea that outdoor climate influences indoor thermal comfort due to human ability to adapt to different temperatures during different times of the year. For adaptive comfort, humidity, air velocity, and clothing, are not taken into account. Many researchers have conducted field research around the world, surveying occupants about their thermal comfort while simulating a building's indoor envirornnent. Analyzing a database of results from 160 of these buildings, including 93 air conditioned buildings, indicated that occupants of buildings with natural ventilation only accept, and even prefer, a wider range of temperature than their counterparts in air conditioned buildings, because their preferred temperature depends on outdoor conditions. An adaptive comfort model was built by ASHRAE based on these results and incorporated in the ASHRAE-55 standard. The adaptive model is related to indoor comfort temperature based on the prevailing outdoor temperature and defmes zones of 80% and 90% satisfaction (Figure 1.3.). 19 34 .. .. .. 32 30 ~ 28 v ~ 26 e ~ 24 E {E -~ 22 e ~ 20 0 18 16 1~0 12 14 16 18 20 22 24 26 28 30 32 Prevailing Mean Outdoor Temperature (°CJ Figure 1.3. Adaptive cha.it according to ASHRAE-55 1.4. Background of current manual control strategies Currently, a dcy-bulb the1mostat is conunonly used in buildings. Thermostat strategy allows users to set a certain dty-bulb temperature manually as indoor air temperature based on their thermal expetience. It is effective when occupants have easy access to the thermostat and in workplaces shating thermal zones with the common thermal requirement (Rose & Dozier, 1997). Howeve1~ the the1mostat control system is only based on one viable, dcy-bulb temperature set by users. It cannot detect other factors significant to human the1mal sensation. including humidity, air velocity, radiant heat gain, etc. For users, supplying conditioned air with different status is needed in different situations, 20 including outside temperature, metabolic rates, and clothing. Manual operation of thermostats can cause trouble and energy waste due to inappropriate thermostat settings. Since there are large variations from person to person in terms of individual satisfaction, both psychologically and physiologically, it is impossible to define an optimal temperature that will satisfy everyone in a given space (ASHRAE-55, 2013). 1.5. Hypothesis Statement According to many researchers, individual control is needed for indoor thermal control (Jaakkola, Heinonen & Seppanen, 1989) in a one occupant space and human thermal sensation is represented by physiological signals, such as skin temperature and heart rate (Choi, 2010). Physiological signals would be the best form of feedback for an automatic system, because it is a human centered response that can be sensed. In an indoor envirornnent, with a sitting occupant, skin temperature in the facial area is generally exposed and, thus, is possible to be detected. So, facial skin temperature has great potential to be investigated as an index of human thermal sensation. Therefore, the hypothesis is as follows: It is possible to create an automatic system by utilizing physiological signals to help control and optimize building indoor air temperature to help maintain thermal neutrality in a one-occupant room. 21 1.6. Potential of utilizing facial skin temperature to be integrated with a thermal control system Skin is the largest organ of the human body. It can protect human body tissues against injury by changing the blood vessel flow and, thus, changing surface temperature. The nerves in skin can receive thermal stimulus and send it to the brain by neural networking. The stimulus can be interpreted by the brain as hot, neutral, and cold. Skin is made up of three layers: epidermis, dermis, and subcutaneous fatty tissue (Egan et al, 2005). Based on this function, skin surface temperature is a human response to ambient thermal conditions. Conversely, human thermal sensation can be represented by skin temperature. When it comes to the potential application, facial skin could be the most applicable part of skin to sense. Typically, the facial area is the most exposed part of body. In an indoor environment, facial temperature is easy to sense by wearing sensing devices or radiant sensing devices. For a computer work-based occupant in a single occupant room, the face is relatively fixed and their surface temperature is possible to sense using remote sensors, such as infrared temperature sensors. Compared to other parts of unexposed skin, measuring facial skin temperature does not require a human to wear any other devices. A conceptual example of measuring facial skin temperature application is shown in the image below (Figure 1.4.). 22 Sensing Facial Skin Temperature ---------------------------------------- --- --- --- --- --- --- --- --- --- · · ···· ·····• Infrared Temperature Sensor · ······ · · ·················• Human User · ························• Computer Monitor Figure 1.4. A conceptual example of measuring facial skin temperature The heat loss from skin can be categorized into sensible heat and evaporative heat (ASHRAE-55, 2013). Sensible heat exchange between human skin and ambient environment must be transported via clothing. These heat exchanges can be described in terms of heat transfer from the skin's surface, through the clothing insulation, to the outer clothing surface, and heat transfer from the outer clothing surface to the environment (ASHRAE, 2013). Heat transfer between the outer clothing surface and the ambient environment mainly includes convection and radiation. Both convective heat losses and radiative heat transfers from the outer surf ace of a clothed body to the ambient environment can be described in terms of 23 a heat transfer coefficient and the difference of temperature between the mean temperature of the clothing outer surface and the envirornnental operative temperature (ASHRAE, 2013). ASHRAE's handbook gives an equation to calculate the total sensible heat loss from skin: (2) where sffi<in ~total sensible losses from skin h~the combination of convective and linear radiative heat transfer coefficient, W/(m 2 -K) fc1~clothing area factor tc 1 ~temperature of outer clothing surface to~temperature of ambient operative temperature Evaporative heat loss (latent heat loss) from human skin depends on the humidity on the skin and the difference of the water vapor pressure between the skin and the ambient envirornnent (ASHRAE, 2013). Eskin w(psk.s - P,) (3) where E,kin ~evaporative heat loss w~skin wettedness, dimensionless p'"_,~water vapor pressure at skin, normally assumed to be that of saturated water vapor at tsk, kPa p,~water vapor pressure in ambient air, kPa R,.c1~evaporative heat transfer resistance of clothing layer, (m2·kPa)/W 24 h,~evaporative heat transfer coefficient, W/(m2 ·kPa) The total heat loss from the skin is the convective heat loss plus radiative heat loss. When it comes to the facial area, the face is not covered by clothing. So, it can be considered that, fc1~ l,tc 1 ~t,kin., R.e.c1~0. Therefore, (4) This equation indicates the heat loss from the skin to ambient envirornnent. Although 70% of losses of body heat are from the head, facial skin temperature is still possible to be utilized to predict human thermal sensation. The biological principle of function of heat loss from the skin will be discussed in Chapter 2. The research included a series of human subject experiments and identified the relationship between skin temperature patterns and varying thermal conditions. Skin temperature is a key factor of human thermoregulation function; change of skin temperature can change the heat transfer rate and, thus, regulate core body temperature. In other words, skin temperature indicates the tendency that a human body needs to be warmed or cooled. The principle of how skin temperature works in the thermoregulation system is discussed in Chapter 2 (background). The proposed control system makes the thermal control system human-centered. It is possible, therefore, to remove satisfaction of occupants by correcting errors automatically. 25 For advanced building systems, an automatic system is highly recommended. Manual operation may cause inconvenience and inaccuracy due to sophisticated building systems; however, an automatic building system may solve some of the problems. An advanced automatic system does not need any unprofessional users to control it, which results in inappropriate settings and adjustments. In addition, semiconductor technology is developing which is reducing the cost of manufacturing sensors, and the development of communication and information technology. It makes building automatic systems valuable to be researched. 1.7. Explanation of terms 1.7.1. Physiological Signals Physiological signals are those signals that can be collected from humans by sensors, such as heart rate, body temperature, and body acceleration. Human facial skin temperature is an important type of physiological signal that can be tested easily because human facial skin is generally exposed. 1. 7 .2. Region oflnterest The Region of Interest (ROI)is the region within the frontal frontier by the edge of the lips, the posterior frontier by the meatus acusticus extemus, the upper frontier by the most caudal part of the nasal ala and the inferior border by the most caudal part of the chin (Christensen, Vaeth, & Wenzel, 2012), as shown Figure 1.5. 26 Figure 1.5. Thermogram of right side of a subject's face with region of interest (Christensen, Vaeth, & Wenzel, 20 12) 1. 7.3. Building Automation Systems Building Automation Systems (BAS) are central! y controlled and interact between hardware and software. A building automation system can monitor and control the environment in residential, office, commercial, and industrial buildings. The automation system can guarantee the operational p erfonnance of the facility as well as the comfort and safety of building occupants while managing various building systems. 'Typically, the building automation systems are installed in new buildings or asp art of a renovation where they replace an outdated control system' (KMC Controls, 2013). 27 For HVAC systems, various equipment can be controlled, such as chillers, boilers, air handling units (AHUs), roof-top units (RTUs), fan coil units (FCUs), heat pump units (HPUs), and variable air volume boxes (VAVs). 1. 7 .4. Infrared Infrared (IR) is a type of invisible radiant energy and electromagnetic radiation. The wavelengths of IR are longer than the wavelengths of visible light. The spectrum ofIR extends from the nominal red edge of the visible spectrum at 700 nanometers (frequency 430 THz) to 1 mm (300 GHz) (Liew, 2006). Most of the thermal radiation emitted by objects near room temperature is infrared. Infrared radiation can be used to detect the temperature of objects remotely. The thermal images collected by infrared sensors are therrnographic (Figure 1.6.). Therrnography is widely used in military and industrial applications to observe objects by reading their heat radiant signals. With the development ofIR sensors and information technology, the low cost makes it possible to use thermography in a broad variety of applications. 28 Figure 1.6. A thermograph of two people (Source: http ://upload.wikimedia.org/wikipedia/comm ons/thum b/c/ cf7Ir _girl. png/ l 024px-Ir _girl. png) 1. 7.5. Core Temperature Core temperature, also called core body temperature, is the operating temperature of the human body. It maintains a narrow range of temperature change to guarantee the essential reaction of the organism. It stands for the temperature of an organism in the deep structure of the body , such as liver, heart, and stomach. 1. 7.6. Thermoregulation Thermoregulation is a physiological function of an organism to keep its body temperature within a narrow range when the surrounding temperature is very diff erent. The human body needs to maintain a relatively solid core temperature. Skin temperature plays an important role in the thermoregulation process. When the core body temperature is higher than normal, blood flow in capillaries will accelerate making skin temperature increase. A relatively high skin temperature tends to lose heat from the body, thus cooling 29 down the core body temperature. However, when the core body temperature becomes lower than normal, blood flow in capillaries will slow down to decrease skin temperature. A relatively low skin temperature reduces the heat loss from the body's surface to the ambient envirornnent, thus increasing the core body temperature. The internal therrnoregulation process is an aspect of homeostasis: a state of dynamic stability in an organism's internal conditions, maintained far from equilibrium with its envirornnent (Hong et al, 2015). If the normal temperature of the human body fails to be maintained, it can increase significantly above normal temperature and a condition known as hyperthermia occurs. For humans, hyperthermia occurs when the body is exposed to constant temperatures of approximately 55 C. If the human body has prolonged exposure (longer than a few hours) at this temperature or is exposed to 75 C, this person is at high risk of dying. Humans may also experience lethal hyperthermia when the wet bulb temperature is sustained above 35 °C (95 °F) for six hours. In the other hand, if body core temperature decreases and is significantly below normal temperature, this situation is called hypothermia (Hong et al, 2015). 1.7.7. Two sample tailed-test A two sample tailed-test (t-test) is an alternative algorithm for calculating the statistical significance of a parameter deduced from a data set in statistical significance testing. A two sample t-test is generally utilized if standard deviations of the estimated parameter in both directions from some benchmark value are considered theoretically possible. The p-value, that reflects the significance level of the test, is an important function of the t-test results. Before executing the test, a critical value is chosen, traditionally 5% 30 is defined as a. If the p-value of a test is equal to or smaller than a, it suggests that the observed data are inconsistent with the null hypothesis and, thus, that hypothesis must be rejected and the alternative hypothesis accepted as true (Nuzzo, 2014). 1.8. Research Objectives This research was conducted to discover a way of utilizing facial skin temperature to control the air temperature settings of an air conditioning unit. A series of human subject experiments was conducted. Subjects' facial skin temperature was measured by skin surface temperature sensors at six locations while the air temperature was changed; in the meantime, thermal sensation questioning surveys were given to subjects to request subjects' opinions about thermal conditions. The primary goal of the proposed research is to better understand the relationship between facial skin temperature and human thermal sensation in order to develop a program to evaluate thermal comfort by reading facial temperature signal patterns and integrating them with building automation systems. The outcome program will be applied into an automatic building system (Figure 1.7.). For functionality and convenience, a user interface of the program will be taken into account. To approach the proposed goal, the objectives are as follows: 31 Human Infrared Sensor indoor air temperature changed signal thermostat with computation capability 0 ate a new set point ~I Figure I. 7. Conceptual flow diagram of the proposed control system 1.8.1. Objective I To compl9:e ~erimental systan widt sensors and cmt1rolled devices. For the experiment, the system for testing was controlled using Labvi ew. The fan and compressor of AC was controlled independently . .All the sensors and devices were connected to data acquisition (DAQ) boards according to the Labviewprogram. The DAQs are connected to the computer by USB cable. 32 1.8.2. Objective 2 To investigate the relationship between human facial skin temperature, ambient thermal conditions, and thermal sensation with thermal conditions. There are several types of bio-signals that could be considered for this research, including skin temperature and heart rate. Skin temperature and heart rate have been well proven as variables reflecting thermal sensation. For building application, the proposed research chose facial skin temperature as a physiological signal to investigate, because facial skin temperature is usually exposed and convenient to be tested. 1.8.3. Objective 3 To determine the optimal sensing and control intervals to minimize any error rate of a thermal sensation estimation while enhancing the comfort condition. Sensing and control intervals are the delay time between two executions of sensing actions or control demand. There is an unavoidable possibility of error coming out during experiments. Determining appropriate sensing and control intervals is significant in reducing the error rate. 1.8.4. Objective 4 To develop a facial temperature sensing based control logic. Estimating thermal sensation and correct set point temperature would be significantly affected by system parameters such as time interval for calculating skin temperature gradients and data sizes for analysis 33 (Choi, 2010). A program with interface was created by using Labview. The data analyses include the parameter decision to enhance the system performance. 1.9. Project scope As discussed in previous sections, thermal comfort is affected by six main factors: air temperature, relative humidity, air speed, mean radiant temperature, metabolic rate, and clothing level. However, the proposed research mainly deals with the relationship between human facial skin temperature and thermal sensation by changing air temperature. Relative humidity, air speed, mean radiant temperature, metabolic rate, and clothing level were kept constant during the experiment. The issue regarding different skin color will not be described in this thesis and the ages of the subject samples are limited to the range of 23-28 years. 34 Chapter 2 state of Art of Thermal Control Technology and Facial Skin Study 2.1. Current tlunnal control teclmology 2.1.1 Individual control teclmologies Individual control technologies have been applied to room temperature control systems, especially for rooms in hotels and residential homes. There are two main possible types of individual control systems: wired remote control and infrared remote control. Both wire remote controllers and infrared remote controllers have similar functions. According to manufacturers, the common functions are temperature setting, fan speed, operation mode, and air flow directions. Wired remote controllers are typically installed on the wall as the thermostat. Infrared remote controllers are movable devices that allow users to control HVAC systems anywhere it is convenient in the room (Daikin, 2012). The images below are examples of those two types of controllers. Figure 2.1. Wired remote controller (Daikin, 2012) Figure 2. 2. Infrared remote controller (Daikin, 2012) 35 2.1.1. Thermal comfort zone strategy In industry, there have been many efforts to develop thermal control systems, in order to reach thermal comfort. 'Thermal comfort is that condition of mind that expresses satisfaction with the thermal enviromnent' (ASHRAE-55, 2013). Since there are large variations of thermal sensation, physiologically and psychologically, from individual to individual, it is hard to compute those of every occupant in a single space. The comfort zone is consequently defined for indicating the thermal condition range acceptable for the majority of people in an indoor enviromnent. The range of operative temperature and humidity shown on the psychrometric chart in Figure 2.3 is defined by ASHRAE as a comfort zone for 80% of people, based on the PMV-PPD index program. It means that 10% of people are dissatisfied with this operative thermal condition, and 10% of people are dissatisfied with local temperature (ASHRAE-55, 2013). Researchers from Finland's National Heath Institute conducted research to evaluate the effect of the measured indoor dry ball temperature with PMV model and the subjective assessment of the indoor temperature on the SBS and thermal comfort (Jaakkola et al, 1989). They found that only a part of the employees can be satisfied by the externally controlled temperature. The greatest proportion of satisfied employees is when the indoor temperature is less than 22°C. They concluded that individual control of indoor temperature can increase the occupants' satisfaction with the indoor temperature and decrease the SBS, because a large part of the SBS was linked to the dissatisfaction with the indoor temperature. Since the PMV !PPD model is not an adaptive model, it cannot meet the requirement of individual 36 control. 10 - No lower tUndfy recommendation for grophical method: 10 15 20 - 1 I 25 RELATIVE HUMIDITY W•J 100 80 80 . . . . ~ 30 / 35 OPERATIVE TEMPERATURE (' CJ Figure 2.3. Psychrometric chrut with comf01t zone (ASHRAE-55, 2013) 2.1.2. Development of control technologies for HVAC systems ,026 .024 .022 .020 40 .016 l .016 ~ 0 ~ ' .0 14 ~to .012 .010 20 .006 .006 .004 .000 ~ Since climate and weather condition includes complex factors which vary significantly, it is not possible to design a unique system suitable for every location. The strategy is to regulate indoor thermal condition within the comf 01t zone. Depending on different aims of the thermal control system, there are two s01ts of solutions to control an indoor theimal environment: conventional control methods and computational intelligence technology (Song, Wu, & Yan 2013). Conventional control methods can be used for 37 multi-control systems, while computational intelligence technology can be used in both multi-control systems and individual control systems. This section describes the character and functions of each type of control. 2.1.2.1. Conventional control system Currently, there are different types of control systems being used for indoor envirornnental quality (IEQ) control. The table below shows the main control systems and functions. Table 2 Comparison of control systems (Song, Wu, & Yan 2013) Control Temperature TC control IAQ control Visual Energy Ventilation system control control consumption control On/Off '1 PID '1 '1 '1 Predictive c '1 '1 '1 FuzzyP '1 '1 '1 '1 '1 Fuzzy PI '1 '1 '1 '1 '1 FuzzyPID '1 '1 '1 '1 '1 NNC '1 '1 '1 Optimal '1 '1 '1 Symbol '1 denotes that the control method can significantly achieve the control objective. Symbol- means that the control method cannot significantly achieve the control objective. 38 Classical controller Thermostats are used to control the indoor temperature as a feedback system. On/off systems usually cause fluctuation of indoor temperature. Proportional-integral-derivative (PID) controllers are used to reduce the fluctuation of On/off systems; however, it is difficult to set the right parameters for a PID controller. Inappropriate setting of gain of the PID controller could result in instability (Song, Wu, & Yan 2013). Optimal, adaptive or predictable control Optimal or adaptive control systems can be used to self-control, and are able to be applied in buildings with different climate conditions. Fuzzy control is one type of adaptive control, which is considered by industry as the most promising control system. For designing optimal or adaptive control systems, a building model is necessary. A predictable control system is designed based on future factors, such as climate change, and load schedule. Because of the difficulty in mathematical analysis, a predictable control system is difficult to design, therefore, a predictable control system is not practical (Song, Wu, & Yan 2013). 2.1.2.2. Computational intelligent control/individual control Computational intelligent control technology is technology that combines fuzzy control with neural networks technology and evolutionary algorithms. With the development of this field of technology, it can 39 be applied to building systems (Song, Wu, & Yan 2013). Fuzzy system and evolutionary algorithm In order to provide thermal comfort and reduce energy consumption, occupants' preferences must be taken into consideration. Fuzzy control is used for this purpose. Fuzzy control and neural technology for HV AC systems have been investigated and discussed by many organizations. The different conditions of buildings, such as weather and orientation, make it impossible to use a unique model to achieve best performance and lowest energy use. However, an intelligent system with fuzzy control is not based on any pre-cast building model, but use adaptive controllers which can keep reducing errors based on actual thermal condition and energy use feedback. It has a great adaption so that it can fix and avoid the impact of weather and orientation. Fuzzy control has been used in a new generation of furnaces for improved performance and reduced energy use for private home heating (Altrock et al, 1994). Synergistic neuro-fuzzy techniques A neuro-fuzzy system is a fuzzy system is in which neuron network technologies are applied. It is an adaptive system with predictable function. A perfect neuro-fuzzy system can reduce energy use and get a high level of indoor comfort by prediction through the human neuron network system. A neural controller with prediction ability has been applied to some hydronic heating control systems and solar powered buildings (Paassen, 1990). 40 Kanarachos and Geramanis (1998) developed and tested an adaptive neural network controller. This controller can be used to control single-zone hydronic heating systems. The parameters of the controller are determined based on the heating device and the set point temperature (Kanarachos & Geramanis, 1998). Nevertheless, the performance of this control system is not good enough, because no prediction of either weather condition or indoor thermal conditions was considered in the control process (Song, Wu, & Yan 2013). To overcome this problem, Yamada and colleagues (1999) developed an algorithm for air-conditioning control that combines neural networks, fuzzy systems, and predictive control technology. Their developed control system contained the prediction of weather conditions and the occupant load, then estimated the building performance of this system, demonstrating that it could guarantee satisfaction about thermal comfort and achieve energy reduction (Yamada et al, 1999). PID-like fuzzy controllers A structure of fuzzy PID controllers is shown in Figure 2.4. Depending on their structures, there are two major types of PID fuzzy controllers: PID-like fuzzy controllers and PID fuzzy controllers (Hu, Mann, & Gosine, 2001). The first type of fuzzy PID controller, the PID-like fuzzy controller, consists of typical FLCs. The reason why it is called a PID-like fuzzy controller is to differentiate it from the second type of PID fuzzy controller. Most current research is about PID-like fuzzy controllers (Song, Wu, & Yan 2013). The second type of fuzzy PID controller combines conventional PID controllers with fuzzy logic. PID controllers are widely used in current industry as feedback loop controllers. These control a system by 41 reducing errors based on the difference between desired output and actual output by utilizing three factors: proportional, integral, and derivative (Nice, 2001). The transfer function of the PID controller is as follows: (5) Where Kp is the propositional gain, K 1 is the integral gain, and Kn is the derivative gain. The fuzzy logic makes PID gains (Kp, Ki, Kn) dynamic according to output from indoor thermal conditions. The main weakness of this type of system is that it is dependent on the model. In order to define the range of the proportional gain, a model with system and human behaviors must be built (Song, Wu, & Yan 2013). The system can regulate by sensing the difference between set point and sensing point, and measure the change ing this difference. Based on these data, the system can define the variables of buildings. A FLC can apply nonlinear control strategies. For example, if the comfort condition, PMV, is 'cold', the increment will be strong, regardless of its tendency, but if the PMV error is small, the tendency is taken into consideration (Song, Wu, & Yan 2013). 42 Set-point --- + Figure 2.4. Structure of PID fuzzy controller Output f----Building Environment f--~--+ 6u + 2.2. The role skin temperature plays in human thermoregulation The1moregulation is a function of the human body to maintain thermal homeostasis. In order to achieve health. a small range of core temperature, 36.8°C ± 0.4°C, must be maintained, otherwise, the human might become ill. To maintain a constant core body temperature, the heat produced by the body must be transported to the ski.n's surface through the blood and released into the ambient environment through conduction. convection, and radiation. Human skin temperature regulation is significant to the maintenance of n01mal core body temperature dming varying environmental conditions. AB shown in Figure 2.5, there are two layers in the human skin: epide1mis and dermis. In the dermis layer, the blood flows from arteriole to venule via capillaries (Carroll, 2007). The structure of capillruies is ve1y similar to a mechanical heat exchanger. According to the heat transfer formula, regru-dless of whether it is conduction. convection. or radiation. the higher tempernture difference results in a higher heat transfer 43 rate; therefore, in a normal temperature envirornnent, 20° Celsius to 30° Celsius, a relatively higher skin temperature accelerates heat loss and a relatively lower skin temperature decreases heat loss. Skin temperature is controlled by blood flow and blood flow is controlled by the sympathetic neural system, which includes the noradrenergic vasoconstrictor system and a sympathetic active vasodilator system. The sympathetic active vasodilator system is responsible for 80% to 90% of the substantial skin vasodilation that occurs with whole body heat stress. With body heating, the magnitude of skin vasodilation dramatically increases: skin blood flow speed can reach 6 to 8 L/min during hyperthermia; as a result, skin temperature can become very high. With body cooling, the magnitude of skin vasodilation decreases; consequently, skin temperature would be at a low level (Charkoudian, 2003). 44 Epidennis-{ Oermis---4 Sweat gland _ ____;...,.. .......... " Venule---- Air ' ' 11o--Arteriole 7.l'JZ7Mli---~i+-- Arteriovenous anastomosis Figure 2.5. Human facial skin dissection and blood flow (Canoll, 2007) 2.3. Review of studies on human thermal sensation Sympathetic activity (vasoconstriction) 2.3.1. An automobile HVAC system with a neural nehvork for controlling thermal sensations ReseaTchers from Toyota created a HVAC system for automobiles to control the interior the1mal environment based on human thermal sensation. The thennal sensation level in this study was estimated by environment conditions and facial skin temperature. Initial facial skin temperature was calculated from exte1ior temperature and solar radiation. Facial skin temperature change was calculated by neural network and the thermal sensation of a passenger was calculated (UEDA et al, 1997). This study is a good example 45 of utilizing network and facial skin temperature for controlling space temperature. However, the system does not keep sensing facial skin temperature. Instead, it utilized facial skin temperature change which is generated by calculation. 2.3.2. A study on regional differences in tern perature sensation and thermal comfort in humans Researchers from Waseda University in Japan investigated regional differences (face, chest, abdomen, and thigh) in temperature-related sensations with special attention to thermal comfort. Subjects entered a climatic chamber that was maintained at 32.5 ±0.5°C (SE) with a relative humidity of 50%. They gave local cooling and heating water-perfused stirnulators after I. 5 hours. Core temperature and mean skin temperature was tested. Temperature sensation and thermal comfort of the stimulated area, and whole body thermal comfort were reported by the subjects (heat exposure). Then, they set the chamber at 21.3±0.l°C and used the same methodology to do the test again (cold exposure) (Nakamura et al, 2008). Finally, the researchers discovered that during heat exposure, facial cooling and heating had the most effect on thermal comfort. But during cold exposure, facial cooling and heating had the least effect on thermal comfort. This study indicates the difference of the regional parts of body thermal sensation. This study is significant because it differentiates the facial area from other body parts. The researchers only investigated the sensation of thermal stimulus to different regions, but the study does not conclude the relationship between thermal sensations and skin or body temperature. 46 2.3.3. Effects of temperature steps on human skin physiology and thermal sensation response Researchers from the National United University in Taiwan discovered the relationship between human physiological response and temperature step when entering an air conditioned room from outside. Twin-climate-controlled chambers were used to create two distinct zones of a!f temperature. One represented the ambient outside environment and the other represented an air-conditioned room (temperature constant at 24 'C). The outside chamber had the temperature steps 20'C, 28'C, 32'C. The temperature transient was replicated when test subjects moved from the outdoor to the indoor chamber. At first, subjects stayed in the outside chamber. After acclimation, thermal sensation vote (TSV) and skin physiology was evaluated, then subjects moved from the outside chamber to the inside chamber. TSV and physiological properties were evaluated at 1, 2, 4, 6, 10, and 20 minutes after entering. Finally, they found that when the temperature step in thermal transition zone was less than 4 'C, thermoregulation was controlled (Chen et al, 2011). This study aimed to determine how many degrees of ambient temperature change are acceptable for human thermoregulation. It investigated thermal sensation and skin temperature when applying an immediate indoor temperature change. The relationship of skin temperature and thermal sensation was not considered in this study. 2.3.4. Thermal sensation and comfort in transient non-uniform thermal environments Researchers from UC Berkeley developed predictive models of local and overall thermal sensation and comfort. At first, they asked subjects to be in a temperature-controlled chamber. Then, they used air-sleeves to attach to one of each subject's body segments (head, face, neck, chest, back, pelvis, arm, leg, 47 hand, and foot) to apply local cooling or heating. The researchers used thermocouples to measure the skin temperature of subjects during experiments. Also, the investigators used an ingestible temperature device to measure core temperature. Researchers collected subjective perceptions (9-point analog scales) of overall and local thermal sensation and comfort (Zhang et al, 2004). According to the data collected, researchers made four models: local thermal sensation model, local thermal comfort model, overall thermal sensation model, and an overall comfort model. Finally, investigators found that the influence of local sensation on overall sensation was different for different body parts. The back and the chest were dominant in influencing overall sensation. But, for hands and feet, the impact of local sensation is much less (Zhang et al, 2004). This project mainly studied the influence of local heating and cooling to overall sensation. Researchers found that humans are sensitive to back and chest local heating and cooling. The project studied how human sensation responds to local temperature change, not how human regional temperature responds to human sensation. 48 Chapter 3 Project Methodology In this research, a thermal control system program was developed using Labview. A series of human subject experiments were conducted in a chamber in the basement of Watt Hall, University of Southern California. Subjects' facial temperatures and environmental conditions were measured during human subject experiments, and the data were analyzed to discover relationships between overall thermal sensation and facial skin temperature. Figure 3 .1 Research flow diagram 49 Lab view pro grams for AC and heaters controlling ambient environment measurement and facial temperatw"e measw-ement were developed for human subject experiments. Based on the data, results of experiments and the control system pro gram were completed as an automatic system. 3.1 Eiqieriment resources 3.1.1 Environmental chamber For the human subject experiment measuring facial temperature, a thermal chamber in the basement of Watt Hall atthe University of Southern California was used. The chamber is 4.5 meters long, 2.9 meters wide, and 2.4 meters high. The plan of the chamber with dimensions is shown in Figw-e 3.2. C7> c:i 4.5 Figure 3. 2. Plan of the chamber The chamber is a concrete-enclosed room. The wall sw-face was covered by foam thermal insulation 50 board. The chamber was surrounded by conditioned space. The chamber was separated by a thermal insulation wall with a window into two parts: a front room and a monitor room. The front room was used as a private office for one subject doing computer-based work. A fish-eye photograph was taken and is shown in Figure 3 .3. Figure 3.3. A fish-eye perspective of the chamber 3.1.2 Heating and cooling system The temperature of the chamber was regulated by a window air conditioning unit (shown in Figure 3.4) with the capacity of 6,000 BTU and two portable heaters (shown in Figure 3.5) at 1000 Watts each. 51 Figure 3.4. Window AC unit Figure 3.5. Portable heater The window AC unit and portable heaters were installed in the monitor room, shown in Figure 3.6, and were connected with flexible ducts. The conditioned air was supplied through four flexible ducts, shown in Figure 3.7, into the front room and connected with four diffusers, shown in Figure 3.8, which were fixed symmetrically on two sides of the desk. Figure 3.6 AC and heaters Figure 3.7 Flexible ducts Two flexible ducts supplied cooling air and two supplied heating air. 52 Figure 3 .8. Diffusers The window AC was hacked and connected to a wire board, shown in Figure 3.9; as a result, the fan speed and compressor could be controlled independently through the computer. The heaters were plugged into a socket, shown in Figure 3 .10 and the socket was connected with a data acquisition so that it could be controlled by the Labview program. Figure 3 .9. Wire board Figure 3 .1 0. Socket 53 3.1.3 Data acquisition tools 3.1.3.1 Objective data acquisition For data acquisition for human facial temperature and environmental conditions, several types of sensors and data acquisition boards were used. Six units of skin surface temperature sensors (Figure 3 .11.) were connected to two data acquisition (DAQ) boards for the measurement of surface temperature at six specific points on the human face. The DAQ board was put into a waist bag (Figure 3.12.) in order for a subject to cany it easily while wearing the sensors . • . ' Fi gure 3.11. Skin surface temperature sensors Figure 3.12. Waist bag with DAQ board An infrared sensor (Figure 3.13)was used to sense the average temperature of the human facial area The infrared sensor was installed at the top of a computer monitor so it was able to continuously sense the subject's facial temperature when the subject was working with the computer. 54 'w Figme 3.13. Infrared sensor Figw·e 3.14. Tupod with environmental sensors All of the sensors for ambient environment measurement, including four air temperatme sensors, one mean radiant temperatme sensor, one C02 sensor, and one relative humidity sensor, were installed on a tripod which is shown in Figure 3.14. The air temperature sensors were installed at the heights of O.lm, 0.6m, 1.lm, and 1.6m from the floor: The height of air temperatme sensors was suggested by ASHRAE 55. The C0 2 and relative humidity sensors were installed at the height of 1.lm which was close to the subject's breathing level when the subject was sitting. The mean radiant temperature sensor was installed at 0.6m height which was suggested by ASHRAE-55 considering the sitting activity level (1.2 Met) required in the experiments (ASHRAE-55, 2013). 55 All of the data acquisition devices are summarized in Table 3. Table 3. List of data acquisition devices Data co llection device for facial temperature measurement Resourcesfrools Purpose Figure Specification Surface temperature Skin Range: -2 5°C to sensor temperature 125°C (Thermocouple) Resolution: 0.03°C measurement (Model STS-BTA) Accuracy: ±0. 2°C Data acquisition Data Three 13 bi~ single board (Model acquisition ended Sensor DAQ) from temp erature analog ..___.,.. channels . sensors 1 V.rnfer One digital sensor ~ot,9=@' channel. • I ] Two general-purpose analog input channels: three 1 3 bits single ended, 14 bits differential Infrared temperature Facial IR Range: -18 to 202C sensor temperature (0 to 400F) Model: measurement Accuracy: 3% os136-1-mv-f 56 Data collection device for environmental condition measurement Resources!fools Temperature sensor (Model: LM35DT) C02 sensor (Model: Telarire6004) Humidity sensor (Model: HIH-4000- 003) Data acquisition board (Model: NI DAQ USB-6008 & 6009) Purpose Air, chamber surface and mean radiant temperature measurements C02 density measurement in the chamber Relative humidity measurement Data acquisition From environmental sensors Figure . ~ ... - ~ -- --~· Specification Range: -55°C to 150°C Resolution: 0.01°C Accuracy: ±0.5°C Range: 0 to 2000 ppm Accuracy: ±40ppm Range: 0 to 100% Resolution:0.5% Accuracy:±3 .5% 8 analog inputs (12-bit, 1 OkS/s) 2 analog output (12-bit, 150 S/s) 12 digital 1/0, 32-bit counter 57 Heating and cooling device Resources I Tools Purpose Portable heater Heating for (model: ceramic? 54 200) Air conditioner (Model: Frigidaire FAA062P7A) the chamber Cooling for the chamber Figure .. Software for sensing, control and data analysis Resources/Tools Purpose Figure Lab VIEW 2014 Data acquisition ElLabVIEw· 2014 from sensors and control logic design Minitab 17 Data analysis m Minitab·17 Specification Two units used (1,000 watt each) Capacity: 6,000 BTU Specification Developer: National Instrument Developer: Minitab Inc Labview 2014 was used to collect and monitor the data from measurements, AC, and heater control. The intetface of the program for facial temperature sensing is shown in Figure 3.15. Based on a sensing interval of 1 second, the interface continuously displays temperature data from six points on the human 58 face. Point 1 Poiht 2 Point3 ro- J :n P.oint 1 ro] 1 0 :: 40- 201 2:iJ P.oint 2 1 0 0- 0- Point4 Point 5 w :J ro- j Point 3 0 40- 40- 20~ 20~ Poipt'4 o~ 0- 1 0 Ppint 6 . ro- j :: Point 5 0 P.oint 6 0 - 1 0 Figure 3.15. futeiface of facial tempei·ature sensing (designed with Labview) For ambient environmental conditions sensing, the program, as shown in Figure 3.1.15, can show and monitor air tempei·ature, relative humidity, C0 2 level, and mean radiant temperature. Air 1.63 Air 1.62 0 Alrl.13 Air 1.12 0 Air0.63 Airo.6 2 0 A ir0.13 60- 40: 20; MRT 2. o; MRT4 0 COZ ln/ out (ppm) 2 C02 In/out 2 0 Figure 3.1.16. futeiface of ambient conditions sensing (designed with Labview) R elative Humidity In/Out (%) 2 20 40 60 80 0 \ \ t , , 00 , , RH3 0 59 3.1.3.2 Subjective data acquisition Human subject recruitment For the human subject testing, voluntary subjects were recruited from students at the University of Southern California (USC) with the help of the Building Science Faculty. Questionnaire for thermal sensation and comfort To collect thermal sensations from subjects during changes in enviromnental conditions, the seven-point scale survey developed for the PMV model (ASHRAE-55, 2013) was used as shown in Table 2. Table 4. Thermal sensation and comfort survey I. What is your overall level of thermal comfort? Very Unsatisfied Slightly Neural Slightly Satisfied Very unsatisfied unsatisfied satisfied satisfied D D D D D D D 2. What is your overall thermal sensation? Very cool Slightly Neural Slightly warm Very Cool cool warm warm D D D D D D D 60 Experimental conditions of human factors In order to get stable experimental conditions, a tutorial was given to subjects. It required subjects not to have food for at least thirty minutes before the experiment. Subjects were asked to do computer based work, such as reading or typing, and they were asked not to move during the testing. 3.2 Expe1iment procedure The human subject experiment tested 25 human subjects. Human subjects were recruited from USC students, aged from 20-30 years. The experiment method was reviewed and approved by the University of Southern California Institutional Review Board (IRB). The process diagram of the experiment is shown below: 15min 20 C Adjusting Survey 20C I Start Survey 2o · c End Survey 22C I Survey 24 · c I Survey 26C I Survey 28C I Survey 30C Heating Process -------- -- - --------------------------------- -- ----- -> - I 10min Survey 22C I 20min Survey 24C I 30min Survey 26C I 40min Survey 2s · c 50min Survey 30C <(- _________ ___ ___ _ Q99~1J9 fr99~_s_s ____________ _______ ~ _ _ __ __ , 90min 80min 70min 60min 50min Figure 3.17. Process diagram of the experiment 61 Each experiment lasted around two hours: one hour for the heating process and one hour for the cooling process. The air temperature range of the chamber during experiments was from 20°C to 30°C. The sensors for ambient envirornnent condition measurement, skin surface sensors, and infrared sensor kept sensing and collect data with an interval of I 0 seconds. The data were input to an Excel spreadsheet. Before testing, the air temperature of the chamber was regulated to 20°C and the subject was asked to enter in the chamber. After 15 minutes adjustment and giving a short tutorial to the subject, the experiment started. The tutorial suggested the preferable pose and required the subject to try not to move during the experiment. When the test started, the heaters started working and were controlled by the interface of the Labview program, to guarantee the air temperature increased smoothly to 30°C for one hour. The air temperature sensor installed at the height of I. Im on the tripod was monitored as the reference indoor air temperature. When the air temperature was at 20°C, 22°C, 24°C, 26°C, 28°C, and 30°C, the questionnaires of thermal sensation and thermal comfort were given to the human subject and an infrared photograph of the human facial area was taken by an infrared camera. After the subject finished the survey at 30°C, the heaters were turned off, and the compressor and fan of the AC were turned on. The air temperature of the chamber was regulated back to 20°C smoothly for one hour. When the air temperature was at 30°C, 28°C, 26°C, 24 °C, 22°C, and 20°C, the questionnaires of thermal sensation and thermal comfort were given to the human subject, and an infrared photograph of the human facial area was taken by an infrared camera for validation. Photographs of a subject wearing the sensing devices during the experiment are shown below: 62 \ . \ -- Figures 3 .18 and 3 .19. Photographs of a subject wearing sensing devices 63 Chapter 4 Human Subjects Experiment Data and Results Twenty subjects participated in the experiment. Subjects' information is provided in an appendix. During human subject experiments, six different points of facial skin temperature were collected with sensing intervals of 10 seconds. For ambient environmental conditions, dry indoor bulb temperature at the heights of O. lm, 0.6m, l. lm, l.6m, C0 2 level, mean radiant temperature, and relative humidity, were measured with the sensing intervals of 10 seconds. For ambient temperature, both dry bulb temperature and operative temperature were considered as environmental components. The air dry bulb temperature was calculated by the following formula: J',;ry = (TO lm + 1a Om + J; lm + J; Om) / 4 (6) Where Truy is air dry bulb temperature, To.im is the temperature sensed value from the sensor at O.lm above the floor, To.om is the temperature sensed value from the sensor at 0.6m above the floor, Tum is the temperature sensed value from the sensor at l. lm above the floor, and Ti.om is the temperature sensed value from the sensor at l.6m above the floor. Operative temperature was calculated as the mean value of the air dry bulb temperature and the MRT. The calculation formula is given below: ~ = (Tdry + MRT) I 2 (7) Where T 0 is operative temperature and T c1ry is the air dry bulb temperature. Overall, facial skin temperature was calculated using the following formula: 64 J;,kin (8) Where T skin is overall facial skin temperature, Tl, T 2, T 3, T 4, Ts, and T 6 are temperature values collected from six points on subjects' faces. The thermal sensation survey was given to the subjects during experiments. Subjects answered the question regarding thermal sensation at every two degree Celsius dry bulb temperature change. It was assumed that within five minutes before and after the subject answered the question, the subject had nearly the same thermal sensation. For thermal sensation data, 3 means very warm, -3 means very cool. The time-series plot data of dry bulb air temperature, operative temperature, and overall facial skin temperature, and thermal sensation for each subject is shown below: Subject#l: 36 ll~ 1 8 30 60 Time (minutes} 90 120 V¥ ..... --- Td<y - • - To - .-+- · Tskin . , " 30 60 90 Time (minutes) Figure 4.1. Time-series plot of T dry, T 0 , and Tskin, and thermal sensation for subject# 1 Subject#2: 120 65 36 21 18 Subject#3: ,. E 10 ~ ~ 27 c. E {!!. 24 21 18 Subject#4: ,. 21 " Subject#5: - -3 30 60 90 120 30 60 Time (minutes) Time (minutes) Figure 4.2. Time-series plot ofT dry, T 0, and Tskin, and thermal sensation for subject#2 -1 _ , 30 60 90 120 30 60 90 Time (minutes) Time {minutes) Figure 4.3. Time-series plot of T dry, T 0 , and T skin, and thermal sensation for subject#3 rMs • 30 60 90 Time (minutes) 120 v .. , ..., - • - l o - +- T $1dn c 0 ·~ 0 ~ ~ -1 _ , _J 30 60 90 Time {minutes) Figure 4.4. Time-series plot ofT dry, T 0, and Tskin, and thermal sensation for subject#4 120 120 120 66 3• .. __ . ..,,_-........ ""'_,,.,.. ... _.~ ........ 33 ...... .._ ~,,,,.,.,. ·1 21 -2 18 _, 30 .. 90 120 lO .. Time (minutes) Time (minutes) Figure 4.5. Time-series plot of T dry, T 0 , and T skin, and thermal sensation for subject#5 Subject#6: 36 " ~·--•11114 _____ ...., ... ~ • .--~ , • ~ 30 ~ .. 27 ~ E ~ 24 21 18 30 .. 90 Time (minutes) 120 v ...... -+-'"" - • - To -+ , .... -2 · 3 30 .. 90 Time (minutes) Figure 4.6. Time-series plot ofT dry, T 0, and T skin, and thermal sensation for subject#6 Subject#7: ,. " ~·-~., ......... - ......... llW ...... ~.._ ...... ~_.... ..... ,.. ... E 10 ~ " ~ 27 8. E ~ 24 21 18 JO .. 90 120 Time {minutes) ·1 ·2 . , 30 .. 90 Time (minutes) Figure 4.7. Time-series plot ofT dry, T 0, and T skin, and thermal sensation for subject#7 Subject#8: 120 120 120 67 ,. ,. ,.,. .... ,... .... ..__,,. ................ ..-.. ......... ... " .,,,, $ ~ ~ 27 8. E ~ 24 21 18 30 .. .. 120 Time (mintues) v..-. ~,..,. - · - To ... '""" -1 -2 - _ , 30 .. .. Time (minutes) Figure 4.8. Time-series plot ofT dry, T 0, and Tskin, and thermal sensation for subject#8 Subject#9: ,. _,, ....... ...,.----""'-ft',.. ... '"7$_ .. ,. .. ~-- 4 "~ ..... . e: 30 ~ ~ 27 ~ E ~ 24 21 18 .. .. Time (minutes) • 120 ._._ ~ ,..,. - · - T o . ..... . Tdlln -2 · 3 30 .. .. Time (minutes) Figure 4.9. Time-series plot of T dry, T 0 , and T skin, and thermal sensation for subject#9 Subject# 10: ,. " ~,.,...._. ............................ .,. ....... . 21 18 30 .. .. 120 Time (minutes) v.. ~ ·..,. - · - fg • ... T.icln c 0 ·~ 0 ~ -1 _ , _, 30 .. .. Time (minutes) Figure 4.10. Time-series plot ofTdry, T 0, and Ts kin, and thermal sensation for subject# lO Subject#ll: 120 120 120 68 " 18 Subject#12: 36 " _, E 30 ~ ~ i! " 8- E ~ 24 21 11 Subject#13: 36 33 21 1 1 Subject#l 4: 30 .. Time (minutes) .. 120 ·· ___ , .., - • - T o . '""" _ , ·2 .3 30 60 90 Time (minutes) Figure 4.11. Time-series plot ofTdry, T °'and Ts kin, and thermal sensation for subject#ll ___ ,.., - - • - T o 4 • , .... c 0 ·~ 0 ~ ~ . f -2 . 3 30 .. .. 120 30 .. 90 Time (minutes) Time (minutes) Figure 4.12. Time-series plot ofTdry, T °'and Tskin, and thermal sensation for subject#12 30 .. 90 Time (minutes) 120 · -+-'"" -• ~ To · + - ,.,. ·2 .3 30 .. .. Time (minutes) Figure 4.13. Time-series plot of Tdry, T °'and Ts kin, and thermal sensation for subject# 13 120 120 120 69 36 ll~ '" ,., 21 18 30 60 90 Time (minutes) 120 v . ____ ,..,. - a -To - + - TIUtl ·3 30 60 90 n me {minutes) Figure 4.14. Time-series plot ofTdry, T 0, and Tskin, and thermal sensation for subject#14 Subject# 15: ,. G 10 '-- ~ ~ 27 Cl. E t!!. 24 21 18 30 60 Time {minutes) 90 1 20 '""' -· - To --+- TKin -2 · 3 30 60 90 Time (minutes) Figure 4.15. Time-series plot ofTdry, T 0, and Tskin, and thermal sensation for subj ect#15 Subject#16: ,. ll ____ .. ,,,..---------------· ~ 10 ~ ~ 27 Cl. E i! 24 21 18 30 60 90 120 Time (minutes) c 0 ·~ 0 & ·1 · 2 . 3 30 60 90 Time (minutes) Figure 4.16. Time-series plot ofTdry, T 0, and Ts kin, and thermal sensation for subject#16 Subject#l 7: L 120 120 120 70 v ...... ~ ·"" - · -To --+ - lskln · 1 _J · 2_J ·l 10 60 Time (minutes) Figure 4.17. Time-series plot ofTc1ry, Ta, and Tskin, and thermal sensation for subject#l 7 Subject#l8: 16 ))~ G 10 ... !! " ~ 27 g_ E {!:. 24 21 11 ....... lO 60 Time (minutes) 90 120 v. __._ ldry - • - lo - -+-- '"°" -2 ·l 10 60 90 Time (minutes) Figure 4.18. Time-series plot ofTc1ry, Ta, and Tskin, and thermal sensation for subject#l8 v. . .., -• ~ lo - + - l*in c 0 ·~ 0 l£ :li ·1 ·l 10 60 90 Time (minutes) Figure 4.19. Time-series plot ofTc1ry, Ta, and Tskin, and thermal sensation for subject#l9 Subject#20: 120 120 120 71 36 ~ v.i..,. _._ r.., - • -To / --+- Tskin 33 30 27 24 ·1 -2 Z1 -3 18 30 60 .. 120 30 60 .. 120 Time (minutes) Time (minutes) Figure 4.20. Time-series plot ofTdry, T 0 , and Tskin, and thermal sensation for subject#20 According to the test results, different subjects generate different facial skin temperatures in their neutral sensations and other sensations. Because facial skin temperatures change based on the human thermoregulation function, individual psychological status and physiological characteristics such as gender and age can generate temperature differences between subjects (Griefahn et al, 2000). The detailed data analysis will be discussed in Chapter 5. 72 Chapter 5 Interaction of Facial Temperature, Ambient Temperature, and Thermal Sensation 5.1 Facial skin sensing location selection for thermal sensation estimation Six different locations in the facial area were sensed during the human subject expeiiments. These six locations are defined as: Point 1 (forehead), Point 2 (upper rim of eye), Point 3 (bottom rim of eye), Point 4 (upper ROI area), Point 5 (bottom ROI area), and Point 6 (the lower jaw). The location of those six sensed points is shown in Figure 5 .1 below. point 6 Figure 5 .1. Location of six sensed points The tempei·ature change at each location presented similar regulation. The facial tempei·ature increased with the air temperature and the1mal sensation score increased. However, the data showed that each location presented cliffei·ent sensitivity and reaction time. The number of skin temperature sensors and 73 their locations are important to the practical application of predicting human thermal sensation. The facial skin temperatures from each sensed facial location were analyzed to discover the location where skin temperature had the most potential to predict overall thermal sensation. The data mining was conducted based on two skin temperature parameters: the absolute temperature and the gradient. Skin temperatures measured at six sensed locations and overall facial temperatures (mean value of data from six locations) were compared. The sampled time series plot of all sensed locations and overall facial skin temperature is shown in Figure 5.2. The similar patterns between those temperatures and the air dry bulb temperature indicate that the selected skin locations seem to be affected by their ambient conditions but with different sensitivity, since the change rate and range are different. 35 G 0 - 34 ~ :J ...., ~ <11 a.. 33 E ~ 32 0 30 60 90 120 Time (minutes) Variable ~ Ponit1 - • - Ponit2 --+ -· Ponit3 ~- Ponit4 -II>- Points ~ Point6 - y - Overal Tskin Figure 5.2. Patterns of skin temperatures at 10 selected body locations as air temperature (Subject#!) 74 5.1.1 Absolute skin temperature in heating and cooling process In order to discover the overall interaction of skin temperature at each location, box plot diagrams of skin temperatures at sensation were generated. In each box plot chart, there is one box plot image above each sensation. The box plot image consists of a box and two vertical lines, one is on the top of the box and another is above the bottom of the box. The vertical lines indicate the range of the data, which means that the end points stand for the maximum and the minimum temperature data. The box stands for the 95% confidence interval. The 95% confidence interval indicates that there is a 95% possibility the mean value is within the proposed range at each sensation. The horizontal line inside the box indicates the median temperature and the point connected by the tendency line indicates the mean value. The box plot charts for each facial skin sensed location and overall facial skin temperature in the heating and cooling process are shown as follows: " l7 16 16 0 ~JS il ~ 8- 34 E ,! t: " 'E 0 ... 0 r + ~ 35 il -+- mp ~ ~ 34 E ,! ~ ll g_ l2 " " -l ·2 -t " -l -2 -1 0 sensation sensation Figures 5.3. and 5.4. Box plot of skin temperature at point #1 in heating process (left) and cooling process (right) 75 37 37 " " G G ~ 35 L " ~ a a ~ ~ 8- 34 8- 34 E E !!- !!- N " ~ 33 ·E ·c g_ 0 0.. " " " " -3 -2 ., -3 -2 -1 sensation sensation Figures 5.5. and 5.6. Box plot of skin temperature at point #2 in heating process (left) and cooling process (right) 37 37 ,. ,. 2 ~ 35 ~ ~ 34 E ~ ~ 33 ·c g_ " " " " -3 -2 -1 -l -2 -1 sensation sensation Figures 5.7. and 5.8. Box plot of skin temperature at point #3 in heating process (left) and cooling process (right) 37 36 G ~ 35 ~ ~ 34 E ~ ... ·E 11 g_ 32 " -3 -2 -1 37 36 2: ~ 35 ~ ~ 34 E ~ :;!; ·c 33 0 0.. 32 " _ , - 2 -1 sensatio n sensation Figures 5.9. and 5.10. Box plot of skin temperature at point #4 m heating process (left) and cooling process (right) 76 l7 l7 " " 9 E • ; lS ~ JS I ' z 3 + .. .. ~ ~ ~ '4 8. 34 E E ...... ~ ~ I ~ ~ 1: 33 1: 3) ·c; ~ .. " 32 l1 l1 -l -2 -t 0 - l -2 " sensation sensation Figures 5.11. and 5.12. Box plot of skin temperature at point #5 in heating process (left) and cooling process (right) 37 " ~ ~ J S 3 ~ 8. 34 E ~ "' ll ~ " l1 ·l -2 -t 0 sensation l7 " 9 ; 35 ~ l! ~ 34 E ~ "' 1' ll ·c; .. 32 l1 -l -2 0 sensation Figures 5.13. and 5.14. Box plot of skin temperature at point #6 in heating process (left) and cooling process (right) 37 l7 - " u .... - " u .... ~ ~ a n ~ • .a lS ~ ' I ~ a. .,. a. ~ 34 ; 34 ..... ' .... ;;; ;;; ·o ·o ::. ll ~ ll "E .,. "E ~ 32 ~ 32 - - + - l1 l1 ·l -2 " _ , ·2 · 1 0 sensation sensation Figures 5.15. and 5.16. Box plot of average facial skin temperature in heating process (left) and cooling process (right) According to the patterns above, temperature data collected from point#5 is the most stable and accurate 77 sensitivity index to mirror ambient condition. The skin temperature collected from point#5 increases with the thermal sensations, changing from cold (-3) to hot (3) sensations both in the heating process and cooling process. It indicates that subjects present a high facial temperature with warm sensation and relatively low temperature with cool sensation. However, even though the ranges of skin temperatures are increasing, there is no statistical significance between 95% confidence intervals in the heating process between neutral (0) and slightly warm (1) sensation, between slightly warm(+ 1) and warm (+2) sensation, and between warm (+2) and hot (+3) sensation; and in the cooling process, there is no statistical significance between 95% confidence intervals of cool (-2) and slightly cool (-1) sensation, between slightly cool (-1) and neutral (0) sensation, between neutral (0) and slightly warm ( + 1) sensation, between slightly warm(+ 1) and warm (+2) sensation, and between warm (+2) and hot (+3) sensation. It means that the intervals are widely different depending on individual characteristics. Table 2 shown below concludes the two sample t-test results of facial skin temperatures between the neutral and slightly warm, and between neutral and slightly cool sensation for six locations and average facial temperature. Because the confidence interval of the skin temperature in each sensation is too wide, the t-test P-value results do not show any statistical significance. It means that the absolute facial skin temperature in each sensation is different depending on the different people. 78 TABLE 6. Absolute levels of facial skin temperature of each location in each thermal sensation score Slightly cool Neutral (n~20) Slightly warm P-value P-value (n~20) (n~20) (neutral vs. (neutral vs. slightly cool) slightly warm) Point#l 35.55 35.87 35.99 0.301 0.433 Point#2 35.33 35.56 35.86 0.435 0.344 Point#3 34.99 35.22 35.77 0.334 0.297 Point#4 34.57 34.89 35.01 0.471 0.312 Point#5 34.04 34.77 34.96 0.125 0.201 Point#6 33.68 33.75 33.98 0.621 0.433 Average skin 34.38 34.92 35.01 0.574 0.361 temperature Consequently, the absolute facial skin temperature is not accurate enough to indicate subjects' thermal sensation independently. In order to better understand the relationship between facial skin temperature and thermal sensation, the gradient of facial skin temperature will be discussed. However, it is worth mentioning that Point#5 has the lowest P-value for both neutral vs. slightly cool and neutral vs. slightly warm. It implies that Point#5 has the most statistically significant temperature data. In order to get a better control advice for this project, the combination of all cool sensations, including 79 cold (-3), cool (-2), and slightly cool (-3); and combination of warm sensations, including slightly warm (+I), warm (+2), and hot (+3) were conducted. Since the purpose was to differentiate the neutral sensation, the combined sensations were analyzed by two sample t-tests. The two sample t-test results of facial skin temperatures between the neutral sensation and combined warm sensation, and between neutral sensation and combined cool sensation for six locations and average facial temperature are shown in Table 3 below. TABLE 7. Absolute levels of facial skin temperature of each location in each thermal sensation score Cool Neutral Warm P-value (neutral P-value (neutral vs. (n~20) (n~20) (n~20) vs. cool) warm) Point#! 34.13 34.76 34.89 0.095 0.352 Point#2 34.26 34.46 34.86 0.375 0.312 Point#3 33.56 34.43 34.83 0.268 0.237 Point#4 33.53 33.96 34.02 0.173 0.294 Point#5 33.11 33.74 33.97 0.063 0.036 Point#6 32.89 33.10 33.67 0.423 0.241 Average skin 33.27 33.98 34.12 0.363 0.146 temperature According to the table above, every P-value decreased after combination, but the only significant value appeared at Point#5 comparing neutral and warm sensation. Therefore, absolute facial temperature is not accurate enough to indicate human sensation. 80 5.1.2 Gradient of skin temperature in heating and cooling process For a more detailed analysis, the gradient of the skin temperature is discussed as a parameter of the facial temperature index. According to the box plots charts shown previously (from Figures 5.3 to 5.14), the skin temperatures show an overall incremental tendency from cool to warm sensation. The gradient of temperature is larger if the rate of change is calculated with a larger time interval. Five minutes was selected to be the gradient interval as the temperature gradient is about 1°C. The equation for calculating facial temperature gradient is shown below: Faci a l temperature gradient = (T 2 - T 1 ) per 5 minutes (9) where, Ti, T 2 =current facial skin temperature and skin temperature for five minutes later. The box plot charts of the gradient of temperature at each sensation score collected from each location in both the heating and cooling process are shown in the following images. 0.5 i 0.4 :; ·= 0.3 E U"'I 0.2 ~ 01 ~ . ~ 0.0 ~ -0.1 i -0.2 ~ -0.3 ] en -o.4 -0.5 · l -2 _ , 0 sensation 0.5 ~ 0.4 :; ·= 0.3 E "' 02 ~ 01 ~ . ~ 0.0 ~ -0.1 i -0.2 • ~ -0.3 i ] en -o.4 -0.5 _ , ·2 _ , 0 sensation Figures 5.17. and 5.18. Box plot of gradient of skin temperature at point #1 in heating process (left) and cooling process (right) 81 0.5 i' 0.4 " .=: 0.3 E I.I\ 0.2 i) t... 01 ~ . ~ 0.0 ~ -0.1 -a -0.2 ~ -0.3 ~ m -o.4 -0.5 + _ , -2 _ , sensation 0.5 i 0 .4 " .=: 0.3 E "' 01 i) <... 01 ~ . ~ 0.0 ~ -0.1 i 01 • ~ -0.3 T :g m -o.4 -0.5 _ , -2 0 sensation Figures 5.19. and 5.20. Box plot of gradient of skin temperature at point #2 in heating process (left) and cooling process (right) 0.5 i 0.4 " £ 0.3 E "' 01 i) t... 01 ~ . ~ 0.0 ~ -0.1 ~ 0 -0.2 ~ -0.3 ] m -o.4 -0.5 _ , -2 _ , 0 sensation 0.5 j 0.4 ~ .: 0.3 E "' 01 ~ 01 ~ . ~ 0.0 ~ -0.1 i -01 .. ~ -0.3 ,. :g m -o.4 -0.5 _ , -2 _ , sensation Figures 5 .21. and 5 .22. Box plot of gradient of skin tern perature at point #3 in heating process (left) and cooling process (right) 0.5 i 0.4 " .: 0.3 E V"I 0.2 ~ 01 ~ . ~ ~ 0.0 ~ -0.1 -a -0.2 ~ -0.3 :g Oi -0.4 -0.5 _ , -2 _ , 0 sensation 0.5 i 0.4 " .: 0.3 E ~ 0.2 ~ 0.1 5 ~ 0.0 ~ -0.1 i -0.2 • ~ -0 .3 T ~ m -o.4 -0.5 -o _ ,--., _ 2 - _ , 0 sensation Figures 5.23. and 5.24. Box plot of gradient of skin temperature at point #4 in heating process (left) and cooling process (right) 82 0.5 0.5 i' 0.4 i 0 .4 " ' " .=: 0.3 ~ .=: 0.3 • E t E • + I.I\ 0.2 "' 01 i) i) t... 01 T <... 01 ~ . • ~ . • ' ~ 0.0 ~ 0.0 ~ -0.1 + ~ -0.1 ' + -a -0.2 i -0.2 ~ -0.3 ~ -0.3 '5 '5 ~ ~ m -o.4 m -o.4 -0.5 -0.5 _ , -2 _ , _ , -2 0 sensation sensation Figures 5.25. and 5.26. Box plot of gradient of skin temperature at point #5 in heating process (left) and cooling process (right) 0.5 i 0.4 " £ 0.3 E "' 01 i) t... 01 ~ . ~ 0.0 ~ -0.1 ~ 0 -0.2 ~ -0.3 ] m -o.4 -0.5 _ , -2 _ , 0 sensation 0.5 j 0.4 ~ .: 0.3 E "' 01 ~ 01 ~ . ~ 0.0 ~ -0.1 i -0.2 ~ -0.3 :g m -o.4 -0.5 _ , -2 sensation Figures 5.27. and 5.28. Box plot of gradient of skin temperature at point #6 in heating process (left) and cooling process (right) 0.5 i 0.4 " .: 0.3 E V"I 0.2 ~ 01 ~ . ~ ~ 0.0 ~ -0.1 -a -0.2 ~ -0.3 ~ Oi -0.4 -0.5 _ , -2 _ , 0 sensation 0.5 i 0.4 " .: 0.3 E ~ 0.2 ~ 0.1 5 ~ 0.0 ~ -0.1 + i -0.2 ~ -0 .3 ~ m -o.4 -0.5 -o _ ,--., _ 2 - _ , 0 sensation • I Figures 5.29. and 5.30. Box plot of gradient of average skin temperature in heating process (left) and cooling process (right) 83 According to the box plot charts above, each facial skin location and overall face generate temperature gradients with increasing tendency as the sensation score increases. The mean gradients of temperature at all six locations are negative when sensations are cold (-3), cool (-2), and slightly cool(-!). The mean temperature gradients are positive when sensations are slightly warm(+ I), warm (+2), and hot (+3). This indicates that the gradients close to zero possibly mean neutral sensation and the deviation depends on the human sample. Besides, based on the cases at hand, the gradient of skin temperature changes more obviously when the subject has slightly warm, warm, or hot sensation. The two sample t-test results of facial skin temperature gradient between the neutral and slightly warm, and between neutral and slightly cool sensation for six locations and average facial temperature are shown below in Table 8. TABLE 8. Gradient of facial skin temperature of each location in each thermal sensation score Slightly cool Neutral (n~20) Slightly warm P-value P-value (n~20) (n~20) (neutral vs. (neutral vs. slightly cool) slightly warm) Point#! -0.22 -0.04 0.23 0.517 0.637 Point#2 -0.18 0.03 0.17 0.434 0.363 Point#3 -0.31 -0.08 0.14 0.234 0.046 Point#4 -0.11 0.03 0.16 0.146 0.007 84 Point#5 -0.26 -0.03 0.21 0.034 0.042 Point#6 -0.16 0.08 0.19 0.743 0.535 Average skin -0.20 -0.02 0.19 0.385 0.462 temperature When P-value is less than 0.05, the data are considered statistically significant. Point#3 and Point#4 show a statistically significant difference between neutral and slightly warm sensation. Point#5 shows both statistically significant differences between neutral and slightly warm sensation and between neutral and slightly cool sensation, meaning neutral sensation can be statistically differentiated by reading the gradient of skin temperature at Point#5. As for the method of absolute facial temperature, all cool sensations, including cold ( -3), cool (-2), and slightly cool (-3) were combined; all warm sensations, including slightly warm(+ 1), warm (+2), and hot ( + 3) were combined. The two sample t-test results of combined sensation are shown below: TABLE 9. Gradient of facial skin temperature of each location in each thermal sensation score Cool (n~20) Neutral (n~20) Warm (n~20) P-value P-value (neutral vs. (neutral vs. cool) warm) Point#l -0.25 -0.04 0.24 0.434 0.536 Point#2 -0.22 0.03 0.19 0322 0.263 85 Point#3 -0.35 -0.08 0.19 0.124 0.024 Point#4 -0.17 0.03 0.21 0.074 0.046 Point#5 -0.31 -0.03 0.27 0.028 0.036 Point#6 -0.19 0.08 0.24 0.453 0.345 Average skin -0.24 -0.02 0.23 0.334 0.353 temperature According to the table above, all P-values decreased. Point#3 and Point#4 show a statistically significant difference between neutral and warm sensation. Point#5 shows both statistically significant differences between neutral and warm sensation and between neutral and cool sensation. As described previously, both before and after combining sensations, Point#5 has the greatest statistically significant difference of gradient of temperature both between neutral and warm sensation and between neutral and cool sensation, therefore, Point#5 will be selected to stand for facial skin temperature, since it generated the most reliable result. 5.2 Discussion of ambient temperatures at each thermal sensation The facial skin temperature levels of all subjects were summarized. For more detailed analysis, the ambient temperature was utilized to form a possible index for controls. The intervals of ambient air dry bulb temperature at each surveyed thermal sensation in the heating and cooling process are shown in Figures 5.31 and 5.32. 86 30 G 28 0 .......... ~ ~ 26 ~ (l) c.. E 24 (l) ....... ..a ::::J ..a 22 ~ ""C 20 -3 -2 -1 0 sensation 2 Figure 5 .31. Box plot of ambient dry bulb temperature at each sensation in heating process 30 - 28 u 0 .......... (l) lo... ::::J 26 ....... ra lo... (l) c.. E 24 (l) ....... ..a ::::J ..a 22 ~ ""C 20 -3 -2 -1 0 sensation 2 Figure 5 .32. Box plot of ambient dry bulb temperature at each sensation in heating process 3 3 The box plot charts indicate a significant regulation, which is that the dry bulb temperature increases with 87 thermal sensation change from -3( cold) to 3(hot). However, there is no significant difference between each location. From the data analysis provided above, neither facial skin temperature nor the gradient of facial temperature independently can determine human thermal sensation accurately. In order to determine human sensation, those three parameters need to be integrated. In the next chapter, a decision algorithm will be established by the data mining software, WEKA. 5.3. Facial tern perature based control model development 5.3.1. Decision tree The human subject experiments provided evidence that facial skin temperature, gradient of facial skin temperature, and ambient air dry bulb temperature, can indicate human thermal sensation. However, each status of those three parameters can possibly relate to more than one sensation score. In order to determine human sensation by utilizing sensed signals, the integration of facial skin temperature, the gradient of facial skin temperature, and ambient dry bulb temperature is used to establish the decision tree. In this project, the sensation with scores from -3 to -1 is defrned as cool sensation, and sensation with scores from + 1 to +3 is defined as warm sensation. In order to create a neutral temperature indoor envirornnent, when the predicted sensation is cool, the AC is in heating mode. When the predicted sensation is warm, the AC turns into a cooling task. 88 \VEKA was used to make the decision tree for determining human thermal sensation and air conditioning task type. \VEKA, developed by the University of Waikato, is a collection of machine learning algorithms for solving real-world data mining problems. \VEKA can generate a decision tree by the J48 Algorithm, which includes a large amount of data, and find the mathematical experience to make final decision. The principle of the J48 Algorithm is introduced in the following paragraph. The decision tree generated by \VEKA in the heating process and cooling process are shown in Figures 5.33 and 5.34, respectively. 9¥? = ferrra~ ------- ~ 1 <-~007 4: Cool (1.7/1.5) 1 2:Cool (13G'1: ~ 25 >= 325 ~ t----->= 26.3 -----, < -OC6 --==r 10 Wovm (1.Ml5) 14 Cool (11004) >= -O. C6 9 Neutra (67.58) 15: Aml::ient temperature < 26.4 >= 26.4 I _ 16 Neutral (147/126) [ 17 Wovm (25~ 1 H eating on Heating on H eating off Heating on Heating on Heating off Figure 5.33. The sensation and AC task decision tree in heating process The accuracy of each decision is shown in the table below: Decision Accuracy Cool (Heating on) 77.9% Neutral (None) 85.5% 89 I Warm (Heating ofi) =male ~47 ~-->= 347 <O~az 4 Coo (17/15) 5 Ambient temperature < 26.3 86.0% female------~ ~~~ < 33.9 r >= 33. 9 8Warm(14.1/100) 10 Ambient temperature < 25.1 1 1:Cool (4171385) < 0.06 >=25.1 I ~ >= 0.06 I 15 Warm (101/81 ) 6 N eutral (26.2!23. 1) 13 Neutral (3.7 !2.1 ) 1 14: Warm (21 !Y1 82)1 l Condenser off Condenser on Condenser on Condenser off Condenser on Condenser on Figure 5.34. The sensation and AC task decision tree in cooling process The accuracy of each decision is shown in the table below: Decision Accuracy Cool (Condenser oft) 77.9% Neutral (None) 85.5% Warm (Condenser on) 86.0% The decision trees above are generated by the J48 Algorithm. The J 48 Algorithm is a method for machine learning data mining. In this method, splitting is performed by using one attribute at internal nodes (Kaur & Chhabra, 2014 ). Firstly, the dataset including gender, facial skin temperature, the gradient of facial skin temperature, ambient dry bulb temperature, and thermal sensation, was used as a set of training instances. 90 J48 uses a top-down decision tree algorithm and merit selection criteria to choose the best splitting attribute to create a branch and, consequently, the system generates two partitions. The algorithm then conducts the same top-down analysis to make further partitions until the entire attribute's values are included in a single class. Many different decision trees could be generated. After pruning, the decision tree with the simplest structure is kept, that is the one with the smallest number of partitions where every decision has the accuracy of 80% and above. A conceptual control system was built by measuring facial skin temperature and indoor air dry bulb temperature. The gradient of facial skin temperature was calculated as another index besides facial skin temperature and indoor dry bulb temperature. The gender of subjects was the fourth parameter. The decision tree presented above tells the algorithm that the system to utilize the measured data to defme control task type. In the decision tree, every possible combination of facial skin temperature, the facial skin temperature gradient, and ambient dry bulb temperature is included in the decision algorithm. For example, suppose there is a male person in the cooling process, with facial skin temperature sensed at 34.4 °C, and the system calculated his facial temperature change rate at 0.06°C/5minutes and air temperature at 26.1°C. In the first step, the system will confirm the person is male. In the second step, the system will recognize this subject's facial skin temperature, 34.4°C, is less than 34.7°C. Then the system will compare the subject's gradient of facial temperature; the 0.06°C/5minutes is larger than 0.02°C/5minutes. Then, air temperature will be compared to 26.3° C; it is less than 26.3°C. Therefore, the system will predict this person's sensation as neutral. The work flow chart for this example is shown below in Figure 5.35. 91 =male female ------~ <:§~ < 33.9 r >= 33. 9 ~47 ---- >= 34.7 < 0~(12 8 Warm (141/10.0) 1 0: Ambient temperature 15 Warm (101/81) < 25.1 4. Cool (1.711.5) 5 Ambient temperature 11 Ccci (41.7138.5) < 26.3 6 Neutral ( 26. 2123 1 ) Condenser off Condenser on Condenser on Condenser off Figure 5.35. Work flow chart of the example >=25.1 I I ~ I >= 0.05 I < 0.05 13 Neutral (3.7f2.1) 1 14 Warm (215'182)1 l Condenser on Condenser on Based on the decision tree, it is easy to make the control logic program use any computer language. The Java code of this control logic in the heating process is provided below: import java.util.Scanner; import org.junit. Test; public class DecisionTree { private Scanner sc; private double prevSkinTemp; private double skinTemp; private double ambientTemp; @Test public void test() { TASK task = getTask(); if (task.equals(TASK.COOLING)) { startCoolingTask(); } else { startHeatingtask(); } 92 } public void startCoolingTask() { } public void startHeatingtask() { GENDER gender ~ getGender(); } if (gender. equals( GENDER.MALE)) { if(getSkinTernp() < 33.1) { if (getGradient() < -0.07) { heatingOn(); } else { } if (getArnbientTernp() < 23. 7) { heatingOn(); } } else { } if (getArnbientTernp() >~ 26.3) { heatingOff(); } } else { } if (getSkinTernp() < 32.5) { heatingOn(); } else { } if (getGradient() < -0.05) { heatingOn(); } else if (getArnbientTernp() >~ 26.4) { heatingOff(); } public void heatingOn() { Systern.out.println("Heating On"); } 93 public void heatingOff() { Systern.out.println("Heating Oft''); } public void condensorOn() { Systern.out.println("Condensor On"); } public void condensorOff() { Systern.out.println("Condensor Oft''); } public enurn GENDER { MALE, FEMALE; } public enurn TASK { COOLING, HEATING; } public GENDER getGender() { } sc ~new Scanner(Systern.in); Systern.out.println("Please input your gender: male or female"); String gender ~ sc.nextLine(); if (gender.equals("rnale")) { return GENDER.MALE; } else if (gender.equals("fernale")) { return GENDER.FEMALE; } else { } Systern.out.println("Input incorrect, please try it again"); return getGender(); public TASK getTask() { sc ~new Scanner(Systern.in); Systern.out.println("Please input the task: cooling or heating"); String gender ~ sc.nextLine(); if (gender.equals("heating")) { return TASK.HEATING; } else if (gender.equals("cooling")) { 94 } } return TASK.COOLING; } else { } System.out. printlu("Input incorrect, please try it again"); return getTask(); public double gefPrevSkinTernp() { return prevSkinTernp; } public void setPrevSkinTernp(int prevSkinTernp) { this.prevSkinTernp ~ prevSkinTernp; } public double getSkinTernp() { return skinTernp; } public void setSkinTernp( double skin Temp) { this.skinTernp ~ skinTernp; } public double getGradient() { return this.skin Temp - this.prevSkinTernp; } public double getArnbientTernp() { return arnbientTernp; } public void setArnbientTernp( double arnbientTernp) { this.arnbientTernp ~ arnbientTernp; } With the above code, the controller can be created and control the condenser unit and heaters according to the decision trees. 95 5.4. Validation experiment based on decision tree In order to prove the reliability of the control logic developed in Chapter 5, a second round of human subject experiments was conducted for validation. Three subjects participated in this round of experiments, with two male students and one female student. Each test lasted for one hour. In each test, the subject was asked to do computer based work sitting in front of a desk. The test started after the subject had entered the room for 10 minutes. The control system, which controls the condenser's On/Off status, was programmed based on the decision tree logic and the control interval was two minutes. As discussed previously, Point#5 was the best location for collecting skin temperature data, so Point#5 was the only location sensed in this round of experiments, shown below in Figure 5.36. \ \ Figure 5.36. Validation test The subjects were asked to answer the thermal sensation survey every 10 minutes; the first survey was done after 10 minutes of the test starting. The thermal sensation scores of the subjects are shown below: 96 Time after starting 10 minutes 20 minutes 30 minutes 40 minutes 50 minutes Subject#l 0 0 0 0 0 Subject#2 0 0 +l 0 0 Subject#3 0 0 0 0 0 As shown above, during the test, only subject#2 answered + l(slightly warm) at the 30 1 h minute. The rest of of the sensation scores were all zero, which stands for thermal neutrality. According to this result, the control system based on the logic developed by the decision tree does have the ability to maintain thermal neutrality. 5.3.2 Neural network analysis In machine learning, artificial neural networks are statistical learning algorithms derived from biological neural networks and are utilized to predict unknown functions based on a large number of inputs. Neural networks are described as systems of interconnected 'neurons' which can compute values from inputs and are able to conduct machine learning and pattern recognition based on their adaptive nature. For example, as shown in Figure 5.36, a neural network for thermal sensation estimation in this thesis is defined by a set of input neurons that include four parameters: skin temperature, gradient of skin temperature, air temperature, and gender. After being weighted and transformed by a function which is a black box (determined by the network's designer), the activations of these neurons are then passed on to other neurons. This process is repeated until, finally, an output neuron, including cool sensation, neutral 97 sensation, and warm sensation, is activated. This determines which character was read . .. ~----~ .~----~ Figure 5.36 Sample structure of neural network In order to find an accurate structure, the number of hidden layers was defined by manual comparison. The accuracy of different hidden layers is shown in the table below. Hidden layers Correctly Classified Incorrectly Classified Accuracy Instances Instances 4,3,2 260101 16011 94.2012% 4,2,2 220535 59932 78.6314% 4,2,1 252530 23582 91.4592% 3,2,2 219577 56535 79.5246% 3,2,1 241545 34567 87.4808% 2,2,2 208167 67945 75.3922% 2,1,1 240326 35786 87.0393% According to the table above, 4,3,2 was chosen as the number of hidden layers. The principle of neural 98 network will not be deeply described. However, the high accuracy shows that there is a big potential to apply neural networks into building technology and thermal control. 99 Chapter 6 Conclusion This research is an experiment based project that mainly investigated and described human facial skin temperature and its interaction with air temperature while relative humidity, mean radiant temperature, air speed, metabolic rate, and clothing insulation were kept constant. The research is based on thermoregulation, a physiological function to maintain core body temperature. Twenty human subject experiments were conducted and used to measure the necessary data, such as skin temperature and ambient temperature. Thermal sensation surveys were given to subjects and responses were collected to indicate their thermal sensation under a certain thermal condition with their measured facial skin temperature. Those data were analyzed by robust data mmmg technology and machine learning technology. The most sensitive facial location was selected and used as the sensing point. Finally, the temperature patterns were generated with sensation scores and the decision tree was generated to determine the control logic. The simple control system was developed based on the decision tree and validation experiments with three people were conducted. Results showed that the system can keep subjects in neutral sensation. The research shows that there is a big potential to utilize facial skin temperature as an index to predict human thermal sensation and to control the HVAC system. 6.1 Project methodologies The greatest difference of methodology of this research to other research about thermal comfort is that, most research studies the mechanism of human gain and loss heat energy and identifies heat balance (heat 100 gain equals to heat loss) as a condition of thermal comfort. Besides, many researchers focus on the enhanced calculation of air temperature, radiant temperature, humidity, air speed, and further factors such as radiant asymmetry. However, this research focused on the relationship between input (facial skin temperature) and output (human sensation). Although air temperature was considered as a parameter to estimate human sensation, the deeper reason why human sensation reacts with ambient condition was not investigated in this thesis. The research began with the thermal chamber installation. The thermal chamber was a room in the basement, surrounded with conditioned space. This character provided the chamber with the constant mean radiant temperature, constant wall surface temperature, and constant relative humidity. The air temperature was the only variable altered during the experiment. The space was conditioned by a window condenser unit and two portable heaters. The constant air flow was supplied into the room. Twenty subjects attended the two-hour human subject experiments. During the experiment, subjects' facial skin temperatures were measured at six facial locations. Ambient air temperature, MRT (constant), and relative humidity (constant) was measured. Labview was the tool used to control the thermal system and data acquisition. The air temperature was heated from 20°C to 30°C, and cooled from 30°C to 20°C. The thermal sensation survey was given to subjects during the experiment and all responses were received. 101 The data mining technology was used to analyze experiment data. According to the box charts, including the information of 95% confidence interval and mean value, the absolute level of facial skin temperature at each sensation score was calculated, as was the gradient of skin temperature (change rate). According to observation of the box charts and two-sample t-tests, Point#5 (bottom ledge of ROI) was selected as the most sensitive location in the facial area and used to apply to the control logic. By utilizing the J48 machine learning algorithm, the thermal sensation and air conditioning decision tree was generated with acceptable accuracy. The controller of the air conditioning system was programmed based on the decision tree. In order to test the reliability of the control logic, 30 minute validation experiments with three participants were conducted. The AC and heaters were controlled by the developed controller and subjects were asked to answer the thermal sensation surveys. According to the survey response, subjects presented neutral sensation during the test and proved that the system is reliable. 6.2 Distribution of the project The purpose of this research is to discover the regulation of interaction of human facial skin temperature, air temperature (dry bulb temperature), and human thermal sensation. Control logic was generated by using data mining and machine learning technology. The research outcomes will help develop an individual sensing, dynamic control system using facial skin temperature to control building mechanical systems to provide people with comfortable thermal conditions. This research can also be used as a 102 reference by other researchers regarding building and human interaction. This control method made it possible to utilize a human body's physiological signal as an index for thermal sensation evaluation. The system used the measured skin temperature that reflects thermal sensation to regulate heating or cooling system operation. Via data mining and machine learning technology, the skin temperature data and facial skin temperature data were analyzed. The relationship between human thermal sensation and facial skin temperature as well as ambient air temperature was discovered. Finally, three parameters were chosen to predict human thermal sensation. They are skin temperature, gradient of skin temperature (temperature change rate), and ambient air dry bulb temperature. By utilizing those three parameters, the human thermal sensation was predicted by machine learning with accuracy above 80%. Although the proposed system needs a great deal of improvement, this opens a new direction in using physiological signals to control thermal systems. 6.3. Characterization of facial skin temperature related to individual thermal sensation The research conducted a series of human subject experiments and discovered a significant relationship between facial skin temperature patterns and human thermal sensation. After evaluation of six facial locations for temperature responses to changing air dry bulb temperatures, Point#5 (bottom ROI area) was identified as one of the most sensitive and robust sensing locations. While net skin surface temperature was tested as an important parameter, this research proved that gradients in facial skin temperature can be used to develop a simplified, but adjustable model for individual thermal preferences to estimate the air 103 conditioning task settings. The results and data analysis investigated in this research contribute to indoor temperature control systems integration by individuals and automatic controllers without a high cost, and contribute to the area of building science related to thermal comfort and system performance with the help of machine learning technology. 6.2. Limitations of the project For this project, there are some limitations which need further investigation. 6.2.1. Small sample size In the research, 20 human subjects participated in the experiment and three subjects participated in the validation test. Limited by fmancial constraint, all subjects were recruited as volunteers from students (aged between 22 and 28 years). Among the subjects, 13 were Chinese, two were Asian, and five White. Because of the sample size, the research cannot deeply discuss different body conditions, such as skin color, weight, age, fat rate, to confirm if the discoveries were consistent across the different physiological conditions. This flaw should be resolved in future research. 6.2.2. Air conditioning type This was conducted with a constant flow window AC unit and constant flow heaters. The only control execution is On/Off of the condenser and heaters. The location of air delivery was constant and there was no ventilation. Future research should explore conditioning system type and function in more depth, including variable air flow and constant air flow. The research also needs to test the reliability of the discovery with 104 water-air systems and all air systems with variable air flow systems. 6.2.2. Lack of other thermal comfort parameters This thesis only focused on the dry bulb temperature as an experiment variable. Because of the experimental condition, it was not possible to study the entire six factors which affect thermal comfort. 105 Chapter 7 Future Work As discussed in Chapter 6, there are some limitations of this thesis; elimination of those limitations will be the main focus of future work. 7.1 Conduct a larger sample size of experiment A larger sample size could improve the accuracy of the findings. It is possible that a larger group of samples can provide enough data to study interaction of different body conditions and skin temperature, such as skin color, age, and fat rate. 7.2 Radiant temperature As described in Chapter 1, the operative temperature which affects human thermal sensation is calculated as a combination of dry bulb temperature and radiant temperature. Since this research did not investigate the radiant temperature and its impact on facial skin temperature and human sensation, radiant temperature will be a big part of future research. In order to control the radiant temperature, a better thermal chamber is needed, in which wall surface temperature can be controlled and monitored. In order to better understand the impact of radiant temperature, the radiant asymmetry, including the vertical radiant temperature asymmetry and the horizontal radiant temperature asymmetry, also needs to be discussed. 106 7.2 Skin color When enough samples and data are collected, different skin types, such as Caucasian, Asian, and African can be discussed. The test will show the different reaction of different skin types when thermal conditions change. 7.3 Integration with passive strategies Finally, the skin temperature based control system should be tested when applied to passive conditioning systems, such as natural ventilation systems, as well as mechanical HV AC, because the thermal chamber used in this thesis does not have any natural ventilation. If considering external conditions, such as solar heat and weather, the facial skin temperature based control system may generate more accurate control decisions and could estimate the most effective thermal control strategy via an energy simulation process. This idea would contribute to energy saving and environmental protection beyond the building occupants' thermal comfort. 107 References ASHRAE. (2013a)ANSllASHRAE Standard 55: Thermal environmental conditions for human occupancy Atlanta, GA, American Society of Heating, Refrigerating, and Air-Conditioning Engineers, Inc. Huizenga, C., Abbaszadeh, S., Zagreus, L. & Arens, E.A (2006) Air quality and thermal comfort in office buildings: Results of a large indoor envirornnental quality survey. Proceeding Healthy Buildings, 3. de Dear, R.J. & Brager, G.S. (2002) Thermal comfort in naturally ventilated buildings: Revisions to ASHRAE Standard 55. Energy & Buildings, 34(6), 549-561. EPA. (1991) Indoor Air Facts No. 4 (revised) Sick Building Syndrome, Air and Radiation (6609J). Research & Development. Hawkes, D. (1982) The theoretical basis of comfort in the 'selective' control of envirornnents. Energy & Buildings, 5(2), 127-134. Ueda, M., Taniguchi, Y, Asano, A, Mochizuki, M., Ikegami, T. & Kawai, T. (1997) An automobile heating, ventilating and air conditioning (HV AC) system with a neural network for controlling the thermal sensations felt by a passenger. JSME International Journal, Series B, Fluids and Thermal Engineering, 40(3), 469. Myhren, J.A, Holmberg, S., Str6mnings- och klirnatteknik (stiingd 20090101), KTH & Skolan for teknik och hiilsa (STH). (2008) Flow patterns and thermal comfort in a room with panel, floor and wall heating. Energy & Buildings, 40( 4), 524-536. Wyon, D. (1996) Individual microclimate control: required range, probable benefits and current feasibility. In: Proceedings of the 7th International Conference on Indoor Air Quality and Climate, Nagoya, Japan. pp. 1067-1072. Wong, L.T, Mui, KW., Fong, N.K. & Hui, PS. (2007) Bayesian adaptive comfort temperature (BACT) of air-conditioning system in subtropical climate. Building & Environment, 42(5), 1983-1988. Stanton, N.A (2005) Handbook of Human Factors and Ergonomics Methods. Boca Raton, CRC Press. Ainsworth, B.E, Haskell, W.L., Whitt, M.C., Irwin, M.L., Swartz, AM., Strath, S.J, O'Brien, W.L., Bassett D.R., Schmitz, K.H., Emplaincourt, PO., Jacobs, D.R.J. & Leon, AS. (2000) Compendium of physical activities: An update of activity codes and MET intensities, Medicine & Science in Sports & Exercise, 32(9), S498-S516. 108 Szokolay, S.V & Brisbin, C. (2004) Introduction to Architectural Science: The Basis of Sustainable Design. Amsterdam, Elsevier. Havenith, G (1999) Heat balance when wearing protective clothing. Annals of Occupational Hygiene, 43(5), 289-296. McCullough, EA, Eckels, S. & Harms, C. (2009) Determining temperature ratings for children's cold weather clothing. Applied Ergonomics, 40(5), 870-877. ISO. (1998)ISO 7726. Ergonomics of the thermal environment - Instrument for measuring physical quantities. Geneva, Switzerland: International Organization for Standardization. Mcintyre, D.A & Griffiths, I.D. (1972) Subjective response to radiant and convective environments. Environmental Research, 5(4), 471-482. Nall, D. H. (2004) Looking across the water: Climate-adaptive buildings in the United States and Europe. Construction Specifier, 57, 50-56. Balaras, C.A, Dascalaki, E & Gaglia, A (2007) HVAC and indoor thermal conditions in hospital operating rooms. Energy & Buildings, 39(4), 454-470. Wolkoff, P. & Kjaorgaard, S.K. (2007) The dichotomy ofrelative humidity on indoor air quality. Environment International, 33(6), 850-857. Fanger, 0. (1972) Thermal Comfort: Analysis and Applications in Environmental Engineering. New York, McGraw-Hill. Cena, K. & de Dear, R. (2001) Thermal comfort and behavioural strategies in office buildings located in a hot-arid climate. Journal of Thermal Biology, 26, 409-414. doi: 10.1016/S0306-4565(01)00052-3 Barlow, S. & Fiala, D. (2007) Occupant comfort in UK offices-How adaptive comfort theories might influence future low energy office refurbishment strategies. Energy Building, Comfort and Energy Use in Buildings - Getting Them Right, 39, 837-846. doi: 10.1016/j.enbuild.2007.02.002 Rose, R.J. & Dozier, J. (1997) EPA program impacts office zoning. ASHRAE Journal, 39(1), 37. Jaakkola, J.J.K., Heinonen, O.P. & Seppanen, 0. (1989) Sick building syndrome, sensation of dryness and thermal comfort in relation to room temperature in an office building: Need for individual control of temperature. Environmentlnternational, 15, 163-168. doi:l0.1016/0160-4120(89)90022-6 Choi, J.H. (2010) CoBi: Bio-Sensing Building Mechanical System Controls for Sustainably Enhancing 109 Individual Thermal Comfort. ProQuest, UMI Dissertations Publishing. Egan, G.F., Johnson, J., Farrell, M., McAllen, R., Zamarripa, F., McKinley, M.J., Lancaster, J., Denton, D. & Fox, P. T. (2005) Cortical, thalamic, and hypothalamic responses to cooling and warming the skin in awake humans: A positron-emission tomography study. Proc. Natl. Acad. Sci., 102, 5262-5267. doi: 10.1073/pnas.0409753102 ASHRAE. (2013b) ASHRAE Handbook Fundamentals. Atlanta, GA, American Society of Heating, Refrigeration and Air-Conditioning Engineers. Christensen, J., Vaeth, M. & Wenzel, A (2012) Therrnographic imaging of facial skin--gender differences and temperature changes over time in healthy subjects. Dentomaxillofacial Radial. 41, 662--067. doi: 10.1259/drnfr/55922484 KMC Controls. (2013) Building Automation System. Hong, B.N., Do, M.H., Her, YR., Lee, YR. & Kang, T.H. (2015) The effects of Panax ginseng and Panax quinquefolius on therrnoregulation in animal models. Evidence-based Complementary & Alternative Medicine: eCAM, 2015, 748041. Nuzzo, R. (2014) Scientific method: statistical errors. Nature, 506(7487), 150-152. Daikin. (2012) Individual Control Technologies. Song, Y, Wu, S. & Yan, YY (2013) Control strategies for indoor envirornnent quality and energy efficiency-a review. Int. J. Low-Carbon Technol. ctt05 l. doi: 10.1093/ijlct/ctt05 l Altrock, C.V., Arend, H.-0., Krause, B., Steffens, C. & Behrens-R6mmler, E. (1994) Adaptive fuzzy control applied to home heating system. Fuzzy Sets Syst. 61, 29-35. doi:l0.1016/0165-0114(94)90281-X Paassen, A.H.C.V. (1990) Digital Control Systems for Passive Solar Buildings. Delft University. Kanarachos, A & Geramanis, K. (1998) Multivariable control of single zone hydronic heating systems with neural networks. Energy Convers. Manag. 39, 1317-1336. doi:l0.1016/S0196-8904(98)00015-6 Yamada, F., Yonezawa, K., Sugawara, S. & Nishimura, N. (1999) Development of air-conditioning control algorithm for building energy saving. In: Proceedings of the 1999 IEEE International Conference on Control Applications, vol. 2, 1999, pp. 1579-1584. doi: 10.1109/CCA.1999.801207 Nice, K. (2001) How Cruise Control Systems Work. [Online] Available from: http://auto.howstuffworks.com/cruise-control.htm [Accessed 14th November 2014]. 110 Hu, B.-G, Mann, GK.I. & Gosine, R.G controllers-function-based evaluation approach. doi: 10.1109/91.963756 (2001) IEEE A systematic Trans. Fuzzy Carroll, R. (2007) Integrated Physiology. Philadelphia, PA, Mosby Elsevier. study of fuzzy PID Syst., 9, 699-712. Charkoudian, N. (2003) Skin blood flow in adult human thermoregulation: How it works, when it does not, and why. Mayo Clin. Proc., 78, 603-612. doi:l0.4065/78.5.603 Nakamura, M., Yoda, T., Crawshaw, L.I., Yasuhara, S., Saito, Y, Kasuga, M., Nagashima, K. & Kanosue, K. (2008) Regional differences in temperature sensation and thermal comfort in humans. J. Appl. Physiol., 105, 1897-1906. doi: 10.1152/japplphysiol.90466.2008 Chen, C., Hwang, R., Chang, S. & Lu, Y (2011) Effects of temperature steps on human skin physiology and thermal sensation response. Building & Environment, 46(11), 2387-2397. Zhang, H., Huizenga, C., Arens, E. & Wang, D. (2004) Thermal sensation and comfort in transient non-uniform thermal environments. Eur. J. Appl. Physiol., 92, 728-733. doi:l0.1007/s00421-004-1137-y Griefahn, B., Kiinemund, C., Gehring, U. & Mehnert, P. (2000) Drafts in cold environrnents--the significance of air temperature and direction. Industrial Health, 38(1), 30. Kaur, G & Chhabra, A (2014) hnproved J48 classification algorithm for the prediction of diabetes. International Journal of Computer Applications, 98(22),. 111
Abstract (if available)
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Enhancing thermal comfort: data-driven approach to control air temperature based on facial skin temperature
PDF
Developing a data-driven model of overall thermal sensation based on the use of human physiological information in a built environment
PDF
Multi-occupancy environmental control for smart connected communities
PDF
Exploration for the prediction of thermal comfort & sensation with application of building HVAC automation
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
Developing environmental controls using a data-driven approach for enhancing environmental comfort and energy performance
PDF
Impacts of indoor environmental quality on occupants environmental comfort: a post occupancy evaluation study
PDF
Human-environmental interaction: potential use of pupil size for office lighting controls
PDF
Indoor air quality for human health in residential buildings
PDF
Indoor environmental quality and comfort: IEQ adaptation and human physiological responses in commercial buildings
PDF
Human-building integration: Investigation of human eye pupil sizes as a measure of visual sensation in the workstation environment
PDF
Economizer performance and verification: effect of human behavior on economizer efficacy and thermal comfort in southern California
PDF
Natural ventilation in tall buildings: development of design guidelines based on climate and building height
PDF
Real-time simulation-based feedback on carbon impacts for user-engaged temperature management
PDF
Mitigating thermal bridging in ventilated rainscreen envelope construction: Methods to reduce thermal transfer in net-zero envelope optimization
PDF
Energy savings by using dynamic environmental controls in the cavity of double skin facades
PDF
Tiny house in the desert: a study in indoor comfort using moveable insulation and thermal storage
PDF
Using bio-signals with smart windows
PDF
Visualizing thermal data in a building information model
PDF
Energy performance of different building forms: HEED simulations of equivalent massing models in diverse building surface aspect ratios and locations in the US
Asset Metadata
Creator
Yi, Bo
(author)
Core Title
Enhancing thermal comfort: air temperature control based on human facial skin temperature
School
School of Architecture
Degree
Master of Building Science
Degree Program
Building Science
Publication Date
04/25/2017
Defense Date
03/23/2015
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
bio-sensing,decision tree,neural network,OAI-PMH Harvest,thermal chamber,thermoregulation
Format
application/pdf
(imt)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Choi, Joon-Ho (
committee chair
), Konis, Kyle (
committee member
), Noble, Douglas (
committee member
)
Creator Email
hory.bo.yi@gmail.com,yib@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c3-562687
Unique identifier
UC11300527
Identifier
etd-YiBo-3405.pdf (filename),usctheses-c3-562687 (legacy record id)
Legacy Identifier
etd-YiBo-3405.pdf
Dmrecord
562687
Document Type
Thesis
Format
application/pdf (imt)
Rights
Yi, Bo
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
Tags
bio-sensing
decision tree
neural network
thermal chamber
thermoregulation