478 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 64, NO. 2, FEBRUARY 2015

A Computer-Aided Modeling and MeasurementSystem for Environmental Thermal

Comfort SensingQingyuan Zhu, Jian Yi, Shiyue Sheng, Chenglu Wen, Member, IEEE, and Huosheng Hu, Senior Member, IEEE

Abstract— Predicted mean vote (PMV) is a well-known thermalcomfort index with four environmental variables (air temper-ature, relative humidity, air velocity, and average radiationtemperature) and two human factors (metabolic rate and clothingthermal resistance). This paper presents a novel computer-aided thermal comfort measurement system with PMV, whichcombined the advanced sensors with the virtual instrumenttechnology. The system software is developed using the LabVIEWplatform. The measured data can be transmitted to the server-computer in a data center and displayed on a web page throughthe Internet. The impact of the measurement error of eachenvironmental variable on PMV is analyzed via MATLAB. Thesystem tests were conducted under the certain environmentalconditions and Monte Carlo method is deployed to analyze thePMV measurement uncertainty. The experimental results showthe feasibility and effectiveness of the proposed system, andconfirm that the measurement uncertainty of PMV is not aconstant, and varies with the environment changes.

Index Terms— Computer-aided measurement system, MonteCarlo method (MCM), predicted mean vote (PMV), thermalcomfort, uncertainty.

I. INTRODUCTION

THERMAL comfort is a subjective evaluation thatexpresses satisfaction with the thermal environment.

At present, thermal comfort is usually calculated according todifferent standards, including ANSI/ASHRAE Standard 55 [1],EN 1525 [2], and ISO 7730 standard [3]. Different standardsuse different thermal comfort indexes. For example, ISO 7730standard employs predicted mean vote (PMV) and predictedpercentage dissatisfied. In contrast, ANSI/ASHRAE Standard55 employs thermal sensation scale. Among these indicators,the PMV index is a well-known and widely used example, andit has been applied for almost 40 years throughout differentbuilding types. It includes all the major variables influencingthermal sensation and gives more objective evaluations. It can

Manuscript received March 20, 2014; revised June 23, 2014; acceptedJuly 9, 2014. Date of publication August 19, 2014; date of current versionDecember 31, 2014. The Associate Editor coordinating the review processwas Dr. Deniz Gurkan.

Q. Zhu is with the Department of Mechanical and Electrical Engineering,Xiamen University, Xiamen 361005, China.

J. Yi and S. Sheng are with the School of Physics, Mechanical and ElectricalEngineering, Xiamen University, Xiamen 361005, China.

C. Wen is with the School of Information Science and Engineering, XiamenUniversity, Xiamen 361005, China (e-mail: clwen@xmu.edu.cn).

H. Hu is with the School of Computer Science and Electronic Engineering,University of Essex, Colchester CO4 3SQ, U.K.

Color versions of one or more of the figures in this paper are availableonline at http://ieeexplore.ieee.org.

Digital Object Identifier 10.1109/TIM.2014.2345922

quantitatively evaluate the thermal sensation, and adopt a rangeof sensation levels, numbered from −3 (cold) to 3 (hot) forquantization [4]–[8].

Fanger [9] proposed the PMV model based on experimentalstudies. They used heat balance equations and empirical stud-ies about skin temperature to define the thermal comfort of thehuman body [9]. Based on this PMV model, Gao [10] estab-lished a PMV index predictor that combined artificial neuralnetworks and modern sensor technology. Alfano et al. [11]analyzed the effects of six independent environmental vari-ables on the PMV index for thermal environment assessment.

Currently, many smart sensing and wireless sensor networkshave been widely deployed in environmental monitoringsystems [12]. Wai and Wai [13], [14] used a distributedsensor network to conduct real-time PMV measurementsin an office environment. Wang et al. [15] established anenvironmental monitoring system based on virtual instrumenttechnology and proposed a forecasting control method forthe regulation and control of a greenhouse environment.Tse et al. [16] developed a real-time measurement systemfor thermal comfort using an open networking technology.Through an embedded subsystem within an existing thermalcomfort detection system, data is collected and calculated,resulting in limited data storage. When a large amount ofdata is collected, data loss is likely to occur. Most measuringinstruments have only data acquisition and display functions;the capacity to analyze data is weak.

Although the calculation of the PMV model is widelyconducted for analyzing its influencing factors, it is still notwidely used in thermal comfort measurement applications.This paper aims to analyze the features of the PMV modeland propose an integrated measurement system for indoorenvironmental thermal comfort assessment based on virtualinstrument technology. To extract historical data for furtheranalysis and predict the future state of the environment, adatabase management subsystem is established to effectivelymanage data inspection. A subsystem of remote data trans-mission and accession is designed to view the measured dataremotely through the Internet. Then, the measuring error ofeach environment variable affects the PMV system accuracy,which is analyzed under typical conditions. With the useof the Monte Carlo method (MCM), the PMV measurementuncertainty is analyzed. It was found that the measurementuncertainty of PMV is not a constant and varies with theenvironment changes.

0018-9456 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

ZHU et al.: COMPUTER-AIDED MODELING AND MEASUREMENT SYSTEM 479

The rest of this paper is organized as follows. Section IIintroduces the calculation theory of PMV. Section III describesthe system design, including system framework, hardwareconfiguration, and the modular software design that offersflexibility and good scalability during the system operation.Experimental results and a discussion are given in Section IVto show the performance of the proposed system, includingthe analysis of measurement uncertainty. Finally, a briefconclusion and future work are given in Section V.

II. CALCULATION THEORY OF PMV

The PMV index is a prediction index that can be usedto predict thermal comfort by a PMV-based thermal comfortmodel. The index integrates four environmental variables [airtemperature, relative humidity (RH), air velocity, and averageradiation temperature] and two human factors (metabolic rateand clothing thermal resistance). Its mathematical expressionis expressed by the following equation [3], [9], [17]:

PMV = [0.303e−0.036M+0.0275]×{M − W − 3.05[5.733 − 0.007(M − W ) − Pa]

− 0.42(M − W − 58.15) − 1.73 × 10−2 M

× (5.867−Pa)−0.0014M

× [34 − ta − 3.96 × 10−8 × fcl|(tcl + 273)4

− (tr + 273)4| − fclhc(tcl − ta)]} (1)

where M is the metabolic rate (W/m2), W is the effectivemechanical power (W/m2), Pa is water vapor partial pressure(KPa), fcl is the clothing surface area factor, tcl is theclothing surface temperature (°C), tr is the mean radianttemperature (°C), Icl is the clothing insulation (m2 · K/W),ta is the air temperature (°C), and hc is the convective heattransfer coefficient [W/(m2 · K)].

In (1), the water vapor partial pressure, Pa , and the effectivemechanical power, W , are calculated by

Pa = RH × exp[16.6536 − 4030.183/(ta + 235)] (2)

W = ηM (3)

where RH is measured in % and η is the effective utilizationcoefficient of the mechanical work. The clothing surface areafactor, fcl, is determined as follows:

fcl ={

1.00 + 1.290Icl, Icl ≤ 0.078 m2 · K/W

1.05 + 0.645Icl, Icl > 0.078 m2 · K/W.(4)

The clothing surface temperature, tcl, is calculated by

tcl = 35.7 − 0.028(M − W ) − Icl{3.96 × 108 fcl

×[(tcl + 273)4 − (tr + 273)4] + fclhc(tcl − ta)}. (5)

The convective heat transfer coefficient, hc, is determinedas follows:

hc ={

2.38(tcl − ta)0.25, 2.38(tcl − ta)0.25 > 12.1v0.5a

12.1v0.5a , 2.38(tcl − ta)0.25 < 12.1v0.5

a(6)

where va is the air velocity (m/s).The environment variables, ta , tr , va , and RH, are obtained

by the proposed measurement system. The parameters, M , Icl,

and η, are constants obtained from [3, Tables B.1 and C.1].Then, PMV is calculated in an iterative process. The PMVindex establishes a quantitative expression about the rate ofbody heat storage and the thermal sensation. The quantitativeexpression represents the sensation of the majority of thepopulation under a same environment.

III. SYSTEM DESIGN

A. System Framework

Fig. 1 shows the whole system, which consists of thePMV portable mobile measurement, remote data transmis-sion, and accessing subsystems. First, the measurements ofair temperature, RH, air velocity, and black globe tempera-ture are converted into a standard voltage or current signalthrough signal conditioning and then sent into the integratedacquisition instrument. The integrated acquisition instrumentwirelessly transmits the analogy signal to the computer-aidedPMV measurement system software for data acquisition, PMVcalculation, and further analysis.

Moreover, the data transmission subsystem transmits themeasured data to the server-computer from a monitoringcomputer through the Internet. Then, a web page sited in aserver-computer is developed using PHP in order to displaythe measured data. Client computers can visit the web pageand access data remotely over the Internet.

B. Hardware Configuration

The thermal comfort measurement system hardware mainlyconsists of an acquisition instrument and a computer. Theacquisition instrument is divided into two layers: 1) the upperlayer consisting of the sensors and signal conditioning modulesand 2) the lower layer consisting of the data acquisitionmodules and power supply. The computer functions as themonitoring terminal. This configuration can reduce the tem-perature change caused by the heating of electronics and thedata acquisition module. Fig. 2 shows the physical hardwareconfiguration of our system.

To measure the indoor air temperature and RH, the tem-perature and humidity sensor is deployed, whose sensitivity is12 °C/V and 20%/V, respectively. Its power supply is 24 Vdc and its output signal is in the range of 0–5 V. Similarly,an air velocity sensor module is used to measure the indoorair velocity, which has a 24 V dc voltage supply, an outputvoltage signal of 0–10 V, and the sensitivity of 0.5 (m/s)/V.

Radiant temperature is related to the amount of radiant heattransferred from a surface. It depends on the material’s abilityto absorb or emit heat, namely emissivity. The mean radianttemperature experienced by a person in a room with sunlightstreaming in is based on how much of his or her body isunder the sunlight. Thus, the mean radiant temperature can bereplaced with the black globe temperature [17]. The black balltemperature sensor consists of a black ball and an integratedtemperature transmitter, and the sensor probe is located in thecenter of the black ball. The sensitivity of the black ball is15 °C/V with a 24 V dc voltage supply and a module outputvoltage signal of 1–5 V.

480 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 64, NO. 2, FEBRUARY 2015

Fig. 1. System framework of the proposed system. The architecture consists of three main parts such as PMV portable mobile measurement, remote datatransmission, and accessing system.

Fig. 2. Hardware configuration of the PMV portable mobile measurement subsystem.

A wireless measurement suite with 32 analog input channelsis used to transmit the collected data to the computer viaWi-Fi. The input analog signals are enhanced by a sharedprogrammable instrumentation amplifier and then fed into anA/D converter. The whole system shares a common-modevoltage and a common-mode noise rejection capability so thatthe quality of the measurement can be effectively improved.

Table I presents the details of the sensors and data acquisitiondevice discussed above.

C. Software Design

The computer-aided PMV measurement system softwareis developed using LabVIEW, which has a parameter setting

ZHU et al.: COMPUTER-AIDED MODELING AND MEASUREMENT SYSTEM 481

TABLE I

SENSORS AND DATA ACQUISITION DEVICE LIST USED IN THE SYSTEM

Fig. 3. Flow chart of the measurement system.

module, a thermal comfort module, an air quality module, aresidential properties module, an emergency stopping module,a historical data querying module, and a system settingmodule. Fig. 3 shows its main program flowchart.

By means of a parameter setting module, the user can input:1) the test location, construction parameters (e.g., area, numberof windows/doors, occupancy, and so forth), and measurementstarting and ending times and 2) the PMV parameter selectionand setting.

In the thermal comfort module, the following curves areshown in the measurement interface in real time: 1) airtemperature; 2) RH; 3) indoor air velocity; 4) black balltemperature; and 5) PMV. The highest, average, and lowestvalues of each curve in the interface are updated at a fixedinterval. If the testing data exceeds the upper or lower limitsset by the user, the system will give an alarm.

If an emergency occurs, the emergency stopping modulewill halt the measurement and return to the main interface.In addition, this program measures not only PMV, but alsothe air quality and residential properties as demanded.

D. Remote Data Transmission

In this system, data transmission between the monitoringterminal and the server-computer is realized based on theInternet. The network communication is realized via theSocket interface, as well as the application programminginterface (API) and the Internet protocol suite (TCP/IPprotocol suite). In our system, a LabVIEW function based onPHP language calls the API of Windows Sockets to achieve

Fig. 4. Flow chart for remote data transmission.

network communication. The underlying network protocoland operating system function is used by Windows Socketsfor actual communications.

On the server-side, we developed the web page for datareceiving using PHP language. The measured data is receivedthrough the socket communication interface. The mobileterminal or other client-computers can remotely view andmonitor the field measured PMV data by accessing the systemweb page.

Fig. 4 shows the program flowchart for remote data trans-mission. The data acquiring program on the server-side is theweb application developed using PHP language. This programbinds the IP address and the port number through a bindfunction and then monitors the port if there is a connectionrequest by activating the listen function. The LabVIEW-basedmeasurement system software simplifies the communicationwith the Internet via the optimization and management ofTCP/IP. The field measurement data can be broadcasted tothe remote terminal through the computer network withoutcomplex underlying TCP programming. In this paper, a TCPnode module is used to establish a connection to the Socketinterface on the server-side and sends data packets by an IPaddress and server port number.

IV. EXPERIMENTS AND DISCUSSION

A. PMV Calculation Process Verification

To verify the correctness of the PMV values calculatedby (1) in our system, we chose six groups of input variables

482 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 64, NO. 2, FEBRUARY 2015

TABLE II

RESULTS FOR THE SIX SETS OF TESTING INPUTS

TO THE NEW SENSING SYSTEM

from [3, Table D.1] to calculate the PMV values by oursystem. The differences between the example output from[3, Table D.1] and the outputs from our system are shownin Table II. It is clear that the difference in the PMV valuesis small, i.e., varying between 0 and 0.01 for all six groupsof data. The results verify the effectiveness of the PMVcalculation in our system.

B. System Error Analysis

The PMV multivariate function expression is expressed asPMV = f (ta, tr , RH, va). According to [11], the system errorof PMV, written as �PMV, is estimated as follows:

�PMV ≈∣∣∣∣∂PMV

∂ ta

∣∣∣∣ �ta +∣∣∣∣∂PMV

∂ tr

∣∣∣∣ �tr +∣∣∣∣∂PMV

∂RH

∣∣∣∣ �RH

+∣∣∣∣∂PMV

∂va

∣∣∣∣�va . (7)

This equation indicates that �PMV is influenced by theenvironmental variables and their measurement errors. Thesetwo kinds of influence are analyzed as follows.

1) Environmental Variables Influence on ΔPMV: To studyhow each environmental variable influences the PMV systemerror, 13 tests were conducted with different values of thefour environmental variables (ta, tr , RH, va). The 13 sets ofenvironmental variable values and the first partial derivativesare shown in Table III.

1) The change of ta in rows 1, 2, 3, and 4 results in thechanges of ∂PMV/∂ ta , ∂PMV/∂ tr , ∂PMV/∂RH, and∂PMV/∂va . Thus, ta affects the values of ∂PMV/∂ ta ,∂PMV/∂ tr , ∂PMV/∂RH, and ∂PMV/∂va .

2) The change of tr in rows 3, 5, 6, and 7, as well as thechange of va in rows 3, 11, 12, and 13 results in thechange of the values of ∂PMV/∂ ta , ∂PMV/∂ tr , and∂PMV/∂va .

3) The changes of RH in rows 8, 9, and 10 affect thecalculation of ∂PMV/∂ ta .

The measurement error of the four environmental variablesprimarily comes from the sensor error and the quantizationerror of the A/D signal conversion. As shown in (8), �T isthe measurement error of the environmental variable; �T1 isthe sensor error; n is the conversion resolution of A/D [18].In our system, n = 16, resulting in �T ≈ �T1. Thus, thequantization error of the A/D signal conversion is ignored inthe following analysis:

�T = (1 + �T1)(1 + 1/2(n+1))−1. (8)

The errors for each of the four variables in our system(ta, tr , RH, va) are �ta = ±0.3 °C, �tr = ±0.12 °C,�RH = ±3%, and �va = ±0.04 m/s, respectively. Asshown in Fig. 5, �PMV decreased linearly from 0.235 to0.135 as air temperature (ta) increased from 19 °C to 32 °C,respectively. As shown in Fig. 6, �PMV increased linearlyfrom 0.184 to 0.202 as the RH increased from 0% to 100%,respectively.

Fig. 7 shows that �PMV remains at ∼0.1 as the airvelocity increases from zero to <0.08 m/s. At an air velocityequal to 0.08 m/s, �PMV dramatically increases to 0.208.As the air velocity increases from 0.08 to 0.4 m/s, �PMVdecreases and asymptotically approaches to 0.15. It is clearthat �PMV is affected by the air velocity in the range from0.08 to 0.4 m/s.

2) Measurement Errors Influence ΔPMV: Different valuesof �PMV with different measurement errors in the environ-mental variables are shown in Table IV.

In the first row of Table IV, the measurement errors inthe environment variables are the required errors as listed in[17, Table 2], and a �PMV of 0.15 is calculated. In the secondrow, the measurement errors of the environment variables areabout the sensors used in our system. �PMV in these tworows are close.

In each of the next four rows, three of the four measurementerrors are set to zero, and the other measurement error is cal-culated under the condition that �PMV = 0.15. Because themeasurement error is always greater than zero, these four rowsof data indicate the minimum requirement of the measurementerrors for the four environmental sensors when �PMV asymp-totically approaches 0.15. Following the principle of equalprecision, the last row of data indicates the requirement for themeasurement errors of the four environmental sensors when�PMV asymptotically approaches 0.15.

C. System Test in an Indoor Environment

Before starting a measurement, a static calibration for eachsensor was carried out so that the displayed value of eachsensor is matched to the value measured by the preciseinstrument under the same condition. Then, the proposed

ZHU et al.: COMPUTER-AIDED MODELING AND MEASUREMENT SYSTEM 483

TABLE III

RESULTS OF PARTIAL DERIVATIVE OF PMV TO EACH ENVIRONMENTAL VARIABLE FOR 13 TYPICAL GROUPS OF ENVIRONMENTAL

VARIABLE VALUES IN THE CONDITION OF M = 58.15 W/m2 AND Icl = 1 CLO

Fig. 5. Curve of the relationship between system error of PMV and airtemperature in the condition of tr = ta , RH = 50%, va = 0.1 m/s,M = 58.15 W/m2, and Icl = 1 clo.

Fig. 6. Curve of the relationship between system error of PMV and RH inthe condition of tr = ta = 24 °C, va = 0.1 m/s, M = 58.15 W/m2, andIcl = 1 clo.

measurement system was tested under the following environ-mental conditions.

1) The indoor area was 50 m2.2) Four people were present.3) The indoor area had four windows and a door.

Fig. 7. Curve of the relationship between system error of PMV and airvelocity in the condition of tr = ta = 24 °C, RH = 50%, M = 58.15 W/m2,and Icl = 1 clo.

The PMV-related measurement parameters were:

4) the PMV sampling rate of one point per second;5) the activity of sit in and typical indoor clothing.

The following parameters were obtained from [3, Tables B.1and C.1].

1) The human body metabolic rate is M = 58.15 W/m2.2) The effective mechanical power is W = 0 W/m2.3) The clothing insulation is Icl = 1 clo.4) The clothing surface area factor is fcl = 1.15 by

calculation.

Under the above conditions, 72 points of data were obtained.The four environmental variables, the calculated PMV, andtheir distribution histograms are shown in Fig. 8.

D. Measurement Uncertainty Analyzing of PMV Using MCM

Guide to the expression of Uncertainty in Measurementuncertainty framework is a common method used to evaluatethe measurement uncertainty in linear systems. It requiresapproximation at each processing stage when the model is

484 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 64, NO. 2, FEBRUARY 2015

TABLE IV

RESULTS OF SYSTEM ERROR OF PMV WITH DIFFERENT MEASUREMENT ERROR OF EACH ENVIRONMENT VARIABLE IN THE

CONDITION OF ta = 24 °C, tr = 24 °C, RH = 50%, va = 0.1 m/s, M = 58.15 W/m2, AND Icl = 1 CLO

Fig. 8. Four environmental variables obtained from an indoor test and thevalues of PMV calculated by our system. (a) Air temperature. (b) RH. (c) Airvelocity. (d) Black globe temperature. (e) PMV.

nonlinear. As a result, the validity of the results may beunreliable. However, in this situation, MCM can generallybe expected to give valid results, since it does not makeapproximation assumptions [19]. Thus, MCM is deployed inthis paper to calculate the uncertainty of PMV that has anonlinear model. Fig. 9 shows its detailed calculation process.

MCM is a numerical method used to assess the measure-ment uncertainty, which can simulate the spread of uncertaintyby repeated sampling [20]. Its application needs to establishprobability models as inputs, and then requires the simulatingcalculation to determine a discrete expression of the output.

Fig. 9. Detailed calculation process of MCM.

Finally, the characteristics of the output are estimated byanalyzing the discrete expression.

According to the shape of the distribution histograms inFig. 8, we assumed that the four PMV inputs are the normaldistribution and their standard deviations equal their combinedstandard uncertainties, respectively. To simplify the uncertaintymodel, the combined standard uncertainty only includes theuncertainties caused by measurement repeatability and thesensor inaccuracy.

The uncertainty caused by the measurement repeatability(u1) belongs to type A uncertainty, which is obtained by

�i = xi − x̄ (9)

σ =

√√√√√n∑

i=1�2

i

n − 1(10)

u1 = σ√n

. (11)

ZHU et al.: COMPUTER-AIDED MODELING AND MEASUREMENT SYSTEM 485

TABLE V

THREE KINDS OF UNCERTAINTIES AND ASSUMED DISTRIBUTIONS

OF THE FOUR ENVIRONMENT VARIABLES

TABLE VI

MEASUREMENT UNCERTAINTIES OF PMV

WITH DIFFERENT SAMPLE SIZE

In the precision measurement, the width of the confidenceinterval is usually set at 3σ . Thus, the uncertainties caused bythe measurement accuracy of sensors (u2) belong to type Buncertainty, which equals one third of the accuracy valuesgiven by their specifications, respectively.

Both of the uncertainty components are independent; there-fore, the combined standard uncertainty is obtained by thefollowing equation [21]:

uc =√

u21 + u2

2. (12)

Table V presents the three kinds of uncertainties and theassumed distributions for the four environment variables.As can be seen, four groups of random numbers were gen-erated to function as the measurement errors of the fourenvironmental variables.

MCM requires a large number of sample data to estimate thecharacteristics of the output. As long as the sample sizes arelarge enough, the estimated result is accurate enough. In orderto choose a reasonable sample size, the PMV measurementuncertainty is calculated with the same input when M = 1000,2000, and 5000, respectively. Here, the input denotes theaverage values of the four environmental variables, expressedas ta , RH, va , tr , respectively.

Table VI shows the results that are calculated when theinput equals the average values (ta = 27.55 °C, RH = 63.3%,va = 0.057 m/s, and tr = 29.84 °C) of the 72 points shownin Fig. 8; then, M ≥ 2000, and the PMV measurement uncer-tainty is almost invariable. Therefore, to save computationtime, the sample size is set at 2000 in subsequent calculations.

Fig. 10 shows the distribution histogram of the output ofMCM under the conditions that ta = 27.55 °C, RH = 63.3%,va = 0.057 m/s, and tr = 29.84 °C. MCM is used to computerthe PMV values and the average value of PMV is 1.536 under

Fig. 10. Distribution histogram of PMV in the conditions of ta = 27.55 °C,RH = 63.3%, va = 0.057 m/s, and tr = 29.84 °C. (a) PMV calculated with72 points referred in Fig. 8. (b) PMV calculated with the use of MCM whenM = 2000.

TABLE VII

MEASUREMENT UNCERTAINTY OF PMV IN DIFFERENT

ENVIRONMENTAL CONDITIONS

these conditions. When the coverage probability P is 95%, thecoverage factor k p is 1.96. Hence, the expanded uncertaintyis ue = k p ∗ u = 0.035, and the PMV measuring result is1.536 ± 0.035. Moreover, the PMV values in Fig. 10(a) fallwithin this range.

Similar to the above process, the PMV measurement uncer-tainties are calculated under three different environmentalconditions and the results shown in Table VII. The PMVmeasurement uncertainty is not a constant; it changes withthe change of environment values.

V. CONCLUSION

This paper presents a new computer-aided thermal comfortmeasurement system with PMV, which consists of five mod-ules, namely multiparameter data acquisition, transmission,calculation, display, and data management. The system wastested under specific environmental conditions, and MCMwas used to analyze the PMV measurement uncertainty. Howthe measurement error of each environmental variable affectsthe PMV values was analyzed by MATLAB, which providesa theoretical basis for selecting the measurement accuracyof sensors. The PMV values calculated by our system arecompared with the PMV values obtained from the ISO 7730standard [3]. The result shows the correctness of our computerprograms for PMV calculation. Moreover, the PMV measured

486 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 64, NO. 2, FEBRUARY 2015

data can be transmitted to the server-side in a data center, andcan be remotely viewed through the Internet.

The main contribution of this paper is in proposing an inte-grated measurement system for indoor environmental thermalcomfort assessment based on virtual instrument technology.The system has two key features: 1) powerful mathematicalanalysis function for the complex calculation of PMV and2) the modular software design that offers flexibility and goodscalability in system operation. Our future work will focuson the evaluation of the system in the assessment of thermalcomfort.

ACKNOWLEDGMENT

This study was supported by NSFC grant (51205331) par-tially awarded to the corresponding author, Chenglu Wen.

REFERENCES

[1] Thermal Environmental Conditions for Human Occupancy,ANSI/ASHRAE Standard 55–2013, 2013.

[2] Indoor Environmental Input Parameters for Design and Assessment ofEnergy Performance of Buildings Addressing Indoor Air Quality, Ther-mal Environment, Lighting and Acoustics, BS EN Standard 15251:2007,2007.

[3] Ergonomics of the Thermal Environment—Analytical Determinationand Interpretation of Thermal Comfort Using Calculation of thePMV and PPD Indices and Local Thermal Comfort Criteria,BS EN Standard ISO 7730:2005, 2005.

[4] S. P. Corgnati, M. Filippi, and S. Viazzo, “Perception of the ther-mal environment in high school and university classrooms: Subjectivepreferences and thermal comfort,” Building Environ., vol. 42, no. 2,pp. 951–959, 2007.

[5] N. H. Wong and S. S. Khoo, “Thermal comfort in classrooms in thetropics,” Energy Buildings, vol. 35, no. 4, pp. 337–351, 2003.

[6] A. Pourshaghaghy and M. Omidvari, “Examination of thermal comfortin a hospital using PMV–PPD model,” Appl. Ergomomics, vol. 43, no. 6,pp. 1089–1095, 2012.

[7] S. Atthajariyakul and T. Leephakpreeda, “Real-time determination ofoptimal indoor-air condition for thermal comfort, air quality and efficientenergy usage,” Energy Buildings, vol. 36, no. 7, pp. 720–733, 2004.

[8] R. Z. Homod, K. S. M. Sahari, H. A. F. Almurib, and F. H. Nagi, “RLFand TS fuzzy model identification of indoor thermal comfort based onPMV/PPD,” Building Environ., vol. 49, pp. 141–153, Mar. 2012.

[9] P. O. Fanger, Thermal Comfort: Analysis and Applications in Environ-mental Engineering. Copenhagen, Denmark: Technical Press, 1970.

[10] L. X. Gao, “Research on the intelligent predictor of indoor thermalcomfort,” Comput. Meas. Control, vol. 10, no. 8, pp. 519–520, 2002.

[11] F. R. d’Ambrosio Alfano, B. I. Palella, and G. Riccio, “The role ofmeasurement accuracy on the thermal environment assessment by meansof PMV index,” Building Environ., vol. 46, no. 7, pp. 1361–1369, 2011.

[12] D. De Donno, L. Catarinucci, and L. Tarricone, “RAMSES: RFIDaugmented module for smart environmental sensing,” IEEE Trans.Instrum. Meas., vol. 63, no. 7, pp. 1701–1708, Jul. 2014.

[13] W. L. Tse and W. L. Chan, “A distributed sensor network for mea-surement of human thermal comfort feelings,” Sens. Actuators A, Phys.,vol. 144, no. 2, pp. 349–402, 2008.

[14] W. L. Tse and W. L. Chan, “Real-time measurement of thermal comfortby using an open networking technology,” Measurement, vol. 40, no. 6,pp. 654–664, 2007.

[15] L. B. Wang, Z. Y. Ge, and X. Y. Kong, “Greenhouse monitoring systembased on virtual instrument technology,” M.S. thesis, Dept. Biosyst.Energy Eng., Kunming Univ. Science Technology, Yunnan, China, 2010.

[16] W. L. Tse, A. T. P. So, W. L. Chan, and I. K. Y. Mak, “The validityof predicted mean vote for air-conditioned offices,” Facilities, vol. 23,nos. 13–14, pp. 558–569, 2005.

[17] Ergonomics of the Thermal Environment—Instruments for MeasuringPhysical Quantities, BS EN ISO Standard 7726:2001, 2001.

[18] M. J. Ma, Data Acquisition and Processing Technology, 3rd ed. Xi’an,China: Xi’an Jiaotong Univ. Press, 2012.

[19] Evaluation of Measurement Data—Supplement 1 to the ‘Guide to theExpression of Uncertainty in Measurement’—Propagation of Distribu-tions Using a Monte Carlo Method, JCGM Standard 101:2008 2008.

[20] J. Wang, J. Guo, H. Wang, and Y. Deng, “The evaluation of measurementuncertainty for laser tracker based on Monte Carlo method,” in Proc.IEEE Int. Conf. Mechatronics Autom., Beijing, China, Aug. 2011,pp. 608–612.

[21] L. Kirkup and R. B. Frenkel, An Introduction to Uncertainty in Mea-surement: Guide to the Expression of Uncertainty in Measurement.New York, NY, USA: Cambridge Univ. Press, 2006.

Qingyuan Zhu received the bachelor’s degree in vehicle engineering in 2002,and the Ph.D. degree in mechanical and electrical engineering from ChinaAgricultural University, Beijing, China, in 2009.

He was a Visiting Researcher with the School of Engineering, University ofIllinois at Urbana-Champaign, Champaign, IL, USA, in 2007. He is currentlyan Assistant Professor with the Department of Mechanical and ElectricalEngineering, Xiamen University, Xiamen, China. His current research interestsinclude vehicle dynamic simulation and modeling, robotics, and mechatronics.

Jian Yi received the B.Sc. degree from the Department of Mechanical Engi-neering, Xiamen University, Xiamen, China, in 2013, where he is currentlypursuing the Degree with the School of Physics and Mechanical and ElectricalEngineering.

His current research interests include vehicle dynamic simulation andmodeling.

Shiyue Sheng received the B.Sc. degree from the Department of MechanicalEngineering, Xiamen University, Xiamen, China, in 2011, where he iscurrently pursuing the Degree with the School of Physics and Mechanicaland Electrical Engineering.

His current research interests include vehicle dynamic simulation andmodeling.

Chenglu Wen (M’13) received the B.Sc. degree in mechanical automationin 2003, and the Ph.D. degree in mechanical and electrical engineering fromChina Agricultural University, Beijing, China, in 2009.

She was a Visiting Researcher with the School of Engineering, MichiganState University, East Lansing, MI, USA, from 2007 to 2009. She is currentlyan Assistant Professor with the School of Information Science and Engineer-ing, Xiamen University, Xiamen, China. Her current research interests includeintelligent robots, machine vision, and learning algorithms.

Huosheng Hu (M’94–SM’01) received the M.Sc. degree in industrial automa-tion from Central South University, Changsha, China, in 1982, and the Ph.D.degree in robotics from the University of Oxford, Oxford, U.K., in 1993.

He is currently a Professor with the School of Computer Science andElectronic Engineering, University of Essex, Colchester, U.K., where he isleading the Robotics Research Group. He has authored over 420 papers. Hiscurrent research interests include robotics, human–robot interaction, embeddedsystems, mechatronics, and pervasive computing.

Prof. Hu is a Founding Member of the IEEE Robotics and AutomationSociety Technical Committee on Networked Robots, a fellow of the Institutionof Engineering and Technology and the Institute of Measurement Control, anda Senior Member of the Association for Computing Machinery. He has beenthe Program Chair and a member of the Advisory/Organizing Committee formany international conferences, such as the IEEE International Conference onRobotics and Automation, the IEEE International Conference on IntelligentRobots and Systems, the IEEE International Conference on Mechatronics andAutomation, the IEEE International Conference on Robotics and Biomimetics,and the IEEE International Conference on Information and Automation.He currently serves as an Editor-in-Chief of the International Journal ofAutomation and Computing and the online Robotics journal, and an ExecutiveEditor of the International Journal of Mechatronics and Automation.