- Email: [email protected]
Local and overall thermal comfort in an aircraft cabin and their interrelations
Building and Environment 46 (2011) 1056e1064
Contents lists available at ScienceDirect
Building and Environment journal homepage: www.elsevier.com/locate/buildenv
Local and overall thermal comfort in an aircraft cabin and their interrelations Sumee Park a, *, Runa T. Hellwig a, b, Gunnar Grün a, Andreas Holm a a b
Fraunhofer Institute for Building Physics, Dept. Indoor Climate and Climatic Impacts, Fraunhofer Str. 10, 83626 Valley, Germany University of Applied Sciences, Dept. Energy Efficient Design and Building Climatology, Germany
a r t i c l e i n f o
a b s t r a c t
Article history: Received 28 May 2010 Received in revised form 8 November 2010 Accepted 11 November 2010 Available online 18 November 2010
In this study the interrelation between local and overall thermal comfort of passengers in aircraft cabins was investigated by thirteen simulated flights. For each of the tests forty test persons filled out questionnaires concerning their perceived overall and local thermal comfort at temperatures of 20 Ce25 C, which were measured at every second seat. With these physical and subjective data PMV (Predicted Mean Vote) and TSMV (Thermal Sensation Mean Vote) of test persons as well as PPD (Predicted Percentage of Dissatisfied) and PD (Percentage of Dissatisfied) were compared. The PMV was consistently similar to the TSMV, while the thermal dissatisfaction in tests was always higher than PPD. The hypothesis at the beginning of this study was that the high ratio of thermal dissatisfaction in the aircraft cabin reported in literature might be caused by local discomfort. Therefore statistical analyses about the interrelations between local and overall thermal comfort were performed and models indicating such interrelations were developed. Some local perceptions are significantly different from overall thermal perception and these body segments alter in dependence of the overall thermal environment. Also body segments rated similarly were detected and these segments were pooled to distinct body regions using principal component analysis. Under the same overall thermal sensation the local thermal perception on a certain body region predominantly influenced the overall thermal comfort. Therefore weighting factors of local body regions on the overall thermal comfort were determined in dependence of the overall thermal sensation by means of multiple linear regression models. Ó 2010 Elsevier Ltd. All rights reserved.
Keywords: Local thermal comfort Overall thermal comfort Statistical analysis Aircraft cabins
1. Introduction There are only a few investigations published on thermal comfort in aircraft cabins, while a comfortable environment in aircraft cabins is a major factor in airlines business competition. The existing investigations [1,2] indicate the character of the indoor environment of aircraft cabins as follows: low relative humidity (typically below 15%), vertical temperature difference (typically 3 K), varying air velocity depending on the location in and the type of cabin (from 0.1 m/s to 0.6 m/s). These inhomogeneous thermal environments can lead to an unwanted local cooling or warming and cause thermal dissatisfaction (approx. 25% of occupants). There are internationally established models (DR and PD due to vertical air temperature gradients) for the assessment of these local thermal discomfort phenomena [3], which are valid only for persons in overall thermal neutral status. However, the local thermal comfort may vary depending on the overall thermal situation or the local thermal comfort feeling might influence the
* Corresponding author. Tel.: þ49 8024 643 237; fax: þ49 8024 643 366. E-mail address: [email protected] (S. Park). 0360-1323/$ e see front matter Ó 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.buildenv.2010.11.003
overall thermal evaluation. In addition, the models do not consider the body segment differences regarding local cooling or warming. Extensive local and overall thermal comfort investigations on inhomogeneous environments have been performed since the 1990s focusing mainly on the assessment of the climate in vehicles. There were two approaches; one is the assessment with equivalent temperature using multi-segmented thermal manikins or heated sensors [4,5]. The result of this approach is established in ISO 14505-2 as annex D2: “Interpretation of equivalent temperature in terms of perception of thermal sensation and thermal comfort” [6], which identifies the acceptable equivalent temperature depending on the body segment. A similar study was performed in a simulated aircraft cabin in recent years [7]. Another approach is the multi-segment physiological model combined with computational fluid dynamics CFD, which predicts the local skin temperature of human body in inhomogeneous environments (e.g. [8]). However, this approach evaluates the overall thermal comfort without considering the local thermal sensation but only with regard to the mean skin temperature. According to these results the cooling of the feet and at the same time the heating of the head cause an identical overall thermal comfort as without local stimulation, although the observed
S. Park et al. / Building and Environment 46 (2011) 1056e1064
perceptions are generally different. Zhang [9] studied this subject and developed a comfort model combined with the physiological model of Huizenga [10]. Zhang concluded that overall thermal comfort cannot be determined from mean skin temperature but from the local discomfort or comfort derived from local thermal sensation depending on the overall thermal sensation. Since the primary interest of the Zhang study was the vehicle cooling system, the test conditions were focused on typical vehicle indoor climate in summer, where impacts of cooling of one or multiple local body segments for 10e20 min on the overall comfort were studied. Thus the experimental conditions are not similar to the typical indoor climate of aircraft cabins. In addition, the influence of local stimulation duration and intensity were not varied in the test, which might cause other results. Furthermore Zhang developed a comprehensive model except for overall thermal comfort. The overall thermal comfort is determined using some rules. The objectives of the present study are to develop a statistical model indicating local and overall thermal comfort and their interrelations based on a subject study with the typical indoor climate of aircraft cabins and long flight duration. This paper concentrates on the methodological approaches of this study and statistical analyses about the interrelations between local and overall thermal comfort (Fig. 1). 2. Methods A subject study was performed in winter 2006 in a simulated aircraft cabin, the Fraunhofer Flight Test Facility (FTF) (Fig. 1). It consists of a 30 m long pressure vessel, which holds the first 16 m of a complete wide-body aircraft. The study consisted of 11 simulated 3.5h-flights and two 7h-flights, in which the air temperature differed from 20 C to 25 C. Forty different test persons were acquired for each of the flights according to specified gender and age profiles. Approximately 20 males and females participated in each test. The mean age profile achieved was as follows: 34% of the test persons were within the range of 18e35 years, 32% of the test persons were within the range of 36e55 years; 34% of the test persons were 55 years and older. The air temperatures were measured at every second seat in the heights of 0.1 m, 0.6 m and 1.1 m (Fig. 2). Relative humidity was also detected in 1.1 m height. At two locations the air velocities were measured in the heights analogous to the air temperatures. To relate the physical environment to the test persons’ answers the mean values of the corresponding 20-minquestionnaire interval were calculated. To analyze the mean air temperature in each test firstly the mean value of three heights was calculated at
Fig. 1. Subject study in the Fraunhofer Flight Test Facility (FTF).
1057
each seat, afterwards the mean value of every seat was estimated. The measured mean air temperatures at three heights as well as standard deviations between seats and mean vertical differences of air temperature between 1.1 m and 0.1 m in each test are reported in Table 1. The test persons were allowed to wear and adapt their own clothing and to use blankets provided as during common flights. Meals and drinks were provided on a regular basis. The typical high noise and vibration level (70 dB(A)) in aircraft cabins are simulated as well as the low air humidity (10e20%) and low air pressure (750e800 Pa). The measured mean air humidity, air pressure and air velocity in each test are listed in Table 1. The clothing values of test persons are calculated using the questionnaire. The test persons were allowed to move inside the cabin and to spend their time on their own way apart from the filling out of questionnaires. The test persons’ perception was asked by questionnaires at 70 min, 120 min and 170 min after the start of each test for 3.5 h test duration (Fig. 3). The questionnaires were presented on seven-point scales. In this paper questions related to the local thermal perception (too coldetoo warm scale) on 11 body segments as well as overall thermal perception, sensation (ASHRAE scale) and satisfaction (satisfiededissatisfied scale) are considered (see Fig. 4). The software package StatisticaÒ 7.0 (Statsoft. Tulsa. USA) was used for statistical analysis. Since the scales used in this study can be interpreted as interval scales [11] and additionally a huge sample number is available parametric statistics were used: the t-test for paired samples to analyze the differences of local thermal perception from overall and principal component analysis to detect similarities and differences between local perceptions of body segments. For the overall thermal comfort analysis the multiple regression model based on the least-squares model was chosen. All statistical analyses and the modeling as well as the validation in this paper are based on individual votes of the test persons unless they are indicated as “MV (Mean Vote)”. The first questionnaire (Q1) carried out during cruise at 70 min or respectively 140 min in 7 h flights after take-off was used for the development of the model and 500 out of 520 responses were
Fig. 2. Air temperature measurement locations in the FTF.
1058
S. Park et al. / Building and Environment 46 (2011) 1056e1064
Table 1 Test condition: Flight duration, set temperature, set altitude and mean values of the measured physical parameters in aircraft cabin during the questionnaire action in first cruise. Test
1
2
3
4
5
6
7
8
9
10
11
12
13
Flight duration Set altitude (km) Mean outdoor temperature ( C) Cabin air pressure [hPa] Air temperature ( C) SD between seats [K] Vertical difference [K] Mean clothing (clo) Relative humidity (%) Air velocity (m/s) Noise (dB(A))
3.5 0.7 8.4 939
3.5 0.7 8.6 939 22.5 0.4 2.5 1.2 15 0.06 70
3.5 1.8 7.1 808 25.1 0.2 2.4 1.0 14 0.06 70
3.5 1.8 9.4 808 24.1 0.4 3.4 1.1 14 0.06 70
7 1.8 9.1 809 24.5 0.4 2.4 1.2 15 0.06 70
3.5 1.8 4.3 808 19.7 0.5 2.3 1.3 18 0.04 70
3.5 2.4 0.7 748 20.6 0.4 1.7 1.0 18 0.05 70
3.5 2.4 1.3 749 21.1 0.6 1.2 1.1 18 0.05 70
3.5 2.4 2.9 749 24.1 0.4 4.2 1.1 14 0.06 70
7 2.4 5.6 749 22.8 0.4 1.7 1.0 16 0.05 70
3.5 2.4 10.2 749 22 0.6 3.6 1.2 15 0.05 70
3.5 2.4 3.9 749 25.1 0.5 3.6 1.0 15 0.06 70
3.5 0.7 2.7 920 25.1 0.3 2.8 1.1 18 0.06 70
a
70
Outdoor temperature: mean outdoor temperature on the test site during the test day, SD ¼ Mean value of standard deviation of air temperature between seats, Vertical difference: Mean of differences of air temperature between 1.1 m and 0.1 m, Grey: mean values of the measured parameters in the cabin. a no physical measurements, but questionnaire data available.
Fig. 3. Test schedules (left: 3.5 h flight duration, right: 7 h flight duration) Q: Questionnaire L: Lunch.
considered as valid. The physical environment during this period was analyzed in the present study. 3. Analysis of the interrelation of local and overall thermal perception in aircraft cabins Before the local and overall thermal comfort model for aircraft cabin is developed, the following questions were analyzed to get an idea of possible interrelations between local and overall thermal comfort and for the modeling approach. 1. Is the local thermal perception generally different from the overall thermal perception in aircraft cabins? If yes, which body segments differ?
2. Could the overall thermal satisfaction be derived from overall thermal sensation? 3. Might the local discomfort or comfort influence the overall thermal comfort under the same overall thermal sensation? 4. How many local segments are necessary for the thermal comfort analysis in typical aircraft cabins?
3.1. The difference between local and overall thermal perception For the analysis the local and overall seven-point bipolar (too cooletoo warm) scales were transformed to (3) to (þ3). Fig. 5 shows the average difference between the local and overall thermal judgments in 11 body segments in test 6 (Ta ¼ 19.7 C,
Fig. 4. Seven-point scales presented to the subjects.
S. Park et al. / Building and Environment 46 (2011) 1056e1064
1059
Fig. 5. Deviation of local thermal judgments from overall thermal judgments in 11 body parts in test 6 (left: MV ¼ 1.6, Ta ¼ 19.7 C, N ¼ 40) and in test 12 (right: MV ¼ 0.5, Ta ¼ 25.1 C, N ¼ 40).
N ¼ 38) and in test 12 (Ta ¼ 25.1 C, N ¼ 40). In test 6 the test persons evaluated the upper body clearly warmer and the lower body a little colder or not different than overall, while in test 12 they judged all body segments except the head much colder than overall. In order to find whether these differences are significant or not, a t-test for paired samples was performed in each test. The thermal perceptions of some body segments are significantly different from the overall thermal perception (see Table 2). Such body segments alter depending on experimental conditions. At cool air temperature (test 6, 7, 8) the test persons assessed only the upper half of the body to be significantly warmer than overall. Thus the overall thermal perception is similar to the thermal perception of the lower half of the body. In contrast to that the overall thermal perception in the neutral or slightly warm environment is significantly different from the local perception of the lower half of the body which is always recognized to be cooler than overall (except for test 5). In this case the overall thermal perception is rather comparable to the upper part of body. The results of test 5 and test 10 with 7 h flight duration differed from the other tests with 3.5 h duration. In both tests the first questionnaires were completed after a warm lunch (see Fig. 3), which might strongly influence the thermal judgments of subjects. The test persons evaluated the local thermal perception not to be
significantly different from the overall perception except the head in test 5 and feet in test 10. They felt warmer and more satisfied than in the comparable tests. However, the missing value ratio in test 10 was the highest, where only 34 people from 40 filled out the satisfaction question. 3.2. The difference of overall thermal sensation and overall thermal satisfaction The thermal comfort assessment in ISO 7730 [3] is based on the assumption that the three central categories in the ASHRAE scale (slightly cool, neutral, slightly warm) indicate thermal comfort, thus the thermal satisfaction and thermal sensation are interchangeable in a homogeneous environment [12]. Is this assumption valid also for an inhomogeneous environment like aircraft cabins? However, it was a controversial issue in the previous experimental studies also in a homogeneous environment. Mayer [13] found in his experimental study in a climatic chamber that test persons who voted “(1) slightly cool” did not feel thermally comfortable, while test persons who judged the thermal environment as “(þ1) slightly warm” evaluated the environment as thermally comfortable. Therefore he suggested another PPD (Predicted Percentage of Dissatisfied) calculation depending on PMV, where the most
Table 2 Differences between overall and local thermal votes on body segments in each test (1. questionnaire). Test
1
2
3
4
5
6
7
8
9
10
11
12
13
N PMV TSMV PD [%] Head Neck Torso Right upper arm Left upper arm Right lower arm Left lower arm Right leg Left leg Right foot Left foot
40
40 0 0.4 48 (**) (*) (*) (*) (*)
38 0.3 0.6 37
39 0.2 0.3 31
40 0.4 0.7 25 (**)
38 0.5 1.5 68 (***) (***) (***) (***) (***) (***) (***)
39 0.7 1 64 (***) (***) (***) (***) (***) (***) (***)
38 0.5 0.7 55 (***) (***) (**) (**) (**) (*) (*)
37 0.2 0.2 22 (*) (*)
34 0.2 0.1 9
39 0.1 0.4 36 (**) (*)
40 0.3 1 25 (***)
40 0.4 0.7 33
*** *** *** ***
** * ** **
*** *** *** ***
* ** *** ***
0.6 25
*** ** *** ***
** ** *** ***
(**) (*)
*** *** *** ***
* **
*** *** *** ***
N: Number of samples PMV: calculated from the measured values in Table 1 (metabolic rate of persons as 1met and radiation temperature in the middle of the cabin as air temperature assumed). TSMV: Thermal Sensation Mean Vote, PD: Percentage of Dissatisfied. *: p < 0.05, grey: mean is cooler than overall. **: p < 0.01. ***: p < 0.001. (*): mean is warmer than overall.
1060
S. Park et al. / Building and Environment 46 (2011) 1056e1064
Fig. 6. Thermal comfort judgments depending on Thermal Sensation Vote (TSV) and Thermal Perception Vote (TPV).
comfortable point is defined as PMV ¼ 0.4 with PD ¼ 15% (see Fig. 8). McIntyre [14] found already in 1978 similar results in his field experiment in winter however the opposite in the summer experiment. Humphreys and Nicol [15] explained such phenomena with subjective association of a word “cold” and “warm” depending on the outdoor climate. People in a cold climate or in winter would probably associate the word “warm” as “comfort” and respectively people in warm a climate for the word “cool”. In this study a seven-point bipolar comfort scale (satisfiededissatisfied) in addition to a seven-point ASHRAE scale and a seven-point bipolar perception scale (too warmetoo cold) were used for overall thermal assessment. Fig. 6 shows the different judgments of the test persons depending on different scales. 10% of the test persons perceiving overall thermally “neutral” judge the climate still dissatisfying as rating (1), (2) or (3) on the thermal satisfaction scale. Only 11% of the test persons rating “slightly warm” vote “dissatisfied” in contrast to 58% dissatisfaction of the test persons rating “slightly cool”. The judgment on the bipolar perception scale also does not match to the judgment of the test persons on the thermal comfort scale (see Fig. 6, right). Did the test persons feel objectively comfortable under physical “slightly warm” conditions like “neutral” conditions or were these judgments rather influenced by the verbal psychology as Humphreys and Nicol assumed? In Fig. 7 the percentage of dissatisfaction and the thermal sensation mean vote are compared for each temperature range of the air temperature measured at the seats. The mean air velocity (Va ¼ 0.06 m/s) and the mean clo value (clo ¼ 1.1) are applied for the calculation of PMV at each air temperature. The subjects evaluated the environment slightly warmer than predicted by the PMV in neutral or warm environments (PMV > 0.5) and cooler in cool environments (PMV < 0.75). The subjective judgments seem to be more sensitive than the physical assessment PMV. Interestingly the ratio of dissatisfied is lowest at 24 C with TSMV ¼ 0.4 and PMV ¼ 0 except at 26 C (Fig. 8). The sample number (N) at 26 C (PMV ¼ 0.5) is with 12 too small compared to other temperatures, for example N ¼ 119 at an air temperature of 25 C in which the ratio of dissatisfied was obviously larger than at 24 C. The result that the test persons feel satisfied at PMV ¼ 0 and at TSMV ¼ 0.4 strengthens the assumption of Humphreys and Nicol. The study was carried out in Germany from 30 January to 17
February 2006 and during this period the mean air temperature on the test site was 5 C. The cold outdoor temperature during that seasonal period might have influenced the thermal comfort assessment of the test persons. They assessed the physical neutral situation (PMV ¼ 0) as slightly warmer (TSMV ¼ 0.4) in which they feel most comfortable. However the percentage of dissatisfied is obviously higher than the estimation of PPD, which might be based on the difference between local and overall thermal perception (see Section 3.1). Fig. 8 shows that PPD according to Fanger can not be applied for the assessment of thermal comfort in aircraft cabins. The revised
Fig. 7. Comparison of PMV and TSMV depending on local air temperature (Tr ¼ Tr, Va ¼ 0.06 m/s, clo ¼ 1.1).
S. Park et al. / Building and Environment 46 (2011) 1056e1064
1061
Table 3 Comparison of the local thermal perception of the satisfied and dissatisfied group under same thermal sensation. Overall thermal sensation
Slightly cool (n ¼ 112)
Neutral (n ¼ 142)
Slightly warm (n ¼ 109)
Warm (n ¼ 67)
n:(Satisfied/dissatisfied) Head Neck Torso Right upper arm Left upper arm Right arm Left arm Right leg Left leg Right foot Left foot
(41/71)
(127/15)
(97/12)
(44/23) **
* * * (***) (***) (***) (**)
(*) (*) (**) (**)
* *
*: p < 0.05, **: p < 0.01, ***: p < 0.001, n ¼ sample number (*): mean is warmer in the satisfied group; *: mean cooler in the satisfied group.
Fig. 8. Percentage of dissatisfied (PD) depending on local air temperature (Tr ¼ Tr, Va ¼ 0.06 m/s, clo ¼ 1.1).
model of suggested by Mayer predicted the proportion of subjects thermally dissatisfied better than the PPD index was able to do, however the optimal thermal condition in this study was not at the PMV ¼ 0.4 as in Mayers model but at PMV ¼ 0 at an air temperature of 24 C (see Section 6.2). 3.3. The relation of local thermal perception and overall thermal comfort In order to define whether local thermal perceptions influence the overall thermal comfort under the same thermal sensation the local thermal perceptions of the two groups (satisfied and dissatisfied) stating the same overall thermal sensation are compared in Table 3 using the t-test for two independent samples. The mean local thermal perception of legs and feet in the satisfied group is significantly warmer than in the dissatisfied group in “neutral” and “slightly cool” overall sensation, while no significant difference is found for the “slightly warm” state. For the overall “warm” state the satisfied group feels locally cooler in the upper part as well as partially in the lower body part. Especially the perception of the head was significantly different between two groups. No test person indicating a “cool” thermal sensation was satisfied. 3.4. The relation of local thermal perceptions between local body segments How sensitive can people perceive thermal differences impacting their body in a typical indoor environment? Which partitions of the human body are sensible for the thermal comfort research? Such questions have not yet been scientifically discussed in depth. Therefore in previous local thermal comfort investigations always many different body parts were investigated. Another important question for modeling in this study is how different or similar local body segments react to the typical climate of aircraft cabins. This can be explained by means of principal
component analysis. The original aim of this statistical analysis is to reduce several variables to a smaller number of factors. This is also required in order to define the independent variables for the development of statistical models since strong linear correlations between variables (multicollinearity) can result in large uncertainties of the estimation coefficients of a regression model [16]. Multicollinearity is a general problem of regression analysis and occurs, if two independent variables have a very strong correlation. This kind of strong correlation was already observed in this study during the analysis of the relation between local and overall perception (see Table 2). The thermal perception of the upper arm and lower arm did not differ from each other in this study. In order to avoid this multicollinearity in a regression model the principle component analysis is recommended in statistical literature. By means of the principal component analysis the eleven local body segments could be extracted to the following three factors: factor 1: upper part; factor 2: lower part and factor 3: head and neck (Fig. 9). An extraction of these three factors still explains more than 90% of the original variance with 11 body segments.
4. Modeling of overall thermal comfort for aircraft cabins The analysis described above provides the following results for the model development. 1. Overall thermal sensation (OTS) and overall thermal comfort (OTC) require different models 2. Three body regions instead of eleven body segments can be used for modeling 3. The local thermal perception might determine the OTC 4. The weighting coefficients of local thermal perception (LTP) determining the OTC may alter in dependence of OTS The local thermal perceptions (LTP) of the three body regions (upper part, lower part and head/neck) have been estimated from the LTP of eleven body segments. In addition, two new variables, the maximal difference between the most extreme LTP and also the most extreme LTP are included as predictors for modeling. Afterwards the overall thermal sensation (OTS) was estimated from these variables and finally overall thermal comfort (OTC) model was developed from the LTP with regard to three groups of OTS (cool, neutral, warm). The approach of modeling can be found in Fig. 10. Where y is criterion variable (in this study: OTS or OTC), x are the predictor variables (in this study: LTPs and the maximal
1062
S. Park et al. / Building and Environment 46 (2011) 1056e1064
Fig. 9. Factor loadings of 11 local thermal perception variables on three factors 1: Head 2: Neck 3: Torso 4: Right upper arm 5: Left upper arm 6: Right lower arm 7: Left lower arm 8: Right leg 9: Left leg 10: Right foot 11: Left foot.
y ¼ b0 þ b1 x1 þ b2 x2 þ b3 x3 þ b4 x4 þ 3
(1)
At first the modeling was performed always with more than five predictors. If the individual coefficient of a predictor in the model was not significant, a new model was formed without such a predictor. If the new model performs better with fewer variables, this predictor will be permanently removed from the model. In the process the extreme LTP proved to be an insignificant predictor for all models. The final result of modeling is shown in Table 4. The OTS is influenced by three predictors, while OTC “cool” is determined by the upper and lower body. The great difference between individual maximum LTP reduces the thermal comfort only in the thermally “neutral” state in which the warm lower body part increases thermal comfort. In warm conditions the head region is responsible for the overall thermal comfort.
Fig. 10. Model approach.
difference between the most extreme LTP), b are the regression coefficients, 3 is the residual. Half of the sample data chosen at random by the split-node method of STATISTICA was used for modeling and another half of the data was used for validation later. This procedure is based on the Equal Probability of Selection Method (EPSEM) in which every case in the population of interest has an equal opportunity of being selected for the sample [17]. With this method the validation can be performed with a sample drawn from the same population with the same exposure conditions as for the modeling. In order to find the influences and the intensities of LTP to the OTC depending on OTS a multiple regression was applied based on the method of least squares in which the sum of the squared differences between modeled data and observed data in all observations must have the least value, i.e. a function with the smallest sum of the residuals P ð ð3i Þ2 Þ is investigated (see Eq. (1)).
5. Validation The OTS and OTC were calculated using the developed models with the other half of sample data. The standard errors of OTS, OTCcool, OTC neutral , OTC warm of modeling and validation are reported in Table 5. OTS and OTC at cool sensation were better predicted than overall thermal comfort at neutral and warm thermal sensation. No major differences between modeling and validation could be observed. 6. Discussion 6.1. Limitations of the model The experimental conditions in this study are different from real aircraft cabin climate in the following aspects. First the outside fuselage temperature in this study was not similar to the
Table 4 Results of the modeling. Model
Adjusted.R2
Coefficients
Table 5 Standard errors of models for modeling and validation.
LTP Head/ LTP Upper LTP lower LTP max. Constant Neck part part diff. OTS 0.68 OTC cool 0 OTC neutral 0 OTC warm 1.00
0 0.74 0 0
0.27 0.51 0.40 0
0.16 0 0.50 0
0.33 0.55 1.79 1.63
OTS 0.52 0.40 0.26 0.15
Modeling Validation
OTCcool
OTCneutral
OTCwarm
N
SE
n
SE
N
SE
N
SE
249 251
1.0 0.9
70 64
1.2 1.3
126 112
1.3 1.4
53 75
1.6 1.7
N ¼ sample number, SE ¼ Standard Error.
S. Park et al. / Building and Environment 46 (2011) 1056e1064
temperatures during real flights (approx. 30 C). The cool surface in real flights could cause different local thermal sensations between outer and inner side of the body observed in [7]. Thus the allocation of body segments could be different inside a real aircraft cabins. However, analysis with another further three experiments under fuselage cooling also showed an insignificant difference between side and inside seat locations. The lack of significance could be due to the small subjective numbers with 40 subjects in each test and the spatial variation in the cabin. Second, the air velocity measured in the simulated aircraft cabin was less than 0.1 m/s and thus lower than reported from real aircraft cabins. Consequently the local discomfort and its influence on the overall thermal comfort could be higher than in this study. Therefore the further test person studies in FTF after this study were performed under physical conditions of fuselage cooling and higher air velocities but have not yet been included into the model. 6.2. Thermal comfort assessment based on subjective and objective data The aim of the thermal comfort modeling in this study is to understand the thermal comfort or discomfort of passengers in aircraft cabins and to provide a tool for the planer. This is only possible based on a combined subjective and objective assessment. For a rough estimation or mean judgment for the planning phase of an air-conditioned room there is already Fangers PMV/PPD approach which is applied and established worldwide. This approach was confirmed in parts again in this study. The ratio of satisfied persons was highest at the calculated PMV ¼ 0 environment, but not at 95% but at 76%. The first hypothesis in this study was that this high ratio of dissatisfaction reported already in literature might be caused by local discomfort. The local thermal perception on the lower body part is significantly different between the satisfied and the dissatisfied group under overall thermal sensation “neutral“ or at an air temperature of 24 C at “PMV ¼ 0”. The satisfied group clearly feels warmer in the foot and leg region than the dissatisfied group. So the preliminary hypothesis could be confirmed by the questionnaire analysis. However, the warmer perception on the lower body part of the satisfied group could not be physically explained in this study. A higher local air or radiation temperature, lower local air velocity or a higher local clothing insulation value may physically result in this warmer perception. The local air temperature (height: 0.1 m) was measured on the test persons’ seats and the local clothing insulation values were gained from individual clothing information The local air temperature between the two groups at 24 C was not different and the local clothing insulation value on the feet was higher in the satisfied group, but not significantly (see Table 6).
Table 6 Comparison of satisfied and dissatisfied group in 24 C air temperature (mean air temperature from 0.1 m, 0.6 m and 1.1 m). Satisfied Dissatisfied P (n ¼ 66) mean (n ¼ 20) mean value value value Questionnaire: local thermal perception Physical measurement
Right leg Left leg Right foot Left foot Local air temperature (0.1 m) ( C) Local clothing insulation value (Feet) (clo)
0.20 0.12 0.25 0.16 21.8
1.30 1.10 1.65 1.55 21.9
0.000 0.001 0.000 0.000 0.794
3.47
2.87
0.137
1063
The local air velocity or local surface temperature might result in this different perception. This cannot be proved in this study due to the missing local measurements. In order to prove this impact indirectly perceptions of the various seat positions of the test persons (side of air craft/aisle/another passenger or according to row) were compared, but the results showed no significant relation between seat position and local thermal perception at the same air temperature. 6.3. Local and overall thermal comfort model for near to neutral temperature with different local conditions It might be difficult to develop an accurate thermal comfort model based on physical data in a typical indoor climate near to neutral temperature because individual physiological or psychological differences are often greater than objective physical differences. However, it is important for the design of most comfortable indoor environments to have the possibility at hand to assess local and overall thermal comfort under near to neutral conditions with different local conditions. The existing local thermal considerations in ISO 7730 focus on the “discomfort” side of the thermal asymmetry. Zhang noted and reflected on a possible positive side of asymmetric thermal environments in her LTC model, however only limited in the OTC consideration. According to her OTC model the intensities of a maximum LTC or two minimum LTCs are deciding an overall thermal comfort independent from where in 19 body segments (dis) comfort is indicated. The model does not address the overall most thermally comfortable condition by means of local asymmetry, while the LTC model indicates it for local segments. In this study under near to neutral conditions the warm leg and foot region and low maximum differences between local thermal perceptions are deciding on OTC, while the cooler head region is more significant for OTC under slightly warm conditions. However, it could be that a warmer torso and a warmer foot region at a slightly cool air temperature can result in an overall neutral sensation and may be perceived overall more thermally comfortable with a cool head and warm feet. 6.4. Future research Due to the limitation of exposure conditions in this study the individual regression coefficients in models could be different in the real aircraft cabin. However this study shows that OTC in an inhomogeneous environment can not be defined from thermal sensation only, since the LTP influences the OTC under same OTS and the assessment of thermal sensation could be affected by the expectations and preconditioning depending on the outdoor temperature (which may even be different for aircraft passengers on the same flight). In an in-homogenous environment, a high level of thermal comfort, maybe higher than in a homogeneous environment, could be achieved by careful control of the local thermal environment. This indicates that further studies are required at near to neutral temperature with a local variation of the thermal environment to find the most comfortable condition. 7. Conclusion In this study while PMV of Fanger predicted overall thermal sensation in the simulated aircraft cabin relatively well, the thermal dissatisfaction (PD) of test persons is obviously higher than the PPD of Fanger, which might be resulted from the local thermal discomfort. The LTP of some body segments can be significantly different from the OTP and also significantly different between satisfied and dissatisfied groups under same OTS or at the same air temperature. These statistical analyses indicated that the OTC in the aircraft cabin can be derived from the LTP and OTS. However which
1064
S. Park et al. / Building and Environment 46 (2011) 1056e1064
partitions of the human body are sensible for the thermal comfort research under inhomogeneous environments? In previous local thermal comfort investigations always many different body parts were investigated. In order to answer this question in this study body segments rated similarly were detected and these segments were pooled to distinct body regions using principal component analysis. Also this procedure can minimize the error based on multicollinearity in the multiple regression models. The eleven body segments investigated in this study were reduced to three body regions due to the similarity of thermal perceptions. However the allocation of body segments could be different inside real aircraft cabins, since the experiments in this study were performed without fuselage cooling and with lower air velocity than in typical aircraft cabin environments. In the neutral sensation case the thermal perception of the lower part of the body dominates the OTC, while the LTP of the head and neck influences OTC under warm conditions. Based on these effects three different OTC (Overall Thermal Comfort) models have been developed depending on OTS (Overall Thermal Sensation) using the LTP (Local Thermal Perception) of body regions and the maximum difference between LTPs. Acknowledgement The data analyzed in this paper was collected within the framework of the EU-FP5 project FACE (contract No. G4RD-CT2002-0076). The authors thank the FACE consortium for the good collaboration. References [1] Spicer CW, Murphy MJ, Holdren MW, et al. Relate air quality and other factors to comfort and health symptoms reported by passengers and crew on commercial transport aircraft 2004; Report No.: Part I.
[2] Ross D, Crump D, Hunter C, Perera E, Sheridan A. Extending cabin air measurements to include older aircraft types utilised in volume short haul operation. UK: BRE 2003; Report No.: 212034. [3] En Iso 7730:2005. Ergonomics of the thermal environment eAnalytical determination and interpretation of thermal comfort using calculation of the PMVand PPD indices and local thermal comfort criteria. EN ISO 7730: 2005. [4] Nilsson HO, Holmer I, Bohm M, Norén O. Equivalent temperature and thermal sensation - comparison with subjective responses. Comfort in the automotive industry. Bologna; 1997, p. 157e162. [5] Schwab R. Einfluss der Sonneneinstrahlung auf die thermische Behaglichkeit in Kraftfahrzeugen. FAT Schriftenreihe; 1994. FAT 109. [6] prEN ISO 14505e2:2004: Ergonomics of the thermal environment. Evaluation of thermal environment in vehicles Part 2: Determination of equivalent temperature. prEN ISO 14505-2:2004. [7] Strøm-Tejsen P, Wyon DP, Zukovska D, Jama A, Fang L. Occupant evaluation of 7-hour exposures in a simulated aircraft cabin - Part 2: Thermal effects. Proceedings of Indoor Air 2005, Beijing, China; 2005, p. 46e51. [8] Tanabe S, Kobayashi K, Nakano J, Ozeki Y, Konishi M. Evaluation of thermal comfort using combined multi-node thermoregulation (65MN) and radiation models and computational fluid dynamics (CFD). Energy and Buildings 2002;34(6):637e46. [9] Zhang H. Human thermal sensation and comfort in Transient and NonUniform thermal environments, PhD Thesis. Berkeley: University of California; 2003. [10] Huizenga C, Zhang H, Arens E. A model of human physiology and comfort for assessing complex thermal environments. Building and Environment 2001;36:691e9. [11] En Iso 10551:2002. Ergonomics of the thermal environment e Assessment of the influence of the thermal environment using subjective judgement scales. EN ISO 10551:2001. [12] Fanger PO. Thermal comfort. New York: McGraw-Hill Book Company; 1970. [13] Mayer E. A new Correlation between Predicted Mean Votes and (PMV) and Predicted Percentages of Dissatisfied (PPD). Healthy Buildings / IAQ ’97-Global issues and regional solutions; 1997 Sep 27; 1997 p. 189e94. [14] McIntyre DA. Seven point scales of Warmth. The Building Services Engineer 1978 Mar;45(12):215e26. [15] Humphreys MA, Nicol JF. Do people like to feel ’Neutral’? Response to the ASHRAE scale of subjective Warmth in relation to thermal Preference, indoor and outdoor temperature. ASHRAE Transactions 2004;110:569e77 (Part 2). [16] Backhaus K, Erichson B, Plinke W, Weiber R. Multivariate Analysemethoden. 12. Auflage. Berlin: Springer; 2006. [17] Kish L. Survey Sampling. New York: John Wiley & Sons Inc.; 1965.
Contents lists available at ScienceDirect
Building and Environment journal homepage: www.elsevier.com/locate/buildenv
Local and overall thermal comfort in an aircraft cabin and their interrelations Sumee Park a, *, Runa T. Hellwig a, b, Gunnar Grün a, Andreas Holm a a b
Fraunhofer Institute for Building Physics, Dept. Indoor Climate and Climatic Impacts, Fraunhofer Str. 10, 83626 Valley, Germany University of Applied Sciences, Dept. Energy Efficient Design and Building Climatology, Germany
a r t i c l e i n f o
a b s t r a c t
Article history: Received 28 May 2010 Received in revised form 8 November 2010 Accepted 11 November 2010 Available online 18 November 2010
In this study the interrelation between local and overall thermal comfort of passengers in aircraft cabins was investigated by thirteen simulated flights. For each of the tests forty test persons filled out questionnaires concerning their perceived overall and local thermal comfort at temperatures of 20 Ce25 C, which were measured at every second seat. With these physical and subjective data PMV (Predicted Mean Vote) and TSMV (Thermal Sensation Mean Vote) of test persons as well as PPD (Predicted Percentage of Dissatisfied) and PD (Percentage of Dissatisfied) were compared. The PMV was consistently similar to the TSMV, while the thermal dissatisfaction in tests was always higher than PPD. The hypothesis at the beginning of this study was that the high ratio of thermal dissatisfaction in the aircraft cabin reported in literature might be caused by local discomfort. Therefore statistical analyses about the interrelations between local and overall thermal comfort were performed and models indicating such interrelations were developed. Some local perceptions are significantly different from overall thermal perception and these body segments alter in dependence of the overall thermal environment. Also body segments rated similarly were detected and these segments were pooled to distinct body regions using principal component analysis. Under the same overall thermal sensation the local thermal perception on a certain body region predominantly influenced the overall thermal comfort. Therefore weighting factors of local body regions on the overall thermal comfort were determined in dependence of the overall thermal sensation by means of multiple linear regression models. Ó 2010 Elsevier Ltd. All rights reserved.
Keywords: Local thermal comfort Overall thermal comfort Statistical analysis Aircraft cabins
1. Introduction There are only a few investigations published on thermal comfort in aircraft cabins, while a comfortable environment in aircraft cabins is a major factor in airlines business competition. The existing investigations [1,2] indicate the character of the indoor environment of aircraft cabins as follows: low relative humidity (typically below 15%), vertical temperature difference (typically 3 K), varying air velocity depending on the location in and the type of cabin (from 0.1 m/s to 0.6 m/s). These inhomogeneous thermal environments can lead to an unwanted local cooling or warming and cause thermal dissatisfaction (approx. 25% of occupants). There are internationally established models (DR and PD due to vertical air temperature gradients) for the assessment of these local thermal discomfort phenomena [3], which are valid only for persons in overall thermal neutral status. However, the local thermal comfort may vary depending on the overall thermal situation or the local thermal comfort feeling might influence the
* Corresponding author. Tel.: þ49 8024 643 237; fax: þ49 8024 643 366. E-mail address: [email protected] (S. Park). 0360-1323/$ e see front matter Ó 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.buildenv.2010.11.003
overall thermal evaluation. In addition, the models do not consider the body segment differences regarding local cooling or warming. Extensive local and overall thermal comfort investigations on inhomogeneous environments have been performed since the 1990s focusing mainly on the assessment of the climate in vehicles. There were two approaches; one is the assessment with equivalent temperature using multi-segmented thermal manikins or heated sensors [4,5]. The result of this approach is established in ISO 14505-2 as annex D2: “Interpretation of equivalent temperature in terms of perception of thermal sensation and thermal comfort” [6], which identifies the acceptable equivalent temperature depending on the body segment. A similar study was performed in a simulated aircraft cabin in recent years [7]. Another approach is the multi-segment physiological model combined with computational fluid dynamics CFD, which predicts the local skin temperature of human body in inhomogeneous environments (e.g. [8]). However, this approach evaluates the overall thermal comfort without considering the local thermal sensation but only with regard to the mean skin temperature. According to these results the cooling of the feet and at the same time the heating of the head cause an identical overall thermal comfort as without local stimulation, although the observed
S. Park et al. / Building and Environment 46 (2011) 1056e1064
perceptions are generally different. Zhang [9] studied this subject and developed a comfort model combined with the physiological model of Huizenga [10]. Zhang concluded that overall thermal comfort cannot be determined from mean skin temperature but from the local discomfort or comfort derived from local thermal sensation depending on the overall thermal sensation. Since the primary interest of the Zhang study was the vehicle cooling system, the test conditions were focused on typical vehicle indoor climate in summer, where impacts of cooling of one or multiple local body segments for 10e20 min on the overall comfort were studied. Thus the experimental conditions are not similar to the typical indoor climate of aircraft cabins. In addition, the influence of local stimulation duration and intensity were not varied in the test, which might cause other results. Furthermore Zhang developed a comprehensive model except for overall thermal comfort. The overall thermal comfort is determined using some rules. The objectives of the present study are to develop a statistical model indicating local and overall thermal comfort and their interrelations based on a subject study with the typical indoor climate of aircraft cabins and long flight duration. This paper concentrates on the methodological approaches of this study and statistical analyses about the interrelations between local and overall thermal comfort (Fig. 1). 2. Methods A subject study was performed in winter 2006 in a simulated aircraft cabin, the Fraunhofer Flight Test Facility (FTF) (Fig. 1). It consists of a 30 m long pressure vessel, which holds the first 16 m of a complete wide-body aircraft. The study consisted of 11 simulated 3.5h-flights and two 7h-flights, in which the air temperature differed from 20 C to 25 C. Forty different test persons were acquired for each of the flights according to specified gender and age profiles. Approximately 20 males and females participated in each test. The mean age profile achieved was as follows: 34% of the test persons were within the range of 18e35 years, 32% of the test persons were within the range of 36e55 years; 34% of the test persons were 55 years and older. The air temperatures were measured at every second seat in the heights of 0.1 m, 0.6 m and 1.1 m (Fig. 2). Relative humidity was also detected in 1.1 m height. At two locations the air velocities were measured in the heights analogous to the air temperatures. To relate the physical environment to the test persons’ answers the mean values of the corresponding 20-minquestionnaire interval were calculated. To analyze the mean air temperature in each test firstly the mean value of three heights was calculated at
Fig. 1. Subject study in the Fraunhofer Flight Test Facility (FTF).
1057
each seat, afterwards the mean value of every seat was estimated. The measured mean air temperatures at three heights as well as standard deviations between seats and mean vertical differences of air temperature between 1.1 m and 0.1 m in each test are reported in Table 1. The test persons were allowed to wear and adapt their own clothing and to use blankets provided as during common flights. Meals and drinks were provided on a regular basis. The typical high noise and vibration level (70 dB(A)) in aircraft cabins are simulated as well as the low air humidity (10e20%) and low air pressure (750e800 Pa). The measured mean air humidity, air pressure and air velocity in each test are listed in Table 1. The clothing values of test persons are calculated using the questionnaire. The test persons were allowed to move inside the cabin and to spend their time on their own way apart from the filling out of questionnaires. The test persons’ perception was asked by questionnaires at 70 min, 120 min and 170 min after the start of each test for 3.5 h test duration (Fig. 3). The questionnaires were presented on seven-point scales. In this paper questions related to the local thermal perception (too coldetoo warm scale) on 11 body segments as well as overall thermal perception, sensation (ASHRAE scale) and satisfaction (satisfiededissatisfied scale) are considered (see Fig. 4). The software package StatisticaÒ 7.0 (Statsoft. Tulsa. USA) was used for statistical analysis. Since the scales used in this study can be interpreted as interval scales [11] and additionally a huge sample number is available parametric statistics were used: the t-test for paired samples to analyze the differences of local thermal perception from overall and principal component analysis to detect similarities and differences between local perceptions of body segments. For the overall thermal comfort analysis the multiple regression model based on the least-squares model was chosen. All statistical analyses and the modeling as well as the validation in this paper are based on individual votes of the test persons unless they are indicated as “MV (Mean Vote)”. The first questionnaire (Q1) carried out during cruise at 70 min or respectively 140 min in 7 h flights after take-off was used for the development of the model and 500 out of 520 responses were
Fig. 2. Air temperature measurement locations in the FTF.
1058
S. Park et al. / Building and Environment 46 (2011) 1056e1064
Table 1 Test condition: Flight duration, set temperature, set altitude and mean values of the measured physical parameters in aircraft cabin during the questionnaire action in first cruise. Test
1
2
3
4
5
6
7
8
9
10
11
12
13
Flight duration Set altitude (km) Mean outdoor temperature ( C) Cabin air pressure [hPa] Air temperature ( C) SD between seats [K] Vertical difference [K] Mean clothing (clo) Relative humidity (%) Air velocity (m/s) Noise (dB(A))
3.5 0.7 8.4 939
3.5 0.7 8.6 939 22.5 0.4 2.5 1.2 15 0.06 70
3.5 1.8 7.1 808 25.1 0.2 2.4 1.0 14 0.06 70
3.5 1.8 9.4 808 24.1 0.4 3.4 1.1 14 0.06 70
7 1.8 9.1 809 24.5 0.4 2.4 1.2 15 0.06 70
3.5 1.8 4.3 808 19.7 0.5 2.3 1.3 18 0.04 70
3.5 2.4 0.7 748 20.6 0.4 1.7 1.0 18 0.05 70
3.5 2.4 1.3 749 21.1 0.6 1.2 1.1 18 0.05 70
3.5 2.4 2.9 749 24.1 0.4 4.2 1.1 14 0.06 70
7 2.4 5.6 749 22.8 0.4 1.7 1.0 16 0.05 70
3.5 2.4 10.2 749 22 0.6 3.6 1.2 15 0.05 70
3.5 2.4 3.9 749 25.1 0.5 3.6 1.0 15 0.06 70
3.5 0.7 2.7 920 25.1 0.3 2.8 1.1 18 0.06 70
a
70
Outdoor temperature: mean outdoor temperature on the test site during the test day, SD ¼ Mean value of standard deviation of air temperature between seats, Vertical difference: Mean of differences of air temperature between 1.1 m and 0.1 m, Grey: mean values of the measured parameters in the cabin. a no physical measurements, but questionnaire data available.
Fig. 3. Test schedules (left: 3.5 h flight duration, right: 7 h flight duration) Q: Questionnaire L: Lunch.
considered as valid. The physical environment during this period was analyzed in the present study. 3. Analysis of the interrelation of local and overall thermal perception in aircraft cabins Before the local and overall thermal comfort model for aircraft cabin is developed, the following questions were analyzed to get an idea of possible interrelations between local and overall thermal comfort and for the modeling approach. 1. Is the local thermal perception generally different from the overall thermal perception in aircraft cabins? If yes, which body segments differ?
2. Could the overall thermal satisfaction be derived from overall thermal sensation? 3. Might the local discomfort or comfort influence the overall thermal comfort under the same overall thermal sensation? 4. How many local segments are necessary for the thermal comfort analysis in typical aircraft cabins?
3.1. The difference between local and overall thermal perception For the analysis the local and overall seven-point bipolar (too cooletoo warm) scales were transformed to (3) to (þ3). Fig. 5 shows the average difference between the local and overall thermal judgments in 11 body segments in test 6 (Ta ¼ 19.7 C,
Fig. 4. Seven-point scales presented to the subjects.
S. Park et al. / Building and Environment 46 (2011) 1056e1064
1059
Fig. 5. Deviation of local thermal judgments from overall thermal judgments in 11 body parts in test 6 (left: MV ¼ 1.6, Ta ¼ 19.7 C, N ¼ 40) and in test 12 (right: MV ¼ 0.5, Ta ¼ 25.1 C, N ¼ 40).
N ¼ 38) and in test 12 (Ta ¼ 25.1 C, N ¼ 40). In test 6 the test persons evaluated the upper body clearly warmer and the lower body a little colder or not different than overall, while in test 12 they judged all body segments except the head much colder than overall. In order to find whether these differences are significant or not, a t-test for paired samples was performed in each test. The thermal perceptions of some body segments are significantly different from the overall thermal perception (see Table 2). Such body segments alter depending on experimental conditions. At cool air temperature (test 6, 7, 8) the test persons assessed only the upper half of the body to be significantly warmer than overall. Thus the overall thermal perception is similar to the thermal perception of the lower half of the body. In contrast to that the overall thermal perception in the neutral or slightly warm environment is significantly different from the local perception of the lower half of the body which is always recognized to be cooler than overall (except for test 5). In this case the overall thermal perception is rather comparable to the upper part of body. The results of test 5 and test 10 with 7 h flight duration differed from the other tests with 3.5 h duration. In both tests the first questionnaires were completed after a warm lunch (see Fig. 3), which might strongly influence the thermal judgments of subjects. The test persons evaluated the local thermal perception not to be
significantly different from the overall perception except the head in test 5 and feet in test 10. They felt warmer and more satisfied than in the comparable tests. However, the missing value ratio in test 10 was the highest, where only 34 people from 40 filled out the satisfaction question. 3.2. The difference of overall thermal sensation and overall thermal satisfaction The thermal comfort assessment in ISO 7730 [3] is based on the assumption that the three central categories in the ASHRAE scale (slightly cool, neutral, slightly warm) indicate thermal comfort, thus the thermal satisfaction and thermal sensation are interchangeable in a homogeneous environment [12]. Is this assumption valid also for an inhomogeneous environment like aircraft cabins? However, it was a controversial issue in the previous experimental studies also in a homogeneous environment. Mayer [13] found in his experimental study in a climatic chamber that test persons who voted “(1) slightly cool” did not feel thermally comfortable, while test persons who judged the thermal environment as “(þ1) slightly warm” evaluated the environment as thermally comfortable. Therefore he suggested another PPD (Predicted Percentage of Dissatisfied) calculation depending on PMV, where the most
Table 2 Differences between overall and local thermal votes on body segments in each test (1. questionnaire). Test
1
2
3
4
5
6
7
8
9
10
11
12
13
N PMV TSMV PD [%] Head Neck Torso Right upper arm Left upper arm Right lower arm Left lower arm Right leg Left leg Right foot Left foot
40
40 0 0.4 48 (**) (*) (*) (*) (*)
38 0.3 0.6 37
39 0.2 0.3 31
40 0.4 0.7 25 (**)
38 0.5 1.5 68 (***) (***) (***) (***) (***) (***) (***)
39 0.7 1 64 (***) (***) (***) (***) (***) (***) (***)
38 0.5 0.7 55 (***) (***) (**) (**) (**) (*) (*)
37 0.2 0.2 22 (*) (*)
34 0.2 0.1 9
39 0.1 0.4 36 (**) (*)
40 0.3 1 25 (***)
40 0.4 0.7 33
*** *** *** ***
** * ** **
*** *** *** ***
* ** *** ***
0.6 25
*** ** *** ***
** ** *** ***
(**) (*)
*** *** *** ***
* **
*** *** *** ***
N: Number of samples PMV: calculated from the measured values in Table 1 (metabolic rate of persons as 1met and radiation temperature in the middle of the cabin as air temperature assumed). TSMV: Thermal Sensation Mean Vote, PD: Percentage of Dissatisfied. *: p < 0.05, grey: mean is cooler than overall. **: p < 0.01. ***: p < 0.001. (*): mean is warmer than overall.
1060
S. Park et al. / Building and Environment 46 (2011) 1056e1064
Fig. 6. Thermal comfort judgments depending on Thermal Sensation Vote (TSV) and Thermal Perception Vote (TPV).
comfortable point is defined as PMV ¼ 0.4 with PD ¼ 15% (see Fig. 8). McIntyre [14] found already in 1978 similar results in his field experiment in winter however the opposite in the summer experiment. Humphreys and Nicol [15] explained such phenomena with subjective association of a word “cold” and “warm” depending on the outdoor climate. People in a cold climate or in winter would probably associate the word “warm” as “comfort” and respectively people in warm a climate for the word “cool”. In this study a seven-point bipolar comfort scale (satisfiededissatisfied) in addition to a seven-point ASHRAE scale and a seven-point bipolar perception scale (too warmetoo cold) were used for overall thermal assessment. Fig. 6 shows the different judgments of the test persons depending on different scales. 10% of the test persons perceiving overall thermally “neutral” judge the climate still dissatisfying as rating (1), (2) or (3) on the thermal satisfaction scale. Only 11% of the test persons rating “slightly warm” vote “dissatisfied” in contrast to 58% dissatisfaction of the test persons rating “slightly cool”. The judgment on the bipolar perception scale also does not match to the judgment of the test persons on the thermal comfort scale (see Fig. 6, right). Did the test persons feel objectively comfortable under physical “slightly warm” conditions like “neutral” conditions or were these judgments rather influenced by the verbal psychology as Humphreys and Nicol assumed? In Fig. 7 the percentage of dissatisfaction and the thermal sensation mean vote are compared for each temperature range of the air temperature measured at the seats. The mean air velocity (Va ¼ 0.06 m/s) and the mean clo value (clo ¼ 1.1) are applied for the calculation of PMV at each air temperature. The subjects evaluated the environment slightly warmer than predicted by the PMV in neutral or warm environments (PMV > 0.5) and cooler in cool environments (PMV < 0.75). The subjective judgments seem to be more sensitive than the physical assessment PMV. Interestingly the ratio of dissatisfied is lowest at 24 C with TSMV ¼ 0.4 and PMV ¼ 0 except at 26 C (Fig. 8). The sample number (N) at 26 C (PMV ¼ 0.5) is with 12 too small compared to other temperatures, for example N ¼ 119 at an air temperature of 25 C in which the ratio of dissatisfied was obviously larger than at 24 C. The result that the test persons feel satisfied at PMV ¼ 0 and at TSMV ¼ 0.4 strengthens the assumption of Humphreys and Nicol. The study was carried out in Germany from 30 January to 17
February 2006 and during this period the mean air temperature on the test site was 5 C. The cold outdoor temperature during that seasonal period might have influenced the thermal comfort assessment of the test persons. They assessed the physical neutral situation (PMV ¼ 0) as slightly warmer (TSMV ¼ 0.4) in which they feel most comfortable. However the percentage of dissatisfied is obviously higher than the estimation of PPD, which might be based on the difference between local and overall thermal perception (see Section 3.1). Fig. 8 shows that PPD according to Fanger can not be applied for the assessment of thermal comfort in aircraft cabins. The revised
Fig. 7. Comparison of PMV and TSMV depending on local air temperature (Tr ¼ Tr, Va ¼ 0.06 m/s, clo ¼ 1.1).
S. Park et al. / Building and Environment 46 (2011) 1056e1064
1061
Table 3 Comparison of the local thermal perception of the satisfied and dissatisfied group under same thermal sensation. Overall thermal sensation
Slightly cool (n ¼ 112)
Neutral (n ¼ 142)
Slightly warm (n ¼ 109)
Warm (n ¼ 67)
n:(Satisfied/dissatisfied) Head Neck Torso Right upper arm Left upper arm Right arm Left arm Right leg Left leg Right foot Left foot
(41/71)
(127/15)
(97/12)
(44/23) **
* * * (***) (***) (***) (**)
(*) (*) (**) (**)
* *
*: p < 0.05, **: p < 0.01, ***: p < 0.001, n ¼ sample number (*): mean is warmer in the satisfied group; *: mean cooler in the satisfied group.
Fig. 8. Percentage of dissatisfied (PD) depending on local air temperature (Tr ¼ Tr, Va ¼ 0.06 m/s, clo ¼ 1.1).
model of suggested by Mayer predicted the proportion of subjects thermally dissatisfied better than the PPD index was able to do, however the optimal thermal condition in this study was not at the PMV ¼ 0.4 as in Mayers model but at PMV ¼ 0 at an air temperature of 24 C (see Section 6.2). 3.3. The relation of local thermal perception and overall thermal comfort In order to define whether local thermal perceptions influence the overall thermal comfort under the same thermal sensation the local thermal perceptions of the two groups (satisfied and dissatisfied) stating the same overall thermal sensation are compared in Table 3 using the t-test for two independent samples. The mean local thermal perception of legs and feet in the satisfied group is significantly warmer than in the dissatisfied group in “neutral” and “slightly cool” overall sensation, while no significant difference is found for the “slightly warm” state. For the overall “warm” state the satisfied group feels locally cooler in the upper part as well as partially in the lower body part. Especially the perception of the head was significantly different between two groups. No test person indicating a “cool” thermal sensation was satisfied. 3.4. The relation of local thermal perceptions between local body segments How sensitive can people perceive thermal differences impacting their body in a typical indoor environment? Which partitions of the human body are sensible for the thermal comfort research? Such questions have not yet been scientifically discussed in depth. Therefore in previous local thermal comfort investigations always many different body parts were investigated. Another important question for modeling in this study is how different or similar local body segments react to the typical climate of aircraft cabins. This can be explained by means of principal
component analysis. The original aim of this statistical analysis is to reduce several variables to a smaller number of factors. This is also required in order to define the independent variables for the development of statistical models since strong linear correlations between variables (multicollinearity) can result in large uncertainties of the estimation coefficients of a regression model [16]. Multicollinearity is a general problem of regression analysis and occurs, if two independent variables have a very strong correlation. This kind of strong correlation was already observed in this study during the analysis of the relation between local and overall perception (see Table 2). The thermal perception of the upper arm and lower arm did not differ from each other in this study. In order to avoid this multicollinearity in a regression model the principle component analysis is recommended in statistical literature. By means of the principal component analysis the eleven local body segments could be extracted to the following three factors: factor 1: upper part; factor 2: lower part and factor 3: head and neck (Fig. 9). An extraction of these three factors still explains more than 90% of the original variance with 11 body segments.
4. Modeling of overall thermal comfort for aircraft cabins The analysis described above provides the following results for the model development. 1. Overall thermal sensation (OTS) and overall thermal comfort (OTC) require different models 2. Three body regions instead of eleven body segments can be used for modeling 3. The local thermal perception might determine the OTC 4. The weighting coefficients of local thermal perception (LTP) determining the OTC may alter in dependence of OTS The local thermal perceptions (LTP) of the three body regions (upper part, lower part and head/neck) have been estimated from the LTP of eleven body segments. In addition, two new variables, the maximal difference between the most extreme LTP and also the most extreme LTP are included as predictors for modeling. Afterwards the overall thermal sensation (OTS) was estimated from these variables and finally overall thermal comfort (OTC) model was developed from the LTP with regard to three groups of OTS (cool, neutral, warm). The approach of modeling can be found in Fig. 10. Where y is criterion variable (in this study: OTS or OTC), x are the predictor variables (in this study: LTPs and the maximal
1062
S. Park et al. / Building and Environment 46 (2011) 1056e1064
Fig. 9. Factor loadings of 11 local thermal perception variables on three factors 1: Head 2: Neck 3: Torso 4: Right upper arm 5: Left upper arm 6: Right lower arm 7: Left lower arm 8: Right leg 9: Left leg 10: Right foot 11: Left foot.
y ¼ b0 þ b1 x1 þ b2 x2 þ b3 x3 þ b4 x4 þ 3
(1)
At first the modeling was performed always with more than five predictors. If the individual coefficient of a predictor in the model was not significant, a new model was formed without such a predictor. If the new model performs better with fewer variables, this predictor will be permanently removed from the model. In the process the extreme LTP proved to be an insignificant predictor for all models. The final result of modeling is shown in Table 4. The OTS is influenced by three predictors, while OTC “cool” is determined by the upper and lower body. The great difference between individual maximum LTP reduces the thermal comfort only in the thermally “neutral” state in which the warm lower body part increases thermal comfort. In warm conditions the head region is responsible for the overall thermal comfort.
Fig. 10. Model approach.
difference between the most extreme LTP), b are the regression coefficients, 3 is the residual. Half of the sample data chosen at random by the split-node method of STATISTICA was used for modeling and another half of the data was used for validation later. This procedure is based on the Equal Probability of Selection Method (EPSEM) in which every case in the population of interest has an equal opportunity of being selected for the sample [17]. With this method the validation can be performed with a sample drawn from the same population with the same exposure conditions as for the modeling. In order to find the influences and the intensities of LTP to the OTC depending on OTS a multiple regression was applied based on the method of least squares in which the sum of the squared differences between modeled data and observed data in all observations must have the least value, i.e. a function with the smallest sum of the residuals P ð ð3i Þ2 Þ is investigated (see Eq. (1)).
5. Validation The OTS and OTC were calculated using the developed models with the other half of sample data. The standard errors of OTS, OTCcool, OTC neutral , OTC warm of modeling and validation are reported in Table 5. OTS and OTC at cool sensation were better predicted than overall thermal comfort at neutral and warm thermal sensation. No major differences between modeling and validation could be observed. 6. Discussion 6.1. Limitations of the model The experimental conditions in this study are different from real aircraft cabin climate in the following aspects. First the outside fuselage temperature in this study was not similar to the
Table 4 Results of the modeling. Model
Adjusted.R2
Coefficients
Table 5 Standard errors of models for modeling and validation.
LTP Head/ LTP Upper LTP lower LTP max. Constant Neck part part diff. OTS 0.68 OTC cool 0 OTC neutral 0 OTC warm 1.00
0 0.74 0 0
0.27 0.51 0.40 0
0.16 0 0.50 0
0.33 0.55 1.79 1.63
OTS 0.52 0.40 0.26 0.15
Modeling Validation
OTCcool
OTCneutral
OTCwarm
N
SE
n
SE
N
SE
N
SE
249 251
1.0 0.9
70 64
1.2 1.3
126 112
1.3 1.4
53 75
1.6 1.7
N ¼ sample number, SE ¼ Standard Error.
S. Park et al. / Building and Environment 46 (2011) 1056e1064
temperatures during real flights (approx. 30 C). The cool surface in real flights could cause different local thermal sensations between outer and inner side of the body observed in [7]. Thus the allocation of body segments could be different inside a real aircraft cabins. However, analysis with another further three experiments under fuselage cooling also showed an insignificant difference between side and inside seat locations. The lack of significance could be due to the small subjective numbers with 40 subjects in each test and the spatial variation in the cabin. Second, the air velocity measured in the simulated aircraft cabin was less than 0.1 m/s and thus lower than reported from real aircraft cabins. Consequently the local discomfort and its influence on the overall thermal comfort could be higher than in this study. Therefore the further test person studies in FTF after this study were performed under physical conditions of fuselage cooling and higher air velocities but have not yet been included into the model. 6.2. Thermal comfort assessment based on subjective and objective data The aim of the thermal comfort modeling in this study is to understand the thermal comfort or discomfort of passengers in aircraft cabins and to provide a tool for the planer. This is only possible based on a combined subjective and objective assessment. For a rough estimation or mean judgment for the planning phase of an air-conditioned room there is already Fangers PMV/PPD approach which is applied and established worldwide. This approach was confirmed in parts again in this study. The ratio of satisfied persons was highest at the calculated PMV ¼ 0 environment, but not at 95% but at 76%. The first hypothesis in this study was that this high ratio of dissatisfaction reported already in literature might be caused by local discomfort. The local thermal perception on the lower body part is significantly different between the satisfied and the dissatisfied group under overall thermal sensation “neutral“ or at an air temperature of 24 C at “PMV ¼ 0”. The satisfied group clearly feels warmer in the foot and leg region than the dissatisfied group. So the preliminary hypothesis could be confirmed by the questionnaire analysis. However, the warmer perception on the lower body part of the satisfied group could not be physically explained in this study. A higher local air or radiation temperature, lower local air velocity or a higher local clothing insulation value may physically result in this warmer perception. The local air temperature (height: 0.1 m) was measured on the test persons’ seats and the local clothing insulation values were gained from individual clothing information The local air temperature between the two groups at 24 C was not different and the local clothing insulation value on the feet was higher in the satisfied group, but not significantly (see Table 6).
Table 6 Comparison of satisfied and dissatisfied group in 24 C air temperature (mean air temperature from 0.1 m, 0.6 m and 1.1 m). Satisfied Dissatisfied P (n ¼ 66) mean (n ¼ 20) mean value value value Questionnaire: local thermal perception Physical measurement
Right leg Left leg Right foot Left foot Local air temperature (0.1 m) ( C) Local clothing insulation value (Feet) (clo)
0.20 0.12 0.25 0.16 21.8
1.30 1.10 1.65 1.55 21.9
0.000 0.001 0.000 0.000 0.794
3.47
2.87
0.137
1063
The local air velocity or local surface temperature might result in this different perception. This cannot be proved in this study due to the missing local measurements. In order to prove this impact indirectly perceptions of the various seat positions of the test persons (side of air craft/aisle/another passenger or according to row) were compared, but the results showed no significant relation between seat position and local thermal perception at the same air temperature. 6.3. Local and overall thermal comfort model for near to neutral temperature with different local conditions It might be difficult to develop an accurate thermal comfort model based on physical data in a typical indoor climate near to neutral temperature because individual physiological or psychological differences are often greater than objective physical differences. However, it is important for the design of most comfortable indoor environments to have the possibility at hand to assess local and overall thermal comfort under near to neutral conditions with different local conditions. The existing local thermal considerations in ISO 7730 focus on the “discomfort” side of the thermal asymmetry. Zhang noted and reflected on a possible positive side of asymmetric thermal environments in her LTC model, however only limited in the OTC consideration. According to her OTC model the intensities of a maximum LTC or two minimum LTCs are deciding an overall thermal comfort independent from where in 19 body segments (dis) comfort is indicated. The model does not address the overall most thermally comfortable condition by means of local asymmetry, while the LTC model indicates it for local segments. In this study under near to neutral conditions the warm leg and foot region and low maximum differences between local thermal perceptions are deciding on OTC, while the cooler head region is more significant for OTC under slightly warm conditions. However, it could be that a warmer torso and a warmer foot region at a slightly cool air temperature can result in an overall neutral sensation and may be perceived overall more thermally comfortable with a cool head and warm feet. 6.4. Future research Due to the limitation of exposure conditions in this study the individual regression coefficients in models could be different in the real aircraft cabin. However this study shows that OTC in an inhomogeneous environment can not be defined from thermal sensation only, since the LTP influences the OTC under same OTS and the assessment of thermal sensation could be affected by the expectations and preconditioning depending on the outdoor temperature (which may even be different for aircraft passengers on the same flight). In an in-homogenous environment, a high level of thermal comfort, maybe higher than in a homogeneous environment, could be achieved by careful control of the local thermal environment. This indicates that further studies are required at near to neutral temperature with a local variation of the thermal environment to find the most comfortable condition. 7. Conclusion In this study while PMV of Fanger predicted overall thermal sensation in the simulated aircraft cabin relatively well, the thermal dissatisfaction (PD) of test persons is obviously higher than the PPD of Fanger, which might be resulted from the local thermal discomfort. The LTP of some body segments can be significantly different from the OTP and also significantly different between satisfied and dissatisfied groups under same OTS or at the same air temperature. These statistical analyses indicated that the OTC in the aircraft cabin can be derived from the LTP and OTS. However which
1064
S. Park et al. / Building and Environment 46 (2011) 1056e1064
partitions of the human body are sensible for the thermal comfort research under inhomogeneous environments? In previous local thermal comfort investigations always many different body parts were investigated. In order to answer this question in this study body segments rated similarly were detected and these segments were pooled to distinct body regions using principal component analysis. Also this procedure can minimize the error based on multicollinearity in the multiple regression models. The eleven body segments investigated in this study were reduced to three body regions due to the similarity of thermal perceptions. However the allocation of body segments could be different inside real aircraft cabins, since the experiments in this study were performed without fuselage cooling and with lower air velocity than in typical aircraft cabin environments. In the neutral sensation case the thermal perception of the lower part of the body dominates the OTC, while the LTP of the head and neck influences OTC under warm conditions. Based on these effects three different OTC (Overall Thermal Comfort) models have been developed depending on OTS (Overall Thermal Sensation) using the LTP (Local Thermal Perception) of body regions and the maximum difference between LTPs. Acknowledgement The data analyzed in this paper was collected within the framework of the EU-FP5 project FACE (contract No. G4RD-CT2002-0076). The authors thank the FACE consortium for the good collaboration. References [1] Spicer CW, Murphy MJ, Holdren MW, et al. Relate air quality and other factors to comfort and health symptoms reported by passengers and crew on commercial transport aircraft 2004; Report No.: Part I.
[2] Ross D, Crump D, Hunter C, Perera E, Sheridan A. Extending cabin air measurements to include older aircraft types utilised in volume short haul operation. UK: BRE 2003; Report No.: 212034. [3] En Iso 7730:2005. Ergonomics of the thermal environment eAnalytical determination and interpretation of thermal comfort using calculation of the PMVand PPD indices and local thermal comfort criteria. EN ISO 7730: 2005. [4] Nilsson HO, Holmer I, Bohm M, Norén O. Equivalent temperature and thermal sensation - comparison with subjective responses. Comfort in the automotive industry. Bologna; 1997, p. 157e162. [5] Schwab R. Einfluss der Sonneneinstrahlung auf die thermische Behaglichkeit in Kraftfahrzeugen. FAT Schriftenreihe; 1994. FAT 109. [6] prEN ISO 14505e2:2004: Ergonomics of the thermal environment. Evaluation of thermal environment in vehicles Part 2: Determination of equivalent temperature. prEN ISO 14505-2:2004. [7] Strøm-Tejsen P, Wyon DP, Zukovska D, Jama A, Fang L. Occupant evaluation of 7-hour exposures in a simulated aircraft cabin - Part 2: Thermal effects. Proceedings of Indoor Air 2005, Beijing, China; 2005, p. 46e51. [8] Tanabe S, Kobayashi K, Nakano J, Ozeki Y, Konishi M. Evaluation of thermal comfort using combined multi-node thermoregulation (65MN) and radiation models and computational fluid dynamics (CFD). Energy and Buildings 2002;34(6):637e46. [9] Zhang H. Human thermal sensation and comfort in Transient and NonUniform thermal environments, PhD Thesis. Berkeley: University of California; 2003. [10] Huizenga C, Zhang H, Arens E. A model of human physiology and comfort for assessing complex thermal environments. Building and Environment 2001;36:691e9. [11] En Iso 10551:2002. Ergonomics of the thermal environment e Assessment of the influence of the thermal environment using subjective judgement scales. EN ISO 10551:2001. [12] Fanger PO. Thermal comfort. New York: McGraw-Hill Book Company; 1970. [13] Mayer E. A new Correlation between Predicted Mean Votes and (PMV) and Predicted Percentages of Dissatisfied (PPD). Healthy Buildings / IAQ ’97-Global issues and regional solutions; 1997 Sep 27; 1997 p. 189e94. [14] McIntyre DA. Seven point scales of Warmth. The Building Services Engineer 1978 Mar;45(12):215e26. [15] Humphreys MA, Nicol JF. Do people like to feel ’Neutral’? Response to the ASHRAE scale of subjective Warmth in relation to thermal Preference, indoor and outdoor temperature. ASHRAE Transactions 2004;110:569e77 (Part 2). [16] Backhaus K, Erichson B, Plinke W, Weiber R. Multivariate Analysemethoden. 12. Auflage. Berlin: Springer; 2006. [17] Kish L. Survey Sampling. New York: John Wiley & Sons Inc.; 1965.