Modeling individual variations in thermal stress response for humans in transient environments.
Researchers have developed thermal computational models using different approaches, such as analytical, statistical, empirical, and physiological, to forecast and to better understand human thermal responses under different environ mental conditions (Bue 1971; Campbell et al. 1994; French et al. 1997; Fu 1995; Gonzalez et al. 1997; Gonzalez 2001; Guan et al. 2003; Hsu et al. 1971; Pennes 1948; Stolwijk 1971; Tanabe et al. 2002; Wissler 1971; Hwang and Konz 1977). However, due to the complexity of human thermoregulation, including interactions with the environment, the suitability of such models to real-world applications has been limited. Better characterization of the uncertainties involved, including physiological mechanisms and parameter variations will be important as we seek improved understanding of human thermophysiology related to heat stress.
One of the earliest human thermal efforts was the development of a steady-state model to analyze heat transfer in a resting human forearm by Pennes (1948). This cylindrical model served as the basis for a more advanced model by Wissler (1971) and is still being used for the prediction of temperature elevation during hyperthermia (Frank et al. 1999). Subsequent advances in computing technology and increased experimental data on human physiology helped researchers in developing more sophisticated human thermal models. In the 1960s, early versions of the well-known Wissler (1964) and Stolwijk (1971) models were being developed. Later human thermal models derive their ideas largely from these three mathematical models.
Various research teams (Fu 1995; Huizenga et al. 2001; Iyoho 2002) have developed models in the past decade to be used in environments that range from steady-state to transient and non-uniform cases. Examples of models that are being improved include the Wissler (1964) model, a detailed finite element model (Fu 1995), a model by Huizenga et al. (2001), and one by the authors' group. All of these models include heat transfer within the body and between the body and its environment, as well as sweating, shivering, and vasomotor functions. The authors' group has also been investigating human thermal dynamic modeling issues (Campbell et al. 1994; Iyoho 2002; Thornton et al. 2002; Thornton and Nair 2000) and are developing an advanced two-dimensional human thermal model as part of a larger goal to design an automatic thermal controller for astronauts during extra vehicular activities (Iyoho 2002; Thornton et al. 2002).
Quantitative comparisons among human thermoregulation models have been difficult due to the individual characteristics of each model-in particular, environmental conditions (Campbell et al. 1994; Gonzalez 2001; Guan et al. 2003). From a user's point of view, it has not been clear which of the models would be best suited for a particular environment and application. From a physiology point of view, many aspects of the human active (thermoregulatory/control) thermal system are still not well understood. The passive thermal system (conduction through body regions, etc.) has been modeled with more success, but the effect of uncertainties, such as intra-subject and inter-subject variations in thermal parameters, continue to be poorly understood. Hence, although human thermal models predict core and mean skin temperature fairly accurately, they fail to predict other thermal responses, such as sweat production and metabolic heat production, particularly in extreme conditions. Important indicators for thermal risk, such as heart rate and dehydration level due to the excess sweating, are not typically considered explicitly in the models. Perturbations, such as fast transient and widely disparate environmental conditions, individual physiological differences, altitude, clothing, and terrain level, also cause problems for such models. As an example, causes for large variations in thermal responses and tolerance time limits among individuals continue to be poorly understood.
A neural network model has been developed by our team to predict heat stress using transient experimental thermal data from a group of subjects, details of which are provided in the next section. The inputs to the model are prior values of core temperature (Tcore), heart rate (HR), skin temperature (Tsk), and seven "individual" parameters, and the model outputs are future predictions of Tcore, HR, and Tsk. The seven parameters considered to characterize individual differences are gender, age, height, weight, maximal oxygen consumption (VO2max), basal metabolic rate (Minitial), and initial heart rate (HRinitial). A methodology is then proposed to identify the relative importance of "individual" parameters on heat stress variables, such as core temperature and heart rate, using a sensitivity analysis. A method to calculate the relative importance of environmental conditions on those variables is also reported.
The data set used to develop the neural network was provided by Dr. Richard Gonzalez of the US Army Research Institute of Environmental Medicine (USARIEM), Natick, Massachusetts, and contains thermal observations for 35 healthy male and female subjects, ranging in age from 8 to 67 years. The data were collected in compliance with appropriate guidelines. For each subject, in a resting supine position through the experiment, 10 variables were measured for 140 min, in still air, with transient environmental conditions (from 9[degrees]C to 50[degrees]C, and from dry to humid environments). After an initial period of 30 min at Ta 30[degrees]C, the chamber temperature was increased at a rate of 1.5[degrees]C [min.sup.-1], rising rapidly for the first 20 min, and leveling off at 50[degrees]C. Following a 30 min hot-dry exposure, the dew-point temperature was raised to 32[degrees]C, and responses were observed for another 25 min, after which the chamber temperature was lowered, falling rapidly and leveling off at 10[degrees]C . For more details, refer to Gonzalez et al. (1981). The data set is divided into eight groups by age and gender, as shown in Table 1. Since the subjects are resting, the metabolic rate is considered constant for each subject, with a term added separately if shivering occurs.
Table 1. Details of the Experimental Data Set Gender Group N[degrees] of Subjects Age (years) Female F1 5 11.7 [+ or -] 1.6 F2 5 22.5 [+ or -] 2.3 F3 3 40.0 [+ or -] 6.0 F4 2 61.8 [+ or -] 2.0 Male M1 5 11.8 [+ or -] 2.8 M2 5 22.3 [+ or -] 2.9 M3 5 34.0 [+ or -] 5.6 M4 5 60.2 [+ or -] 5.0 Gender Surface Area to Mass Ratio Maximum Aerobic Capacity (VO2 ([Cm.sup.2][Kg.sup-1]) max)ml(mim-kg) Female 321.3 [+ or -] 21.4 43.8 [+ or -] 1.3 307.3 [+ or -] 36.4 49.8 [+ or -] 1.3 279.3 [+ or -] 11.1 35.8 [+ or -] 1.3 290.3 [+ or -] 9.4 30.9 [+ or -] 1.3 Male 319.3 [+ or -] 35.9 47.8 [+ or -] 2.4 270.3 [+ or -] 14.2 47.1 [+ or -] 2.1 252.3 [+ or -] 11.9 44.4 [+ or -] 2.2 253.3 [+ or -] 18.6 27.7 [+ or -] 1.5
Three separate models are developed-two gender-based models (male and female using male and female data sets, respectively) and one "general" combined model with gender as an input (using the entire data set). Starting with the basic thermal equations for human modeling, the important "inputs" are identified using a systematic procedure for artificial neural network modeling (Thornton et al. 1997; Mistry and Nair 1993), including their dynamic characteristics, i.e., whether to consider one or two past values of the variable. After such a careful analysis, nine inputs were selected for the general combined model. These were as follows: two environmental parameters (Ta, Pa) and seven individual characteristics (age, gender, VO2max, weight, height, Minitial, and HRinitial; see Table 2 for symbols and description) for the subjects with the only gender variable being omitted in the two gender-based models. The potential user thus has to provide only the environmental conditions and the individual characteristics as inputs. The outputs of the model are Tsk, Tcore, and HR. Esophageal temperature, Tes, and rectal temperature, Trec, were used to represent Tcore in the training process. A feedforward neural network architecture is used for the models with an input layer, two hidden layers, and an output layer, using a back-propagation algorithm for update, with a mean-square error (MSE) performance function. The architecture has feedback of one time-delayed value for each of the three outputs, making it a dynamic model. The structure of the net for the combined model is 12 x 12 x 14 x 3, while for gender-based models it is 11 x 10 x 10 x 3. The data set provided 4759 patterns for training, each separated in time by one minute. An early stopping method (Amari et al. 1997; Pennes 1948) is used to prevent overfitting and ensure generalization, with the data divided into three subsets-50% for training, 25% for validation, and 25% for testing. From the six models, the one with the least MSE for the testing data set is selected. The weights, biases, and number of epochs for the network are then used to train the final model with the entire data set.
Table 2. Description of the Parameters Variables Description Unit Individual Subject Identifier Characteristics Group 1 to 4 Age Age of a Years subject [A.sub.d] Surface area [m.sup.2] Gender Male or female [VO.sub.2] Max.aerobic ml/(min.kg) Max capacity Height in. Weight kg M initial Initial W*[m.sup.-2] metabolic rate HR initial Initial heart Beat/min rate Controlled Time A total of Min Variables 140 min Ta Air [degrees]C temperature Pa Ambient vapor in.-Hg pressure Measured Tsk Skin [degrees]C Variables temperature Tes Core [degrees]C temperature (ecophageal) HR Heart rate Beat/min Evg Evaporative g/min loss [m.sub.s] Sweating rate mg/(min*[cm.sup.2]) sweat capsule [K.sub.chest] Skin W/([m.sup.2].[degrees]C) conductance in chest [K.sub.arm] Skin W/[m.sup.2].[degrees]C conductance M Metabolic W*[m.sup.-2] free energy production Disc Discomfort index Tsens Temperature sensation magnitude Calculated Esk Heat Transfer W*[m.sup.-2] Variables via evaportion from the skin surface Tb Mean body [degrees]C temp. w Skin wettedness R+C Dry heat W*[m.sup.-2] loss [DELTA]Tb/min Mean body W*[m.sup.-2] temp.rate S Body heat W*[m.sup.-2] storage rate Req Esk Required W*[m.sup.-2] evaporative heat loss ET Effective [degrees]C temp. HRR Heart rate ratio
We perform a sensitivity analysis with mean skin temperature (Tsk) and core temperature (Tcore) as outputs, and the individual parameters (age, VO2max, HRinitial, Minitial, HRinitial, height, weight) as inputs. The sensitivity derivative is defined as [delta] (output) / [delta] (parameter), and this derivative is calculated by perturbing each input by one standard deviation of its value within the group, and noting its effect on the steady-state value of the output. Taking the ratio of the change in output to the change in input provides the sensitivity derivative. This analysis is carried out for each input, one at a time, keeping the others constant, in four environmental conditions. An "average" subject from each group is chosen as the representative subject for the sensitivity analyses (Table 3). This value is then scaled by the largest magnitude in the column to provide a ranked estimate of its relative importance. This results in the largest value in each column having a magnitude of 1, and the magnitudes and signs for each parameter provide an indication of its relative importance. Sensitivity analysis provides further insights into system characteristics by permitting us to rank the importance of the "inputs" on influencing the "outputs." Sensitivity analysis also serves as an additional tool to check whether the model has generalized adequately since it can be checked against known trends reported in the literature.
Table 3. Average Subject for Each Group with Mean Individual Parameters Male Average Male of SD Average Male of SD group 1 group 2 Age (yrs) 11 2.58 22 3.69 VO2 Max (W*[m.sup.-2]) 48.55 6.29 47.11 5.93 M initial (ml/(min-kg) 54.027 4.42 49.06 5.23 HR initial (bpm) 72 8.64 73.6 5.16 Height (inches) 57.38 6.34 70.9 2.68 Weight (Kg) 39.1 12.65 70.24 8.04 Female Average Female SD Average Female SD of group 1 of group 2 Age (yrs) 12 1.14 23 2.5 VO2 Max (W*[m.sup.-2]) 43.86 3.05 47.35 7.22 M initial (ml/(min-kg) 50.88 6.92 46.32 11.48 HR initial (bpm) 84 3.31 74 8.64 Height (inches) 60.75 7.87 64.37 2.04 Weight (Kg) 43.25 4.68 51.13 4.17 Male Average Male of SD Average Male of SD group 3 group 4 Age (yrs) 34 6.22 60 5.56 VO2 Max (W*[m.sup.-2]) 46.08 7.61 27.6 3.94 M initial (ml/(min-kg) 54.6 6.42 51.05 5.60 HR initial (bpm) 76.5 4.56 83.46 3.82 Height (inches) 73.13 3.32 66.85 2.40 Weight (Kg) 77.21 4.63 72.61 2.76 Female Average Female SD Average Female SD of group 3 of group 4 Age (yrs) 39 4.94 62 2.82 VO2 Max (W*[m.sup.-2]) 37 8.98 30.4 10.60 M initial (ml/(min-kg) 47.66 1.26 41.25 7.99 HR initial (bpm) 74.66 2.82 80 5.65 Height (inches) 64.70 0.61 61.12 0.17 Weight (Kg) 59.21 8.36 51 3.53 SD: standard deviation of each group
RESULTS AND DISCUSSION
The neural network structures for our models were chosen based on extensive statistical analysis of the data and a review of the literature. These analyses included correlations, ANOVA, and development of linear regression models. Havenith and van Middendorp (1990) have shown that Ta, vapor pressure, and metabolic rate can explain 96% of the variance in mean skin temperature, with Ta being the largest contributor. This finding coincides with our multiple regression analysis of the data set, which gives a value 92% for the same check. Havenith and van Middendorp (1990) show that a multiple regression analysis using Ta, Pa, M, % Fat, and sweating set point can explain 71% of the variance in Tcore. A regression analysis performed with our data set showed that 61% of the variance could be explained when Tsk, HR, Evg, and age group were included along with Ta and Pa. Moran et al. (1995) showed that 77% to 88% of the variance in heart rate can be explained by initial heart rate, metabolic rate, maximum evaporative rate, required evaporative rate, and time of exposure in minutes. Our data analysis showed that Tsk, Tes, Evg, Ta, Pa, and age group could explain 64% of the variance (Table 4). For the variance in heat storage rate, S, 92% could be explained by Ta, Pa, M, %Fat, [A.sup.d], VO2max, and sweating gain (Havenith and van Middendorp 1990).
Table 4. Regression Coefficients for Various Physiological Heat Stress Reactions with Environmental, Heat Production and Individual's Parameters from Havenith et al. Study (13) Dependent Variables Ta [PH.sub.2]O M % Fat ([degrees]C) (kPa) (kW) S (J/g) x x x S (J/g) x x x x Rectal Temperature ([degree]C) x x x x Rectal Temperature ([degree]C) x x x x Skin Temperature ([degree]C) x x x x HR (bpm) x x x x HR (bpm) x x x x Sweating Dependent Surf:mass [VO.sub.2max] % set Variables ([m.sup.2]kg-1) [VO.sub.2max] point S (J/g) S (J/g) x x Rectal Temperature ([degree]C) Rectal x x x Temperature ([degree]C) Skin Temperature ([degree]C) HR (bpm) HR (bpm) x Dependent Variables gain S [R.sup.2] (%) [R'.sup.2] (%) S (J/g) 88 S (J/g) x 92 33 Rectal Temperature ([degree]C) 38 Rectal Temperature ([degree]C) 60 37 Skin Temperature ([degree]C) 96 HR (bpm) 75 HR (bpm) 88 46 Ta, Ambient temperature; [[PH.sub.2]O], water vapor pressure; [VO.sub.2] max, maximal oxygen uptake; [r.sup.2], total variance explained by the equation; [r.sup.'2], the fraction of the residual variance left after introduction of climatic and work rate parameters in the prediction equation, which is explained by individual's parameters.
Predictions of the models developed using a representative "average" for each group (Table 3) compared well with actual data, some of which are shown in Figures 1-3. Actual data for all of the subjects are superimposed in the figures with the prediction for a representative subject in the group with average properties. The correlation coefficients between predicted values and experimental values are shown in Table 5 for each of the subjects, for all the three outputs (Tcore, HR, and Tskin), showing good overall performance using gender-based models. Several other similar investigations were performed that are not described due to space limitations, and the observations from the study are summarized next. The model is found to successfully predict Tcore, HR, and Tskin, given only two environmental variables and seven individual parameters as inputs. Tcore and HR are two of the most important thermal predictors of thermal risk, and they are products of complex internal and external interactions with the surrounding environment and within the body.
[FIGURE 1 OMITTED]
[FIGURE 2 OMITTED]
[FIGURE 3 OMITTED]
Table 5. Coefficients of Correlation (R) of Predicted Values Against Experimental Values for Each Output for All Groups Using Gender-Based Models Group1 Male1 Male2 Male3 Male4 Male5 Tsk 0.985 0.983 0.9728 0.993 0.999 Tcore 0.742 0.853 0.8720 0.778 0.942 HR 0.9207 0.8995 0.8757 0.9019 0.8732 Group2 Male1 Male2 Male3 Male4 Male5 Tsk 0.9850 0.9922 0.9881 0.9930 0.9850 Tcore 0.8705 0.7580 0.9322 0.7390 0.7772 HR 0.9552 0.8983 0.9669 0.8838 0.8428 Group3 Male1 Male2 Male3 Male4 Male5 Tsk 0.9971 0.9963 0.9920 0.9900 0.9912 Tcore 0.9140 0.8381 0.8452 0.8141 0.8274 HR 0.9285 0.8334 0.7425 0.8433 0.8246 Group4 Male1 Male2 Male3 Male4 Male5 Tsk 0.9843 0.9661 0.9783 0.9520 0.9800 Tcore 0.8812 0.8203 0.8672 0.8460 0.8392 HR 0.9470 0.8437 0.8011 0.9054 0.8283 Group1 Female1 Female2 Female3 Female4 Female5 Tsk 0.9262 0.9715 0.9566 0.9825 0.9819 Tcore 0.7843 0.8188 0.9344 0.8504 0.8811 HR 0.9041 0.9225 0.8776 0.8998 0.9086 Group2 Female1 Female2 Female3 Female4 Female5 Tsk 0.9851 0.9843 0.9917 0.9478 Tcore 0.8920 0.8561 0.9193 0.7759 HR 0.9135 0.8886 0.8893 0.7959 Group3 Female1 Female2 Female3 Female4 Female5 Tsk 0.9870 0.9835 0.9748 Tcore 0.7921 0.8912 0.8798 HR 0.9262 0.8925 0.8574 Group4 Female1 Female2 Female3 Female4 Female5 Tsk 0.9761 0.9750 Tcore 0.947 0.9192 HR 0.9318 0.8373
Observations from Sensitivity Analysis
Sensitivity analysis is systematically performed using all three models under four different environmental conditions: normal (Ta = 20, Pa = 24), jungle (Ta = 40, Pa = 30), desert (Ta = 43, Pa = 10), and cold (Ta = 5, Pa = 10). The data set used for the first model has been categorized into four age groups for each gender. From each age group an average subject is generated to test how well the model generalizes. This average subject in each group has a mean value for each input (Table 3). As mentioned earlier, Figures 1-3 show that the model has generalized well without overfitting, since the thermal responses of these average subjects lie within the range of thermal outputs of the group in transient environmental conditions. Results of the sensitivity analyses are provided as ranked values of the scaled sensitivity derivatives ([delta] (output) / [delta] (parameter)) in Tables 6 and 7.
Many interesting observations can be noted, some of which are described next, and they agree with findings reported in the literature. For instance, Havenith et al. (1998) suggested that part of the variance in Tcore could be explained by the [A.sub.d]/mass relationship for the subjects, with "big" subjects who are characterized by large values of [A.sub.d] (body surface area as a function of height and weight) and low [A.sub.d]/mass having a lower Tcore. One physiological explanation is that subjects with high Ad have a larger surface area for heat loss to the environment and have higher heat storage capacity. Small subjects (low mass; high [A.sub.d]/mass) are at a disadvantage when coping with heat stress due to their smaller heat storage capacity, as noted by Havenith et al (1998). Havenith et al. (1998) found a positive correlation between [A.sub.d]/mass ratio and core temperature not only for a warm-humid or jungle, but also for a neutral and a hot-dry (desert) climate. The positive correlation was highest in the warm-humid climate or jungle, however, implying that big subjects have the most marked advantage in the jungle or warm-humid climate. Our sensitivity analysis yielded the same results. As seen in Tables 6 and 7, an increase in weight, implying a smaller [A.sub.d]/mass ratio, causes Tcore to decrease in most of subjects (negative value for the scaled derivative), with the largest effect visible in jungle and desert environments. And the increase in height implying a larger Ad causes an increase in Tcore in most of the male groups, but this was not true for female group 1. With regard to the effect of fitness on thermoregulation, Gonzalez et al. (1997) showed that a higher value of gain in the rate of the total change in evaporative heat loss ([[DELTA]E.sub.SK]) to the change in mean body temperature ([[DELTA].sub.Tb]) indicates a greater ease in handling heat stress by increased evaporation of sweat without incurring excessive excursions of internal body temperature. And there was a significant positive correlation between [[[DELTA]E.sub.SK]/[[DELTA]T.sub.b]] and the maximum aerobic capacity (VO2max). The effect of VO2max on Tcore was confirmed in our sensitivity analysis, where the increase of VO2max leads to a decrease in Tcore in all models in most environmental conditions. There were two cases with a positive increase in Tcore in desert conditions for male groups 1 and 3, using the male model. Age by itself does not seem to be a factor in thermoregulation since the sensitivity analysis shows no major effect on Tcore, Tskin, and HR due to age.
Havenith et al. (1998) report that [VO.sub.2max], [A.sub.d], and % body fat were responsible for 62 % of the variance in linear regression model for Tskin. Their data set was obtained from 27 subjects (19 men and 8 women) engaged in ergometer exercise in a warm humid climate (35[degrees]C, 80% humidity) for one hour. The data are different from ours where the subjects are seminude and in a supine resting position in transient environmental conditions. However, our sensitivity study is consistent with this result. Also, there was a gender-based difference in terms of parameter contribution to Tskin. In males, [VO.sub.2max] was a main parameter with a negative impact on Tskin, implying partial acclimation, improved sweating, and skin cooling. However, in females, height and weight were the main parameters representing the heat loss capacity of the female body. But overall contributions of these parameters to Tskin were very small, as Tair and Pa are the main driving factors in our data set.
Heart rate variation accounts for some of the gender-based differences in heat stress responses. Women are reported to have higher heart rates than men (Stephenson and Kolka 1993) since, on average, they have a greater surface area to body mass ratio (Nunneley 1978), lower sweat rates (Frye and Kamon 1981), higher fat percentage, and lower aerobic fitness (Shapiro et al. 1980). In a study presented by Wells and Horvath (1973), the authors noted that independent of other stresses, heat raised HR by 6.5 beats per minute for women. Possibly the higher heart rate, elevated respiration, and metabolic requirements occurring in female subjects implies an inadequacy of circulatory transport of oxygen for women in hot conditions (Havenith and van Middendorp 1990). Havenith and van Middendorp (1990) note that the increase in HR needed is smaller for a higher [A.sub.d], as it provides more heat loss surface, resulting in improved cooling and, thus, efficient heat removal from core to skin (McArdle et al. 1991). Our findings indicate that the group with old women shows a negative effect on HR with increase of [A.sub.d] (by increasing height by one standard deviation) in all four conditions, but the other female groups do not. There are many such trends that can be determined from the sensitivity results in Tables 6 and 7, and we have discussed only some representative ones here. It is important to note that the model reported is a regression model that is valid only for the conditions tested, and not generalizable beyond that range. Also, the model has been developed with a limited data set, both as far as the number of subjects and the different transient conditions, which may influence the trends reported. Furthermore, future studies should consider the effect of different exercise levels, of mechanisms in cold conditions such as non-shivering thermogenesis, of factors such as hydration level, and of non-thermal factors such as smoking, medication, and menstrual cycle, depending on the conditions modeled and the prediction accuracy desired.
A novel transient predictive thermal model is reported for resting semi-nude supine individuals, to reliably predict Tcore, Tskin, and HR in transient environmental conditions for individual subjects. The model was then used to determine the relative importance of individual parameters on thermal stress variables, using a proposed sensitivity method. An important contribution of the reported study is to demonstrate that the predictive thermal model development methodology using neural networks is viable for human thermal response studies and can identify the relative importance of the individual parameters and environmental inputs. The model is validated by demonstrating prediction trends that are consistent with the literature. The ranking of the relative importance of the inputs in Tables 6 and 7 for various conditions provide 'fine-grained' insights into the thermal stress response phenomena. Some such insights are: fitness level is very important for predicting thermal risk but age is not; subjects with large [A.sub.d] will have an advantage in handling heat stress in jungle and desert conditions; the effect of gender on the handling of thermal stress was not obvious from the model, but it seems that women usually have higher core temperature even if their individual physiological characteristics are matched with that of male, and this is something that needs further study. The sensitivity results in Tables 6 and 7 can be used for numerous other comparison studies, depending on the application. For instance, one can compare between environments (jungle, desert, neural, cold), between outputs, and between groups of individuals (Table 3).
The data set we have used was limited to 35 semi-nude subjects resting in a supine position; it will be interesting to determine trends with larger data sets, in varying environmental conditions and exercise levels. Our study does, however, demonstrate that the proposed methodology for modeling and sensitivity analysis using neural networks has the potential to model and also discern the relative importance of the environ mental and individual parameters on thermal stress responses in different conditions.
This research was supported in part by grant # DAAD16-03-P-0119, from US Army Research Institute of Environmental Medicine, Natick, MA. The authors also gratefully acknowledge the numerous insights provided by Richard Gonzalez of USARIEM.
Table 6a. Ranked and Scaled Sensitivity Derivatives in Jungle and Desert Conditions for "Average" Male from Each Group JUNGLE Male Group 1 Male Group 2 'Height' -1.0000 (0.0307) 'VO2max' -1.0000 (0.0499) 'VO2max' -0.8168 'Height' -0.6403 Skin 'Weight' -0.5168 'HR' 0.5607 'HR' 0.4580 'age' -0.3570 'M' 0.1808 'M' 0.2425 'age' -0.0729 'Weight' -0.1791 'Height' -1.0000 (0.0208) 'VO2max' -1.0000 (0.0456) 'VO2max' 0.6244 'M' 0.6592 Core 'Weight' -0.5352 'Height' 0.5426 'HR' -0.5329 'Weight' -0.4719 'M' 0.3199 'HR' 0.3152 'age ' -0.0128 'age' 0.0543 'Height' 1.0000 (0.6435) 'VO2max' -1.0000 (1.6084) 'VO2max' 0.6887 'M' 0.6686 HR 'Weight' -0.6825 'Weight' -0.5244 'HR' -0.5687 'Height' 0.4324 'M' 0.2964 'HR' 0.3781 'age ' 0.0407 'age' 0.0388 JUNGLE Male Group 3 Male Group 4 'VO2max' -1.0000 (0.0402) 'Height' -1.0000 (0.0358) 'Height' -0.9099 'HR' 0.4420 Skin 'HR' 0.5445 'age' -0.3815 'age' -0.3629 'VO2max' -0.3348 'Weight' -0.3210 'M' -0.1268 'M' 0.1272 'Weight' 0.1177 'VO2max' -1.0000 (0.0345) 'Height' 1.0000 (0.0186) 'M' 0.6557 'M' 0.4067 Core 'Height' 0.6164 'age' 0.3757 'Weight' -0.4187 'Weight' -0.3748 'HR' 0.3391 'HR' -0.1680 'age' 0.1024 'VO2max' -0.0047 'VO2max' -1.0000 (1.1731) 'Height' 1.0000 (0.4831) 'M' 0.7125 'Weight' -0.6204 HR 'Weight' -0.5661 'M' 0.5774 'HR' 0.4910 'age' 0.4706 'Height' 0.4096 'VO2max' 0.0639 'age' 0.1027 'HR' -0.0028 DESERT Male Group 1 Male Group 2 'Height' -1.0000 (0.1226) 'Weight' -1.0000 (0.0384) VO2max' -0.6771 'Height' -0.9929 Skin 'age' -0.1032 VO2max' -0.7021 'M' -0.0824 'age' -0.1229 'Weight' -0.0797 'M' -0.0891 'HR ' 0.0289 'HR' -0.0148 'Height' 1.0000 (0.0667) 'Height' 1.0000 (0.0192) 'VO2max' -0.6832 'VO2max' 0.7685 Core 'HR' 0.0712 'Weight' 0.6059 'M' 0.0539 'HR' -0.1172 'Weight' 0.0512 'age' 0.0431 'age' -0.0173 'M' 0.0119 'VO2max' -1.0000 (0.8082) 'Weight' -1.0000 (0.8205) 'Height' -0.9513 'age' 0.6184 HR 'age' 0.8146 VO2max' 0.5288 'Weight' -0.5613 'HR' -0.2391 'HR' -0.2875 'M' 0.0714 'M ' 0.0845 'Height' 0.0368 DESERT Male Group 3 Male Group 4 'Weight' -1.0000 (0.1588) 'Height' 1.0000 (0.1146) 'HR' -0.8961 'VO2max' 0.7765 Skin 'VO2max' -0.4687 'age' -0.5112 'M' -0.3690 'M' -0.1242 'age' 0.1485 'HR' -0.0624 'Height' -0.0679 'Weight' 0.0616 'Weight' 1.0000 (0.0536) 'VO2max' -1.0000 (0.0127) 'HR' 0.8775 'Height' -0.9101 Core 'VO2max' 0.6373 'HR' 0.3905 'M' 0.3704 'M' 0.3741 'Height' 0.2671 'Weight' 0.3580 'age' -0.1984 'age' 0.1439 'Weight' -1.0000 (1.2349) 'Height' 1.0000 (0.7449) 'age' 0.6460 'VO2max' 0.8327 HR 'Height' -0.2815 'Weight' -0.3591 VO2max' 0.1717 'M' 0.2020 'HR' -0.1112 'age' -0.1281 'M' 0.0678 'HR' 0.0783 Table 6b. Ranked and Scaled Sensitivity Derivatives in Jungle and Desert Conditions for "Average" Female from Each Group JUNGLE Female Group 1 Female Group 2 'Height' 1.0000 (0.2052) 'HR ' 1.0000 (0.0839) 'VO2max' -0.0792 'Height' 0.2866 Skin 'Weight' -0.0422 VO2max' -0.2551 'HR' -0.0383 'Weight' -0.0998 'M' 0.0152 'M' 0.0254 'age' 0.0070 'age' 0.0046 'Height' -1.0000 (0.0074) 'VO2max' -1.0000 (0.0767) 'VO2max' -0.7913 'Height' 0.7765 Core 'Weight' -0.4197 'Weight' -0.3896 'HR' -0.3538 'HR' 0.0527 'M' 0.1738 'age' 0.0461 'age' 0.0582 'M' 0.0047 'Height' 1.0000 (4.0794) 'HR ' 1.0000 (2.9869) 'VO2max' -0.0464 'VO2max' 0.8947 HR 'Weight' -0.0257 'Height' -0.6043 'HR' -0.0183 'Weight' 0.3455 'M' 0.0110 'age' -0.0492 'age' 0.0034 'M' 0.0217 JUNGLE Female Group 3 Female Group 4 'Height' 1.0000 (0.2052) 'HR' 1.0000 (0.0839) 'VO2max' -0.0792 'Height' 0.2866 Skin 'Weight' -0.0422 VO2max' -0.2551 'HR' -0.0383 'Weight' -0.0998 'M' 0.0152 'M' 0.0254 'age' 0.0070 'age' 0.0046 'Height' -1.0000 (0.0074) 'VO2max' -1.0000 (0.0767) 'VO2max' -0.7913 'Height' 0.7765 Core 'Weight' -0.4197 'Weight' -0.3896 'HR' -0.3538 'HR' 0.0527 'M' 0.1738 'age' 0.0461 'age' 0.0582 'M' 0.0047 'Height' 1.0000 (4.0794) 'HR' 1.0000 (2.9869) 'VO2max' -0.0464 'VO2max' 0.8947 HR 'Weight' -0.0257 'Height' -0.6043 'HR' -0.0183 'Weight' 0.3455 'M' 0.0110 'age' -0.0492 'age' 0.0034 'M' 0.0217 DESERT Female Group 1 Female Group 2 'Height' -1.0000 (0.1737) 'HR' -1.0000 (0.0969) 'Weight' -0.5117 'Height' -0.6465 Skin 'M' 0.4560 'Weight' -0.4915 'HR' 0.4544 'M' -0.3519 'VO2max' -0.3486 'age' 0.1297 'age' 0.0033 'VO2max' -0.0638 'Height' -1.0000 (0.2101) 'Height' 1.0000 (0.0888) 'VO2max' -0.3828 'HR' -0.6387 Core 'M' 0.3810 'VO2max' -0.5965 'Weight' -0.2437 'Weight' -0.4501 'HR' 0.2225 'M' 0.0513 'age' -0.0396 'age' 0.0464 'M' 1.0000 (0.9551) 'Height' -1.0000 (3.7397) 'HR' 0.7674 'HR' 0.5831 HR 'VO2max' -0.5548 VO2max' 0.4943 'Height' -0.3551 'Weight' 0.2906 'Weight' -0.3133 'M' -0.0351 'age' -0.0079 'age' -0.0186 DESERT Female Group 3 Female Group 4 'Weight' -1.0000 (0.2013) 'HR' -1.0000 (0.1002) 'M' -0.1653 'VO2max' -0.6901 Skin 'HR' 0.1031 'M' -0.3525 'Height' 0.0783 'Weight' -0.0950 VO2max' 0.0686 'age' 0.0926 'age' 0.0681 'Height' 0.0846 'Weight' -1.0000 (0.0507) 'Height' 1.0000 (0.2851) 'Height' 0.8240 'HR ' -0.1692 Core 'VO2max' -0.5391 'Weight' -0.1195 'HR ' -0.2055 'VO2max' -0.1001 'M ' 0.0592 'age ' 0.0207 'age ' 0.0520 'M ' 0.0202 'Height' -1.0000 (1.2761) 'Height' -1.0000 (10.9790) 'VO2max' 0.8468 'HR' 0.1449 HR 'HR' 0.7491 'Weight' 0.1163 'Weight' -0.1037 'age' -0.0186 'M' -0.0921 VO2max' 0.0183 'age' 0.0340 'M' -0.0015 Table 7a. Ranked and Scaled Sensitivity Derivatives in Neutral and Cold Conditions for an "Average" Male from Each Group NEUTRAL Male Group 1 Male Group 2 Skin 'VO2max' -1.0000 'VO2max' -1.0000 (0.0596) (0.0943) 'M' 0.7546 'M' 0.5339 'HR' 0.6593 'HR' 0.4549 'age' 0.3512 'Weight' -0.3209 'Weight' -0.2115 'age' -0.1890 'Height' -0.0608 'Height' -0.0288 Core 'VO2max' -1.0000 'M' 1.0000 (0.0494) (0.0346) 'M' 0.9818 'HR' 0.7116 'HR' 0.7320 'VO2max' -0.5670 'Weight' -0.2322 'Height' 0.2015 'age' -0.1995 'Weight' -0.1186 'Height' 0.0992 'age' -0.1130 HR 'VO2max' -1.0000 'M' 1.0000 (1.7602) (1.1042) 'M' 0.9260 'VO2max' -0.8093 'HR' 0.6919 'HR ' 0.7340 'Weight' -0.2702 'Weight' -0.2925 'age' -0.2027 'Height' 0.1750 'Height' 0.1044 'age ' -0.1177 COLD Male Group 1 Male Group 2 Skin 'VO2max' -1.0000 'VO2max' -1.0000 (0.1254) (0.0916) 'M' 0.6508 'M' 0.9661 'HR' 0.4718 'HR' 0.7239 'Weight' -0.3122 'Weight' -0.3813 'age ' -0.2065 'age' -0.2169 'Height' 0.1588 'Height' 0.1440 Core 'M' 1.0000 'M' 1.0000 (0.0108) 'VO2max' (0.0044) 'HR' 0.8668 'HR' -0.8247 VO2max' -0.8210 'Height' 0.6706 'Weight' -0.2639 'Weight' 0.4811 'Height' 0.1094 'age' 0.0732 'age' -0.0298 -0.0485 HR 'VO2max' -1.0000 'VO2max' -1.0000 (0.9350) (0.6525) 'M' 0.6692 'M' 0.9854 'HR' 0.5025 'HR' 0.7294 'Weight' -0.3224 'Weight' -0.3266 'Height' 0.1524 'Height' 0.2299 'age' -0.1220 'age' -0.1688 NEUTRAL Male Group 3 Male Group 4 Skin 'VO2max' -1.0000 'VO2max' -1.0000 (0.0915) (0.0761) 'Weight' -0.5862 'HR' 0.7357 'M' 0.4256 'M' 0.6903 'age ' -0.1335 'Height' -0.2913 'HR' 0.1155 'Weight' -0.2654 'Height' 0.0658 'age' -0.1324 Core 'M' 1.0000 'Height' 1.0000 (0.0535) (0.0619) 'HR' 0.8582 'age' -0.3773 'VO2max' -0.5472 'M' 0.2728 'age' -0.1770 'VO2max' 0.2440 'Height' 0.1097 'Weight' 0.1029 'Weight' 0.0093 'HR' 0.0431 HR 'M' 1.0000 'VO2max' -1.0000 (1.6003) (1.6833) 'HR' 0.8076 'M' 0.6670 'VO2max' -0.6502 'HR' 0.6355 'Weight' -0.2326 'Weight' -0.2567 'age' -0.1512 'Height' -0.2002 'Height' 0.0669 'age' -0.0601 COLD Male Group 3 Male Group 4 Skin 'VO2max' -1.0000 'Height' 1.0000 (0.1099) (0.0864) 'M' 0.8237 'M' 0.7971 'Weight' -0.5119 'VO2max' -0.5506 'HR' 0.4979 'HR' 0.5298 'Height' 0.2050 'age' -0.4803 'age' -0.1843 'Weight' -0.3115 Core 'HR' 1.0000 'VO2max' -1.0000 (0.0093) (0.0260) 'M' 0.8360 'M' 0.7896 'Weight' 0.5040 'HR' 0.5192 'VO2max' -0.3730 'Height' -0.2338 'age' -0.1399 'Weight' -0.1972 'Height' 0.0718 'age' 0.0144 HR 'M' 1.0000 'VO2max' -1.0000 (0.7283) (0.9646) 'VO2max' -0.9832 'M' 0.8780 'HR' 0.8712 'HR' 0.6396 'Weight' -0.2380 'Weight' -0.3306 'age' -0.1957 'Height' -0.0959 'Height' 0.1582 'age' -0.0489 Table 7b. Ranked and Scaled Sensitivity Derivatives in Neutral and Cold Conditions for an "Average" Female from Each Group NEUTRAL Female Group 1 Female Group 2 Skin 'Height' -1.0000 'HR' 1.0000 (0.4401) (0.0379) 'HR' 0.1941 'Height' 0.4880 'M' 0.1908 'M' -0.4250 'VO2max' -0.0468 'VO2max' -0.3696 'age' 0.0060 'Weight' 0.1242 'Weight' -0.0024 'age' 0.0596 Core 'Height' -1.0000 'Height' -1.0000 (0.2805) (0.0919) 'HR' 0.1273 'VO2max' 0.6874 'M' 0.1246 'HR' -0.3885 'VO2max' -0.0501 'Weight' 0.3716 'Weight' -0.0267 'age' -0.0661 'age' -0.0001 'M' 0.0546 HR 'HR' 1.0000 'Height' 1.0000 (0.5884) (3.4201) 'Height' -0.9977 'HR' 0.8945 'M' 0.9580 'VO2max' -0.7080 'VO2max' -0.2107 'Weight' 0.4223 'Weight' -0.1902 'M' -0.0703 'age' 0.1173 'age' 0.0537 COLD Female Group 1 Female Group 2 Skin 'Height' -1.0000 'Height' -1.0000 (0.5384) (0.3751) 'Weight' 0.1941 'HR' -0.5662 'M' 0.1908 'VO2max' -0.2842 'HR' 0.0468 'Weight' -0.1759 'age' -0.0060 'M' 0.0929 0.0024 'age' -0.0589 'VO2max' Core 'Height' -1.0000 'VO2max' -1.0000 (0.2581) (0.0906) 'HR' 0.1273 'HR' -0.9480 'M' 0.1246 'Height' 0.4039 'Weight' 0.0501 'Weight' -0.3758 'age' -0.0267 'M' 0.0584 -0.0001 'age' 0.0445 'VO2max' HR 'HR' 1.0000 'Height' -1.0000 (0.7461) (3.5811) 'M' 0.9977 'VO2max' 0.6936 'Height' 0.9580 'HR' 0.6396 'Weight' 0.2107 'Weight' 0.4979 -0.1902 'M' 0.1032 'VO2max' -0.1173 'age' -0.0772 'age' NEUTRAL Female Group 3 Female Group 4 Skin 'M' 1.0000 'Height' 1.0000 (0.0145) (0.0691) 'Height' -0.9927 'VO2max' -0.7113 'HR' 0.9741 'HR' 0.4191 'Weight' 0.3912 'M' 0.2255 'age' 0.1264 'Weight' -0.1747 'VO2max' 0.1055 'age' 0.0765 Core 'Height' 1.0000 'Height' 1.0000 (0.0331) (0.3721) 'VO2max' -0.6904 'Weight' -0.1253 'HR' -0.5064 'VO2max' -0.0920 'Weight' -0.2698 'HR' -0.0688 'M' 0.1636 'age ' 0.0284 'age' 0.0627 'M' 0.0103 HR 'HR' 1.0000 'Height' -1.0000 (1.1482) (14.1845) 'Height' -0.9973 'HR' 0.1828 'VO2max' 0.8577 'Weight' 0.1232 'Weight' 0.3942 'VO2max' 0.0477 'M' 0.1622 'M' 0.0248 'age' -0.0050 'age' -0.0240 COLD Female Group 3 Female Group 4 Skin 'Height' -1.0000 'HR' -1.0000 (0.3742) (0.2593) 'Weight' 0.9927 'VO2max' -0.9295 'M' 0.9741 'Height' 0.1834 'age' -0.3912 'M' 0.1763 'VO2max' -0.1264 'Weight' 0.1172 'HR' -0.1055 'age' -0.0612 Core 'VO2max' -1.0000 'Height' 1.0000 (0.0249) (0.4716) 'HR' -0.6904 'HR' -0.1713 'Height' -0.5064 'VO2max' -0.1411 'M' 0.2698 'Weight' -0.1178 'Weight' -0.1636 'age' 0.0245 'age' -0.0627 'M' 0.0117 HR 'Height' -1.0000 'Height' -1.0000 (1.7732) (17.0163) 'HR' 0.9973 'Weight' 0.1329 'VO2max' 0.8577 'HR' 0.0898 'Weight' 0.3942 'M' 0.0310 'M' 0.1622 'age' -0.0309 'age' -0.0050 'VO2max' -0.0064
Amari, S., N. Murata, K.-R. Muller, M. Finke, and H.H. Yang. 1997. Asymptotic statistical theory of overtraining and cross-validation. IEEE Transactions on Neural Networks 8(5):985-96.
Bue, G.C. Computer program documentation: 41-node transient metabolic man program. Lockheed Engineering and Sciences Company, Houston, TX, Tech. Rep.
Campbell, A.B., S.S. Nair, and J.B. Miles. 1994. Modeling the sweat regulation mechanism. Proceedings of the 24th International Conference on Environmental Systems and 5th European Symposium on Space Environ mental Control Systems. Friedrichsafen, Germany, June 20-23.
Frank, S.M., S.N. Raja, C.F. Bulcao, and D.S. Goldstein. 1999. Relative contribution of core and cutaneous temperatures to thermal comfort and autonomic responses in humans. Journal of Applied Physiology 86(5):1588-93.
French, J.D., A.D. Viswanath, S.S. Nair, J.B. Miles, and C.H. Lin. 1997. Parameters and assumptions in human thermal modeling for EVA applications. SAE Transactions, Journal of Aerospace 106:579-86.
Frye, A.J., and E. Kamon. 1981. Responses to dry heat of men and women with similar aerobic capacities. Journal of Applied Physiology 50(1):65-70.
Fu, G. 1995. A transient, three-dimensional mathematical thermal model for the clothed human. PhD thesis, Kan sas State University.
Gonzalez, R.R., L.G. Berglund, and J.A.J. Stolwijk. 1981. Thermoregulation in humans of different ages during thermal transients. Advances in Physiological Sciences. Chapter 32, Contributions to Thermal Physiology. Pergamon Press. pp. 560.
Gonzalez, R.R., T.M. McLellan, W.R. Withey, S.K. Chang, and K.B. Pandolf. 1997. Heat strain models applicable for protective clothing systems: Comparison of core temperature response. Journal of Applied Physiology 83(3):1017-32.
Gonzalez, R.R. 2001. Problems of heat exchange and exercise during long term space operations: Use of a thermo-regulatory model to describe physiologic response. Society of Automotive Engineers.
Guan, Y., M. Hosni, B. Jones, and T. Gielda. 2003. Literature review of the advances in thermal comfort modeling. ASHRAE Transactions.
Havenith, G., and H. van Middendorp. 1990. The relative influence of physical fitness, acclimatization state, anthropometrics measures and gender on individual reactions to heat stress. European Journal of Applied Physiology 61:419-27.
Havenith, G., J.M.L. Coenes, L. Kistemaker, and W.L. Ken-ney. 1998. Relevance of individual characteristics for human heat stress response is dependent on exercise intensity and climate type. European Journal of Applied Physiology 77:231-41.
Hsu, F. 1971. Modeling, simulation, and optimal control of the human thermal system. PhD thesis, Kansas State University.
Huizenga, C., H. Zhang, and E. Arens. 2001. A model of human physiology and comfort for assessing complex thermal environments. Building and Environment 36:691-99.
Hwang, C.L., and S.A. Konz. 1977. Engineering models of the human thermoregulatory system: A review. IEEE Transactions on Biomedical Engineering 24(4):309-25.
Iyoho, A. 2002. Development of a two-dimensional human thermal model for E VA Applications. MS thesis, Department of Mechanical and Aerospace Engineering, University of Missouri-Columbia.
McArdle, W.D., F.I. Katch, and V.L. Katch. 1991. Exercise Physiology: Energy, Nutrition, and Human Performance. Philadelphia: Lea & Febiger.
Mistry, S.I., and S.S. Nair. 1993. Nonlinear HVAC computations using neural networks. ASHRAE Transactions 99(1):775-84.
Moran, D.Y., A. Epstein, A. Laor, A. Vitalis, and Y. Shapiro. 1995. Predicting heart rate response to various meta bolic rates, environments, and clothing. Journal of Applied Physiology 78(1):318-22.
Nunneley, S. 1978. Physiological responses of women to thermal stress: A review. Medicine and Science in Sports and Exercise 10:250-55.
Pennes, H.H. 1948. Analysis of tissue and arterial blood temperatures in the resting human forearm. Journal of Applied Physiology 1(2):93-122.
Shapiro, Y., K.B. Pandolf, B.A. Avellini, N.A. Pimental, and R.F. Goldman. 1980. Physiological responses of men and women to humid and dry heat. Journal of Applied Physiology 49:1-8.
Stephenson, L.A., and M.A. Kolka. 1993. Thermoregulation in women. Exercise and Sports Science Reviews 21:231-62.
Stolwijk, J.A.J. 1971. A mathematical model of physiological temperature regulation in man. Tech. Rep. CR-1855, NASA, Washington, DC.
Tanabe, S., K. Kobayashi, J. Nakano, Y. Ozeki, and M. Konishi. 2002. Evaluation of thermal comfort using com bined multi-node thermoregulation (65MN) and radiation models and computational fluid dynamics (CFD). Energy and Buildings 34:637-46.
Thornton, S.B., W. Xu, and S.S. Nair. 2002. Thermal control of an advanced space suit. Transactions of the Society of Automotive Engineers, Journal of Aerospace, pp. 69-77.
Thornton, S.B., and S.S. Nair. 2000. Parametric studies of human thermal mechanisms and measurements. IEEE Transactions on Biomedical Engineering 47(4):444-51.
Thornton, S.B., S.I. Mistry, and S.S. Nair. 1997. Sensitivity analysis for building thermal loads. ASHRAE Transactions 103(2):165-74.
Wells, C.L., and S.M. Horvath. 1973. Metabolic and thermoregulatory responses of women to exercise in two thermal environments. Medicine and Science in Sports 6(1):8-13.
Wissler, E.H. 1964. A mathematical model of the human thermal system. Bulletin of Mathematical Biophysics 26:147-66.
Wissler, E.H. 1971. Comparison of computed results obtained from two mathematical models: A simple 14-node model and a complex 250-node model. Journal de Physiologie 63(3)P:455-58.
Tai S. Jang
Anthony E. Iyoho
Larry G. Berglund, PhD, PE
Satish S. Nair, PhD, PE
Tai S. Jang and Anthony E. Iyoho are PhD students and Satish S. Nair is a professor in Electrical and Computer Engineering at the University of Missouri, Columbia, MO. Larry G. Berglund works in the Biophysics and Biomodeling Division, U.S. Army Research Institute of Environmental Medicine, Natick, MA.
|Printer friendly Cite/link Email Feedback|
|Author:||Jang, Tai S; Berglund, Larry G; Iyoho, Anthony E; Nair, Satish S|
|Date:||Jan 1, 2009|
|Previous Article:||Modeling for predicting frost behavior of a fin-tube heat exchanger with thermal contact resistance.|
|Next Article:||Optimization of cooling-dominated hybrid ground-coupled heat pump systems.|