Options
2009
Conference Paper
Titel
Are indices for thermal comfort adequate for modelling of overall comfort?
Abstract
Today several indices, such as the Predict Mean Vote (PMV), are accepted widely and are used to rate a situation's thermal comfort. In the last years there have been some studies with a focus on a more holistic approach: to find the interactions between comfort votes influenced by the main environmental parameters, e.g. temperature, background noise, air quality, relative humidity, illumination level. Based on data from studies in a simulated aircraft cabin, the Fraunhofer Flight Test Facility (FTF), two models have been developed to investigate whether it is advantageous to use indices for thermal comfort instead of directly measured data to formulate a more holistic comfort model. The technique of structural equation models has been used to combine statistically the measured and/or calculated data with the votes of more than a thousand subjects with respect to the latent (i.e. not directly measurable) variables of comfort. Parameters for thermal and aural environment as well as air quality have been considered for the modelling and linked together with the aim to find some model for the prediction of overall comfort. Both models are based on the same data and the same structure except for the main exogeneous ("input") variables. One model uses measured variables of the physical environment, such as dry bulb temperature, while the other models use climate indices, such as PMV and ET. Since climate indices generally are premised on some model they imply a certain modelling error per se. On the other hand they reduce the degrees of freedom for further modelling and thus a source of unexplainable variance. After evaluating the admissibility of the models the approaches have been compared based on goodness-of-fit indices. Based on these fit measures it is judged whether the models using indices for comfort are suitable to depict the subjects' perception of the indoor environment and thus can compensate the disadvantages because of implicit errors through the benefit of data reduction. The structural equation models hypothesized are admissible and fit the data well. The models using environmental indices have better fit criteria than the one based on environmental comfort models, since indices already carry some information to explain the implied covariance. There is no remarkable difference of the models using PMV or ET.