ニューラルネットワークを用いた車両空調時の温感評価方法

概要

論文の詳細を見る
We have been studying a new automobile air conditioner that controls the thermal sensation of the occupants. We believe that in air conditioners, the most important controlling factor is not air temperature but the thermal amenity of occupants. For control of thermal amenity, it is necessary to quantitatively evaluate the thermal sensation of the occupant. We had proposed that the thermal sensation of an occupant corresponds well with skin temperature in previous work, but this relationship is not always linear. The proposed method to quantitatively evaluate the thermal sensation of the occupant is based on the neural networks which have often been applied to those concerning nonlinear problems^<(1)>. The input data are the occupant's skin temperature and the passenger compartment air temperature; the output is the thermal sensation of the automobile occupant. We make the neural networks learn the relation between skin temperature and thermal sensation by back-propagation. The effect of the number of hidden cells is also shown. The results show that the accuracy is within 1.0 for the voting scales used. This accuracy is sufficient to control our new air conditioner.
社団法人日本機械学会の論文
1992-01-25