213 サポートベクトルマシンを用いたエアコンの診断手法について(セッション7 診断システム(2))

概要

論文の詳細を見る
To avoid failures of an air-conditioner, we need suitable condition monitoring technique. Unfortunately, an air-conditioner has many parameters in it, i.e., inlet temperature, outlet temperature of a condenser or an evaporator, pressure at each component and input power to compressor and so on. Furthermore, we need expensive sensors to measure pressure. It is very difficult to monitor so many parameters from viewpoint of cost, setup in an air-conditioner and calculation cost of diagnosis technique. This is one of reasons why we must select suitable parameters for diagnosis. In this paper, we checked suitable parameters to monitor the condition of an air-conditioner by experiment. And then we applied neural network, i.e., NN, and support vector machine, i.e., SVM, with two types of kernel function to diagnose its failure. We also checked the effect of parameters on diagnosis performance to reduce the number of input parameter for NN or SVM. Finally, we selected several parameters for input data and showed good diagnosis ability of SVM with polynomial kernel.
2010-12-15