Fault Diagnosis For Air Conditioning Sensors Based On Kernel Fisher Discriminant Analysis

With extensive application, HVAC systems have been improved and become much complicated. In an increasingly complex and large HVAC System, It's beyond the ability of operators to detect and diagnose all faults immediately. So the stability and reliability of sensors play a significant role in air conditioning systems and the fault detection and diagnosis technology is becoming more and more important at present and in the future.Fisher discriminant analysis is presented to diagnose fault of temperature and flow rate sensors in air conditioning systems. Fisher discriminant analysis studies and projects the selected fault samples by using statistic method to separate them first, and then calculate and compare the Mahalanobis(M) distances between projections of the real measures and fault swatch to classify real measurement vector. The first step of fisher discriminant analysis is to deal with the selected sample, so the accuray of sample can strongly affect the ability of fisher discriminant analysis. Then a statistic method combining principal component analysis (PCA) with Fisher discriminant analysis(FDA) is presented to the application of fault detection and diagnosis(FDD).PCA model is set up to detect the peculiar vectors by comparing the...