STUDY ON FAULT DETECTION AND DIAGNOSIS OF HVAC SYS

When some faults take place in HVAC system changing pattern of the state variables of using some statistical analysis, Fourier Trans form, Cluster analysis and variable selection to achieve optimal fault detection parameter. and using recognizing patterns with Mahanobis' generalized distance to detect any difference between normal state and abnormal state .In the last paper ( Zheng ,et ,al .,1998) , the high performance of fault detection and diagnosis for an actual water thermal storage HVAC system using this method was verified . The present paper reports an application on the fault detection and diagnosis for the air side system in an actual HVAC/VAV system, and the following five matters have been discussed .a) achieving the system character from the mesured data of the object system using Fourier Transform, b) designing parameters ,c) selecting useful parameters with cluster analysis ,d) process of selecting optimal parameters , or vectors and e) analyzing physical meaning of them.