A Fault Diagnosis System for the Continuous Process with Load Fluctuations

Abstract

A method for fault diagnosis of a continuous process with load fluctuations has been developed. It consists of two algorithms, one an efficient algorithm for data rectification developed by introducing Akaike's Information Criterion and the other a fault diagnosis algorithm utilizing the signed directed graph. The former detects abnormal values of the state variables in the process with load fluctuations and supplies them to the latter. The latter searches candidates for the origin of failure which can explain all the abnormal values detected by the former. The efficency of this method is demonstrated by applying it to three types of miniature pipelines with different configurations.