Nonstationary Fault Diagnosis by Dual Analysis of Common and Specific Fault Variations with Cointegration Analysis

Abstract

The fault cases of complex industrial processes in general show typical nonstationary variations which reveal time-varying means or time-varying variances. The stationary fault information may be buried in the nonstationary fault variations and hard to be extracted. Besides, the existing fault diagnosis methods do not consider the underlying relations among different fault classes, which may lose important classification information. Here, it is recognized that different faults may not only share some common information but also have some specific characteristics. A fault diagnosis strategy with dual analysis of common and specific fault variations is proposed in this work. The nonstationary variables are first distinguished from the stationary variables by using Augmented Dickey-Fuller test. Then common and specific fault information are separated by developing two models for fault diagnosis. The fault-common model is constructed by cointegration analysis to capture the common nonstationary fault variations, and fault-specific model is built to explain the specific fault variations of each fault. With dual consideration of common and specific fault characteristics, the classification accuracy and fault diagnosis performance can be greatly improved. The performance of the proposed method is illustrated with both a real industrial process and a well-known benchmark process.