Engineering System Condition Monitoring And Fault Diagnosis Using Method Of Cointegration

At 2003, Robert F. Engle and Clive WJ. Granger received the Nobel Prize, because their Cointegration theory had solved two difficult problems, the Time-Varying Volatility and Non-stationary in the time series analysis field.In this paper, will introduce the normal knowledge of Cointegration theory, and emphatically depict the possibility of applying Cointegration algorithm to condition monitoring and fault diagnosis for engineering systems, which are non-stationary processes. It is well known that non-stationary system behavior causes grave difficulties on system modeling and condition monitoring due to the time-dependent statistics. ARIMA model and dynamic PCA or PLS methods have been employed to deal with the non-stationary issues and made a good progress. However, there are severe limitations about those dynamic models. Cointegration method is a new algorithm developed in economics area for modelling non-stationary economical and financial processes. It can generate a stationary process based on a linear combination of a set of non-stationary processes if these non-stationary processes have long-term balanced relation. This feature provides a possibility for researchers in engineering sectors to apply cointegration method to condition monitoring and fault diagnosis of engineering systems. To verify this possibility, an example based on a vehicle engine and FCCU simulation system was discussed in details. The data from the engine and FCCU simulator are processed and modeled using a cointegration testing method, and, propose the sub-cointegration model to diagnosis the systems'fault. The results demonstrate that the cointegration test method has a bright future in engineering area for condition monitoring and fault diagnosis for dynamic engineering systems.