A Novel Nonstationary Process Monitoring Model Based on Cointegration Analysis

The complex industrial processes are often characterized by nonstationarity due to operational change, measurement aging, or catalyst passivation. Traditional process monitoring methods may fail to effectively detect the fault because they are based on the assumption of independently and identically distributed without considering the nonstationarity. To solve this issue, a novel monitoring approach is developed to monitor the processes with the coexistence of stationary and nonstationary variables. A variable partitioning strategy is first proposed to distinguish stationary and nonstationary variables, and the cointegration analysis is applied to exploit the long-term equilibrium relationship among nonstationary variables. Then, an approach for extracting global and local information is developed by combining principal component analysis and local tangent space alignment, which can explore the informative representations of data. A case study on the Tennessee Eastman process is used to demonstrate the validity of the proposed approach.