Fault Detection for Multibatch Dynamic Nonstationary Processes Using Dynamic CCA and Multibatch Benchmark Distance

ABSTRACT Modern chemical production processes are complex and diverse, and fault detection methods have always been an important part of complex chemical production process research. The variance and mean of the multibatch dynamic nonstationary process vary with time, which makes the dynamic canonical correlation analysis (DCCA) method deficient in calculating statistics and modeling the multibatch process. Therefore, the DCCA with multibatch benchmark distance (DCCA–MBD) for fault detection method is proposed in this paper. Based on the use of DCCA, the detection of multibatch dynamic nonstationary process is based on DCCA improved by the multibatch modeling scheme and statistic calculation method of the MBD method. Through numerical simulation and monitoring experiments of continuous stirred tank reactor production, it is verified that the method of this paper is more effective in dealing with multibatch dynamic nonstationary process compared with conventional methods.