Application of blind source analysis to multivariate statistical process monitoring

Multivariate statistical process control (MSPC) has been applied to performance monitoring for chemical processes. However, traditional methods of MSPC are based on the noise-corrupted data, which will make the performance of MSPC become worse. In this paper, a novel multivariate statistical projection analysis based on data de-noised with blind signal analysis and wavelet transform is presented, which can detect fault more quickly, so improves monitoring performance of the process. Through a simulation with a binary distillation column for benzene-toluene, we verify the more effectiveness and better performance of the new method than conventional MSPC.