Monitoring independent components for fault detection



A chemical process has a large number of measured variables, but it is usually driven by fewer essential variables, which may or may not be measured. Extracting such essential variables and monitoring them will improve the process-monitoring performance. Independent component analysis (ICA) is an emerging technique for finding several independent variables as linear combinations of measured variables. In this work, a new statistical process control method based on ICA is proposed. For investigating the feasibility of its method, its fault-detection performance is evaluated and compared with that of the conventional multivariate statistical process control (cMSPC) method using principal-component analysis by applying those methods to monitoring problems of a simple four-variable system and a continuous-stirred-tank-reactor process. The simulated results show the superiority of ICA-based SPC over cMSPC.