A novel quality-related process monitoring based on multi-block slow feature analysis of mutual information-index

As an emerging feature extraction method, slow feature analysis (SFA) abstracts the intrinsic properties characterizing the system from the original input signal, and therefore it is considered to have the potential to uncover the true laws of industrial processes. In recent years, slow feature analysis has developed into an important method in the field of quality-related process monitoring. In this paper, a multi-block strategy is combined with slow feature analysis, and mutual information (MI) indicators are used for the division of quality-related subspaces. Besides, statistical indicators are used to further divide the quality-related subspaces. Monitoring statistics and control limits are constructed in the three subspaces for process monitoring. To verify the effectiveness of the method, simulation experiments were conducted in the context of the Tennessee Eastman process.