Wasserstein local slow feature analysis and its application to process monitoring

Abstract

Abstract Complex industrial processes are commonly characterized by dynamics, which results from compensation of the closed-loop control and complex reflux during process operations. In this work, we propose a novel method termed the Wasserstein local slow feature analysis approach (WLSFA), used to monitor the dynamic process, which can learn slowly varying information to capture the trend of process variations and characterize dynamics. Specifically, Wasserstein graph embedding based on optimal transport is proposed to preserve the local geometrical structure of raw data so that the reduced-dimensionality representations can more faithfully reveal the underlying information of data, enhancing monitoring performance. Moreover, the l 2 -norm orthogonal constraint is incorporated into the objective function to improve the generalization ability and alleviate the false alarm rate. Based on the WLSFA model, a contribution plot denoting reconstruction is developed to locate the fault variables, which is beneficial for regulating abnormal conditions toward normal levels. Finally, the efficiency of the proposed approach is illustrated by a benchmark process and application to a real-world fractionation process.