Concurrent quality and process monitoring with a probabilistic sparse nonlinear dynamic method

Abstract

In industrial processes, significant attention has been directed toward developing monitoring frameworks that effectively capture the interactions between process variables and quality-related variables. This paper presents a novel probabilistic sparse nonlinear dynamic method, CPSINDy, for concurrent quality and process monitoring in industrial systems. The proposed method incorporates nonlinear dynamics by formulating a comprehensive framework based on a probabilistic state–space model. Subsequently, leveraging the particle filtering technique, parameter estimation is performed using the Expectation–Maximization algorithm. After that, four dynamic indices are introduced to detect abnormal operating conditions. Both feasibility and superiority of the presented model are confirmed through three realistic industrial fault cases. Results demonstrate that CPSINDy based model outperforms traditional approaches in terms of fault detection rates and false alarm rates.