Prediction and isolation of process faults using operable adaptive sparse identification of systems (OASIS) and contribution plots

Fault prognosis is of great significance for improving the safety and stability of process operation. In this article, we propose a data-driven adaptive fault prognosis strategy. Specifically, the proposed method utilizes operable adaptive sparse identification of systems (OASIS) to adaptively predict the future behavior of the process based on its current dynamics. Further, a dynamic risk evaluation scheme is employed using the obtained future behavior for predicting impending faults. The OASIS algorithm can well-approximate the process dynamics by coping with abrupt system changes, including the changing dynamics under the initial faulty conditions, and therefore provide an early fault prediction. Once a fault is predicted, the fault isolation step is executed to identify and isolate the faulty variables using contribution plots. One of the main goals of our work is to explore the utility of contribution plots for fault prognosis applications. Implementing contribution plots at the initial stage of a fault before it affects the other parts of the process overcomes the limitation of ambiguity in analyzing the contribution plots due to fault propagation, which otherwise is a challenge for traditional fault diagnosis methods. Thus, achieving an early fault prediction using OASIS enhances the performance of contribution plots and improves fault isolation accuracy. The effectiveness of the proposed framework is evaluated by implementing it for the fault prognosis of a polyethylene reactor.