Feature Based Fault Detection for Pressure Swing Adsorption Processes

Abstract Over the past few decades, there has been widespread development of pressure swing adsorption (PSA) systems, with their applications expanding from traditional bulk gas separation and drying, to CO2 sequestration, trace contaminant removal, and many others. With extensive industrial applications, there is a significant need for effective monitoring methods to detect and diagnose process abnormalities in realtime, as well as to facilitate predictive maintenance for avoiding major production disruptions ahead. Although periodic operations such as PSA have been used widely in chemical and petrochemical industries, the process monitoring of these operations has received limited attention compared to non-periodic continuous or batch processes. A potential reason is that the monitoring of periodic processes is significantly more challenging than that of processes operated at steady-state. In this work, we propose a data-driven feature space monitoring (FSM) approach for PSA processes. We show that the FSM based fault detection naturally addresses the challenges in monitoring periodic processes, such as unequal step and/or cycle time that requires trajectory alignment or synchronization for the traditional statistical process monitoring (SPM) methods. In addition, we demonstrate the superior fault detection performance of the proposed method compared to the conventional SPM methods using both simulated faults and real faults from an industrial PSA process.