Feature Based Fault Detection for Pressure Swing Adsorption Processes

摘要: 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, many others. With extensive industrial applications, is a significant need for effective monitoring methods detect diagnose process abnormalities in realtime, as well facilitate predictive maintenance avoiding major production disruptions ahead. Although periodic operations such PSA have used widely chemical petrochemical industries, these received limited attention compared non-periodic continuous or batch processes. A potential reason that processes significantly more challenging than operated at steady-state. In this work, we propose data-driven feature space (FSM) approach We show FSM based fault detection naturally addresses challenges processes, unequal step and/or cycle time requires trajectory alignment synchronization statistical (SPM) methods. addition, demonstrate superior performance proposed method conventional SPM using both simulated faults real an process.