Physics-Inspired Feature Engineering for Condition Monitoring of Alternating Current-Powered Solenoid-Operated Valves

The interest to develop a reliable condition monitoring (CM) and prognostics and health management (PHM) program for solenoid-operated valves (SOVs), being one of the critical components of machines/equipment in the industry, has increased since the last decades. The literature comprises methods including physics-based, data-driven modeling or hybrid, which rely on the analysis of current and voltage signals measured on SOVs. The majority of the literature covers CM and PHM methods for direct-current (DC)-powered SOVs. On the contrary, the literature on CM and PHM methods for alternating-current (AC)-powered SOVs is still relatively limited despite a recent continuing growth of interest. This paper focuses on the development of a CM program that can serve as a foundation of a PHM program for AC-powered SOVs using the measurements of AC current and voltage signals. The basis for the development lies in a physical understanding of the effect of major failure modes on the current signatures using an improved physical model that is recently developed in our previous works. The gained physical insights help identify relevant current signatures and define a signal processing strategy to extract condition indicators (features) from these signatures. To demonstrate the effectiveness of our method, we performed an accelerated life test (ALT) campaign on 48 identical AC-powered SOVs. Besides acquiring the current and voltage signals of each valve during the test campaign, the temperature and flow of the SOV outlet were also measured as a means to generate the “ground truth data” for labeling the end of life in an objective way. The acquired data have been analyzed offline, and some physics-inspired features show consistent trends for the majority of the tested SOVs indicating that the features are effective to track the health status of AC-powered SOVs.