Quantitative evaluation of electric features for health monitoring and assessment of AC-powered solenoid operated valves

Quantitative assessment of feature performance for health monitoring is key to feature selection. This paper illustrates the application of well-established metrics in the research community - namely, monotonicity, robustness and prognosability - to the quantitative performance assessment of features for health monitoring of alternating-current (AC) powered solenoid operated valves (SOVs). These features are extracted from voltage and current signals measured on the valves and builds on previous work of the authors. Based on these metrics, the appropriate features are selected to be used as condition indicators. The selected features are inputs to a logistic regression model to predict a health index ranging from 0 to 1, which can be easily monitored and assessed by non-experts. We demonstrated the developed methodology on the experimental data acquired from accelerated life tests on 48 identical AC-powered SOVs.