The use of data signatures in Condition Based Maintenance Plus

This paper examines a CBM+ capability for medium power diesel generators by showing the results of an analysis of data that was collected from sensors that were placed on various components on a 30 kW generator; the paper focuses on fuel flow sensors that are sampled at a high rate. Condition Based Maintenance Plus (CBM+) contributes to optimal supply chain management of parts on a system platform by providing a prognosis of their remaining useful life (RUL). A data stream from sensors that are placed on a system to measure failure characteristics of its critical components is analyzed to develop a prognosis of system health. To that end, this paper demonstrates that detailed measurement at high sampling rates of physical effects of a component produces a predictive and reductive data signature of the performance of the component. Signatures evolve in time and a dynamical metric is proposed that evolves to produce a prognostic assessment of the condition of a component. Thus, prognostic assessments are done not only from time-series data but also the time-varying spectral analysis of the data. This means that a prognostic result is reached more quickly by using dynamic data signatures than it is from a time-series trend analysis.