Pattern analysis in real time with smart power sensor

Abstract

The current state of the art in electronic prognostic health management systems does not fully support detection, collection, and remediation of real-time faults. As a result, knowledge has not been captured from an actual platform failure mechanism. Thus, point-of-failure feedback cannot be applied by system designers or operators to improve lifecycle weak links in replacement platforms, or to strengthen effectiveness of mission-critical platforms. Our innovation makes it possible to extract and analyze the power system's eigenvalues, which are related to the intrinsic frequencies of the power system that determine correlations between extracted features and state of health (SoH). In-situ electronic prognostics for power systems are crucial for attaining a sound theoretical basis of health status. To provide correlation information such as state of health (SOH) using pattern analysis with real-time data from a non-intrusive smart power sensor, Ridgetop researched using data-driven modeling with a proposed health distance and support vector machines (SVMs) with signatures in a standard IEEE 1451-enabled smart power sensor. Results of this study indicate that a fault pattern analysis methodology overcomes certain disadvantages of the standard failure modes and effects analysis (FMEA) approach, which does not account for the contribution of unobserved failure to a degradation trajectory. The efficacy of the proposed pattern analysis approach is illustrated with test results showing critical distinction in pattern analysis and test data acquired from a real-time IEEE 1451-enabled smart power sensor testbed, and monitored via a testbed with appropriate instrumentation.