High-accuracy diagnostics from HUMS in noisy environments

摘要

Condition Based Maintenance (CBM) of military helicopters are tracked by Condition Indicators (CI) calculated from Health Usage and Monitoring Systems (HUMS) vibration sensors. Many CIs have been developed and implemented, yet success has been partial at best, owing to their sensitivity to noises and artifacts that invariably corrupt measurements under real-life operations. Here we report a sequential Monte Carlo algorithm operating a stochastic non-linear model that includes a description of fault evolution. This algorithm estimates fault magnitudes and probabilities, which were compared to component removals validated by tear down analyses. We obtained a high accuracy rate (～95%) over all available data. Data encompassed a 6-year operational history of an entire US military helicopter fleet. These results demonstrate the excellent artifact rejection enabled by this approach, which handles probabilities rigorously to detect fault processes from noise-limited signals. Consequently, our decision support tool could detect faults early and accurately. This technology could drive a significant reduction in maintenance costs by making faults evident and reducing NEOFs (false positives).