Rolling element bearings (REBs) perform a fundamental role in rotating machinery, with accurate and robust knowledge of their health state, as well as prognostics for remaining useful life (RUL) being critical to optimising maintenance strategies. A key challenge to such health indication lies in reducing defect detection latency (DDL): the time between the onset of an incipient defect and its identification. Existing trend-based signal processing approaches employ various forms of temporal and spectral analysis to identify characteristic defect frequencies of the REBs but are often limited by their reliance on monotonic trend evolution, retrospectively recognising defect onset once the trend has developed sufficiently. This study introduces a novel event-based approach; the UNFIT (contracted from Unstable Negentropy Fluctuations with Indication Trace) methodology for early defect monitoring and detection, which extends prior research on informational entropic change as a means of detecting impulsive signals and proposes an alternative to traditional trend-based techniques. The proposed method applies a spectral-temporal assessment that both emphasises and characterises impulsive events shown to ultimately lead to failure, and does so at the point of defect emergence rather than retrospectively, thereby theoretically minimising the DDL. The methodology is validated using the industry-recognised IMS dataset published by NASA, providing fault detection results across all three (3) accelerated run-to-failure tests and benchmarking them against fault detections reported in existing signal-processing literature for trend-based methodologies. Supplementary datasets from XJTU-SY and Ferrara are presented for further comparison, with further investigation required to confirm robustness under noisy operating conditions. Results from a subset of test cases demonstrate that the UNFIT methodology can reduce DDL and enhance defect identification, particularly in cases where conventional trend-based indicators are less effective, thereby improving maintenance optimisation and supporting Prognostics and Health Management (PHM) workflows by providing earlier, actionable diagnostic information, amenable to integration within Digital twin and Industry 4.0 predictive maintenance frameworks.
Watson et al. (2026) studied this question.