Silence Is Not Safety: Minute-Level Forensics of Spur Gear Spall Progression from Low-Cost Condition Indicators

Abstract

The sensitivity loss of classical condition indicators during damage progression has been on record since the early NASA gear test programs: FM4 peaks early in pitting tests and returns toward its nominal value as damage spreads (Zakrajsek et al., 1993; Decker et al., 1994), and Dempsey (2000) drew the practical conclusion that vibration indicators alone are an unreliable basis for damage assessment, arguing for fusion with oil-debris measurement. Lewicki et al. (2009) rated FM4 and RMS below average for detecting surface pitting on a helicopter-scale test program. That the indicator families are not interchangeable is also established: Bechhoefer and Butterworth (2019), ranking eighty-eight analyses by separability on a cracked tooth, retained the energy operator despite its weak showing there on the reasoning that no single indicator serves every gear fault, and closed by cautioning that a pitting fault would call for a different set. What the present work adds is an axis: their comparison runs across failure modes and across acquisitions, whereas the separation reported here is temporal -- within a single progression of a single mode, resolving inside one six-hour run.