Bending-fatigue damage-detection on notched-tooth spiral-bevel gears using the average-log-ratio, ALR, algorithm

Abstract The frequency-domain ALR (average-log-ratio) damage-detection algorithm [MSSP 24 (2010) 2807–2823] is utilized to illustrate damage detection and progression on notched-tooth spiral-bevel gears. Use of equal weighting of increases or decreases of individual rotational-harmonic amplitudes caused by damage, for early ALR detections, is substantiated. Continuously improving statistical reliability of ALR is documented by using increasing numbers of rotational-harmonic amplitude-ratios and increasing numbers of waveforms in the synchronous averaging. Sensitivity of the ALR algorithm to incipient damage is observed to be comparable to that obtained from the kurtosis-based Figure of Merit 4 (FM4). In contrast to FM4, ALR is shown to monotonically increase with increasing damage and running time. Interestingly, this diagnostic technique can be implemented with remarkably low analog-to-digital conversion rates. Computation of ALR for differing torque levels shows strong indications of weakening tooth-stiffness and increasing tooth-plastic-deformation. ALR computation utilizing tooth-rotational-location windowing also is illustrated.