Early diagnostic of concurrent gear degradation processes progressing under time-varying loads

This study develops a gear diagnostic procedure for the detection of multi- and concurrent degradation processes evolving under time-varying loads. Instead of a conventional comparison between a descriptor and an alarm level, this procedure bases its detection strategy on a descriptor evolution tracking; a lasting descriptor increase denotes the presence of ongoing degradation mechanisms. The procedure works from time domain residual signals prepared in the frequency domain, and accepts any gear conditions as reference signature. To extract the load fluctuation repercussions, the procedure integrates a scaling factor. The investigation first examines a simplification assuming a linear connection between the load and the dynamic response amplitudes. However, while generally valuable, the precision losses associated with large load variations may mask the contribution of tiny flaws. To better reflect the real non-linear relation, the paper reformulates the scaling factor; a power law with an exponent value of 0.85 produces noticeable improvements of the load effect extraction. To reduce the consequences of remaining oscillations, the procedure also includes a filtering phase. During the validation program, a synthetic wear progression assuming a commensurate relation between the wear depth and friction assured controlled evolutions of the surface degradation influence, whereas the fillet crack growth remained entirely determined by the operation conditions. Globally, the tested conditions attest that the final strategy provides accurate monitoring of coexisting isolated damages and general surface deterioration, and that its tracking-detection capacities are unaffected by severe time variations of external loads. The procedure promptly detects the presence of evolving abnormal phenomena. The tests show that the descriptor curve shapes virtually describe the constant wear progression superimposed on the crack length evolution. At the tooth fracture, the mean values of the residual signal evince strong perturbations, while after this episode, the monitoring curves continue signaling the ongoing wear process.