Statistically Interpretable Health Indicator Based on Dynamic Interactions Decoupling for Gear Wear Assessment

With high transmission efficiency and compact structures, gears are indubitably critical components of rotating machineries. Generally served in non-stationary and complex environment, gear surfaces are inevitable to wear. This surface performance degradation can potentially lead to very serious gear failures and also threaten the operating safety. Therefore, it is essential to make a quantitative assessment for gear surface wear progression and accordingly provide a foundation for the subsequently implementable preventative maintenance. To achieve this industrial target, a statistically interpretable health indicator (HI) based on dynamic interactions decoupling is proposed in this paper for gear wear progression assessment. Specifically, the interactions between the gear wear and the dynamic responses are firstly decoupled by the discrete/random separation technique. Afterwards, characteristic parameters of the extracted sliding vibrations are appropriately estimated by an optimized multi-parameter regression method, based on which the corresponding signal probability distributions to characterize the statistics of the gear wear can be established. Then a novel HI, by quantifying the statistic deviations at different operating stages, is accordingly developed for gear wear assessment. Experimental tests indicate that the proposed HI outperforms conventional indicators in gear surface wear assessment under different lubrication conditions.