A digital twin methodology for vibration-based monitoring and prediction of gear wear

The process of gear wear is a complex phenomenon, and multiple effects will lead to non-uniform wear propagation rates in the tooth profile alone. As a result, current model-based methods cannot produce features that account for non-uniform wear propagation rates, leading to an inaccurate representation of the actual tooth profile. This limitation negatively impacts the accuracy of predicting gear wear progression. This paper proposes a digital twin methodology for monitoring and predicting gear wear to overcome this challenge. More specifically, in the proposed digital twin methodology, the geographical distribution of the wear coefficient is introduced with the utilization of a series of scale coefficients and a shape function, and then feed the generated wear coefficient into the well-known Archard wear model to indicate the non-uniform wear propagation rate alone the tooth profile. Even though the actual tooth profile can be structured with the introduced geographical distribution, the wear coefficient varies throughout the gear wear process. To achieve accurate predictions of gear wear, it is essential to update the wear coefficient timely as the wear progresses. Thus, a “grey box” approach is used to estimate the wear coefficient, updating the scale coefficients using measurements obtained from the physical systems. The gear wear process can be effectively monitored by continuously updating the gear wear model. This approach allows for accurately predicting the remaining useful life at any given time. The paper illustrates the ability and effectiveness of the proposed digital twin methodology in wear progradation prediction using measurements from a laboratory gear rig.