Physic-based vs data-based digital twins for bush bearing wear diagnostic

Abstract

Industry is increasingly turning to predictive maintenance by using digital twins (DTs) to follow and predict evolution of mechanical system. This article presents, compares and discusses two DTs to diagnose wear of bush bearings under dynamic loads. The first DT is driven by a model based on data analysis using statistical process control (SPC). The second DT is based on physical laws: Boussinesq's and Archard's equations. Both DTs are fed by data recorded on a test bench instrumented with sensors of temperature, acceleration and displacement. Rules for fault detection were identified, explained and applied to the two DTs implemented. The two implemented DTs detected abnormal wear behaviours. The data-based DT using the SPC is easier to implement and it detects change in wear behaviour earlier. In contrast, the physic-based DT has the advantage of being predictive, so it can be used when only the operating conditions of the system are known. This work is a contribution for new wear diagnostic tools.