An algorithm based on digital twin sets of solutions for rolling-element bearing prognostics

The present work proposes a physics-based approach based on lumped-parameter Digital Twins (DTs) solving the dynamics of rolling element bearings with defects. A novel indicator for rolling-element bearing prognostics is defined on the basis of related DT solutions. The indicator, namely the Equivalent Damaged Volume, is extrapolated from sets of pre-computed solutions referring to the NVH behavior of the bearing under consideration, affected by a wide range of defect dimensions, on the basis of experimental observations. Then, experimental RMS values are compared with DT estimations to deduce the volume of the defect producing the same RMS level. A decision tree is defined to extrapolate the value of the defect volume when multiple computational defects lead to the same experimental RMS value. A first experimental study based on bearings with artificial defects is carried out to tune the DT by using measured RMS values. Secondly, run-to-failure tests are used to assess the prognostic method.