Improved Box-Cox Transformation Based Residual Life Predictions for Bearings

Rolling element bearing is one of the most critical components in rotating machines. Effectively predicting the Remaining Useful Life (RUL) of the bearing can prevent the occurrence of sudden equipment failure. In this paper, the time-domain features of bearing vibration signals are extracted and then these features are fused. In order to establish a more accurate prediction model, the health state of bearing is divided into health stage and degradation stage by 3σ principles. In degradation stage, the Box-Cox transformation method is applied on the data. The transformation parameter λ of the sample number curve is further predicted by Support Vector Regression (SVR). Then, Box-Cox transformation is performed using the predicted parameter λ and a linear degradation model is established to predict the RUL of the bearing. Finally, the effectiveness of the proposed approach is verified with the experimental data on bearings' accelerated life tests provided by FEMTO-ST institute.