A Novel Whale Optimization Algorithm Based on Music Theory Knowledge for RUL Prediction of Motor Bearing

There are two contradictions between the occurrence of multiple failure types in motor bearings and traditional remaining useful life (RUL) prediction methods: 1) the contradiction between time series data and prediction accuracy among neurons and 2) the contradiction between the large amount of data calculation and prediction speed. Aiming at the above problems, inspired by the music theory knowledge, the method for RUL prediction of motor bearings based on a novel whale optimization algorithm (WOA) is proposed. First, based on the conventional WOA, nonlinear time-varying adaptive weights and convergence factors are proposed. The method solves the problem of decoupling the optimization of time series data due to the overlapping of multiple fault features and improves the accuracy of RUL prediction. In addition, the optimization process for multiple types of faults is complex and time-consuming, so a feature mapping method based on the principle of chord variation is proposed for the first time. The method renders single-fault features into multiple-fault features by mimicking the chord transformations of stringed instruments playing. Finally, the validity of the method is verified on the PRONOTIA public dataset and self-built bearing aging experimental platform. The method improves the prediction speed by 35% compared with the existing advanced methods under a variety of experimental conditions.