In order to ensure system availability and reduce maintenance cost, a bearing RUL (remaining useful life) prediction model based on performance degradation index is proposed to predict the RUL of bearings. The original vibration data is decomposed into several multiplicative functions (PFs) by local mean decomposition (LMD), and the effective PFs are selected to reconstruct the original signal according to the kurtosis criterion; Time domain degradation features of the reconstructed original signal are extracted, and high-quality features are selected by using an attention mechanism model based on artificial neural network (ANN) training; The K _ mean clustering algorithm and piecewise fitting were introduced to obtain the health degradation indicator (HI), and the grey regression model (GM) was used to evaluate the reliability range of bearing degradation. And the HI-PSO LS-SVM model (particle swarm optimization least squares support vector machine, PSO _ LSSVM) is established to predict the bearing RUL. The experimental results show that the method has achieved good results in the prediction of reliability.