关键词: 支持向量描述, 粒子群优化, 轴承, 故障预测, 剩余寿命

Abstract: To solve the failure prognostics problem of industrial equipment,based on Support Vector Data Description (SVDD),a Degradation Index (DI) for representing the operation state of target component was defined,and a method for remaining useful life prediction was proposed.In this method,feature vectors were extracted from historical monitoring data of sensors by applying Wavelet Packet Decomposition (WPD).The particle swarm algorithm was employed to select an appropriate parameter of kernel function,which could make the trend of DI of training set closer to the exponential function.A SVDD model was trained to obtain the hypersphere by using features extracted from the healthy state.The distance between sample points and hypersphere was used to calculate DI,which was adopted to determine the operating state of component and furthermore to estimate its remaining useful life.The effectiveness of the proposed method was verified with the experimental data on bearings' accelerated life tests provided by FEMTO-ST institute.

Key words: support vector data description, particle swarm optimization, bearing, failure prognostics, remaining useful life