Research And Implementation Of An AC Contactor Remaining Useful Life Predication Method

With the development of industry,remaining useful life(RUL)prediction has become an important technology to ensure the safe operation of equipment.AC contactor is a control device used in power system and industry.Its structural properties and working state make it difficult to implement traditional maintenance.A remaining useful life prediction method for AC contactors is proposed to solve the above problems.The work and contributions are summarized as follows:(1)A feature selection method for AC contactor monitoring data is proposed to improve the predictive performance and interpretability of the model,which including data preprocessing,feature construction,feature analysis and feature selection.(2)The health index construction method is proposed.Firstly,the process events are detected based on the adaptive perceptually important points segmentation algorithm and the unsupervised clustering strategy.Furthermore,the multi-dimensional health index is constructed.In addition,an evaluation method is proposed,and health states are divided according to the health index.(3)A model Combining Transformer and Bi-LSTM network is established for predicting the remaining useful life of AC contactors,which effectively solves the long-term dependence and association of the degradation state in the remaining life prediction.The Transformer is used to extract the historical information in the health index.In addition,a relative distance encoding method and a relative position encoding method are proposed to realize the sequential inheritance of historical degradation information.Based on the AC contactor dataset,relevant experiments are carried out to compare the performance of the existing remaining useful life prediction method and the method proposed in this thesis,and discuss the effect of those methods on the prediction results.The experimental results show that the proposed method achieves the expected goal.