Remaining Useful Life Prediction Of Hydraulic Pump Based On Deep Learning DSAE And LSTM

Since the birth of hydraulic system because of its small size,light weight,large load force,accurate and stable action and other advantages in the industry,wind power,construction machinery and aerospace and other fields have been widely used as the power source of the hydraulic system,the running state of the hydraulic pump directly affects the normal work of the whole mechanical equipment,so it is of great significance to predict the remaining useful life(RUL)of the hydraulic pump.Thanks to the rapid development of sensor and data storage technology,it has entered the era of big data,and the acquisition of massive monitoring data has become more and more easy.Many data-driven rul prediction methods emerge as the times require.However,traditional prediction methods have been unable to meet the needs of simple and efficient equipment health management due to over reliance on domain experts’ experience and complex signal processing technology.Therefore,it is becoming more and more important to seek to build a simple,convenient and general mechanical equipment performance degradation modeling and residual life prediction method.Based on this problem,combined with deep learning technology,this paper studies the performance degradation modeling and residual life prediction method of gear pump.In this paper,a new method based on deep sparse autoencoders(DSAE)and support vector data description(SVDD)is proposed in order to prove the effectiveness of this method,an experimental verification and comparative analysis are carried out on the bearing open data set phm2012,and then the method is applied to the gear pump life data set.Finally,combined with the advantages of long short-term memory(LSTM)network in processing time series,a multi-layer bidirectional LSTM based residual service life prediction method for gear pump is proposed.The results show that the performance degradation modeling method proposed in this paper has better monotonicity and time correlation than the performance degradation curve constructed by traditional methods.Based on the performance degradation curve of the hydraulic pump,the multi-layer bidirectional LSTM prediction model can effectively predict the remaining service life of the gear pump in the later life.And the whole prediction process is basically in an unsupervised state,which greatly reduces the manual participation and has good versatility.