Research On Health State Recognition Of Piston Pump Under Variable Speed Based On Extreme Learning Machine With Snake Optimizer

As the power heart of the hydraulic system,the axial piston pump has the characteristics of compact structure,high volumetric efficiency and adjustable flow rate.It is often used in various construction machinery,and its health state affects the stability of the whole system.Due to the complexity of working conditions,the vibration signal frequency of the piston pump shell has non-stationary characteristics when the speed changes,which brings problems to signal processing and health diagnosis.The time domain index and frequency domain index extracted by traditional methods are no longer suitable for the analysis of non-stationary signals.Therefore,it has practical significance to diagnose the health status of the piston pump at the variable speed.Firstly,the internal structure of the axial piston pump is analyzed to explain the causes and transmission path of vibration,and the theoretical basis for identifying the health state is provided.By analyzing the common faults of the piston pump,the port plate wear is one of the causes of the most damage to the piston pump,and the port plate wear is selected to simulate the different health state of the piston pump.Secondly,in view of the serious energy divergence of traditional time-frequency analysis methods,the Synchroextracting Transform(SET)method is introduced.It can only retain useful time-frequency information,greatly increase the time-frequency energy concentration,and accurately extract instantaneous frequency.Compared with Synchrosqueezing Transform(SST)and Reassignment method(RS),the superiority of SET method is proved.It provides a theoretical basis for piston pump signal analysis.Thirdly,features are extracted by testing.The piston pump health diagnosis test platform is built,and four port plate with different wear amount are selected.The SET method and angle resampling method reconstruct the angle domain signal and order the signal and eliminate the speed interference.Eight kinds of distinguishing features of Angle domain and order domain are selected as characteristic parameters.Finally,the SO-ELM classification algorithm is proposed,and the above characteristic parameters are taken as input.The snake optimization(SO)algorithm is used to optimize the initial weight and bias of the extreme learning machine,which is compared with the PSO-ELM model and ELM model.The results show that the SO-ELM model has good global optimization ability and the highest classification accuracy.In this paper,health status identification of piston pump under variable speed is completed,and good results are achieved.It enriches the theoretical system of signal processing of hydraulic pump.