Multiple Fault Diagnosis For Axial Piston Pump Based On Extreme Learning Machine

With the rapid development of science and technology and the substantial increase in productivity,hydraulic systems play an irreplaceable role in engineering applications.The axial piston pump has the advantages of long life,high energy density,compact structure and small leakage,so it often used as a power component of the hydraulic system.However,its working environment is rather harsh,and it is in a working state of high pressure and high speed for a long time,so various failures of the hydraulic pump will inevitably occur.Therefore,the status monitoring and diagnosis of the piston pump is an important guarantee for the normal operation of the system.A large number of engineering practices have shown that the positions of faults do not only exist in one place for axial piston pumps,often manifesting multiple failures at the same time.The difference of locations,forms,and degrees of failure will lead to various effects on the pump body,and these fault signals usually is covered by the strong noise signal of background.Hence,it is difficult to detect and bring challenges for fault diagnosis.In this paper,the separation of fault features have been studyed deeply coupled with the algorithm intelligent classification in view of the above problems.We adopt the rational method of noise reduction to extract the fault characteristics from the vibration signal,flow signal and pressure signal of the axial piston pump aiming to this research object under the environment of strong noise interference.The research will be conducted as the below:(1)Study and analyze the structure and working principle of the axial piston pump.Meanwhile,understand the common faults of the axial piston pump which includes shoe wear,valve plate wear,center spring failure and loose shoes.For the above faults,it is found out for the mechanism of each fault and the influence on the external characteristics of the axial piston pump,such as body vibration,system pressure,and outlet flow.According to the existing test platform,a reasonable solution of fault testing has been designed and the acquisition of above signals completed.(2)The signals obtained by the acceleration sensor are usually non-linear,non-stationary signals,and more importantly interfered by various background noises.The fault information of axial piston pump is weak under strong noise environment,and the fault feature extraction is difficult problem.Therefore,this paper proposes a fault diagnosis method of axial piston pump based on the maximum correlated kurtosis deconvolution(MCKD),complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)and extreme learning machine(ELM).Firstly,use the MCKD to filter and denoise the vibration signal to enhance the weak impact component.Then decompose the denoising signal into several intrinsic mode function components by CEEMDAN,and extract the fault feature vector.Finally,diagnose the working state and fault type of the piston pump by using ELM classification algorithm.(3)Aiming at the separation problem of the compound fault of the axial piston pump,a composite fault diagnosis method based on multi-label classification for axial piston pump is proposed.Firstly,the vibration signal is denoised by wavelet.Then use the CEEMDAN decomposition to extract the entropy spectrum entropy of each component.The time domain characteristic index is extracted from the pressure signal and the flow signal directly.After finally combining the above characteristic indicators,the idea of multi-label classification is used to realize the separation of the compound failure of the axial piston pump.