Application of Convolutional Neural Networks in Fault Diagnosis of Axial Piston Pump
As an important component of the hydraulic system,the performance of the piston pump directly affects the operation of the hydraulic system,so the fault diagnosis of the piston pump has always been a hot spot in the fault diagnosis of construction machin-ery.Aiming at the problems of traditional fault diagnosis methods that need to design and extract the signal features manually,and the signal feature extraction is not perfect,it is proposed to use Convolutional Neural Network to diagnose axial piston pump faults.The vibration signals of the piston pump are collected under five working conditions:normal state,loose shoe,valve plate wear,shoe wear and central spring failure.The vibration signal is converted into frequency spectrum and time-frequency diagram,and labeled.The sample data is generated and input into Convolutional Neural Network.Deep Belief Networks and Stacked Auto Encoder to com-pare the performance of different networks.The results show that the Convolutional Neural Network has a higher accuracy rate of 92.17%in fault diagnosis of axial piston pump than the Deep Belief Networks and Stacked Auto Encoder when the sample data select the wavelet transform time-frequency diagram.
The Piston PumpFault DiagnosisDeep LearningConvolutional Neural Network
NETL
NSTL
万方数据
