| Axial piston pump is widely used in hydraulic system because of its compact structure,high volumetric efficiency,convenient variable adjustment,and suitable for high pressure and ultra-high pressure system.Its reliability plays a key role in the stable operation of hydraulic system.However,due to the complex structure of the axial piston pump,there are mechanical vibration and fluid vibration in the work,resulting in the axial piston pump fault diagnosis and identification accuracy is not high.Based on this,this paper uses the method of Variational Mode Decomposition(VMD)and multi axis information fusion to extract the fault features of axial piston pump,and combines with Convolution Neural Network(CNN)to identify the fault accurately.The research contents are as follows(1)In this paper,the A4VSO-125 axial piston pump is selected for research.According to the structure and working principle of the axial piston pump,the vibration mechanism and fault causes of the axial piston pump are analyzed.Aiming at the common faults of the axial piston pump,such as slipper swash plate wear,loose shoe fault,valve plate wear,etc.,the force between the friction pairs in the axial piston pump is analyzed,The vibration transmission mode is found,and the vibration characteristics and fault vibration characteristic frequencies of different faults are obtained.(2)For the selection of decomposition level K and penalty factor of VMD,the existing Particle Swarm Optimization Variational Mode Decomposition(PSO-VMD)has a large amount of computation and slow speed in the optimization process.Therefore,this paper proposes k-value optimization based on Shannon entropy and multi parameter collaborative optimization based on Bayesian optimization,and selects the best parameter combination[K,α],Compared with pso-vmd,it is proved that Bayesian Optimization Variational Mode Decomposition(BO-VMD)has great advantages in parameter optimization efficiency.(3)Aiming at the disadvantage that traditional feature extraction can not extract more comprehensive features of the signal,this paper introduces RGB color mode to fuse the piston pump signal.The red,green and blue in RGB color space correspond to the vibration information of X-Y-Z axis of the piston pump respectively,and the vibration information of the piston pump is converted into a color picture,It does not need to calculate the characteristic parameters of the signal,which can avoid the loss of a large number of fault information and characterize the fault characteristics more comprehensively.The convolution neural network is established to learn and identify the fault feature map of axial piston pump.Compared with the traditional fault diagnosis model,the model has higher accuracy for fault diagnosis of axial piston pump.Taking the common faults of axial piston pump as the research object,this paper proposes a signal processing method based on VMD,and introduces RGB color mode into multi axis information fusion.Experiments show that the diagnosis accuracy of this model for common faults of axial piston pump is higher than that of traditional fault diagnosis model under different working conditions,It is proved that it has good feasibility and generalization. |
