Fault Diagnosis Of Axial Piston Pump Based On Multi-source Sensor Feature Fusion And UMAP Dimension Reduction

The hydraulic system has a wide range of applications in production and life.As the energy device of the hydraulic system,the hydraulic pump provides pressure and flow for it.This paper proposes a fault diagnosis method for axial piston pumps based on multi-source sensor feature fusion and UMAP dimensionality reduction,and builds an experiment bench to simulate common fault types of axial piston pumps and collect operating signals of multi-source sensors to realize fault detection of axial piston pumps.Firstly,analyze the common fault mechanism of axial piston pumps,reasonably carry out fault simulation experiments of axial piston pumps to approximate the real faults in actual production,and design experimental schemes to collect operating signals of multi-source sensors with different fault types,including X,Y,Z axis vibration sensor signal,pressure sensor signal,sound sensor signal.Secondly,considering the limitations of artificial selection of IMF components,an EMD-based comprehensive selection method for IMF components is proposed to realize automatic selection and reconstruction of IMF components,so as to achieve the purpose of noise reduction of operating signals.Extract multi-dimensional features of multi-source sensors,including time domain,frequency domain,and parameter-optimized MPE features,and make feature data sets corresponding to different sensors,including X,Y,and Z-axis vibration feature data sets,pressure feature data sets,and sound feature dataset.Thirdly,considering the inaccuracy of single diagnostic information,five sensor feature fusion methods is proposed to obtain a 120-dimensional feature fusion matrix.Aiming at the problem that too many feature dimensions affect the running time and accuracy of fault diagnosis,compared with the common dimensionality reduction algorithms,the UMAP algorithm is selected to reduce the dimensionality of the multi-source sensor fusion features of the axial piston pump.Finally,using the axial piston pump fault diagnosis method based on multi-source sensor feature fusion and UMAP dimension reduction,the single and compound faults of the axial piston pump are identified,and the accuracy rate is 100%.From four aspects of multi-source sensor feature fusion,feature dimensionality reduction,different feature dimensionality reduction algorithms,and training sets and test sets with different proportion settings,the impact on axial piston pump fault diagnosis is comprehensively compared and analyzed,which verifies the accuracy,generalization and robustness of the proposed method.