Fault diagnosis of axial piston pump based on multi-source subdomain adaptation and sensor data fusion

Abstract

Abstract The axial piston pump is a pivotal component within hydraulic systems, and its condition monitoring and fault diagnosis are crucial for ensuring the safe and reliable operation of the hydraulic system. However, existing fault diagnosis methods for axial piston pumps mostly train and test using data from the same operating conditions. In actual operation, axial piston pumps often experience varying loads, and the collected data is typically correlated but has different distributions. As a response to this, a transfer learning method of multi-source subdomain adaptation and sensor fusion (MSASF) is proposed for the fault diagnosis of axial piston pumps. MSASF integrates three modules: a shared feature extraction module, a domain-specific feature extraction module,and an output decision module. Firstly, heterogeneous data from multiple sources are simultaneously inputted into the shared domain feature extraction module, composed of the parallel adaptive weighted multi-scale convolutional neural network (PAWMSCNN), to extract data features and achieve multi-sensor feature fusion. Then, in the domain-specific feature extraction module, multi-branch networks are introduced to extract features between each source and target domain, utilizing the Local Maximum Mean Discrepancy to align each feature's subdomain distribution of the source domain and target domain. Finally, the distribution distance between each source domain and the target domain is measured using Maximum Mean Discrepancy to assign corresponding weights to each source domain. These weights are then combined with the outputs of classifiers from each source domain to make the final diagnosis decision for the equipment. A dataset was constructed for the axial piston pump design fault experiment, and four sets of transfer tasks were designed to compare with seven classic methods. The experimental results indicate that the proposed MSASF method exhibits a superior domain adaptation effect and fault diagnosis performance.