Domain adversarial transfer fault diagnosis method of an axial piston pump based on a multi-scale attention mechanism
Zhikui Dong, Huijiang An, Siyuan Liu, Shihao Ma, Yuxuan Guo, Hongxin Pan, Chao Ai
Abstract
Effective failure diagnosis of axial piston pumps is crucial to guaranteeing the security and steady run of hydraulic equipment systems. However, the current domain-adaptive methods have insufficient feature extraction and domain-adaptive capability for axial piston pumps across conditions and some unlabeled sample conditions. Therefore, this paper proposes a method of domain adversarial transfer fault diagnosis based on multi-scale attention mechanisms. The method enhances the ability of the model to adapt to the differences in various fault information by constructing a multi-scale residual feature extraction network. The domain adaptive capability of the network is improved by constructing a joint loss function to decrease the distributional discrepancies between the source and the target domain data. The experimental consequences demonstrate that the method can differentiate various fault types of axial piston pumps across conditions and can achieve a high fault identification rate without labeled samples in the target domain.