Abstract:Many transfer learning methods have been proposed to implement fault transfer diagnosis, and their loss functions are usually composed of task-related losses, distribution distance losses, and correlation regularization losses. The intrinsic parameters and trade-off parameters between losses, however, need to be tuned according to the specific diagnosis tasks; thus, the generalization abilities of these methods in multiple tasks are limited. Besides, the alignment goal of most domain adaptation (DA) mechanisms dynamically changes during the training process, which will result in loss oscillation, slow convergence and poor robustness. To overcome the above-mentioned issues, a novel and simple transfer learning diagnosis method named adaptive intermediate class-wise distribution alignment (AICDA) model is proposed, and it is established via the proposed AICDA mechanism, dynamic intermediate alignment (DIA) adaptive layer and AdaSoftmax loss. The AICDA mechanism develops an adaptive intermediate distribution as the alignment goal of multiple source domains and target domains, and it can simultaneously align the global and class-wise distributions of these domains. The DIA layer is designed to adaptively achieve domain confusion without the distribution distance loss and the correlation regularization loss. Meanwhile, to ensure the classification performance of the AICDA mechanism, AdaSoftmax loss is proposed for boosting the separability of Softmax loss. Finally, in order to evaluate the effectiveness and universality of the AICDA diagnosis model to the most degree, various multisource mixed fault transfer diagnosis tasks of wind turbine planetary gearboxes, including DA and domain generalization (DG), are implemented, and the experimental results indicate that our proposed AICDA model has a higher diagnosis accuracy and a stronger generalization ability than other state-of-the-art transfer learning methods.

DCDAN-Based Incipient Fault Diagnosis for Satellite ACS Under Variable Operating Conditions

Automated Design of Fault Diagnosis CNN Network for Satellite Attitude Control Systems

Neural network-based fault diagnosis scheme for satellite attitude control system

Data-based incipient fault detection and prediction for satellite's attitude control system

Fault Diagnosis on Satellite Attitude Control with Dynamic Neural Network

A Random Forest and Model-Based Hybrid Method of Fault Diagnosis for Satellite Attitude Control Systems

Deep Learning Based Fault Diagnosis Methods for Satellite Power System

A Hybrid Adversarial Domain Adaptation Network for Bearing Fault Diagnosis Under Varying Working Conditions

A Novel Joint Adversarial Domain Adaptation Method for Rotary Machine Fault Diagnosis under Different Working Conditions

Dual Classifier-Discriminator Adversarial Networks for Open Set Fault Diagnosis of Train Bearings

Fault Diagnosis Methods of Deep Convolutional Dynamic Adversarial Networks

Dual Weighted-Class Adversarial Network for Rotary Machine Fault Diagnosis Using Multisource Domain with Class-Inconsistent Data

An auto-regulated universal domain adaptation network for uncertain diagnostic scenarios of rotating machinery

Bearing fault diagnosis based on partial domain adaptation adversarial network

Adaptive Intermediate Class-Wise Distribution Alignment: A Universal Domain Adaptation and Generalization Method for Machine Fault Diagnosis

A New Deep Convolutional Domain Adaptation Network for Bearing Fault Diagnosis under Different Working Conditions

Multiweight Adversarial Open-Set Domain Adaptation Network for Machinery Fault Diagnosis With Unknown Faults

A Deep Convolution Multi-Adversarial adaptation network with Correlation Alignment for fault diagnosis of rotating machinery under different working conditions

Dual Structural Consistent Partial Domain Adaptation Network for Intelligent Machinery Fault Diagnosis

Cross-Machine Transfer Fault Diagnosis by Ensemble Weighting Subdomain Adaptation Network