A Novel Domain Adaptation-Based Intelligent Fault Diagnosis Model to Handle Sample Class Imbalanced Problem

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Machine Learning

Background:

• Rotating machinery fault diagnosis is critical for operational reliability.
• Sample class imbalance in industrial data degrades the performance of existing domain adaptation (DA) methods.
• Cross-domain distribution discrepancies hinder accurate fault diagnosis.

Purpose of the Study:

• To propose a novel DA approach (MRMI) that addresses sample class imbalance in rotating machinery fault diagnosis.
• To simultaneously reduce cross-domain distribution and geometric differences.
• To enhance the reliability and accuracy of fault diagnosis systems.

Main Methods:

• Developed a novel distance metric method (MVD) for marginal distribution adaptation.
• Integrated manifold regularization with instance reweighting to leverage manifold structure and adaptively remove irrelevant source samples.
• Applied ℓ2-norm regularization as a data preprocessing step to improve model generalization.

Main Results:

• MRMI demonstrated significant performance improvements on gear and rolling bearing datasets with imbalanced samples.
• The proposed approach effectively mitigates the negative impact of sample class imbalance on DA.
• MRMI outperformed existing competitive approaches in fault diagnosis under imbalanced conditions.

Conclusions:

• MRMI offers a robust solution for rotating machinery fault diagnosis in the presence of sample class imbalance.
• The combined approach of distribution and geometric difference reduction is effective for DA.
• This work advances the field of intelligent fault diagnosis for industrial equipment.