Multi-output Classification using a Cross-talk Architecture for Compound Fault Diagnosis of Motors in Partially Labeled Condition

2025-05-29

AI Summary

To address the domain adaptation challenge in motor compound fault diagnosis under label-scarce target domains and varying operating conditions (e.g., speed and torque fluctuations), this paper proposes a multi-output classification framework that jointly predicts both fault types and severity levels. Methodologically, we design a cross-task interactive “cross-talk” layer to enable selective information sharing among subtasks; introduce frequency-domain layer normalization to enhance domain-invariant representation learning from vibration signals across operating conditions; and integrate partial-label learning with a shared backbone network to improve few-shot generalization. Evaluated on six domain adaptation scenarios, our approach achieves significantly higher macro-F1 scores than baselines—especially for compound fault identification. Ablation studies confirm that performance gains stem primarily from the task interaction mechanism, not increased parameter count.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning
• Intelligent Robots: Learning & Optimization for ROB

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Search and Retrieval-Augmented AI: Vertical and domain-specific search

Abstract

The increasing complexity of rotating machinery and the diversity of operating conditions, such as rotating speed and varying torques, have amplified the challenges in fault diagnosis in scenarios requiring domain adaptation, particularly involving compound faults. This study addresses these challenges by introducing a novel multi-output classification (MOC) framework tailored for domain adaptation in partially labeled (PL) target datasets. Unlike conventional multi-class classification (MCC) approaches, the proposed MOC framework classifies the severity levels of compound faults simultaneously. Furthermore, we explore various single-task and multi-task architectures applicable to the MOC formulation-including shared trunk and cross-talk-based designs-for compound fault diagnosis under PL conditions. Based on this investigation, we propose a novel cross-talk layer structure that enables selective information sharing across diagnostic tasks, effectively enhancing classification performance in compound fault scenarios. In addition, frequency-layer normalization was incorporated to improve domain adaptation performance on motor vibration data. Compound fault conditions were implemented using a motor-based test setup, and the proposed model was evaluated across six domain adaptation scenarios. The experimental results demonstrate its superior macro F1 performance compared to baseline models. We further showed that the proposed mode's structural advantage is more pronounced in compound fault settings through a single-fault comparison. We also found that frequency-layer normalization fits the fault diagnosis task better than conventional methods. Lastly, we discuss that this improvement primarily stems from the model's structural ability to leverage inter-fault classification task interactions, rather than from a simple increase in model parameters.

Problem

• Addressing compound fault diagnosis in motors under partially labeled conditions
• Proposing a multi-output classification framework for domain adaptation
• Enhancing fault classification via cross-talk architecture and frequency-layer normalization

Innovation

• Multi-output classification for compound faults
• Cross-talk layer for selective information sharing
• Frequency-layer normalization for domain adaptation