Data-driven machinery fault diagnosis: A comprehensive review

2024-05-29

Neurocomputing

AI Summary

Existing data-driven mechanical fault diagnosis (MFD) research lacks a systematic analysis of industrial deployment challenges—particularly regarding noise robustness, feature transferability, and adaptability to unknown faults. To address this gap, this paper proposes the first comprehensive, multi-dimensional taxonomy covering signal processing, feature learning, model generalization, and industrial deployment, enabling unified evaluation of deep learning, graph neural networks, and physics-informed methods across their applicability boundaries. The survey systematically reviews key techniques—including time-frequency analysis, 1D-CNNs, LSTMs, Transformers, self-supervised pretraining, and digital twin–assisted diagnosis—and synthesizes optimal method combinations for twelve representative fault scenarios. It identifies three critical bottlenecks hindering real-world adoption: few-shot learning, cross-operating-condition transfer, and model interpretability.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning
• Multiagent Systems: Other Foundations of Multi Agent Systems

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Experiences and lessons learnt from Web-based algorithms and system deployments
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Problem

• Comprehensive review of machine learning approaches
• Challenges in data-driven machinery fault diagnosis
• Future research prospects in fault detection

Innovation

• Machine learning for fault detection
• Deep learning in industrial diagnostics
• Comprehensive review of fault datasets