Progressive Hypergraph Structure Learning for Fault Diagnosis of Industrial Robots

Abstract

Accurate fault diagnosis in industrial robots used in automotive assembly lines, electronics manufacturing, and precision machining operations is critical for maintaining operational safety and productivity. However, the complexity of industrial robots has led to more diverse and unpredictable faulty modes, which poses significant challenges for accurate fault diagnosis. The existing deep learning algorithms often struggle with the capture of high-level features and the resistance to noise, resulting in inaccurate fault diagnosis. To address these limitations, this study proposes a new Hypergraph Neural Network (HGNN) for fault diagnosis. The proposed algorithm integrates three key modules: Feature Convolution, Progressive Hypergraph Structure Learning (PHSL), and HGNN classifier. By integrating PHSL and feature convolution, the proposed algorithm dynamically refines hypergraph representations to capture high-order correlations in sensor measurements while removing noise interference. Finally, the HGNN module performs robust fault diagnosis using the optimized hypergraph structure. Experimental results on a six-axis industrial robot fault diagnosis dataset demonstrate that the proposed algorithm achieves higher accuracy and robustness compared to state-of-the-art benchmark algorithms. The proposed algorithm bridges the gap between advanced graph learning and industrial instrumentation, which can offer a scalable solution for the condition monitoring of industrial robots. The source code of this study is available at: https://github.com/Wade914/PHSL.