Fault Diagnosis Based on Graph Convolutional Network for Industrial Robot Harmonic Reducers

With the increasing integration of industrial robots in automated production processes, the need to enhance health monitoring and ensure the reliability and precision of these robots through effective fault diagnosis has become paramount. Harmonic reducers, crucial transmission components in industrial robots, are susceptible to faults that can lead to performance degradation and downtime, underscoring the importance of accurate fault diagnosis. In this research, we introduce an innovative approach based on graph convolutional networks (GCNs) to address this challenge. Our method involves constructing a graph structure using vibration signal data from harmonic reducers and subsequently applying GCNs for end-to-end fault diagnosis. To validate the effectiveness of our approach, we conducted experiments using practical data and compared it against alternative graph construction methods. Our findings not only confirm the feasibility and efficacy of our approach but also open up avenues for future research in this domain.