Abstract

Automated manufacturing process monitoring using an intelligent system is powerfully ongoing to improve the productivity and quality of the product. However, most of machines are working in normal condition, data imbalance problem is emerging issue for training the AI system. To solve this problem, a frequency-focused sound data generator was developed for the in-situ fault sound diagnosis of industrial robot harmonic reducer. The sound data generator, based on a conditional generative adversarial network, selects a target frequency range without relying on domain knowledge. A sound dataset of normal and faulty harmonic drive rotations of in-situ industrial robots was collected using an attachable wireless sound sensor. The generated sound data were evaluated based on the fault diagnosis accuracy of a simple classifier trained using the generated data and tested using real data. The proposed method well defined the frequency feature clusters and produced high-quality data, exhibiting up to 16.0 % higher precision score on normal and 13.0 % higher accuracy on weak-fault harmonic drive compared to conventional over-sampling methods, achieving fault diagnosis accuracy of 95.6 % even in situation of fault data comprising only 5% of the normal data.