ProbSparse Attention-based Fault Diagnosis for Industrial Robots Under Different Working Conditions

Industrial robots are a complex and critical equipment in intelligent manufacturing systems that require ensuring safe and stable operation. The in-situ fault diagnosis of industrial robots faces serious industrial environment interference, and usually operates in different working conditions. In this paper, we propose a ProbSparse Attention-based Transformer (PSAT) for in-situ fault diagnosis for industrial robots under different working conditions. The sparsity measurement in ProbSparse attention module is specially designed to reduce industrial noise and improve the key general features extraction. In addition, we explore the length of input signals and proposed a lightweight diagnostic framework based on shorter time series. It can not only reduce the complexity of the model with better real-time performance, but also reduce the storage cost of diagnostic samples, which is very valuable and particularly suitable for industrial robots. The superiority of the proposed method with 99.5% accuracy is validated on real industrial signals industrial robots, which is higher than advanced models.