Research on the fault diagnosis method of robot network based on artificial intelligence
Chao Wu · 2025
With the wide application of robots in various fields, robot network fault diagnosis is increasingly critical to ensure the stability of the system. Traditional diagnostic methods have limitations in complex dynamic environment. This paper proposes a hybrid diagnostic framework integrating gating cycle unit (GRU) and trace Kalman filter (UKF). By constructing the robot network data security model, the detection framework based on GRU neural network is designed, and the robot network state information is deeply mined by using the multimodal data fusion technology. In the experimental session, the RoboNet dataset containing multiple fault scenarios was used to evaluate the algorithm performance from multiple indicators such as accuracy, RMSE, F1-score, and running time, and compared with the UKF, GRU, and LSTM-UKF algorithms. The results showed that the GRU-UKF algorithm has 98.3% accuracy, RMSE is 1.61103, F1-score is 0.97, and the running time is only 0.039s, showing excellent performance in comprehensive performance, real-time and noise resistance. This study provides a more efficient and accurate solution for the robot network fault diagnosis.