Fault Model Establishment and Hybrid Neural Network Diagnosis for Manipulator

Due to solve the problem of imbalanced and insufficient fault diagnosis data for manipulator, an effective joint fault model is established and the fault diagnosis analyses are proposed with an optimized hybrid neural network in the paper. the terminal acceleration data is gained by the Co-simulation analysis in Ansys and Adams software. The two-population algorithm is improved to optimize the hybrid network model with three shallow networks. In addition, after preprocessing the acceleration data, the hybrid network is used for training, learning and diagnosing fault. Finally, the network achieves a high fault-diagnosis accuracy with an average of 94.8%. The result indicates that not only the simulation model is established correctly, but also the fault features of acceleration data for the optimized network model are effective.