Fault Diagnosis Based on Integrated Multi-Neural Network

摘要

Based on task decomposition principle, multi-fault diagnosis is decomposed to multiple single-fault diagnosis tasks. One Single Sub-NN is utilized to solve one single-diagnosis problem. All the Single Sub-NNs are combined organically. Two integrated multi-neural network schemes are proposed in the paper. A unitized fault location algorithm is given. The scheme is utilized to diagnose the missile inertial system's servo-stable loop fault and the result shows that the scheme is effective.