Customized Research on Fault Diagnosis of Civil Aircraft Onboard Maintenance System Based on Big Data Technology

Abstract

Abstract This paper dissects the BP neural network and RNN on the basis of existing research, obtains the improved long and short-term memory network, introduces the attention mechanism into it, and designs the fault diagnosis model of airborne maintenance system based on AM-LITM with big data technology. Focusing on the use of the BP learning algorithm to obtain the reference error of each unit needed to adjust the weight of each connection, real-time updating of information through the long- and short-term memory network to solve the problem of gradient dispersion of RNN, and at the same time, the method designed in this paper is compared with the traditional method from the aspects of fault diagnosis and prediction. The results show that the average prediction error of the traditional fault diagnosis algorithm is around 0.06, and it only has good diagnosis performance for the A-type fault, except for the classification accuracy rate of 0.9845. The rest of the performance index is 1. The diagnosis accuracy of the AM-LSTM fault diagnosis algorithm for three types of faults can reach 1, and all the indexes for the A, B, and C faults are 1, which shows that it greatly improves the performance of the traditional fault diagnosis algorithm. Diagnostic accuracy of civil aircraft on-board maintenance system faults is verified to verify the effectiveness and superiority of the designed model.