Xu, Yixiang (Zhejiang University), Yang, Chunning (Zhejiang University)

Fault recognition of electric servo steering gear based on long and short-term memory neural network

Scheduled for presentation during the Regular Session "Biologically Inspired and Energy Efficient UAS" (FrC2), Friday, September 4, 2020, 14:50−15:10, Kozani

2020 International Conference on Unmanned Aircraft Systems (ICUAS), September 1-4, 2020 (Postponed from June 9-12, 2020), Athens, Greece

This information is tentative and subject to change. Compiled on April 24, 2024

Abstract

In order to recognize different kinds of faults of helicopters' three types of electric servos, this paper proposes a fault state recognition algorithm of electric steering gears based on long short-term memory(LSTM) neural network. Firstly, the original three types of electric servo data are preprocessed, the data set is transformed into a supervised learning problem, and the input variables are normalized. Secondly, different optimization algorithms are introduced to the LSTM neural network model to optimize the parameters. Finally, the number of different hidden layer neural network and the time step are set for comparison experiments, and the optimal neural network configuration is selected from them. The test results show that this model can effectively identify single and multiple faults. The training results show that the model can effectively identify single and multiple faults, and the recognition rate of pitch and left roll is high and has good practicability.