A hybrid CNN-LSTM model based actuator fault diagnosis for six-rotor UAVs

Abstract

With the development and popularity of multi-rotor UAVs, actuator fault diagnosis in multi-rotor UAVs has become more and more important. This paper proposes a deep-learning-based method to accurately locate actuator faults by using flight data of a real UAV. The proposed method splits the UAV's data into smaller pieces and then extracts features by one-dimensional convolutional neural network (1D-CNN), and explores internal connections of the UAV's time series data by adding the long short-term memory (LSTM). So, a hybrid CNN-LSTM model is developed for the fault diagnosis of actuator faults. Experiments show that the average accuracy of fault diagnosis of the hybrid CNN-LSTM model is 92.74%, which is better than that of other models, such as the CNN model, the LSTM model, and the deep neural network (DNN) model.