Deep Learning-Based Fault Diagnosis in Quadcopter Actuators: A CNN-LSTM Network With Bayesian Optimization

In recent years, deep neural networks have shown promising results in modern fault diagnosis. This paper focuses on diagnosing actuator faults in quadcopters using a deep learning strategy. Considering the dynamic and temporal characteristics of quadcopters, a deep neural network model consists of the convolution neural network and the long short-term memory network (CNN-LSTM) is designed with the exponential linear unit (ELU) activation function. To enhance the diagnostic capability, Bayesian optimization (BO) algorithm is utilized for selecting optimal hyperparameters of the designed deep neural network model. Experimental results demonstrate that the proposed method can achieve high accuracy in actuator fault diagnosis of quadcopters.