UAV Actuator Fault Detection using Maximal Information Coefficient and 1-D Convolutional Neural Network

Abstract

Actuator is a critical part of the unmanned aerial vehicle (UAV), for which accurate and speedy fault detection is of great significance in practical application. Data-driven method becomes more appealing due to its feasibility and high performance. However, the current fault detection method based on machine learning cannot realize feature selection and real-time detection, and its feature extraction and learning ability of time series is not high enough. To solve the above problems, we propose a new fault detection method based on maximal information coefficient and one dimensional convolutional neural network (MIC-1DCNN) approach. It combines the high feature extraction ability of one dimensional convolutional neural network (1DCNN) for time series and the good feature selection ability of maximal information coefficient (MIC) for nonlinear data, which complete UAV actuator fault detection well and improve its efficiency greatly. The benchmark flight data set of the UAV is adopted for conducting experimental verification. The experimental results indicate that the proposed method can achieve satisfied performance in UAV actuator fault detection regarding speed and accuracy indices.