Using Deep Learning Network for Fault Detection in UAV

Unmanned aerial vehicles (UAV), especially quadrotors, have received much attention in recent years. The most important challenge in a quadrotor system is fault detection. therefore, in this article, we used the Bidirectional Long short-term memory algorithm (BLSTM), deep learning subset, to detect actuator fault in quadrotor. in other words, we were able to detect the actuator fault of quadrotor using BLSTM algorithm, used in the field of deep learning. In this research, we first simulated the quadrotor model, using PID controller and then extracted the data and through this we were able to train the network and eventually detect the actuator fault by collecting data from various state variables and inputs at different frequencies. The simulation results show this method can be effective in detecting actuator fault in quadrotor and its accuracy reaches 99%.