The invention discloses a nuclear emergency quadrotor unmanned aerial vehicle multi-
actuator fault detection method based on a Conv1D-LSTM
hybrid model. The method comprises the steps of collecting
flight data of a quadrotor unmanned aerial vehicle; preprocessing the collected
flight data of the quad-rotor unmanned aerial vehicle, and dividing the preprocessed data into a
training set, a
verification set and a
test set according to a random mode; constructing a
hybrid model based on a one-dimensional
convolutional neural network Conv1D and a long short-
term memory (LSTM) network; training the
hybrid model by using the
training set, then evaluating the trained
hybrid model by using the
verification set to select the optimal hyper-parameter configuration, and performing comprehensive performance evaluation on the independent
test set; and performing real-time monitoring on the flight state of the quad-rotor unmanned aerial vehicle in the nuclear emergency environment by using the finally obtained
hybrid model, and quantitatively predicting the possible fault degree of each
actuator of the quad-rotor unmanned aerial vehicle according to newly collected
flight data of the quad-rotor unmanned aerial vehicle in the real-time monitoring process. According to the method, the detection reliability of the multi-
actuator fault of the quad-rotor unmanned aerial vehicle in the nuclear emergency task is remarkably improved.