Research On Fault Detection Algrithm Of Uav Flight Attitude

The task of the stability of the Unmanned Aerial Vehicle(UAV)flight attitude affects the UAV safety and work efficiency,UAV flight mission in the implementation of the process,when the attitude is wrong during the flight,the measured data of sensor will deviate from normal.Therefore,the fault detection of sensor signals is an important indicator of the failure judgment of UAV flight attitude.This paper takes the flying attitude of the UAV as the subject of research,and uses wavelet analysis,principal component analysis(PCA)and back propagation(BP)neural netwo,rk methods to achieve fault detection.The following aspects are the main research contents of this paper:(1)Aiming at the noise of the acquisition data is large,the error rate is high,and the wavelet analysis method is used to signal and fault detection of the measurement signal.By setting three states of normal,mutation,and slow change,the use of wavelet threshold noise reduction reduces the impact of noise on signal,and the combined signal-to-noise ratio is judged to determine the noise reduction effect,the high frequency component of the denoising signal is then recognized by the maximum principle,and the fault detection of the flying attitude of the UAV is completed.(2)Aiming at the problem that only single variable fault detection may cause detection false alarms,a fault detection method combining wavelet denoising and principal component analysis is proposed,and the main information of the data is left while reducing the number of variables.Wavelet denoising is used to improve the detection accuracy.The fault detection algorithm of principal component analysis is used to detect the sensor data in real time.The change of statistics is used to judge whether the operation is normal and detect the time of failure.The contribution of variables to statistics is used to identify the fault variables.(3)Aiming at the limitations of traditional fault detection methods to detect nonlinear systems,a fault detection method combining wavelet denoising and BP neural network is proposed.Because there is a complex non-linear relationship between the flight attitude of UAV and the measurement data of UAV sensors during the flight,the detection accuracy is improved by combining wavelet denoising.The three-layer BP network is trained with the experimental data as samples,and the results are analyzed by comparing the output values with the actual values.The fault detection of UAV flight attitude is realized according to the calculation results of the algorithm.(4)Because the fault detection method combined with wavelet denoising and BP neural network has the disadvantages of long time consuming and easy to fall into the local optimum.On the basis of the principle of fault detection neural network,and presents fault detection algorithm based on genetic algorithm optimization BP neural network.In this paper,genetic algorithm is used to optimize the weights and thresholds of neural networks,and learning training for the same sample,the test results show that the fault detection accuracy rate has reached 95.6%.Two fault detection methods are compared,as can be seen,and the number of training iterations achieves convergence standards in 820 and 473 respectively.After optimization of genetic algorithm,the accuracy of algorithm fault detection is improved,and the time consumption is reduced.