Research On Early Warning Of Turbulence Based On Doppler Lidar

The randomness of atmospheric turbulence itself is difficult to capture by detection equipment,which seriously affects flight safety.Its violent energy exchange and wind shear can damage the structural performance of the aircraft and cause violent turbulence of the aircraft.More severe turbulence can even lead to loss of control of the aircraft,causing flight accidents such as air crashes.Therefore,there are urgent problems in the civil aviation need to be solved that safeguard the flight safety and achieve the high effective early warning of turbulence.In order to provide fine wind field data for civil aviation airport turbulence warning,by collecting relevant data measured by Lanzhou Zhongchuan Airport Lidar,a wind speed prediction model based on particle swarm(PSO)modified Elman neural network(PSO-Elman)is established,which by collecting relevant data measured by Lanzhou Zhongchuan Airport Lidar.Make the input data of the model with the echo distance of the echo signal and the signal-to-noise ratio spectrum width,and the output data of the model with the radial velocity.After the input data is normalized,the internal weights and thresholds of the Elman network perform optimization by PSO algorithm,which improve the prediction accuracy and convergence speed of the network.The non-linear function mapping is determined by the convergent network model,and the wind speed between the distance gates is predicted.The model simulation results show that the correlation coefficient between the wind speed predicted by the algorithm and the actual wind speed measured by the radar reaches 0.919,and the relative error is 6%,as well as the coefficient of determination is 0.8425.All the results prove the wind speed prediction model is valid and feasible,which achieve the purpose to refine the wind field.In order to achieve high-efficiency early warning of turbulence and realize the purpose of unmanned early warning,the pixel data set is constructed by further processing the refined wind field data into eddy current dissipation rate.A convolutional neural network classification model composed of two convolutional layers,two fully connected layers,a softmax layer,and several activation functions is proposed.After the network converges,its loss is as low as 3%.Comparative experiments show that the accuracy of the network reaches 85%.Using the 2016 Zhongchuan Airport crew report for comparative analysis,the results show that the established neural network model has an early warning hit rate of 80% for atmospheric turbulence,of which the false alarm rate is 13.3%,and the false alarm rate is 6.7%.These results prove that it has strong generalization ability in turbulence early warning,which improves the early warning efficiency to a certain extent.It can provide a judgment basis for relevant weather forecasters and provide some reference value for the research of turbulence early warning.