Design Of Variable Rate Spray Control System And Spray Characteristics Evaluation For Agricultural Unmanned Aerial Vehicle

The variable spraying technology and equipment of agricultural unmanned aerial vehicle(UAV)is an important way to achieve zero pesticide growth and improve the pesticide application effeciency,acting as the frontier direction and core content of agricultural aviation plant protection and intelligent agricultural machinery equipment.A UAV variable spraying control system was built based on the variable spraying technology of agricultural UAV in this study.The spray characteristics of the nozzle in the system were quantitatively evaluated and models were established by using machine learning methods.The operation effect of the system was verified by the UAV simulation platform test and the single wing UAV test.The main results are as follows:(1)The variable spraying control system of agricultural UAV was designed and built.In the system,Arduino UNO R3 was used as the system controller.TEEJET XR110015 VS nozzle was selected as the system nozzle.The quantitative relationship between PWM(Pulse Width Modulation)signal duty cycle,pressure and flow rate was analyzed and calibrated.The change rule of droplet size at different horizontal positions of the system at 0.2 MPa and 3 nozzle heights was determined.(2)Quantitative evaluation methods and models for spray characteristics of a single nozzle variable spraying control system under different system pressure,spray sheet position and nozzle height were established.TEEJET XR110015 VS single nozzle variable spraying system was designed and built.Machine learning methods(multiple nonlinear regression(REGRESS),least squares support vector machines(LS-SVM),extreme learning machine,and radial basis function neural network(RBFNN))were applied to establish multi-parameter prediction models.The results showed that RBFNN achieved the best performance for the quantitative evaluation of droplet size and deposition distribution of different spray sheet positions(nozzle heights and horizontal positions)at 0.2 MPa,the coefficients of determination in the calibration set and prediction set both above 0.9.The results showed LS-SVM was a good candidate for deposition evaluation at system pressure of 0.2 MPa and nozzle height of 1 m with the coefficient of determination in the calibration set and prediction set both above 0.99.The quantitative evaluation models of droplet size under the single nozzle at different pressure and nozzle height were obtained by using machine learning methods,RBFNN achieved the best performance with coefficients of determination in the calibration set and prediction set both above 0.99.(3)Quantitative evaluation methods and models for spray characteristics of different nozzle spacing,spray sheet positions and spray coverage positions at 0.2 MPa were established,the droplet size prediction of overlapping area of twin nozzles was realized.Machine learning methods including REGRESS,LS-SVM,ELM and RBFNN were used to establish the quantitative evaluation models of droplet size and deposition distribution of different nozzle spacing and spray sheet positions(nozzle heights and horizontal positions)under multi-nozzle condition.The best modeling effect was achieved by REGRESS model and RBFNN model respectively,REGRESS model with the coefficient of determination in the calibration set and prediction set both above 0.92,RBFNN model with the coefficient of determination in the calibration set of 0.8806 and the coefficient of determination in the prediction set of 0.7994.The droplet sizes obtained by REGRESS model of TEEJET XR110015 VS nozzle were used to establish machine learning prediction model of the droplet size in the overlapping area of twin nozzles,RBFNN model achieved the best performance with the coefficient of determination in the calibration set of 0.9567 and the coefficient of determination in the prediction set of 0.8616.The quantitative evaluation models of deposition for the whole system(different coverage positions)were established when the nozzle height was 1 m,LS-SVM model achieved the best performance with the coefficients of determination in the calibration set and prediction set both above 0.9.(4)The UAV simulation platform was developed and the single wing UAV field flight test was realized.The indoor UAV simulation platform was developed and the test was operated at 0.2MPa and three different nozzle heights(0.8 m,1 m,1.2m)and the system achieved a good spraying performance and met the requirements of variable spraying experiment of UAV.The filed flight test of single wing UAV based on the prescription map was carried out.The variable spraying effect data of different fields were analyzed according to the prescription map and the results showed that the system could achieve precision spraying.