Performance evaluation of a variable rate sprayer by artificial neural network models

Abstract

To evaluate the performance of a variable rate boom sprayer, an artificial neural network (ANN) was employed. To model output flow of nozzles, 727 nets by four neural net models, namely, linear, MLP, RBF and GRNN were tested. For each nozzle, 45, 22 and 23 experimental data were used for train, verification and test, respectively. The results indicated that RBF model was selected as the best by regression ratio of 0.198 and R² of 0.98. To investigate the capability of RBF model in prediction of nozzles flow, statistical analysis was used. Based on the results, average value of R² for statistical and RBF models were 0.98 and 0.99, respectively. So, the average value of CV for RBF and statistical models were 18.96% and 19.05%, respectively. From the results, it is concluded that ANN model could be a good predictor to evaluate the performance of a variable rate application system.