Evaluating a variable rate sprayer performance by artificial neural networks

Subject Areas :

Received: 2017-01-23 Accepted : 2017-01-23 Published : 2016-12-20

Keywords: Artificial Neural Networks, precision agriculture, Statistical models, Variable rate spraying,

Abstract :

To evaluate the application of a variable rate sprayer, artificial neural network (ANN) was used. Data were collected from field experiments. To model the output flow of nozzles, 727 nets were tested by 4 neural network models include linear, MLP, RBF and GRNN. For each nozzle 45, 22 and 23 data were used for train, verification and test, respectively. Among tested models, RBF model with one input, 4 hidden and one output layers was selected as the best model. To investigate the capability of ANN for prediction of sprayer flow, this model was compared by statistical model. According to the results, average value of R2 for statistical model was 0.980, 0.979, 0.981 and 0.980 and for ANN was 0.994, 0.988, 0.997 and 0.990, respectively. Also, averages CV for statistical and neural network models were 18.96% and 19.05%, respectively. Totally, results indicated that ANN model is more accurate than statistical model for prediction of sprayer flow in variable rate spraying