لیلا پیمان

Considering safety and environmental issues are very important in all domains of agriculture, industry and services in different countries. In the agricultural domain despite of numerous efforts to find alternative methods, millions of liters of toxic chemicals are used by chemical methods to control plant pests every year. Certainly, the most important issue in spraying is the size of drops which is influenced by several factors including pressure, nozzle hole diameter, viscosity of the chemical solution and wind speed in the area. In this study, MLP network Feed Forward modeling was used. Input consisted of two layers including nozzle diameter (three sizes) and spraying pressure (three pressure levels). Output of the artificial neural network determined by volume median diameter. In order to choose the best procedure, five methods including gradient descending, descending gradient with momentum, Levenberg-Marquart, conjugate gradient and Delta Bar Delta were used. Considering both minimum mean square error and coefficient of determination, the descending gradient with momentum was chosen. After training and validation of the network, MSE and coefficient of determination were 0.0176 and 0.90, respectively. In order to verify the results from neural network several tests were carried out and observed particle diameters were compared with values obtained from neural networks by chi-square test. The difference was not significant. These results indicate that neural networks can estimate properly the size of the droplets.