Modeling Drop Size Emitted from Irrigation Impact Sprinklers using Gene Expression Programing and Multiple Linear and Nonlinear Regression Methods

Abstract

The characteristics of drops produced by sprinklers have important role in designing, evaluation of sprinkler irrigation, determining wind drift and evaporation losses and soil compaction. Predicting the size of emitted drops can improve the accuracy of above mentioned issues. In this research Gene Expression Programing (GEP) as one of artificial intelligent methods and multiple linear and nonlinear regression methods (MLR and MNLR) were applied for modeling the size of the drops produced by sprinklers. The input data were included nozzle diameter, operation pressure, and the distance from sprinkler and the outputs was the average size of landed drops in a given distances from sprinkler. The experiments were conducted in 22 combinations of nozzle diameters and operation pressure and in a windless condition. In each experiment 9 to 14 measurement stations were considered with 1.5 meters spacing intervals from sprinkler. Using digital photography method and analyzing the taken photos, hydro dynamical properties of drops in photos were determined. Obtained data were classified in the nozzles diameters, operation pressures, distances from sprinkler in one side and average size of landed drops in each station in the other side. Finally the GEP method and MLR and MNLR methods were applied to develop models for predicting landed drop size in certain distances. Comparisons between models outputs and experimental data were done to evaluate models performances. The results showed that in GEP method, F5 model with R=0.9599 and RMSE=0.4060 mm, and in MNLR method L1 model with R=0.9333 and RMSE=0.5442 mm, have good accuracy to be proposed as proper models for predicting emitted drop size from irrigation impact sprinklers

Keywords

• Drop size, impact sprinkler, modeling, nozzle diameter, operation pressure