Estimation of Optimum Vacuum Pressure of Air-Suction Seed-Metering Device of Precision Seeders Using Artificial Neural Network Models

Abstract

The success of the seed-metering device of a seeder determines the quality seeding and final plant stand. The adjustment of the optimal vacuum pressure of air-suction-type seed-metering devices is a key factor affecting the success of seed-metering devices. The optimal value of vacuum of the seed-metering device should be adjusted in relation to the physical properties of the seed before seeding in the field. This research was carried out to estimate the optimal value of vacuum pressure of an air-suction seed-metering device of a precision seeder by using an artificial neural network method. Training of the network was performed by using a Levenberg–Marquardt (LM) learning algorithm. Training and testing were carried out using Matlab software. The inputs were physical properties of seeds such as surface area, thousand kernel weight, kernel density and sphericity. Optimum vacuum pressures were determined for soybean, maize, cucumber, melon, watermelon, sugarbeet and onion seeds in laboratory. Surface area, thousand kernel weight, kernel density and sphericity of seeds varied from 0.05 to 0.638 cm2, 4.4 to 322.4 g, 0.43 to 1.29 g cm−3 and 42.8 to 85.75%, respectively. The optimal vacuum pressure was determined as 1.5 kPa for onion; 2.0 kPa for sugarbeet; 2.5 kPa for melon and watermelon; 3.0 kPa for soybean; and 4.0 kPa for maize seeds. A trained program using an artificial neural network could satisfactorily estimate the optimum value of vacuum pressure of the air-suction type seed-metering device of precision seeders with a prediction success (R2) of 0.9949 for both linear and polynomial regressions.