Taguchi–ANN-based performance optimization of a retrofit electronic seed metering unit

Abstract

Uniform intra-row seed spacing is essential for maintaining an optimum plant population in green gram, where the recommended agronomic spacing is about 10 cm. Conventional ground-driven seed drills often produce irregular spacing because seed release depends on wheel motion and slip. To improve placement accuracy, a microcontroller-based electronic seed metering mechanism was developed as a retrofit attachment and evaluated using a laboratory sticky-belt test rig. Experiments were conducted at three forward speeds (1.5, 2.0 and 2.5 km h⁻¹) and three groove diameters corresponding to 120% (D1 = 4.45 mm), 110% (D2 = 4.08 mm) and 100% (D3 = 3.71 mm) of the geometric mean seed size. Seed distribution performance was assessed using quality of feed index (QFI), miss index and multiple index. Second-order quadratic regression and Taguchi design were applied for parameter optimization, and an artificial neural network (ANN) model was developed for performance prediction. Spacing increased with forward speed for all groove sizes due to reduced seed pickup duration. The D2 groove produced spacing closest to the agronomic requirement, giving 9.85–10.1 cm at 1.5 km h⁻¹. Under this condition, QFI exceeded 88.5%, while miss and multiple indices remained below 5.2% and 6.3%, respectively. The ANN model showed strong prediction capability ( R  > 0.98). The developed retrofit electronic metering unit achieved reliable seed placement suitable for precision sowing of green gram.