Validation and calibration of parameters sunflower seeds-soil

Abstract

To enhance the mechanical seeding performance of sunflowers and improve seed delivery efficiency, this study measured the adhesion force and angle of repose between sunflower seeds and soil at varying moisture contents through physical experiments. A discrete element model (DEM) was developed to analyze the interaction between soil and sunflower seeds, with the angle of repose as the response variable. The Plackett–Burman (PB) Design was utilized to identify significant influencing factors, and a combination of Response Surface Methodology (RSM) and Feedforward Neural Network (FNN) was employed for optimization. The results indicated that FNN provided higher prediction accuracy and stability. Specifically, at soil moisture contents of 10%, 14%, 18%, and 20%, the static friction coefficients were 0.67, 0.74, 0.66, and 0.63; dynamic friction coefficients were 0.45, 0.46, 0.38, and 0.36; surface energies were 1.18, 2.11, 3.6, and 4.99; and angles of repose were 37.58°, 40.22°, 41.56°, and 41.81°. The absolute errors from physical experiments were 0.59%, 0.6%, 0.82%, and 0.46%, respectively. These findings demonstrate that the FNN model can effectively predict simulation parameters for sunflower seeds and soil under varying moisture conditions, providing a theoretical foundation for field crop seeding processes.