Mechanism and parameter optimization method for ordered vibration conveying of maize ears

Ordered conveying of maize ears is critical for non-destructive handling of seed samples in plant breeding. However, the traditional response surface methodology (RSM) struggles to fit the strongly nonlinear characteristics of this process, and the underlying mechanism of ordered conveying remains unclear, with a weak validation system. In this study, taking maize ears as a case, we constructed a “mechanism simulation–intelligent modeling–global optimization” framework to reveal the mechanism by which vibration parameters regulate the ordered conveying process via influencing the convergence speed of ear posture and queue uniformity. Specifically, frequency dominates posture convergence and thereby determines feeding efficiency. Amplitude and direction angle couple to affect the smoothness of posture adjustment and thus determine conveying stability. This study provides a robust and transferable methodology for optimising the ordered conveying process, which holds significant potential for improving the quality of seed sample handling and reducing mechanical damage in plant breeding programmes. A discrete element model (DEM) of vibration conveying of maize ears was established. Using a hybrid sampling strategy combining Box-Behnken and Latin hypercube sampling, 337 sets of high-fidelity data were obtained. Validation under multiple working conditions showed that the average relative error between simulations and experiments was less than 5%. Three surrogate models, namely response surface methodology (RSM), extreme learning machine (ELM), and grey wolf optimizer–extreme learning machine (GWO-ELM), were constructed. Five-fold cross-validation indicated that the GWO-ELM achieved the best prediction accuracy ( R 2 > 0.98). Through global optimization, the optimal parameters were determined as vibration frequency 16 Hz, amplitude 6 mm, and vibration direction angle 30°, yielding a predicted feeding rate of 0.9981 ears/s and a coefficient of variation (CV) of 43.59%. Three-level bench validation demonstrated that under the optimal parameters, the measured feeding rate was 1.0056 ears/s and the CV was 43.92%. Compared with the optimization results of RSM, the feeding rate increased by 41.6% while the CV decreased by 29.6%, indicating improved conveying stability. The results reveal that frequency dominates ordering efficiency while the coupling of amplitude and direction angle regulates ordering quality. The constructed intelligent optimization framework effectively addresses the accurate modeling challenge of strongly nonlinear systems. The GWO-ELM model exhibits significantly higher prediction accuracy than the traditional RSM. This method provides a technical reference for ordered conveying of maize ears and other irregularly shaped seeds in plant breeding, and holds potential for improving seed sample handling quality and reducing mechanical damage.