A study on the calibration method of bonding parameters for soybean threshing materials based on GA-BP model and whale optimization algorithm

Abstract

The simulation of grain and stem breakage mechanisms during the soybean threshing process can effectively reveal the interaction mechanisms between threshing materials and operating components; however, its accuracy is highly dependent on the precision of the bonding parameters. To address the lack of accurate bonding parameters for soybean threshing materials, this study proposes a calibration method that integrates a GA-BP neural network with the whale optimization algorithm. First, shear tests of stems and compression tests of grains were conducted to obtain the ultimate load, which served as the reference index for parameter calibration. Subsequently, a predictive model for soybean threshing materials was constructed, and parameter optimization was achieved in combination with the whale optimization algorithm. Finally, the effectiveness of the calibrated parameters was verified by comparing simulation results with experimental data. The results demonstrated that the relative errors between the simulated and measured failure loads of stems and grains under the calibrated parameters were 0.38 % and 1.41 %, respectively, confirming the accuracy and reliability of the proposed method. This study not only provides precise bonding parameters for soybean threshing materials but also offers important references for discrete element method-based threshing mechanism studies and parameter calibration of other materials.

Science4Impact statement

This study addresses the lack of accurate bonding parameters for soybean threshing materials by proposing a calibration method that integrates a genetic algorithm-optimized BP neural network with the Whale Optimization Algorithm for bonding parameter optimization. Compared with RF, SVR, and RBF models, the BP neural network demonstrated superior prediction accuracy and generalization ability. Additionally, the Whale Optimization Algorithm outperformed the genetic algorithm in terms of convergence speed and optimization efficiency. Validation tests of the optimal bonding parameters showed that the mechanical response of the calibrated soybean threshing materials exhibited minimal error compared to the actual load, indicating high accuracy and reliability.