改进的遗传神经网络算法对花生种子带式清选设备关键作业参数的优化

Optimization of Key Working Parameters of Belt Separator for Peanut Seeds by Improved Genetic Neural Network

针对花生种子带式清选设备关键作业参数最优组合问题,本文提出了一种改进的遗传神经网络优化方法,利用该方法自适应动态调整神经网络权值阈值的特点,彻底取代神经网络算法中“误差逆传播”过程,可使训练误差达到2.2425×10^-8,拟合精度明显优于二次回归方法。将训练好的算法用于花生种子带式清选设备作业参数优化,当纵向倾角A=23.66°,横向倾角B=24.26°,帆布带带速C=0.73m/s时,可使合格率达到98.36%,带出率1.85%,设备性能最佳。该方法为花生种子带式清选设备关键作业参数的设定提供了新思路,为农业机械设备智能化控制提供了途径。

For the optimal combination of key working parameters of belt separator for peanut seeds, we proposeds an improved genetic neural network optimization method, which can completely replace the Error Back-Propagation process of BP neural network algorithm by using the characteristics of adaptive dynamic adjusting the value of neural network weight and threshold. The training error value can reach 2.2425×10 -8 and the fitting accuracy is obviously better than the quadratic regression. If the trained algorithm was used to optimize the working parameters of belt separator for peanut seeds, the qualified rate reached 98.36%, the take-out rate reached 1.85%, and the equipment performance was the best, when longitudinal angle A =23.66°, heeling angle B =24.26° and velocity of canvas belt C =0.73 m/s. Our research provides a new idea way for the setting of key working parameters of belt separator for peanut seeds, and a way for intelligent control of agricultural machinery equipment.