基于神经网络建模的种子精选机遗传最优控制

Genetic algorithm optimal control of gravity seed cleaner by radial baisis function neural network modeling

为解决原有的精选机控制系统难以实现不同类型、品种、批次的种子的分级精选问题 ,提出了一种新的控制算法。该算法通过采用 RBF神经网络实现对种子精选过程的离线建模和在线修正 ,然后利用遗传算法实现对种子精选过程模型寻优 ,从而实现最优控制。使用 5 XZW-1.5型重力精选机及微机控制系统 ,将人工手动调节的分级结果与该文的控制策略作对比实验。结果表明 ,提高了种子的总获选率 ,该算法对同类的控制系统也具有指导意义。

For solving the problems that seed cleaner control system cannot classify seeds varying with different types, varieties and batch, a method is presented for controlling nonlinear static systems with an example of gravity seed cleaner. In this controlling scheme, nonlinear static system is modeled by using Radial Basis Function(RBF) neural network, and then genetic algorithm uses the model to optimize the control system, meanwhile, the actual system got the data of input and output which were used to train the RBF neural network repeatedly for better mapping to nonlinear system. The proposed approach is applied to the gravity seed cleaner control system. In comparison with the results of manually adjusted classification, the total seed selection percentage by GA optimal control was raised. This genetic algorithm provides some references for the same kind of control system.