基于递阶遗传神经网络的某扫雷犁电液伺服系统建模研究  

Modeling of Mine Sweeping Plough Using Hierarchical Genetic Neural Networks

作　　者：张媛[1] 邢宗义[1] 高强[1] 秦勇[2] 贾利民[2] 

机构地区：[1]南京理工大学,南京210094 [2]北京交通大学轨道交通控制与安全国家重点实验室,北京100044

出　　处：《中国机械工程》2010年第1期89-93,共5页China Mechanical Engineering

基　　金：国家自然科学基金资助项目(60674001);北京交通大学轨道交通控制与安全国家重点实验室开放基金资助项目(SKL2008K010);南京理工大学科技发展基金资助项目(XKF09003)

摘　　要：提出了一种基于递阶遗传算法的径向基神经网络建模方法,实现了具有复杂非线性特征的某扫雷犁电液伺服系统的精确建模。在递阶遗传-神经网络算法中,采用二进制编码的控制基因实现了径向基神经网络的结构辨识,即隐节点数目的确定,采用实数编码的参数基因实现了径向基神经网络的参数辨识,即隐节点中心参数的确定,采用测试数据的赤池信息量准则计算染色体的适应度函数值,减少了隐节点数目,提高了神经网络的泛化能力。实验结果及与其他建模方法的比较结果验证了所提方法的有效性。A method for designing a radial basis function neural network was proposed based on the hierarchical genetic algorithm, and the technique was employed to settle the modelling problem of the electrohydraulic system of a certain mine sweeping plough with complex nonlinear characteristics. In the proposed hierarchical genetic algorithm, the control genes with binary code were used to identify the structure of neural networks, i.e. the number of hidden units, and the parameter genes with real code were employed to identify the parameters of neural networks, i.e. the center parameters of hidden units. In order to havea better generalization performance and simpler network structure, the Akaike＇s information criterion was adopted to evaluate the fitness of individual chromosomes. The experimental results and comparisons with other modeling methods indicate that the proposed approach can be more effectively to construct precise models for the mine sweeping plough.

关 键 词：扫雷犁 递阶遗传算法 神经网络 建模 

分 类 号：TJ812.5[兵器科学与技术—武器系统与运用工程] TP183[自动化与计算机技术—控制理论与控制工程]