基于RBF网络的步行式底盘内力计算优化方法

AN OPTIMIZED METHOD FOR SOLVING INTERNAL FORCE OF WALKING MOBILE CHASSIS BASED ON RBF NEURAL NETWORK

为了确定步行式底盘局部结构在作业时的最大受力状态,提出了一种基于RBF神经网络的两级优化模型求解方法,第一级优化模型用逐步二次规划法找到局部结构在给定位置参数下的最大受力状态,通过正交试验设计,利用RBF网络构造出局部结构界面最大受力状态与位置参数之间的非线性映射关系;第二级优化模型用GA求解RBF网络的最大值,并通过二分法不断缩小位置参数的搜索空间,提高RBF网络的逼近水平。研究表明,计算结果可为步行式底盘设计提供理论依据,该方法是解决复杂结构系统中非线性、多变量优化问题的有效手段。

For the purpose of determining the maximum forces acting on local structural interface of the walking mobile chassis in the structural design, a method using the bi-level optimization model based on the neural network of radial basis function （RBF） is proposed. In first level, the maximum forces acting on local structure of the chassis is, based on a set of given position parameters, optimized by the sequential quadratic programming （SQP） method. The RBF neural network, as a nonlinear mapping relationship between maximum force state in the local structure and position parameters, is trained by using orthogonal design. In second level, the genetic algorithm （GA） is used to optimize the RBF neural network. A high precision RBF neural network is obtained through an iteration procedure, in which the searching volume is reduced using the bisection method. The results show that the calculated results can provide theoretical data for the walking mobile excavator design, and the optimum method is an effective way for nonlinear and multi-variable optimization of the complex mechanism.