神经网络与遗传算法结合在结构优化设计中的应用

Application of a Combined Neural Network and Genetic Algorithm to the Optimum Design of Structures

利用神经网络对有限元计算的样本数据建立起结构设计参数（输入）与位移、应力等（输出）的全局性映射关系，以获得遗传算法求解结构优化问题所需的目标函数值，从而进一步获得问题的优化解．5杆桁架算例表明，此方法可以在较少的有限元分析次数下获得较好的优化解．

A non-linear mapping function from multiple input data (design variables) to multiple output data (displacements and stresses calculated by FEM) is constructed within BP neural networks in order to obtain the objective function values that are necessary in optimum design of structures using genetic algorithms and get the optimum solutions. An example of a 5-bar truss is provided to indicate that a better optimum solution can be obtained using less finite element analyses than using simple genetic algorithm.