基于神经网络与遗传算法的结构优化设计方法

从神经网络和遗传算法的原理出发,利用遗传算法和神经网络相结合的策略对结构参数进行优化.在确定结构优化的目标函数和设计变量集合的基础上,用神经网络学习算法建立货架结构设计参数与结构重量、结构最大应力、最大位移等的非线性全局映射关系,获得遗传算法求解结构优化问题所需的目标函数,用遗传算法进行优胜劣汰的寻优搜索运算,从而求出所需最优解.以货架结构的优化为例说明了上述方法的应用.遗传算法和神经网络的优化结果是在正交设计法确定的训练样本足够大的基础上得出的,具有较强的可靠性.

Optimum Design Method of Structure Based on Neural Network and Genetic Algorithm

With the principles of neural network and genetic algorithm,design parameters of pallet structural are optimized by strategy of combining genetic algorithm and neural network.The structural optimum object function and design variables set were firstly defined,then non-linear global mapping relationship between design parameters with structural weight,max-stress,max-displacement and so on was built with learning algorithm of neural algorithm.Then object function of structural optimum needed by genetic algorithm could be given.Through searching calculating of dominant character of genetic algorithm, the optimal solution could be found.To demonstrate the application of above method,an example of optimum of pallet structure is given.The optimal result,derived by neural network and genetic algorithm on basis of sufficient training samples determined by orthogonal design method,is very reliable.