试验设计法与BP-GA算法在复合材料结构优化中的应用

试验设计法具有均衡分散性和整齐可比性,通过试验设计进行响应面的样本选取可以大大减少样本的数目,同时保证响应面的训练精度。响应面与遗传算法相结合在工程应用中已经形成一套适用于复杂结构设计的高效优化方法。本文以复合材料加筋蒙皮的质量优化问题为例,通过正交试验进行响应面训练样本选取,通过PATRAN/NASTRAN进行有限元分析得出样本的数值,应用神经网络(BP)训练、构建优化的目标函数和强度、稳定性约束响应面,结合其他常规约束条件通过遗传算法(GA)进行优化,验证了这套算法的有效性和实用性。

The Application of Experimental Design Method with BP-GA in the Optimization of Composite Structures

The sample selection based Experimental Design Method can greatly reduce the number of training samples for structural response surfaces,while keeping the training accuracies at satisfactory levels.When such response surfaces are combined with Genetic Algorithm(GA),it can form an efficient framework for optimization of complex composite structures.Taking a composite stiffened panel as example,the orthotropic experiment method was used to select structural samples,then finite element analysis of these samples were performed using PATRAN/NASTRAN.The BP neural networks were constructed and trained for strength and stability constraints as well as objective function of weight.The genetic algorithm based optimization gives the convincing results and shows the effectiveness of such approach.