论文部分内容阅读
根据Kolmogorov多层神经网络映射存在定理,利用进化神经网络来实现结构设计参数(输入)与结构响应参数(输出)的全局非线性映射关系,以此来代替实际结构优化过程中存在的大量有限元计算,从而提高优化效率。以遗传算法为优化求解器,神经网络屈曲稳定性响应面为主要约束,对复合材料格栅加筋结构的优化问题进行了分析研究。算例表明,在相同(有限元)样本数据的情况下,进化神经网络通过自适应调节网络结构和权值,可获得比BP神经网络更高精度的映射模型,具有很强的泛化能力。该方法可为解决大型复合材料结构优化设计提供一条高效途径。
According to the existence theorem of Kolmogorov multi-layer neural network mapping, the global nonlinear mapping relationship between structural design parameters (input) and structural response parameters (output) is realized by using evolutionary neural network instead of a large number of finite elements Calculation, thereby improving the optimization efficiency. The genetic algorithm optimization solver and the neural network buckling stability response surface are the main constraints, and the optimization of the composite grid stiffened structure is analyzed and studied. The example shows that in the same (finite element) sample data, the evolutionary neural network can obtain a more accurate mapping model than the BP neural network by adaptively adjusting the network structure and weights, and has strong generalization ability. This method can provide an efficient way to solve the optimal design of large-scale composite structures.