论文部分内容阅读
该文基于遗传算法对复合材料带加强筋板中加强筋的铺层顺序进行了优化,使结构在质量一定的情况下结构屈曲载荷最大。为了减少优化过程中有限元模型的调用次数,引入径向基神经网络作为代理模型对结构屈曲载荷进行估计,并且将铺层参数作为其输入以降低目标函数的非线性。由于设计空间形状不规则,采用D-optimal实验设计方法确定训练径向基神经网络的采样点集。考虑到代理模型存在估计误差,提出了加强代理模型在暂定最优区域估计精度的方法。算例表明:以铺层参数为输入的径向基神经网络在建立代理模型方面具有较高的精度和效率;代理模型的局部精度加强可进一步提高代理模型在暂定最优区域的精度。
Based on the genetic algorithm, this paper optimizes the order of the ply stiffeners in the stiffened plates with composite materials so that the buckling load of the structure under the condition of certain mass is the highest. In order to reduce the number of calls of the finite element model during the optimization process, the RBF neural network is used as a proxy model to estimate the buckling load of the structure, and the ply parameter is taken as the input to reduce the nonlinearity of the objective function. Due to the irregular shape of the design space, the D-optimal experimental design method is used to determine the sampling point set for training RBF neural network. Considering the existence of estimation error in the proxy model, a method of enhancing the accuracy of the proxy model in the tentative optimal region is proposed. The results show that the radial basis function neural network with ply parameters as input has high accuracy and efficiency in establishing proxy model. The local accuracy of proxy model can be further enhanced to improve the accuracy of the proxy model in the tentative optimal region.