论文部分内容阅读
运用正交试验设计选择设计样本,建立神经网络响应面,以代替复合材料结构优化中的大量的有限元分析;将神经网络响应面作为目标函数或者约束条件,加上其他常规约束条件进行优化模型的建立,再应用遗传算法(GA)进行优化,这可以实现设计分析与设计优化的分离。以复合材料帽型加筋板的重量优化问题为例,建立了重量响应面目标函数、强度和翘曲稳定性响应面约束条件;并通过NASTRAN进行有限元计算,以获取用于响应面训练的样本点数据。研究表明,该方法能以较少的结构分析次数,取得高精度的响应面近似模型,从而使优化效率大为提高。神经网络响应面能够获得与传统响应面同等,甚至更好的精度。
Orthogonal experimental design was used to select the design samples and the neural network response surface was established to replace a large number of finite element analysis in the structural optimization of composite materials. The neural network response surface was used as the objective function or constraint, combined with other conventional constraints to optimize the model The establishment, and then applied genetic algorithm (GA) to optimize, which can achieve the separation of design analysis and design optimization. Taking the weight optimization problem of the composite cap-hat stiffened plate as an example, the objective function of the weight response surface, the constraint conditions of the strength and warpage stability response surface were set up. The finite element method was used to obtain the response surface training Sample point data. The research shows that this method can get a high-precision response surface approximation model with fewer structural analysis times, so that the optimization efficiency is greatly improved. The neural network response surface can obtain the same or even better accuracy than the traditional response surface.