论文部分内容阅读
采用综合处理方法,如用织物结构或单向纤维丝,能够对注塑或模压热塑结构进行局部加强。网状结构可以通过周期性典型的注塑或模压方式生产,具有较好的力学性能。由于局部增强在数量和位置方面都增加了设计过程的复杂性,数学优化技术与有限元模拟的结合能够设计重量和成本均较优的结构。这项工作评价了与遗传算法结合的近似模型,案例采用普通的而不是工业化的梁结构。有限元模型基于层单元,分4种不同材料,用9个设计变量控制不同区域材料的厚度。材料的非线性性质包含了超塑聚合物和一种特殊织物材料的压缩响应。对有和没有定期选择、径向基函数及Kriging的4种不同近似方法的精度进行了评估。不是孤立地检查平均近似误差,而是提出了一个如何处理单一近似误差的方法。结果表明,在连续的、连续与离散混合的64个设计变量约束条件下,多峰遗传算法可以成功地用于优化结构重量。
The use of integrated processing methods, such as fabric or unidirectional fiber yarns, enables localized reinforcement of injection molded or molded thermoplastic structures. The mesh structure can be produced through periodic injection molding or molding, which has good mechanical properties. Because local enhancements increase the complexity of the design process in terms of number and location, the combination of mathematical optimization techniques and finite element simulations enables the design of structures with superior weight and cost. This work assesses an approximation model combined with a genetic algorithm that uses a common but not industrial beam structure. The finite element model is based on layer elements and is divided into 4 different materials. Nine design variables are used to control the thickness of the material in different regions. The nonlinear nature of the material contains the compressive response of the superplastic polymer and a special fabric material. The accuracy of four different approximation methods with and without periodic selection, radial basis functions and Kriging was evaluated. Rather than examining the average approximation error in isolation, a method of dealing with a single approximation error is presented. The results show that the multi-peak genetic algorithm can be successfully used to optimize the structural weight with 64 continuous, continuous and discrete mixed design constraints.