基于遗传神经算法的压力钢管外压稳定性分析

压力钢管外压稳定性分析是重要而难以解决的问题。用遗传-神经网络模型对压力钢管外压稳定问题进行仿真计算,模型具有神经网络强大的函数逼近功能,同时利用遗传算法克服传统神经网络方法易陷入局部极小点的缺陷。实例仿真结果表明,用实数编码遗传神经网络模型分析加劲压力钢管的外压稳定性,具有效率高、鲁棒性好的优点,精度满足工程要求。     

Global and local real-coded genetic algorithms based on parent-centric crossover operators

Parent-centric real-parameter crossover operators create the offspring in the neighbourhood of one of the parents, the female parent. The other parent, the male one, defines the range of the neighbourhood. With the aim of improving the behaviour of these crossover operators, we present three processes that are performed before their application. First, a female and male differentiation process determines the individuals in the population that may become female or/and male parents. Then, two different selection mechanisms choose the female and male parents from each group. In addition, we tackle the election of the most adequate evolution model to take out profit from these parent selection mechanisms. The experimental results confirm that these three processes may enhance the operation of the parent-centric crossover operators.     

用于光子相关光谱颗粒粒度测量的实数编码混合遗传算法

提出一种自动缩小反演范围的实数编码混合遗传算法,并用于光子相关光谱颗粒粒度测量中.该算法利用混沌遍历性质,融入模拟退火算法,具有较强的局部搜索能力.通过数值模拟,讨论在无噪声和多种噪声水平影响下单分散颗粒系的反演结果;分析90 nm聚苯乙烯标准颗粒样品的实验测量数据.结果表明,反演的粒径分布在峰值位置、峰宽度、峰对称性等方面很接近颗粒的真实分布,具有较强的抗噪声干扰能力.     

基于加权实数编码遗传算法的超材料优化设计

针对超材料优化设计参数多、权重各异的特点,提出一种基于加权实数编码的遗传算法,该算法采用等位基因或双倍基因实现对遗传基因的加权编码,是二进制编码中加权思想的推广,与普通遗传算法相比,加权实数编码遗传算法加入了人工选择因素,既可以加快种群收敛速度,又可以提高算法的求解质量,尤其适用于种群规模较大的遗传优化计算,文中以一个超材料吸波体的优化设计为例,对算法进行了验证。