正常晶粒长大的遗传算法模拟 - 中国有色金属学报.PDFVIP

第 18 卷第 5 期 中国有色金属学报 2008 年 5 月 Vol.18 No.5 The Chinese Journal of Nonferrous Metals May 2008 文章编号：1004-0609(2008)05-0903-06 正常晶粒长大的遗传算法模拟 张根元，徐迈里，文 芳，陈洪莲 (河海大学 机电工程学院，常州 213022) 摘 要：引入遗传算法模拟正常晶粒长大过程，基于正常晶粒长大动力学和能量最小原理，建立晶粒长大的对应 遗传规则和能量最小的适应度函数。该算法模拟结果表明：系统的晶界总周长和系统晶界能随着遗传代数的传递 不断减少，系统的热力学状态趋于稳定；在 5 000 GAS 前后阶段的晶粒长大指数 n 分别为 0.477 和 0.414，与理论 值 0.500 相近，与正常晶粒长大的动力学规律较符合；晶粒半径分布具有 Weibull 函数拓朴结构形式，平均晶粒 边数为 5.923 。遗传算法可以根据实际晶粒长大过程的约束条件对遗传规则和适应度函数进行设置，因而具有良 好的柔性。 关键词：遗传算法；遗传规则；适应度函数；晶粒；模拟 中图分类号：TG 111 文献标识码：A Genetic algorithm simulation of normal grain growth ZHANG Gen-yuan, XU Mai-li, WEN Fang, CHEN Hong-lian (College of Mechanical and Electrical Engineering, Hohai University, Changzhou 213022, China) Abstract: The genetic algorithm method was introduced in the simulation process of grain growth. Genetic rules and fitness functions were used in the simulation based on the kinetics and the lowest energy principle of normal grain growth. The simulation results indicate that the system boundary length and boundary energy reduce unceasingly along with the increase of GAS, system’s thermodynamics state tends to be stable. In the stage before and after the 5 000 GAS the grain growth exponent n is 0.477 and 0.414, respectively, close to the theoretical value 0.5, satisfying well to the kinetics of normal grain growth. The grain radius distribution agrees well with the Weibull function topology structural style, the average grain side is 5.923. In the genetic algorithm, the genetic rules and fitness functions may be established according to constraint condition in re