多约束条件下自动配棉的混合遗传算法  

Hybrid genetic algorithm for multi-constrained automatic cotton blending

作　　者：林兰芬[1] 欧冠男[1] 陆俊虎[1] 

机构地区：[1]浙江大学人工智能研究所,浙江杭州310027

出　　处：《浙江大学学报（工学版）》2009年第5期801-806,共6页Journal of Zhejiang University：Engineering Science

基　　金：浙江省科技计划资助项目(2006C11236)

摘　　要：现有自动配棉方法求解的问题规模不大或者不易找到最优解,为此建立了一个多约束条件下自动配棉问题的数学模型.基于基本遗传算法,采用罚函数法处理多个约束条件,通过对种群进化程度进行监控并适时地增大选择压力,形成求解自动配棉问题的基于自适应罚函数法的混合遗传算法(MGA1).进一步提出了求解自动配棉问题的基于模拟退火算法和隔代相传策略的混合遗传算法(MGA2).以企业实际棉批库存与质量数据为例进行实验验证,结果表明,对于中小规模的配棉问题,MGA2具备较好的寻找最优解和较优解的能力,并且其解表现出多样性的特点;而对于大规模的配棉问题,MGA1保持良好的收敛性,能够找到比MGA2更好的最优解和较优解.The existing methods for automatic cotton blending cannot handle large-size problems or find the best solutions easily. A mathematic model for multi-constrained automatic cotton blending was set up. A hybrid genetic algorithm （MGA1） based on the adaptive penalty function was proposed, which used the penalty function to handle multi-constraints, and improved the rate of population revolution by monitoring the extent of population revolution and increasing the selection pressure properly. Furthermore, another hybrid genetic algorithm （MGA2） based on the simulated annealing algorithm and the atavistic strategy was proposed. Experiments were carried out to verify the proposed algorithms using the actual raw cotton stock and quality data from an enterprise. For the automatic cotton blending problem with middle and small size, MGA2 is better to find the optimum and diversiform solutions. For the automatic cotton blending problem with larger size, MGA1 can keep better astringency. And the solutions found by MGA1 are better than those of MGA2. Using MGA1 and MGA2 synthetically can satisfy different demands of cotton enterprises.

关 键 词：配棉 自适应罚函数 模拟退火 混合遗传算法 

分 类 号：TP391.75[自动化与计算机技术—计算机应用技术]