一类流水型知识化制造单元调度的自进化算法  

A Self-evolution Algorithm for Scheduling a Flow-shop-like Knowledgeable Manufacturing Cell

作　　者：李文超[1,2] 严洪森[1] 

机构地区：[1]东南大学复杂工程系统测量与控制教育部重点实验室,南京210096 [2]江苏大学,镇江212013

出　　处：《中国机械工程》2011年第7期830-835,共6页China Mechanical Engineering

基　　金：国家自然科学基金资助项目(60934008;50875046)

摘　　要：机器数大于2的流水型知识化制造单元调度属于NP完全问题,至今尚无完全有效的算法。针对该类问题的自身结构特性,提出一种具备学习能力的自进化算法。该算法采用强化学习中值迭代策略,在运行中能够从环境中获取相应知识,提高其搜索能力。提出一种混合核支持向量机对值函数进行逼近,以解决学习过程中状态过多的问题。数值仿真实验表明,算法对该类问题具有很好的学习进化能力。The flow-shop-like knowledgeable manufacturing cell（whose machine number is greater than 2） scheduling is a NP-complete problem and has not a completely valid algorithm for it until now.A self-evolution algorithm with learning ability was proposed according to its structure and characteristics.By adopting the value iterative strategies of reinforcement learning,the algorithm can absorb the corresponding knowledge from its environment during its running and improve its search ability.The approximation of value function is completed by use of the SVM with a hybrid kernel to avoid the too many states in learning process.Numerical experiments show that the algorithm has excellent performance of learning and evolution for the problem.

关 键 词：知识化制造单元 自进化 强化学习 支持向量机 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]