中国机械工程
中國機械工程
중국궤계공정
CHINA MECHANICAl ENGINEERING
2012年
10期
1194-1199
,共6页
鲜奶采购计划%外包%遗传算法%自适应罚函数
鮮奶採購計劃%外包%遺傳算法%自適應罰函數
선내채구계화%외포%유전산법%자괄응벌함수
raw milk purchasing plan%outsourcing%genetic algorithm%adaptive penalty function
奶制品加工厂主要从总部采购鲜奶(生产),也会从当地奶农采购部分鲜奶(外包)。出于能力和策略的考虑,两种来源的鲜奶采购量均有限;考虑到客户允许奶制品延期交货,相应的鲜奶也可延期交货,生产、外包和库存/延期交货成本均为一般凹函数,问题是以最小的总成本来满足规划期的鲜奶需求。为求解该问题,设计了一种新的基于群体可行状态和个体约束违背程度的自适应惩罚方案,据此设计了求解该问题的遗传算法。为测试算法的性能,先进行了算子组合和遗传算子概率的选择实验,选出最适合的算子和算子概率;在此基础上,针对4个问题实例,通过50次运行,测试了所提自适应罚函数相对4种常见罚函数的优势。
奶製品加工廠主要從總部採購鮮奶(生產),也會從噹地奶農採購部分鮮奶(外包)。齣于能力和策略的攷慮,兩種來源的鮮奶採購量均有限;攷慮到客戶允許奶製品延期交貨,相應的鮮奶也可延期交貨,生產、外包和庫存/延期交貨成本均為一般凹函數,問題是以最小的總成本來滿足規劃期的鮮奶需求。為求解該問題,設計瞭一種新的基于群體可行狀態和箇體約束違揹程度的自適應懲罰方案,據此設計瞭求解該問題的遺傳算法。為測試算法的性能,先進行瞭算子組閤和遺傳算子概率的選擇實驗,選齣最適閤的算子和算子概率;在此基礎上,針對4箇問題實例,通過50次運行,測試瞭所提自適應罰函數相對4種常見罰函數的優勢。
내제품가공엄주요종총부채구선내(생산),야회종당지내농채구부분선내(외포)。출우능력화책략적고필,량충래원적선내채구량균유한;고필도객호윤허내제품연기교화,상응적선내야가연기교화,생산、외포화고존/연기교화성본균위일반요함수,문제시이최소적총성본래만족규화기적선내수구。위구해해문제,설계료일충신적기우군체가행상태화개체약속위배정도적자괄응징벌방안,거차설계료구해해문제적유전산법。위측시산법적성능,선진행료산자조합화유전산자개솔적선택실험,선출최괄합적산자화산자개솔;재차기출상,침대4개문제실례,통과50차운행,측시료소제자괄응벌함수상대4충상견벌함수적우세。
The dairy processing plant purchases from local market, where the former can be viewed raw milk mainly from its headquarters, sometimes as production and the latter as outsourcing. Owing to the production capacity and management policy,both production and outsourcing levels are bound ed. Raw milk supply can be backlogged because of dairy postponement allowed by customers. All of production,outsourcing and inventory/backlogging costs were as general concave functions. The prob- lem was to satisfy all demands in the planning horizon by minimum total costs. To solve the problem, a novel adaptive penalty scheme was devised based on the feasible status of population and the degree of constraint violation for individuals. Furthermore,a new adaptive genetic algorithm was developed to solve the constrained optimization problem by using the penalty scheme. In order to test the perform- ance of the proposed algorithm, we first made a selection examination to obtain the fittest operator combination and genetic operators' probabilities. Then, for four planning problems, by 50 runs of the algorithm, the advantages of the penalty scheme were verified by comparing with other common four penalty methods.