Optimizing multiple qualifications of products on non-identical parallel machines
研究了半导体制造中如何为机器分配产品资格以最大化工作中心利用率平衡，提出了基于对偶变量的启发式算法，并在实际工厂数据上验证了有效性。
In some manufacturing contexts, such as semiconductor manufacturing, machines must be qualified, or eligible, to process a product, and machines cannot be qualified for all products. This paper investigates the problem of optimizing a given number of new qualifications of products to machines to maximize a flexibility measure that evaluates the balance of the qualification configuration of a work center in terms of utilization rate of machines on a set of non-identical parallel machines. Motivated by empirical observations, new solution approaches, notably inspired by heuristics for discrete location problems and based on the analysis of dual variables, are proposed and compared on industrial data from a semiconductor manufacturing facility and on randomly instances. The use of dual variables leads to heuristics that are effective both in terms of solution quality and computational time. The best proposed approach is currently used in the decision support system of a semiconductor manufacturing facility.