Many present markets for goods and services have highly volatile demand due to short life cycles and strong competition in saturated environments. Determination of capacity levels is difficult because capacities often need to be set long before demand realizes. In order to avoid capacity-demand mismatches, operations managers employ mix-flexible resources which allow them to shift excess demands to unused capacities. The Flexibility Design Problem (FDP) models the decision on the optimal configuration of a flexible (manufacturing) network. FDP is a difficult stochastic optimization problem, for which traditional exact approaches are not able to solve but the smallest instances in reasonable time. We develop a Flexibility Design Genetic Algorithm (FGA) that exploits qualitative insights into the structure of good solutions, such as the well-established chaining principle, to enhance its performance. FGA is compared to a commercial solver, a simple GA, and a Simulated Annealing local search on instances of up to 15 demand types and resources. Experimental evidence shows that the proposed approach outperforms the competing methods with respect to both computing time and solution quality.
A problem-adjusted genetic algorithm for flexibility design
VIGO, DANIELE
2013
Abstract
Many present markets for goods and services have highly volatile demand due to short life cycles and strong competition in saturated environments. Determination of capacity levels is difficult because capacities often need to be set long before demand realizes. In order to avoid capacity-demand mismatches, operations managers employ mix-flexible resources which allow them to shift excess demands to unused capacities. The Flexibility Design Problem (FDP) models the decision on the optimal configuration of a flexible (manufacturing) network. FDP is a difficult stochastic optimization problem, for which traditional exact approaches are not able to solve but the smallest instances in reasonable time. We develop a Flexibility Design Genetic Algorithm (FGA) that exploits qualitative insights into the structure of good solutions, such as the well-established chaining principle, to enhance its performance. FGA is compared to a commercial solver, a simple GA, and a Simulated Annealing local search on instances of up to 15 demand types and resources. Experimental evidence shows that the proposed approach outperforms the competing methods with respect to both computing time and solution quality.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
