A two-stage simulation optimization heuristic for robust design

The optimization of simulation models is generally performed with fixed environmental conditions. However, in practice disturbances can modify the expected environment (modification of the part mix or of the customer orders In a manufacturing systems, etc.). In order to provide solutions that are robust w. r. t. such modifications, we propose a new methodology that is closer to the decision-makers´ reasoning than more classical approaches, such as the use of signal to noise ratio. In order to be able to compare solutions in terms of cost, a first heuristic search determines the (near) best solution and its performance. In a second stage, we consider several possible production environments and a second simulation optimization is performed. Solutions are evaluated according the performance (expressed in terms of costs) of what they would lead to save or to loose on other environments than the expected one. Referee curves allows these solutions to be compared in accordance with the decision-makers´ requirements.