Enabling simheuristics through designs for tens of variables: Costing models and online availability

Experiments are key to characterize, model and optimize engineering systems. The use of computer models and hence computer simulations, have allowed engineers to predict the effect of dozen and sometimes hundreds of variables at a specific time in a particular system. The combinatorial explosion that results from using classical techniques to generate experimental designs, however, has hampered such capability. Many analysis tasks, such as simulation optimization and simheuristics, will be importantly enhanced with the possibility of dealing with dozens of variables at a time in a convenient manner. In previous work we identified a series strategies to this end. The objective of the present study is to propose a costing approach to compare these strategies. In addition, designs for 10, 20 or 50 variables and their assessment are made readily available online to different users interested in simulation-optimization based on experimental design, as illustrated here with 50 variables.