work we propose a new
tool for process enginners to compare
different approaches for the design of multiproduct batch plants: the performance profiles.
This tool was introduced in 2002 by Dolan and Moré who claimed that "performance profiles combine the best feauters of other tools for comparing optimization solvers" and as other tools, make use of a battery of a big number -let say a hundred- of instances of the same type.
As in our work we wanted to compare different approaches to minimize equipment costs in the design of biotechnological multiproduct batch plants and as we did not have available a sufficient amount of instances to use these comparison tools we generated with random, but yet reasonable data, two batteries of instances that allowed us to fulfill our goal. Then this page was created to supply this information so is available to the community if want to test our results or come out with new contributions.
Figure1: Comparison of performance profiles of the logarithm of running time of ``sizing instances'' solved using models (P1) and (P3) with an optimality relative gap of 0.1% for the linear solver and 2% for non-linear solvers.
example of the use of performance profiles we present a comparison of
these curves obtained solving different approaches studied in our
article "MILP reformulations for the desing of biotechnological
multi-product batch plants using continuous equipment sizes and
discrete host selection".
On Figure 1 we can see that the 100% of the instances were solved faster using model (P3) (blue lines) while just the 75% of the instances could be solved using model (P1) with SCIP solver (red line) at the asked optimality gap on limit time (48 hours).