Desenvolvimento de um framework para otimização robusta e aplicação para o planejamento tático de uma cadeia de suprimentos de argônio

Abstract

Robust optimization is an area of great applicability and is relativily new in its methodologies and philosopies, and because of that it has been getting a lot of attention and esteem in academia, but also in business and industrial sectors that cope with uncertainties in their processes. The present work bring as a new approach, a framework of linear and non-linear robust optimization, under certain axioms, in which considers the use of probability distributions for the representation of uncertainties, and that it is not necessary to generate scenarios in tree-based nor the assignment of the probability of occurrence of each one of them. The principle of the present framework is to perform a robust optimization by a new approach that regulates the robustness and the conservatism by planner’s will and that also incorporates a regret model specific of the present framework. To perform this principle, it is necessary to transform the original deterministic model into another through the framework’s mathematical formulation and then tune the values of the normal uncertainties’ standard deviations and the penalty factors’ values by an algorithm. Several examples were solved using the developed framework, including a case study regarding a tactical planning of an argon supply chain of real industrial scale. The results show that the framework ensure robustness for the deterministic model that is about to be tuned, at the same time that the uncertainties are taken into account by the optimization, and also show that the framework algorithm’s tuning strategy decrease the penalty in the desired results (as well as its conservatism) by getting closer to the ideal state.