Modelos e algoritmos para planejamento integrado na indústria da mineração

Abstract

In this Thesis, we develop models and algorithms applied to integrated production and logistics problems in the mining industry. Based on an extensive literature review, we address the Global Mining Supply Chain concept and discuss the main related Operations Research problems under a integrated planning perspective. Strategic decisions are evaluated in a novel multistage stochastic integer programmingmodel to address the capacity planning problem in a Global Mining Supply Chain. The model integrates capacitated facility location and network design decisions with economies of scale on the capacity costs. We analyze the characteristics of the problem by means of an empirical study of different settings for the parameters of the CPLEX solver. Such analysis provides pointers to the development of specific algorithmsand solution approaches. We then develop a Lagrangian Heuristic as a means to determine, for large problem instances, good feasible solutions in a reasonable amount of time when compared to CPLEX. Furthermore, the ability of determining good feasible solutions in the early stages of the computation is addressed in a soft-fixing local search framework, which is evaluated against the other solution approaches discussed. Tactical decisions are tackled in a mixed-integer programming approach to the integrated sales and operations tactical planning problem in a Global Mining Supply Chain. The model has characteristics of a lot sizing problem in a network environment, but with challenging aspects related to integer flows, discrete production levels and mass losses in concentration and transportation processes. We develop a series of Relax&Fix strategies in order to address realistic sized problem instances. Those strategies are able to outperform CPLEX for most of the several problem instances considered, and with greater success in longer planning horizons. The soft-fixing local search is also evaluated for its ability of determining good feasible solutions in the early stages of the computation. Operational decisions are briefly discussed in a mixed-integer goal programming model to address the integrated short-term programming of iron ore open pits, processing plants, stockyards and shipping operations. We propose the concept of the Value of the Integrated Solution, which determines how valuable is solving a more complex integrated decision problem given the potential losses incurred when individual decisions are undertaken.