Otimização via simulação aplicada à logística reversa

Abstract

The growth of waste electrical and electronic equipment (WEEE) poses economic, environmental, and social challenges to the planning of reverse logistics (RL). This study proposes and applies a Simulation-based Optimization (OvS) approach to support vehicle routing and fleet allocation in a Heterogeneous Fleet Vehicle Routing Problem (HFVRP), balancing cost, CO₂ emissions, and collection volume. The method integrates: (i) data preparation (distance matrices, demands, and costs, as well as vehicle characteristics); (ii) generation of solutions by a multi-objective Genetic Algorithm (GA); and (iii) stochastic evaluation through discrete-event simulation (FlexSim), without re-optimization. The social dimension was operationalized exclusively as the total collected quantity (kg) and service coverage, with an additional link to “perishable food”: for every kilogram of waste collected, one kilogram of perishable food would be donated to charitable organizations. A total of 20 scenarios from the GA’s non-dominated frontier were evaluated. In Case 1, which involved municipalities of the Metropolitan Region of São Paulo, one scenario (9) proved invalid during simulation and was excluded from the final comparison. Among the valid ones, informative extremes were observed: the lowest cost was R$ 626.53 (Scenarios 13 and 14), with 78.01 kg CO₂ and 7,366.74 kg collected; the lowest emissions were 74.24 kg CO₂ (Scenarios 8 and 10), both with R$ 631.38 and 7,366.74 kg collected; and the largest collection reached 15,979.03 kg (Scenario 20), with a cost of R$ 2,137.76 and 270.21 kg CO₂. In the weighted sum analysis (min–max normalization), Scenario 20 stood out when emissions and collection were jointly prioritized, whereas Scenario 10 emerged as a balanced solution (low cost, minimum emissions, and intermediate collection). In Case 2, designed with distinct fleet parameters, capacities, and distance matrices, the goal was to evaluate the robustness and scalability of the model. The results confirmed that the tool preserved the consistency of the trade-offs identified in the first case, adapting to alternative operational conditions without loss of coherence. The comparative analysis demonstrated that the method is applicable across diverse operational realities, thereby expanding its practical potential in reverse logistics. It is concluded that OvS combined with a multi-objective GA responds positively to the problem: it filters infeasible solutions, makes trade-offs explicit, and delivers an auditable portfolio of alternatives for WEEE reverse logistics, aligning economic, environmental, and service coverage goals. The computational tool developed constitutes a replicable basis for practical adoption and for future extensions.