ABSTRACT
The impacts on the landscape after forest harvesting in reforestation are visible, but the cutting is a necessary process to ensure a sustained yield and introduce new technologies. An alternative of control is to use the adjacency constraints in the mathematical models. Thus, the aim of the study was to assess the ability of the metaheuristic SA to solve mathematical models with adjacency constraints type URM, and to check its action with the increasing of the problem complexity. The study was conducted in a forest project containing 52 stands, and created 8 scenarios, where the Johnson and Scheurmann (1977) model I was used as reference. The adjacency constraint type URM was used to control the cutting of adjacent stands. The models were solved by the ILP and metaheuristic SA, which was sued 100 times per scenario. The results showed that the scenario 8 has consumed 137,530 seconds via PLI, which represented 2,023.09 times more than the average time processing of the SA metaheuristic (67.98 seconds). The best solutions were 4.71 % (scenario 1) to 11.40 % (scenario 8) far from the optimal (ILP). The metaheuristic SA is capable to solve the forest problem, meeting the targets in the most cases. The increasing of complexity produced a higher deviation from the optimal. Concludes that the metaheuristic SA should not be processed a single time, because there are hazards in obtain inferior solutions, but doing it is recommended to increase the stop criterion.
Keywords:
artificial intelligence; integer linear programming; forest harvest