Otimização aplicada à engenharia florestal

• Autor(a) principal: Monti, Cássio Augusto Ussi
• Data de Publicação: 2018
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da UFLA
• Texto Completo: http://repositorio.ufla.br/jspui/handle/1/29006
• topic: Algoritmo genético; Programação linear inteira mista; Sistema integrado; Parametrização não linear; Algorithm; Mixed integer linear programming; Integrated system; Non-linear parameterization; Recursos Florestais e Engenharia Florestal

System modeling is applied in various branches of engineering to represent any type of operational situation through stochastic techniques, such as linear and non-linear regression, or deterministic, as a model of vehicle routing. Lately it is common application of computational intelligence as an alternative to these techniques, such as heuristics and metaheuristics. Based on this information, planning harvest and transport forestry were design, in order to supply the factory. Such planning uses the deterministic modeling of systems through mathematical programming models from the operational research discipline. Thus, the structure of the dissertation contemplated three chapters, the first part being a bibliographical survey of the state of the art on the fitting of statistical models and theory of graphs applied to the transport and optimization of problems. In the second part, we tried to approach methods of fitting nonlinear regression models, proposing the development of an alternative method to the algorithms usually applied. The same predicts the estimation of the initial parameters automatically using genetic algorithm as an option. Thus, three classical fitting methods, Gauss-Newton, Levenberg-Marquardt and NL2SOL, were tested. The hypsometric relation was adopted as the basis for the computational tests, and a combination of 12 distinct forest databases and 12 nonlinear models were tested in 100 replicates to guarantee information reliability. Among the tested methods, Levenberg-Marquardt showed a greater frequency of fitting success, followed by NL2SOL and Gauss-Newton. The genetic algorithm did not produce significant difference in the prediction of the parameters, and can be used in the parameterization of statistical regression models. The last part of the dissertation composed a study involving a deterministic modeling for the vehicle routing problem applied to forest transportation (VRPFT). The model involving integer and continuous variables considered in its multiobjective function the reduction in the number of trucks, distance traveled, overtime, load efficiency and crane movement as an integrated process. The results indicated absence of overtime and a reduction of 72.92% in the use of trucks, consequently promoting an increase in the number of trips / vehicles. In only 3.17% of the trips to the load efficiency was not obtained, being a fact predicted in the final removal of the wood in the plots. It was also identified points of improvement in the movement of cranes between blocks due to operational idle. The VRPFT model acts as an efficient and effective integrated management system applied to forest transportation.