Modelos matemáticos para auxílio à tomada de decisão no processo produtivo de Pinus caribaea Morelet var. caribaea Barr. & Golf. na Empresa Florestal Integral Macurije, Pinar del Río, Cuba

• Autor(a) principal: GUERA, Ouorou Ganni Mariel
• Data de Publicação: 2017
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da UFRPE
• Texto Completo: http://www.tede2.ufrpe.br:8080/tede2/handle/tede2/7386
• topic: Modelo matemático; Pinus caribaea; Produção madeireira; Manejo florestal; CIENCIAS AGRARIAS::RECURSOS FLORESTAIS E ENGENHARIA FLORESTAL

The objective of this study was to propose models that aid decision making in productive process of Pinus caribaea var. caribaea Barr. & Golf. through the application of multivariate techniques, regression analysis, multicriteria decision analysis techniques (MCDA) and Artificial Neural Networks (ANNs) in different stages of said process. The three stages of the forest production process (PPF) involved in the present study were: (1) growth, yield and forest survival stage; (2) wood extraction and transport stage, and (3) wood primary transformation stage. Pinus caribaea var. caribaea growth, yield and survival modeling required data from temporary and permanent circular plots of 500 m² of the Macurije Integral Forest Company, in which the following variables were measured: : Diameter at Breast Height - DBH (cm), total height - H (m) and survival - (num. of trees/ha). At this stage, the specie productive capacity classification was carried, Artificial Neural Networks (ANNs) were trained and regression models were adjusted for growth prediction and yield and survival prognosis. At wood extraction and transport stage, the performance of different wood extraction systems and means was evaluated through univariate and multivariate factorial experiments, being cost and productivity the dependent variables obtained by time and movement studies. At the same stage, a Lexicographic Goals Programming model was proposed to assist decision making in harvesting and forest transport planning. At the stage of wood primary transformation in Combate de Tenerías sawmill, regression models were adjusted and ANNs were trained, both for lumber recovery factor prediction and lumber classification. Lumber quality being a discrete ordinal variable, ordinal logistic regression was used for its modeling. The database required for lumber recovery factor modeling was composed by the variables Diameter at Breast Height (DBH), Smallest log diameter (D) and conicity (Con.) obtained from real-time monitoring of wood sawing at the sawmill Combate de Tenerías. The 24 variables predicting lumber quality were measured in pieces obtained at the end the end of sawing process in the same sawmill. The results obtained during the research indicated that multivariate, multicriteria and Artificial Neural Networks techniques are efficient in assisting decision-making in FPP stages considered. ANNs models presented similar or superior performances to the traditional regression models both in prediction (volumetric growth, lumber recovery factor) or prognosis (survival, growth and yield) and in lumber grading. From the results, it was concluded that it is not prudent to assume absolute superiority of ANNs and that opting for the complementarity of both approaches rather than the exclusive use of ANNs, as most comparative research tends to suggest, is far more prudent. Multivariate evaluation of wood extraction machineries performances and the Lexicographic Goal Programming model proposed for timber extraction and transport planning provided a multicriteria support translated into solutions with greater practicality and functionality.