Proposta de modelagem da resistência à deformação permanente de misturas asfálticas densas

Abstract

Permanent deformation is one of the main pathological manifestations presented by asphalt pavements. The Flow Number (FN) is a parameter obtained in the uniaxial repeated load test that allows to hierarchize asphalt mixtures as to the capacity of resistance to this defect, or in other words, resistance to the accumulation of plastic deformations. MeDiNa, the current national method for asphalt pavements design, establishes that this defect must be prevented even in the asphalt mix design, which opens up the possibility of using FN prediction models to guide the selection of materials and their proportion. In the international literature, there are already several attempts to develop models capable of estimating the FN, however, it is not possible to ensure that such models are suitable to national conditions. Therefore, given the scarcity of predictive models for the FN adapted to the conditions of Brazilian pavements, we sought to develop and analyze models with different predictor variables and different degrees of complexity, aiming with this, the proposal of an auxiliary tool in the design of asphalt mixtures. The explanatory capacities of linear regression, nonlinear regression, and derived from Artificial Neural Networks (ANNs) were analyzed, concluding that the former is not able to satisfactorily explain the variations of the FN and that the use of the latter is only justified for investigations into larger databases. Consequently, nonlinear models are the most recommended for the prediction of this parameter, highlighting those from the Bailey method.