Proposição de nova metodologia e modelo constitutivo para caracterização da deformação permanente de solos e materiais granulares com base na análise de estados de tensão

Abstract

The implementation of the MeDiNa design method in Brazil requires an accurate characterization of the permanent deformation (PD) of soils and granular materials, which currently follows the DNIT 179/2018 standard and utilizes the Guimarães model, which predicts the accumulated permanent deformation as a function of the stress state and the number of load cycles. The present study investigates issues within current calibration processes and proposes improvements aimed at reducing the impacts of the identified problems and increasing confidence in design predictions. The research was divided into four main stages presented in article format. In the first stage, 130 combinations of stress states were analyzed for model calibration across 6 different soils. Results showed that seeking only the best mathematical fit (highest R2) does not guarantee the model's physical representativeness. This often generates parameters with signs contrary to the expected mechanical behavior of the materials, leading to significant errors when predicting stresses outside the tested range (extrapolation). In the second stage, a standardized two-part calibration methodology was proposed, using Multiple Linear Regression followed by the Solver tool. This method eliminated subjectivity from the process but confirmed that the current model has limitations and frequently generates parameters that are physically inconsistent with the expected behavior of the materials. The third stage measured the practical impact of these issues by simulating 48 pavement scenarios in the MeDiNa software. It was evidenced that variations in the calibration of a single layer can change rutting predictions by more than 90%, which is critical for heavy traffic highways. Finally, to address these issues, the study proposed a rational method for selecting stress states based on the actual structural response in the field and developed a new model called MAPD (Improved Permanent Deformation Model). This new model presented superior statistical results compared to classic and current models, combining mathematical precision with physical consistency and allowing for a reduction in the number of required tests. Thus, this master's dissertation contributed to the improvement of the characterization of permanent deformation of soils and granular paving materials, providing methodological tools and a new constitutive model that ensure greater safety and economy for road projects.