Design and validation of a regression model for soil resilience modulus applied in empirical-mechanistic pavement design

ABSTRACT

The early appearance of pathologies in sidewalks and failure to meet the design horizon not only cause discomfort for users due to poor road conditions and lack of safety during traffic, but also demand a greater amount of raw materials and capital for restoration, recovery or reconstruction of damaged structures. Empirical-mechanistic flexible sidewalk design methods, such as the new Brazilian design method (MeDiNa), take into account the mechanical behavior of the layers and the sidewalk as a whole and more appropriately represent the relationship between the traffic loads and the sidewalk’s response. MeDiNa uses stiffness parameters such as the modulus of resilience (MR) to carry out an elastic-plastic analysis of the structure. This parameter, in turn, suffers some limitations in use, since it is obtained in the laboratory through dynamic triaxial tests that require the use of complex, expensive equipment, specialized labor, in addition to demanding a significant amount of material and being difficult to implement in practical situations, making the test restricted outside universities and academic institutions. In view of this problem, this article proposes a neural regression model for predicting the resilience modulus obtained from basic and usual geotechnical parameters. In addition, to prove the viability of the model generated, simulations were carried out considering different types of roads, as well as linear and non-linear elastic soil constitutive models. This demonstrates that it is a satisfactory model for predicting the modulus of resilience that can be used on roads with moderate to heavy traffic volumes, when it is not possible to carry out the test.

Keywords
Pavements; Empirical-mechanistic methods; Neural model; Simulations