Study of retroanalysis of asphalc pavements resilience modules with the use of arficial neural networks

Abstract

Knowledge of pavement Resilience Modules (RM) is an important element for the struc-tural assessment of existing infrastructures. One way to determine them is through dy-namic repeated load testing in the laboratory; another way is to use the technique of retroanalysis, which consists in obtaining the RM from the thickness of the layers and the deflec ons measured on the pavement surface. In this sense, this paper aims to present the RM retroanalysis through the technique of Ar ficial Neural Networks (ANN) as an alterna ve to the tradi onal retroanalysis. The results demonstrate that the ANN could predict the RM with results of coefficients of determina on (R²) above 99.9% be-tween the reference and predicted values. Thus, ANN are a poten al alterna ve to ob-tain this important mechanical property of paving materials compared to tradi onal methods.