Estimating the resilient modulus of subgrade materials using visual inspection

Abstract

The definition of the Resilient Modulus (MR) of subgrade soils is essential for the reliable implementation of mechanistic-empirical pavement design. The MR of the soil is measured through repeated triaxial load tests which require expensive equipment and complex analyses. This reinforces the need to develop accurate statistical models for the prediction of the MR of the subgrade soil to be used for paving highways, especially in developing countries, such as Brazil, where financial resources are limited. The present study used artificial neural networks (ANNs) to create a model for the prediction of the MR of subgrade soils based on a visual-manual classification. For this, the results of MR tests conducted on samples of different soils from northeastern Brazil were used to develop an ANNs model for the prediction of the MR. The results demonstrate that ANNs can predict reliably the MR of soils, with a good degree of correlation in comparison with the laboratory test data. These findings support the use of the ANN model as a cost-effective approach for the preliminary evaluation of subgrade soils for highway pavement design in northeastern Brazil.

Keywords

• Artificial neural networks
• Visual-manual classification
• Mechanistic-empirical design of pavement
• Low volume roads