Prediction of California Bearing Ratio of Subgrade Soils Using Artificial Neural Network Principles

Abstract

AbstractIn soil mechanics, prediction of soil properties is necessary due to the large-scale construction activities and time-consuming testing. California Bearing Ratio (CBR) is one of the soul parameters used as strength and stiffness indicator for subgrade soil. However, for investigating soil subgrade in the field, there is a need of more soil samples to be tested; it may be time-consuming and cumbersome task. Moreover, certain issues like lack of funding, unavailability of skilled labour and poor laboratory infrastructure to handle large number of samples put thrust on development of models to predict strength with reference to certain amount of data. Nowadays, the potentiality of prediction models has been gaining importance in every discipline. Numerous tools and techniques were evolved focusing on model development; which will be able to perform iteration-based techniques. In this study, CBR values of subgrade along a proposed road are collected. Nearly, 480 samples were collected in which 15 samples were used for comparison (control value). The results revealed that the artificial neural networks (ANN) prediction models were significant promising tool for predicting CBR of subgrade soil by using index properties as input parameters.KeywordsCalifornia Bearing Ratio (CBR)Artificial neural networks (ANN)BackpropagationPavements