Prediction of Soaked CBR for Subgrade Layer by Using Artificial Neutral Network Model

Abstract

Among the different layers of the pavement, subgrade is the bottom most layer, and this layer plays a very important role in the design of pavements. However proper characterization of the said layer involves different types of test data like maximum dry density (MDD), optimum moisture content (OMC), plasticity index, and California bearing ratio (CBR). Among which, CBR is the most important parameter used effectively for assessment of the strength parameter of the subgrade soil. For proper characterization of the subgrade, a large number of CBR tests are required. The CBR tests are elaborative, costly and also time consuming which may delay the progress of constructional activity of the road. In such situation, the value of soaked CBR may be predicted with the help of artificial neural network (ANN), which may help to run the project smoothly. The word “network” in ANN refers to the interconnections between the neurons in the different layers of each system. Those preliminary testing data i.e. percent fines, MDD, OMC, LL, PL, PI etc., are required to find out the regression value and mean square error of the training, validation, and test data to create a model by using neural network algorithm. The tested data are taken during testing the model. The method is based on ANN is found to be reliable, cost-effective, and a quick tool for reasonably accurate estimation of CBR from the basic soil properties.