Prediction of Strength parameters of Stabilized Sand Soil with Nonoclay Using Artificial Neural Network

Nowadays, soil stabilization and improving its strength properties has captured researchers’ attention. To achieve the objective, a variety of methods and materials are applied. One of the novel methods of soil stabilization is to combine it with nanoclay. This research seeks to predict the soil strength parameters of stabilized sand with nanoclay, using artificial neural networks. The experiments on samples mixed with different percentages of nanoclay and different processing length indicate that in a few of samples, increase in nonoclay in the soil leads to increase in the soil shear strength and in angle of friction. By the help of experimental results and a series of input (the geotechnical properties of soil samples), output (cohesion and friction angle of soil samples) and using MATLAB software several models were investigated. After training several neural networks and evaluating, and comparing the values of determination coefficients of MSE, T and R2 in each of the networks, we can conclude the neural networks that calculate parameters of cohesion coefficients and friction angles separately have better behavior and accuracy than neural networks that predicts these parameters simultaneously.

Keywords:

Montmorillonite clay , soil shear strength , artificial neural networks , friction angle , cohesion coefficients