Search

Locally available expansive clays (bentonites) are planned to be used as buffer material in deep geological repositories. The current study highlighted the significance of various geotechnical properties such as specific gravity, liquid limit, plastic limit, specific surface area, cation exchange capacity, percentage of clay content, smectite content and dry density on the swelling pressure of the compacted expansive clays using soft computing techniques such as multiple linear regression (MLR) and artificial neural network (ANN). In total, 185 experimental results of 14 different expansive clays of 27 batches were considered from literature. The investigation was carried out by 2 computational approaches. Initially, MLR technique was used and the swelling pressure was assumed to vary linearly with clay properties. In the second approach, nonlinearity was included using ANN technique. Levenberg–Marquardt (LM) and Scaled Conjugate Gradient (SCG) algorithms were used in ANN approach. The sensitivity analysis of the soil properties was carried out and the parameters were ranked according to their influence on the development of swelling pressure. The nonlinear method was found to be more accurate than the linear method. The LM algorithm was noted to be more accurate in predicting swelling pressure of compacted expansive soil than that of SCG. Among all the influential parameters, the dry density was found to be the most critical parameter to predict the swelling pressure of compacted expansive clays. The outcome of current study models can be useful to predict swelling pressure of compacted bentonite that would be helpful to select buffer material in deep geological repositories. [ABSTRACT FROM AUTHOR]