Prediction of Fine-Grained Soils Shrinkage Limits Using Artificial Neural Networks

Shrinkage limit can be used to estimate other critical factors such as shrinkage ratio, shrinkage index and volumetric shrinkage. Experimental determination of shrinkage limit used to be performed following ASTM D427-04, and BSI 1377-2:1990:6.3, however, both standards were withdrawn without replacement due to the hazardousness of mercury used in the testing procedures. The use of numerical modeling for shrinkage limit was absent in the literature. In this study, deep artificial neural networks (ANN) were utilized to predict shrinkage limit from plastic and liquid limits. The results have shown relatively high determination coefficient of 0.71 compared to 0.30 in regular Casagrande approach. In addition, a heat map was developed to visualize the results of developed ANN.