Prediction of unconfined compressive strength of soft grounds using computational intelligence techniques: A comparative study

Cement stabilization is one of the commonly used techniques to improve the strength of soft ground/clays, generally found along coastal and low land areas. The strength development in cement stabilization technique depends on the soil properties, cement content, curing period and environmental conditions. For optimal and effective utilization of cement, there is a need to develop a mathematical model relating the gain in strength in terms of the variables responsible. The existing empirical model in the literature assumes linear variation of normalized strength with the logarithm of curing period and hence, different empirical models are required for different conditions of the same soil. Also, the accuracy of strength prediction is unsatisfactory. Due to unknown functional relationships and nonlinearity in strength development, in this paper the computational intelligence techniques such as multilayer perceptron (MLP), radial basis function (RBF) and genetic programming (GP) are used to develop a mathematical model. To generate the mathematical model, an experimental study is conducted to obtain the strength of three inland soils namely, red earth (CL), brown earth (CH) and black cotton soil (CH) for different water contents, cement contents and curing periods. In order to generate a generic mathematical model using computational intelligence techniques, two saline soils (Ariake clay-3 and Ariake clay-4) and three inland soils are used. A detailed study of the relative performance of the computational intelligence techniques and the empirical model has been carried out.