Utilization of geopolymerisation process for the simultaneous stabilization/solidification of lead/zinc mine tailings and enhancement of expansive subgrade soil

In the present study, geopolymerisation process was employed for the simultaneous solidification/stabilization of lead zine mine tailings (LZMT) and improvement of the engineering properties of an expansive subgrade soil. Taguchi L9 orthogonal array was used for the generation of the mix ratios of additives which comprises LZMT, Metabentonite (MB) and Sodium hydroxide (NaOH), and the combination of the additives that yielded optimal improvement in the properties of the expansive subgrade soil was 20% LZMT, 10% MB and 7 M NaOH. Next, six statistical regression/machine learning tools which included multilinear regression (MLR), ridge regression, LASSO regression, elastic-net regression, Gaussian process regression and relevant vector regression were applied to develop models capable of predicting the unconfined compressive strength (UCS) of the modified expansive subgrade soil blended with the optimum combination of additives. Statistical analysis performed to evaluate the developed models reveals that, the elastic-net regression tool, with coefficient of determination values of 0.992 and 0.914 for the training and testing dataset respectively, and with the least statistical metrics error values used for comparison, gave the best UCS prediction model. This was attributed to ability of the elastic-net inherent ability to balance the selection of variables and shrinkage of coefficients.