Estimation of UCS-FT of Dispersive Soil Stabilized with Fly Ash, Cement Clinker and GGBS by Artificial Intelligence

In this investigation, the unconfined compressive strength-freeze–thaw (UCS-FT) of dispersive soil stabilized with cement clinker, ground granulated blast furnace slag (GGBS) and fly ash was modeled and predicted using group method of data handling, multivariate adaptive regression splines (MARS) and M5P-based techniques. To this end, a related dataset (250 observations) was collected from the experimental program. The model results are evaluated and compared to decide the best model. Results indicated that the proposed MARS model works better than the GMDH- and M5P-based model for predicting the UCS-FT. The correlation coefficient obtained for the best model is 0.9952, 9.9914 while the mean absolute error and root mean square error are 0.1656 (MPa), 0.2174 (MPa) and 0.2205 (MPa), 0.3028 (MPa) for calibration and validation stages, respectively. To estimate each input variable’s effect on the proposed best model, sensitivity analysis is also done using the best performing MARS model. Outcomes of the sensitivity analysis indicate that curing time is the most effective parameter for estimating the unconfined compressive strength-freeze–thaw of dispersive soil stabilized with cement clinker, GGBS and fly ash.