Predictive modelling of compression strength of waste GP/FA blended expansive soils using multi-expression programming

Abstract

Expansive soils (ES) are unsuitable for construction because of their low shear strength, bearing capacity, unconfined compression strength (UCS), and high shrink-swell potential. The traditional experimental techniques for obtaining the optimized mixes of stabilized ES using a variety of waste additives are laborious, time-consuming and uneconomical. Therefore, this study determines the optimized mixes by evaluating the UCS of waste glass powder (WGP) and Fly Ash (FA) modified ES by deploying a robust computational technique, i.e., multi-expression programming (MEP). The ES was modified using WGP at 0, 5, 10, 15, 20, 25, and 30%, in order to determine the optimum dosage of WGP. Furthermore, FA was incorporated at 0, 5, 8, 11, 14, and 17% in the ES at the optimum WGP content, and their synergistic effect was determined. An experimental dataset was compiled with 180 test results on the untreated ES to develop the MEP model. The proposed optimal MEP model relates the UCS at 28 days (UCS28d) to its plasticity index, specific gravity, swell potential, UCS3d, and UCS7d. It was also evaluated using different statistical indices, such as correlation coefficient (R = 0.995), Nash-Sutcliffe efficiency (NSE = 0.989), coefficient of determination (R2 = 0.989), and error analysis indices (MAE = 6.87 kPa, RMSE = 11.42 kPa, and performance index ρ = 0.016) for the test dataset. These values suggest the robust predictive capability of the MEP model, thus enabling the designers and practitioners to readily use these equations where cost-effectiveness and time are to be considered in engineering projects. In addition, the inclusion of these industrial waste-based additives would lead to sustainability in the field of geoenvironmental engineering.