GEP tree-based computational AI approach to evaluate unconfined compression strength characteristics of Fly ash treated alkali contaminated soils

Abstract

The structural failures induced by alkali-contamination of soils are well established by earlier researchers. In the present study, the effect of alkali concentration (2 M and 4 M) and duration of curing on the unconfined compressive strength (UCS) of two types of soils (Kaolin and BC soil) were evaluated. Further, the effect of fly ash (FA) addition (10%, 15%, and 20%) on the UCS of alkali-contaminated soils was also determined. The results revealed that alkali contamination resulted in reducing UCS (by 36% and 46%), observed for kaolin and BC soil, respectively. The addition of FA has resulted in a linear increase in UCS of both soils. The optimum content was noted at 15%, attributed to the pozzolanic reaction initiated by alkali and subsequent gain of UCS with the formation of calcium (with FA addition) based hydration compounds. In addition, three attributes, i.e., FA content, curing period, and alkali concentration, have been used to develop GEP tree-based models to furnish simple prediction equations for calculating the UCS of Kaolin and BC soils. The accuracy of the developed models were initially validated using various statistical checks such as R, MAE, RMSE and RSE. For the 2nd level validation, parametric and sensitivity analyses were carried out for both soils, which yielded a comparable variation and contribution of each input conforming to the literature.

Keywords

• Flyash
• Alkali Concentration
• GEP tree-based
• Unconfined compression strength
• Contaminated soils
• Artificial Intelligence