Consolidation parameters conceptualization using regression analysis and genetic programming for Addis Ababa’s red clay soils

Abstract

The determination of compression and recompression indices in the laboratory is cumbersome and time-consuming. The objective of this paper is to compare a genetic programing model and a multilinear regression modeling to formulate the most precise model for predicting the compression and recompression indices from basic index properties. A total of five laboratory tests namely oedometer consolidation test, specific gravity, natural moisture content, grain size analysis, and Atterberg limit tests have been conducted on sixteen undisturbed and disturbed red clay soil samples of Addis Ababa. To develop a genetic programming model, a software named Eureqa has been utilized and for regression model, a software called SPSS-25 was used. In developing the models, the independent variables considered are liquid limit LL, natural moisture content ${w}_{n}$ and initial void ratio ${e}_{o}$ and the variables to be predicted are compression index C_C and recompression index C_r. The newly developed equations were compared with the existing ones and it was found that both conventional and GP model equations were found to be better than the existing ones. Moreover, the comparison was made between the conventional and GP model equations and it was found that the GP model equation of C_C with ${R}^{2}=0.99$ outperforms the regression model equation of C_C. On the other hand, for recompression index, the GP-based model equation yields the value of C_r with a small variation than the regression model equation.