GEP-based empirical models for estimation of soil nail bond strength in weathered soil

This study presents two newly developed predictive models for calculation of gravity grouted soil nail pullout bond strength (q) in Completely Decomposed Granite (CDG) and Volcanic (CDV) soils classified as coarse and fine grained soils respectively. Gene expression programming (GEP) is utilized, employing measured field data gathered from the literature. The field pullout test database comprises a total of 521 tests, with 363 tests conducted in CDG and 158 tests in CDV soils. The GEP models take into account the impact of various factors including nail configurations (diameter (D), length (L)), installation angle of nail (i)), overburden stress (sigma) and parameters related to soil strength (cohesion (c) and friction angle (phi)). Comparisons with existing design equations in engineering practice provide evidence for the reliability of the proposed models. The proposed GEP model shows superior performance with correlation coefficients (R) of 0.83 for CDV and 0.75 for CDG soils, and RMSE values of 73 and 120, respectively, compared to significantly higher RMSE values of previous models. Additionally, a novel parametric study is conducted for first time to examine the influence of input parameters and assess their respective significance on pullout bond strength. The proposed GEP-based models offer an efficient alternative to labor-intensive field pullout tests, contributing to sustainability in geotechnical engineering. Overall, the findings endorse the adoption of these GEP models for improved estimation of gravity grouted soil nail bond strength.