Modeling Pressuremeter Modulus with Genetic Expression Programming: A Predictive Framework

The pressuremeter test is one of the most reliable in situ methods in geotechnical engineering. It provides key parameters for evaluating bearing capacity and settlement behavior of foundations. This study proposes a gene expression programming (GEP) model to predict the pressuremeter modulus (Ep) based on three easily measurable soil properties, including depth from the ground surface, natural moisture content, and soil unit weight. The developed model achieved strong predictive performance: R2 = 0.9672, normalized RMSE (NRMSE) = 0.0911, and performance index (PI) = 0.0459 for the testing dataset. These measures confirm the high level of reliability of the model. To better understand the role of input variables, sensitivity analyses using both Sobol and SHAP techniques were performed. These analyses consistently identified depth as the most influential factor, followed by moisture content and soil unit weight. A comparative evaluation with similar works in the literature further demonstrated that the proposed GEP model strikes an effective balance between accuracy, dataset adequacy, and model simplicity. These results highlight the model’s potential as a practical and reliable tool for preliminary geotechnical design.