Data-driven optimization and pressuremeter modulus prediction using response surface methodology for smarter geotechnical design

Accurate evaluation of soil mechanical behavior remains a cornerstone of reliable geotechnical engineering and optimized foundation design. Among the key parameters that define soil deformability, the pressuremeter modulus (Ep) plays a crucial role in assessing the ground response and structural stability. However, conventional approaches for determining Ep through empirical correlations or in situ measurements are often labor intensive, costly, and subject to significant uncertainty. These limitations underscore the necessity for advanced predictive frameworks that can provide both accuracy and efficiency in geotechnical characterization. In the present work, response surface methodology (RSM) is applied as an advanced statistical modeling and optimization technique to improve the predictive capability of the pressuremeter modulus. The experimental investigation is structured via a central composite design (L35) framework, enabling the systematic evaluation of nonlinear interactions among the principal geotechnical variables, namely, depth (D), cohesion (C), internal friction angle (φ), and unit weight (γ). The developed model was rigorously validated through analysis of variance, which revealed a high level of statistical significance with an excellent coefficient of determination (R2 = 96.50%). The desirability function approach is subsequently applied to determine the optimal combination of soil parameters that maximize Ep, providing valuable guidance for design and site investigation strategies. The findings highlight the transformative potential of data-driven statistical modeling in refining soil property estimation and advancing geotechnical design practices. By demonstrating the robustness and predictive capacity of the RSM, this research establishes a methodological foundation for intelligent, efficient, and cost-effective geotechnical engineering applications.