Data-driven identification of swell potential of clayey soils for engineering surveying using genetic projection pursuit

Accurately classifying swell potential grades of clayey soils remains a significant challenge in geotechnical engineering, primarily due to the complex interplay of multiple factors that govern swelling-shrinkage behavior. To address this, this study develops a hybrid intelligent model by integrating a Genetic Algorithm (GA) with Projection Pursuit (PP). The GA optimizes the critical projection direction within the PP framework, transforming multi-dimensional soil indicator data into representative one-dimensional projection values. A logistic curve function is then employed to establish the nonlinear mapping between these optimized projection values and the corresponding empirical swelling-shrinkage grades, forming a GA-optimized Projection Pursuit (GAPP) discriminant model. Five key soil parameters-liquid limit, plasticity index, natural water content, free swell rate, and total swelling-shrinkage rate-were selected as the input indicators based on their mechanistic relevance. Model validation confirms its high predictive accuracy. Furthermore, application to a real-world highway engineering case demonstrates that the model’s predictions align closely with observed field conditions, effectively verifying its practical feasibility and reliability. The proposed GAPP model thus provides a robust and effective methodological tool for the quantitative identification and grading of expansive soil swelling-shrinkage potential.