Continuous Predictive Modeling of Geotechnical Parameters: A Probabilistic Alternative to Discrete Characteristic Value Estimation

While geotechnical parameter determination is fundamental to foundation engineering, traditional approaches often suffer from data fragmentation and subjective safety assessments. This research introduces an integrated framework that synthesizes multivariable regression with the Effective Random Dimension (ERD) method to bridge the gap between raw laboratory indices and structural design. By analyzing datasets from the stable Suceava Moldavian Platform (68 samples) and the tectonized Subcarpathian Flysch (50 samples), the study demonstrates that granulometric fractions, moisture content, and carbonate content can predict consistency limits with high statistical fidelity, achieving R2 = 0.98 for the Liquid Limit at Suceava and R2 ≈ 0.90 for the Plasticity Index at Doftana. The novelty of the approach lies in the generation of continuous vertical profiles transformed into code-compliant characteristic values (Xk) via Taylor series linearization and the ERD framework. The derived characteristic interval for the Plasticity Index (58.66–70.15%) quantitatively demonstrates the reduction in hyper-conservative bias compared with discrete sampling. This methodology eliminates subjective judgment and ensures a mathematically rigorous transition to Eurocode 7 and NP 122:2010 standards, optimizing both safety and economic efficiency in variable geological strata.