Site Variability Assessment and LRFD Resistance Factor Calibration Using Artificial CPT Data Incorporating Historical and Site-Specific Variation

Accurate calibration of load and resistance factor design (LRFD) for deep foundations depends on the reliable characterization of both the design method variability and site-specific spatial variation in soil properties. This study presents a comprehensive framework that uses an interpolation-based cone penetration test (CPT) enrichment to improve characterization of horizontal spatial variability in sparse CPT data through inverse distance weighting (IDW), empirical Bayesian kriging (EBK), and Bayesian neural networks (BNN). Four variability metrics, including novel direct pairwise distance analysis and depth declustered random rectangle, are adopted to quantify soil heterogeneity without reliance on parametric variogram models. The combined variability framework integrates method and site variation to calibrate LRFD resistance factors across 10 Louisiana sites with varying geotechnical conditions. The results demonstrate that the interpolation-based CPT enrichment improves resolution of horizontal spatial variability, with BNN capturing fine-scale heterogeneity more effectively than IDW and EBK. Incorporating these enriched datasets into the spatial-variability analysis refines the estimated site-consistent variability used in LRFD calibration, supporting risk-consistent, efficient foundation design in data-scarce environments.