SGS-based Reliability Assessment of Driven Piles at Solar Farm Sites

Abstract

Utility-scale solar farms require reliable design of tens of thousands of driven steel pile foundations, where wind-induced uplift governs. Spatial variability of the shallow subsurface and limited cone penetration test (CPT) coverage mean pile reliability depends on distance from the nearest test – an uncertainty Eurocode 7 does not capture. This paper proposes a probabilistic methodology combining sequential Gaussian simulation (SGS) of CPT data with three capacity methods (LCPC, AFNOR NF P 94-262 and ICP-05) and FOSM/FORM reliability analysis to quantify uplift reliability along linear Chains connecting adjacent CPT locations. The methodology is applied to a solar farm in Pleistocene sandy deposits with 70 CPTs. Two Chains are analyzed: a 309 m Chain through five CPTs in a uniform zone and a 302 m Chain through four CPTs in a variable zone. Pile reliability is governed by two independent contributions. The first is conditioning geometry: the COV of uplift resistance is smallest at CPT locations and grows with distance, forming a wave-like profile with minima of 1.4–2.7% at CPTs and maxima of 3.2–6.1% at midpoints. The second is geological variability: differences in soil strength shift the mean resistance and dominate the reliability index wherever the deposit is non-uniform. Along Chain 1, the FORM reliability index β varies only moderately, driven almost entirely by conditioning geometry. Along Chain 2, β spans a range over three times wider, reflecting dominant geological heterogeneity. The Chain-based framework provides a site-specific basis for separating these contributions, for optimizing CPT spacing and delineating geological zones.