Characterization of Model Uncertainty in the Static Pile Design Formula

Abstract

Level 1 reliability methods have been internationally accepted as the basis for development of the new generation of geotechnical design codes. A key requirement of this design approach is the identification and quantification of uncertainties associated with the geotechnical design under consideration. This paper presents four load test databases from South Africa for driven piles in noncohesive soils (29 tests), bored piles in noncohesive soils (33 tests), driven piles in cohesive soils (59 tests), and bored piles in cohesive soils (53 tests). The capacity model factor is defined as the ratio of the interpreted capacity (Chin-Davisson approach) and the predicted capacity (static pile design formula). The uncertainty in the capacity model factor is modeled as a lognormal random variable. The model factor statistics reported in this study are required for reliability-based ultimate limit state design. The uncertainty in the load-settlement behavior is characterized by fitting measured load-settlement data to a hyperbolic equation and then normalizing the hyperbolic curve with the interpreted capacity. The resulting exercise reduces uncertainties in a set of nonlinear continuous curves to uncertainties in two hyperbolic curve-fitting parameters. This approach is practical and grounded realistically on the load test database with minimal assumptions. The hyperbolic parameter statistics reported in this study are required for reliability-based serviceability limit state design.