Database Assessment of Calculation Model Uncertainties for Axial Resistance and Displacement of Driven Cast-in-Situ Piles

Driven cast-in-situ (DCIS) piles (e.g., Franki, Simplex, and Vibro) are classified as displacement piles and their interaction with the surrounding soil is often analyzed using the methods originally developed for preformed displacement piles. A database of 507 DCIS pile load tests predominately in cohesionless soils was compiled in this paper, aimed at evaluating the calculation model uncertainties of twelve methods for axial resistance and five load transfer models for load-displacement response analyses. For axial resistance, the model uncertainties associated with these methods are characterized as the mean (bias λ) and coefficient of variation (scatter θ) of the ratio of the load defined by the Davisson offset limit to the prediction. It was observed that (1) the API method considerably underestimates the capacity of Franki piles (cased: λ=1.51, θ=0.54 and uncased: λ=1.97, θ=0.55); (2) the API method is more applicable for Simplex and Vibro piles (λ=1.22, θ=0.34), while the predictions by five recent CPT methods (λ=0.81–1.01, θ=0.36–0.65) that account for several influential factors on the behavior of displacement piles in sand are slightly better than six simplified α−CPT methods (λ=0.69–1.19, θ=0.37–0.56); and (3) the predicted load-displacement curves diverge greatly from the load test data with increasing displacements. The values of λ and θ for capacity prediction are used to calibrate the resistance factors for the Strength I limit state in the AASHTO LRFD bridge design specifications.