A risk-based approach for optimizing proof-load test programs for driven piles.

Abstract

There is currently an inconsistency in the recommendations that are available in pile design codes and practices regarding the required number of proof-load tests and the level of the proof loads for piles. Najjar et al. (2017) proposed a pre-posterior decision making framework to allow for selecting the optimal pile load test program that would result in the maximum expected benefit to a project while maintaining a target level of reliability in the pile design at the site. The proposed methodology was based on a robust Bayesian approach that allows for updating the capacity distribution of piles at a site given the results of the proof-load test program. In the proposed methodology, Najjar et al. (2017) adopted a simplified statistical model for the pile capacity, whereby the uncertainty in the pile capacity due to spatial variability in a site was assumed to be known and modeled by a fixed coefficient of variation of 0.2. In addition, the application of the proposed decision-making framework was limited to an illustrative design example that targeted driven steel pipelines in sands. The objectives of this thesis are to (1) extend the statistical model that is proposed by Najjar et al. (2017) for the pile capacity by modeling the uncertainty in the pile capacity at the site (coefficient of variation due to spatial variability) as an uncertain variable that is updated with pile load test results, (2) study the sensitivity of the decision making framework to the parameters describing the uncertainty in the capacity and the load, and (3) apply the pre-posterior decision making framework to a number of practical design scenarios that involve driven piles of different characteristics and soils of different nature.