Expected Design Factor of Safety from a Pile Load Test Program

Abstract

Load tests are often carried out to verify design assumptions and to reduce the design factor of safety (FOS). In this paper, a method based on Bayesian preposterior analysis is suggested to evaluate the expected design FOS when a load test program is planned but not yet implemented. Two algorithms, i.e., Monte Carlo simulation (MCS) and a mean first order method (MFOM) are employed to calculate the expected design FOS. The accuracy of Monte Carlo simulation can be sufficiently accurate when enough samples are drawn. MFOM slightly underestimates the expected design FOS and it becomes more accurate as the number of piles in a test program increases. Results in this study reveal that the inherent variability within a site controls the amount of uncertainty that can be reduced by pile load tests, and that a few tests are often sufficient to remove the majority of uncertainty caused by cross-site variability. If a given number of piles are to be tested, the expected design FOS increases almost linearly with the target reliability index.