Within-site variability and reliability of foundation designs based on load tests

Probabilistic evaluations of the reliability of foundation designs based on site-specific load test information are somewhat limited in number. Published evaluations have generally relied on Bayesian techniques. A primary input for Bayesian analysis is within-site variability, which describes the variability of foundation resistance across a site. Within-site variability is attributed to geologic variation across a site and to differences in construction outcomes among foundation elements. Published evaluations have generally used a deterministic value of within-site variability wherein within-site variability is treated as a known parameter and is not subject to updating based on load test results. In contrast, probabilistic within-site variability treats within-site variability as an uncertain parameter with its own probability distribution that is updated based on load test results. Probabilistic within-site variability has not been applied commonly. This research examines differences in reliability outcomes between deterministic and probabilistic within-site variability. Analysis of micropile load test results from five different sites was used to develop a distribution of within-site variability. The resulting distribution is relatively variable (i.e. the value of within-site variability is, itself, variable), which demonstrates that there is, in fact, considerable variability and uncertainty in the value of within-site variability.