Reliability Based Design And Analysis Of Deep Foundations At The Service Limit State PDF Download

Deep foundations are necessary for the construction of many structures, such as bridges and buildings, located in areas unsuitable for shallow foundations. Owing to the inherent variability of soil and the complex changes that occur in the soil adjacent to deep foundations as they are installed, the ability to accurately predict axial pile capacity is difficult. As a result of their schedule and perceived cost, site-specific full-scale instrumented pile loading tests are not often performed, rather, empirical or semi-empirical static analyses that require simplifications and indirect consideration of true pile-soil response are often used to estimate pile capacity. Many pile-specific and -nonspecific axial capacity estimation methodologies are available; however, most of them are largely inaccurate. The uncertainty associated with foundation design is well recognized, and has been traditionally addressed using deterministic design procedures and global factors of safety. The shortcomings associated with deterministic design approaches are well-documented, and the use of reliability theory to provide safe and cost-effective design solutions is preferred. However, the transition to reliability-based design (RBD) remains an ongoing process, and several challenges remain. This dissertation uses high quality data to investigate and identify pertinent factors that control reliability. Correlations between design variables that were previously overlooked are identified, and improvements are made in order to provide accurate and unbiased pile design models. Robust statistical models are developed; and guidelines are established for incorporating more realistic assessments of the probability of exceeding two particular limit states relevant for piles under axial loading conditions. First, dynamic formulas for estimating axial pile capacity at the ultimate limit state (ULS) are recalibrated for use within a probabilistic design framework using ordinary least squares regression and a geologic-specific database for a variety of driving conditions; Monte Carlo simulations (MCS) are employed to calibrate resistance factors for use with the new and unbiased dynamic formulas. Accurate and unbiased models for estimating the capacity of auger cast-in-place (ACIP) piles at the ULS are developed since current recommendations were shown to be largely unsuitable. A parametric study was conducted using a first-order reliability method approach in order to identify the parameters and statistical modeling decisions that govern the reliability of ACIP piles at the serviceability limit state (SLS). The shortcomings of existing correlation models are identified, and new design models for ACIP piles at the ULS are incorporated into assessments of reliability at the SLS using more robust copula theory and a MCS approach. Because estimates of reliability using conventional techniques have been shown to be overly conservative, resistance distributions are truncated based on theoretical lower-bound limits, resulting in more cost-effective design solutions.