Drilled Shaft Load Tests in Atlanta: Back-Analysis and Calibration with DeepFND
Updated: Oct 2
What Full-Scale Load Testing Can Teach Us About Deep Foundation Design
Predicting the axial response of drilled shafts requires more than estimating an ultimate capacity. For many foundation problems, engineers also need to understand how load is transferred between the shaft and surrounding ground, how much settlement develops as resistance is mobilized, and how sensitive the predicted response is to the selected soil and rock parameters.
Full-scale load tests provide an invaluable opportunity to evaluate these assumptions against measured behavior.
A well-documented load testing program conducted in Atlanta, Georgia, provides an excellent case study. Using the published test results, we performed a series of back-analyses in DeepFND to investigate how different parameter-selection and interpretation methods influence the predicted axial load-displacement response.
The Atlanta Drilled Shaft Load Test Program
In 1993, Georgia Tech Research Corporation prepared the research report Axial Load-Displacement Behavior of Drilled Shaft Foundations in Piedmont Residuum for the Federal Highway Administration (FHWA). The research program, conducted in association with ADSC and ASCE, investigated the axial behavior of drilled shafts installed in the Piedmont geologic profile.
Two approximately 30-inch-diameter drilled shafts were load tested:
A 55-ft-deep floating shaft, founded primarily within the Piedmont residuum.
A deeper shaft extending approximately 72 ft to bedrock, allowing the influence of end bearing and the underlying rock to be examined.
The site was extensively characterized using laboratory and in-situ testing. The generalized subsurface profile consisted of approximately 15 ft of fill, followed by Piedmont residuum—predominantly silty sand (SM)—to approximately 60 ft, partially weathered rock extending to about 72 ft, and Granitic Gneiss below. Groundwater was reported at approximately 56 to 62 ft below grade.
This combination of detailed site characterization and full-scale foundation testing makes the study particularly useful for evaluating analytical methods.
Why Back-Analysis Matters
Foundation design parameters are commonly obtained from laboratory testing, in-situ testing such as the SPT, empirical correlations, published recommendations, and engineering judgment.
However, two parameter sets that appear reasonable independently may not necessarily produce the same foundation response.
For drilled shafts, this becomes especially important because the calculated load-displacement behavior depends not only on the ultimate side and base resistance, but also on the stiffness and mobilization of those resistance components.
The Atlanta tests therefore provide an opportunity to ask a practical engineering question:
How well can commonly used design procedures reproduce measured full-scale drilled shaft behavior?
Using DeepFND, we performed several back-analyses of the published load tests while incorporating FHWA GEC-10 recommendations where applicable.
Case A – Floating Shaft Using Laboratory-Derived Soil Properties
The first analysis examined the 55-ft floating shaft using strength parameters obtained from laboratory testing of the Piedmont residuum.
A friction angle of approximately 35.5° produced a close match to the measured load-test response.
This compares well with the approximately 35.8° friction angle indicated by the reported geotechnical investigation.
An important observation from the back-analysis, however, was that matching the soil strength alone was not sufficient.
The assumed stiffness of the soil springs mobilizing side resistance also had a significant influence on the calculated load-displacement curve. Two analyses can therefore predict similar ultimate capacities while producing noticeably different settlement behavior.
This distinction is important in practice: capacity and stiffness are related design considerations, but they are not interchangeable.
Case B – Shaft Bearing on Piedmont Gneiss
The second analysis considered the deeper drilled shaft extending to the Piedmont Gneiss.
For this case, the rock modulus of elasticity was calibrated to reproduce the observed load-displacement response.
The exercise demonstrates an important aspect of drilled shaft analysis in soil-rock profiles. Once significant base resistance develops, the predicted response can become highly sensitive to assumptions regarding the deformation characteristics of the bearing material.
A capacity calculation alone may indicate that sufficient resistance is available. A load-displacement analysis goes further by examining how that resistance is mobilized and the movement required to mobilize it.
This is particularly relevant for foundations bearing on or socketed into weathered and competent rock, where uncertainty in rock-mass stiffness can significantly affect predicted settlement.
Case C – Estimating Soil Parameters from SPT Data
The third analysis returned to the 55-ft floating shaft, but instead of relying primarily on laboratory-derived strength parameters, soil properties were estimated from Standard Penetration Test (SPT) results.
Several approaches for estimating the effective friction angle of the Piedmont residuum were evaluated.
For this particular case, our analyses indicated that the Triaxial Compression correlation for estimating friction angle produced a response that more closely reproduced the measured behavior than the Kulhawy-Chen approach incorporated within the evaluated FHWA procedures.
The comparison also produced a notable difference in predicted end-bearing resistance. For this specific analysis, the Kulhawy-Chen approach resulted in an axial capacity approximately 15% greater than the alternative evaluated method.
This result should not be interpreted as a universal preference for one correlation over another. Rather, it illustrates a broader engineering lesson:
Empirical correlations should be evaluated in the context of the local geology, available site investigation data, foundation type, and expected load-transfer mechanism.
A correlation that performs well for one geologic environment may not provide the same level of agreement in another.
Engineering Lessons from the Load Tests
The Atlanta drilled shaft tests highlight several considerations that are directly applicable to deep foundation design.
First, ultimate axial capacity is only part of the problem. A realistic analysis should also consider the load-displacement response and the relative mobilization of shaft and base resistance.
Second, soil and rock stiffness matter. Strength parameters may control calculated resistance, but stiffness assumptions strongly influence the predicted displacement required to mobilize that resistance.
Third, SPT-based correlations should not be treated as exact soil properties. They are empirical tools, and different correlations can lead to materially different design predictions.
Finally, full-scale load tests are valuable calibration tools. When test data are available, back-analysis can help engineers evaluate whether the assumptions used in their analytical model reasonably reproduce actual foundation behavior.
From Measured Behavior to Better Foundation Models
The purpose of back-analysis is not simply to force a numerical model to reproduce a test curve. Its greater value is in identifying which assumptions control the predicted response and whether those assumptions are consistent with the available geotechnical information.
In the Atlanta case study, the measured drilled shaft behavior provides a benchmark for examining laboratory-derived parameters, SPT correlations, soil-spring stiffness, and rock deformation properties within the same engineering framework.
DeepFND allows engineers to perform these comparisons systematically, evaluate alternative parameter sets, and examine both capacity and load-displacement behavior.
Ultimately, reliable deep foundation analysis depends on the combination of sound site characterization, appropriate analytical methods, realistic parameter selection, and engineering judgment. Full-scale load testing—and thoughtful back-analysis of those results—provides one of the most direct ways to connect these elements and improve confidence in foundation design.
Let us show you how to reduce your design time by up to 90%!
