Comparison of Analysis Methods against Measured Performance for the Courthouse Station Deep Excavation in Boston
Construction of the Silverline Courthouse Station in South Boston required an 18m deep excavation supported by an internally braced 27m long reinforced concrete diaphragm wall.
The performance of the excavation system was reported initially by Corral (2013), and by Whittle, Corral, Jen, and Rawnsley.
The excavation was underlain by 24m of normally and lightly overconsolidated Boston Blue Clay. Because of the small basal stability safety factors, notable wall displacements were expected and observed at the wall base.
This article compares analysis results between finite element, limit equilibrium, and non-linear Winkler spring models with the DeepEX software.
In the current FEM modelling the Hardening Soil Strain model was used vs. the MIT E3 in the original publication.
Figure 1 compares measured displacements against two finite element models. The full model captured the external surcharges from a nearby structure and included layer variations while the other model assumed simplified ground conditions.
Both simulations reasonably matched observed wall displacements, but most importantly, this effort highlighted that when basal stability is an issue, small modelling differences can have a significant impact on the produced deformations.
Figures 2 & 3 present lateral deformation results from the non-linear analysis within DeepEX with Winkler Springs.
While the full analysis model predicts much larger displacements, the simplified model produces reasonably closer estimates to the measured and original design estimates.
Producing these results though required that a high adhesion value was assumed in the analysis in the order of 90% of the undrained clay strength.
Figure 4 presents limit-equilibrium results for the last construction stage with FHWA apparent earth pressures and the CALTRANS shoring and trenching manual beam approach.
Because the wall does not achieve moment equilibrium below the excavation, the base is essentially free.
As a result, large moments and support reactions are produced.
Figure 5 compares reported maximum support reactions against results from original and current analyses.
FEM and NL methods reasonably matched measured support reactions. Support reactions from the current DeepEX finite element models were also in good agreement with the Class A & C predictions from the original publications.
In the NL analysis the simplified clay model approach produced support reactions closer for the 3rd bracing level and higher reactions on the 5th strut, whereas the exact theoretical model closer matched the 5th level loads but underpredicted the 3rd level bracing forces. LEM support reactions started diverging significantly once the wall base became free in the model.
Concluding, we could summarize the following points from this exercise:
a. Measured data taught us that bracing levels should be designed with extra caution and perhaps extra capacity if one was to rely solely on Class A predictions.
b. Estimated displacements when the basal stability safety factor is approaching unity can be highly sensitive to small modelling changes.
c. FHWA limit equilibrium methods will not be able to properly capture wall bending, and support reactions when the wall base becomes free to move.
Figure 1: Comparison of reported lateral wall movements against new FEM models
Figure 2: Non-linear analysis results for simplified model
Figure 3: Non-linear analysis results for complete model
Figure 4: LEM results for last excavation stage
Figure 5: Comparison of Maximum Support Reactions vs. Measured Data
References:
Whittle A. J., Corral G., Jen L. C., Rawnsley R. P., “Prediction and Performance of Deep Excavations for Courthouse Station, Boston”, MIT Open Access Articles
Corral, G. (2013). “Methodology for Updating Numerical Predictions of Excavation Performance.” Sc.D. Thesis, Department of Civil and Environmental Engineering, Cambridge, MIT. 515p.
