Stability of Embankments Founded on Soft Soil Improved with Deep-Mixing-Method Columns

Abstract

Foundations constructed by the deep mixing method have been used to successfully support embankments, structures, and excavations in Japan, Scandinavia, the U.S., and other countries. The current state of practice is that design is based on deterministic analyses of settlement and stability, even though deep mixed materials are highly variable. Conservative deterministic design procedures have evolved to limit failures. Disadvantages of this approach include (1) designs with an unknown degree of conservatism and (2) contract administration problems resulting from unrealistic specifications for deep mixed materials.

This dissertation describes research conducted to develop reliability-based design procedures for foundations constructed using the deep mixing method. The emphasis of the research and the included examples are for embankment support applications, but the principles are applicable to foundations constructed for other purposes.

Reliability analyses for foundations created by the deep mixing method are described and illustrated using an example embankment. The deterministic stability analyses for the example embankment were performed using two methods: limit equilibrium analyses and numerical stress-strain analyses. An important finding from the research is that both numerical analyses and reliability analyses are needed to properly design embankments supported on deep mixed columns. Numerical analyses are necessary to address failure modes, such as column bending and tilting, that are not addressed by limit equilibrium analyses, which only cover composite shearing. Reliability analyses are necessary to address the impacts of variability of the deep mixed materials and other system components.

Reliability analyses also provide a rational basis for establishing statistical specifications for deep mixed materials. Such specifications will simplify administration of construction contracts and reduce claims while still providing assurance that the design intent is satisfied.

It is recommended that reliability-based design and statistically-based specifications be implemented in practice now.