Theory Manual for the Load Displacement Compatibility Method (LDC) for Design of Column-Supported Embankments: A Companion to GeogridBridge 3.0

Column-supported embankments (CSEs) are a method of ground improvement commonly used when an embankment is required over soft, compressive soil that is likely to experience expansive settlements because of the added load in the absence of soil improvement. CSEs are an alternative to other methods to mitigate settlement in compressible soils such as the use of prefabricated vertical drains and surcharge fills. Because a CSE can greatly reduce the ultimate consolidation settlement, the embankment can be constructed immediately and there is less potential for damage to nearby facilities. CSEs are often constructed with a square, rectangular, or triangular column arrangement. In general, column-supported embankments should be designed and constructed with a total embankment height greater than the critical height. Embankments with a total height at least as tall as the critical height are expected to experience uniform surface settlement even though there will be differential settlement at the base.

The load-displacement compatibility (LDC) method considers three different load transfer mechanisms simultaneously: soil arching in the embankment, vertical load transfer owing to tension developed in geosynthetic reinforcement, and negative skin friction that results in a drag force. Consistent loads and deformations are considered in the calculations for each load transfer mechanism. The reduction in vertical stress caused by the load transfer mechanisms in the LDC method is expressed as a normalized parameter called the stress reduction ratio. Differences in the mobilization of load transfer mechanisms in the time between the short term (i.e. end of embankment construction) and the long-term conditions are not considered in the estimation of the settlement and the time rate of consolidation. Negative skin friction is the load transfer mechanism in which load is distributed from the foundation soil to the columns when soil moves down relative to the columns above the neutral plane.

The load-displacement compatibility (LDC) method as implemented in GeoGridBridge3 (GGB3) calculates the compression of clays, sands, and columns above the neutral plane relative to the neutral plane. Compression of the embankment fill and of foundation soils below the neutral plane are not considered in GGB3. For the consolidation of clays, GGB3 uses traditional one-dimensional consolidation methods. The magnitude at which the clay consolidates is dependent on the initial vertical effective stress, the final vertical effective stress, and the preconsolidation pressure of the clay. The developers of GGB3 adopted a one-dimensional elastic analysis of compression for sand in lieu of the two-dimensional elastic analysis of compression applied in GGB2 based on Poulos and Davis elastic solutions for concentric elastic cylinders. This two-dimensional relationship was chosen to account for radial bulging of columns with a relatively low stiffness such as stone columns, which were thought to be an attractive option for CSEs when Geogrid Bridge was first developed.

To address soil parameter uncertainty and variability, GeoGridBridge3 (GGB3) includes a probabilistic analysis capability that utilizes a Monte Carlo simulation (MCS). The MCS is used to model the probability of an outcome based on the uncertainty of any number of variables. Another advantage of an MCS is that it enables the direct analysis of parameter distributions. Because soil parameters are subject to high levels of uncertainty and variability, an MCS is a common tool for reliability analyses in geotechnical engineering. Because of the complex nature of the LDC methodology, the MCS approach to evaluating the probability of an undesired condition is illustrated using the example of the short-term stability of a vertical cut in saturated clay in lieu of a CSE design. An important note is that the smallest probability of failure observable by an MCS depends on the number of iterations performed.