You are on page 1of 7

Geotechnics of Soft Soils Focus on Ground Improvement Karstunen & Leoni (eds) 2009 Taylor & Francis Group,

up, London, ISBN 978-0-415-47591-4

Numerical modelling of stone columns in soft clay under an embankment


T.M. Weber & S.M. Springman
Institute for Geotechnical Engineering, ETH Zrich, Switzerland

M. Gb, V. Racansky, H.F. Schweiger


Graz University of Technology, Graz, Austria

ABSTRACT: A series of numerical calculations have been performed to investigate the behaviour of floating stone columns in soft clay under an embankment at working load level. Three different approaches are studied using finite element calculations with 2D and 3D idealisations. Challenges in numerical modelling of stone columns include estimation of the effect of smear around stone columns, and consideration of construction effects due to stone column installation. Conversion of the 2D numerical model from a stone column grid into a 2D stone trench structure, demands that an adjustment is made to the soil parameters within the column grid. One objective is to establish whether the response of an embankment on stone columns can be modelled in 2D with less computational effort compared to a full 3D analysis. The results from geotechnical centrifuge tests of an embankment on soft clay improved with floating stone columns provide the basis for this numerical comparison. The main characteristics of the embankment deformation and the development of pore water pressures within the column grid give an indication about the quality of the numerical models investigated.

INTRODUCTION

Construction of stone columns in soft clay under an embankment is a common economical ground improvement method when shear strength increase, settlement reduction and acceleration of consolidation are needed. The prediction of embankment settlements is a challenging task, since the behaviour is governed by complicated interaction between soil, stone columns and structure. Simplification is often needed, particularly when using 2D numerical analysis. However, assumptions have to be made for more sophisticated 3D numerical models to reproduce the stress history and realistic system response. The installation of stone columns changes the stress state and the structure in the ground dramatically. Initially, similar processes of column installation were investigated numerically for driven piles and Cone Penetration Tests (CPT, e.g. Desai 1978, De Borst & Vermeer 1984). Later, the installation of sand compaction piles and stone columns was investigated for example by Asaoka et al. (1994), Debats et al. (2003), Kirsch (2004) and Farias et al. (2005). It could be shown that the horizontal stresses increase to values above the vertical stresses, which remain approximately constant.

Representing a 3D stone column grid in 2D is usually represented by a series of parallel trenches. The stiffness as well as the permeability of both soft soil and coarse grained inclusion needs to be adapted in order to model the deformation behaviour and drainage conditions for consolidation correctly. Hird et al. (1992) and Indraratna & Redana (1997, 2000) recommend how to perform a conversion of permeability. These transformations are also applicable to smear effects, which need to be considered since the drainage conditions and consolidation behaviour will be significantly affected by smear. Mestat et al. (2006) and Wehr & Herle (2006) showed in a prediction and comparison exercise that an accurate class A prediction of embankment settlement is difficult to achieve. Results from 17 different contributions using analytical and finite element methods showed deviations up to 300% against prior measurements from a test embankment. It is not necessarily the case that a complex 3D finite element calculation leads to more precise results than a simplified analytical approach. Uncertainties in modelling the interaction behaviour and in choosing appropriate soil parameters were very high. Schweiger & Gb (2006) give an overview about issues in modelling of stone columns and simplifications, when a full 3D analysis is not performed.

305

Improving prediction from finite element models, either in 2D or 3D, will lead to better understanding of the key mechanisms and their governing parameters in terms of the deformation behaviour of stone columns. This paper offers basic references to the finite element modelling of stone columns in soft clay under an embankment load, via 2D and 3D analyses. Two types of numerical models are compared with each other and to a centrifuge model test (Weber et al. 2006, Weber 2008). The stone columns were excavated, filled and compacted in soft clay in flight in the drum centrifuge under 50 times gravity using a special installation tool (Weber et al. 2005). Then the embankment was constructed in flight. Time dependent settlement as well as development and dissipation of excess pore water pressure in the soil were measured and provided physical data for validation of the numerical models and calibration of parameters when reproducing the behaviour of an embankment on clay improved by stone columns. 2 2.1 PLANE STRAIN FINITE ELEMENT MODEL Geometry

A smear zone is introduced around the stone columns in order to take the effects of column installation on the drainage situation into account. The extension of this zone is estimated to extend to double stone column diameter (Weber 2008) and a geosynthetic base reinforcement is implemented into the model as a geogrid with interface elements. 2.2 Model parameters

Due to the change from stone columns to stone trenches and keeping the column diameter as the trench width, the area ratio of ground improvement is considerably increased from 10% in the prototype situation to 35% in the 2D model. In order to maintain consistency between prototype and model, the equivalent vertical stiffness of the column material should be equal, as shown in Equation 1. ( EA )3D = ( EA )2 D (1)

The 2D plane strain finite element calculation is performed using the program PLAXIS V8 (Brinkgreve 2002) with 15 node elements. The Hardening-Soil Model is used for the clay and the Mohr-Coulomb Model is applied for the embankment and the stone column material. A section of the FE-model is shown in Figure 1 together with nodes where pore pressures were logged. All geometrical dimensions of the numerical model resemble the centrifuge model at prototype scale. The width of stone trenches in the 2D model is chosen to be the same as the diameter of the stone columns in the prototype situation with 0.6 m. The distance between the stone column axes and the stone trenches are kept constant at 1.7 m. In plan view, this results in a higher area ratio of ground improvement and different drainage conditions, which need to be considered in the determination of other constitutive soil parameters.

Consequently, the stiffness of the columns and the clay has to be adapted in the 2D numerical model by a factor of 3.5, representing the area of the trenches compared to the columns. The 1D modulus of the column material was determined from oedometer tests to be 14000 kN/m2 between vertical effective stresses of 200 to 400 kN/m2 although equivalent field values will be affected by compaction of the column material, which changes the stress field and the density of the material. Some experience is required to identify representative values. Similar amendments are applicable to the stiffness of the clay, derived from oedometer tests as a Youngs modulus of 1000 kN/m2. The clay stiffness has to be increased by a factor of 1.4, corresponding to the lower clay area, from 90% to 65%. The drainage condition also needs to be adapted from a prototype to the 2D situation. The consolidation behaviour should be similar in both systemsthe radial drainage system and the plane drainage system, so that the degree of consolidation at any time should be equal, Equation 2. U r = U pl (2)

Figure 1.

Finite element mesh of the plane strain model.

Figure 2 shows the change from preferred radial drainage towards the stone column to a linear drainage path towards the stone trenches in a 2D model (Hird et al., 1992). In this prototype case, the maximum drainage path reduces from 1.8 m to 1.1 m by a factor of about 1.6 and the surface area of the stone trenches (2.0 m2/m/m) are a factor of 1.8 bigger than the surface area of each stone column (1.1 m2/m/m). The permeability must be adapted to represent the modified drainage conditions. Hird et al. (1992)

306

Figure 3. Effective horizontal stresses after simulation of column installation, with a maximum value of 80.4 kPa.
0 -1

before column installation after column installation

Figure 2. Conversion from radial drainage to plane strain conditions (Hird et al. 1992).
depth [m]

-2 -3 -4 -5 -6 -7 0 0.5 1 1.5 2 2.5 3

derived Equation 3 for conversion of the horizontal permeability of the prototype kh into the horizontal permeability of a plane strain model khp, when the width of the trench, 2bw and spacing, 2B is equal to column radius rw and spacing, D respectively. k hp = 2 1 kh 3 R k h rs ln ln + rs k h ,s rw 3 4

earth pressure coefficient K [-]

(3)
Figure 4. Coefficient of earth pressure in the soft clay, before and after simulation of column installation.

Indraratna & Redana (1997, 2000) extended this relationship to estimate the plane strain permeability of the smear zone khp,s, Equation 4 to 6. k hp,s k hp = k hp kh R k h rs ln ln + rs k h ,s rw 3 4 (4)

2.3

Calculation steps

2 2bs bs b2 s2 1 3 B B 3B b 1 ( bs b w )2 + s 3 (3b w 2 bs 2 ) B2 3B

(5)

(6)

It is important to make a good estimation of the horizontal permeability of the soil kh and the horizontal permeability of the smear zone kh,s in the prototype situation. This ratio is assumed to be 2, because reconstituted clay was used in the centrifuge models. This can increase substantially in varved clays. khp can be calculated from Equation 3, and khp,s can be estimated from Equation 4.

The stress history of the soil is modelled according to that imposed in the centrifuge experiment. Firstly, stepwise consolidation takes place under 100 kPa loading, then the model consolidates under increased gravity conditions in the centrifuge. The stone column installation is modelled with a volumetric increase of the stone trenches in order to introduce a higher horizontal stress in the ground, illustrated by Figure 3 (horizontal effective stresses) and Figure 4 (change of earth pressure coefficient K along axis A in Figure 3 within the stone column grid). Consequently, the initial K (at rest) is 2.5 near the surface, reducing with depth, and increases following simulation of column installation. The subsequent calculation step models placement of the geosynthetic base reinforcement. After that the embankment is constructed in one stage, followed by a consolidation phase. 3 3D FINITE ELEMENT MODEL

3D-modelling has been done using PLAXIS 3D Foundation Version 2.1 (Brinkgreve 2007). First, the

307

Figure 6. Direction of total displacement vectors on the deformed mesh after consolidation, maximum displacement: 379 mm. Figure 5. 3D-model of embankment with columns, with coordinate system, x: longitudinal, y: vertical, z: transverse.

2D-model was simply transformed into a 3D-model using also trenches instead of real columns, i.e. a plane strain model was created in 3D. This has been done in order to have a basis for comparison because some different assumptions in 2D and 3D had to be made because of restrictions of the code, e.g. the updated mesh option is not available in the 3D code as yet. Therefore a 2D analysis without updated mesh was performed in addition. The second model resembled a representative row of columns with their actual geometry (Figure 5). The calculation procedure follows the one described for the 2D-model. No adjustment to the soil parameters were made, so the differences in results from 2D and the 3D trench model are due to differences in discretisation (element type and coarseness of mesh). For the 3D model representing one row of columns, the same parameters are taken, as for the 2D analyses, without the geometry adjustment. As the embankment is now supported by a grid of stone columns, the bearing capacity is reduced and therefore higher deformations can be expected for higher load levels. Smear effects have been ignored in these analyses. 4 RESULTS OF NUMERICAL ANALYSIS AND COMPARISON WITH THE CENTRIFUGE MODEL TEST Results of plane strain FE-Model

Figure 7. Principal effective stresses (direction and magnitude) after consolidation, maximum value: 411.9 kPa.

Figure 8. Effective vertical stresses after consolidation, maximum value: 330.2 kPa.

4.1

The following Figures 6 to 13 show results from the 2D plane strain analysis. Figure 6 illustrates the total displacement vectors after consolidation was completed. The embankment settles mainly vertically with some rotation around the embankment toe and with slight heave in front of the embankment. Figure 7 shows the principal effective stresses after embankment construction in the consolidated state.

Together with Figure 8, showing the contours of vertical effective stress, it can be seen that the embankment load is transferred by the stone trenches to deeper soil layers and the top soil layer experiences less vertical stresses than the base layer of the embankment. This is exacerbated by arching over the stone trenches within the embankment, which leads to further stress concentration in the stone trenches. The distribution of shear stress is shown in Figure 9. Maximum shear stresses develop under the slope within the embankment over the stone trenches and at their base. Figure 10 shows the development of excess pore water pressure in the clay right after embankment construction. The load transfer of the embankment is visible because the maximum excess pore water pressure appears under the toe of the stone trenches.

308

70

excess pore water pressure [kPa]

60 50 40 30 20 10 0 -10 0 50

PPT1.25m-centrifuge PPT3.5 m-centrifuge PPT6.0 m-centrifuge PPT1.25 m - FEM PPT 3.5 m - FEM PPT 6.0 m - FEM

Figure 9. Shear stresses after consolidation, maximum value: 90.6 kPa.

100 150 200 250 300 350 400 450 500

time [d]

Figure 11. Development of excess pore water pressure underneath the embankment (Figure 1)comparison between numerical and centrifuge modelling.
0 0.05 0.1

centrifuge FEM

settlement [m]

0.15 0.2 0.25 0.3

Figure 10. Excess pore water pressure directly after embankment construction, maximum value: 84.9 kPa.

Figures 11 to 13 show the comparison between the numerical calculation and measurements from the centrifuge model test. A match is obtained with little adjustment of stiffness and permeability of the clay for the magnitude as well as curvature of the excess pore water pressure underneath the embankment (Figure 11) and the settlement of the embankment crest (Figure 12). The measurement points of excess pore water pressure are marked in Figure 1. Figure 13 shows, in a superimposed perspective, the deformation of the numerical model and the deformed soil model after centrifuge testing. Due to the test set up in the drum centrifuge, the embankment falls off the soil model after the centrifuge has been stopped at the end of the test. 4.2 Results of 3D FE-Model

0.35 0.4 0 50 100 150 200 250 300 350 400 450 500

time [d]

Figure 12. Development of settlement at embankment crestcomparison between numerical and centrifuge model.

Figure 14 compares the settlement trough of both 2D and 3D calculations at the end of a fully drained analysis and the original construction process (21 days of construction). The curve 2D, 21 days, Upd mesh is the reference solution as discussed in section 3. The following can be observed: a. the updated mesh option reduces final settlements, b. the fully drained analysis has the same effect, c. application of the full embankment under undrained conditions (instead of a 21 days construction

Figure 13. settlement trough and column deformation of numerical and centrifuge model.

period) with subsequent consolidation increases calculated settlements, d. 2D plane strain analysis and 3D trenches compare reasonably well (differences can be attributed to element type and discretisation), the match is almost perfect for the drained analysis. However, if the 3D trench model is compared to the model with the true 3D geometry a significant

309

0,40 0,30 0,20

settlement [m]

0,10 0,00 0m -0,10 -0,20 -0,30 -0,40 -0,50 -0,60

4m

8m

12 m

16 m 2D, 21 days, Upd Mesh 2D, 21 days, no Upd Mesh 2D, undr-cons, no Upd Mesh 2D, drained, no Upd Mesh 3D, 21 days, trenches 3D, undr-cons, trenches 3D, drained, trenches

20 m

differences are significant. The real 3D model fails when an undrained analysis with subsequent consolidation is performed with the parameters chosen for the 2D analysis, reflecting the accelerated pore pressure dissipation occurring in the centrifuge models due to the scaling. 5 CONCLUSION

Figure 14. Development of settlement trough underneath the embankmentcomparison between 2D and 3D-modelling using trenches.
0,40 0,30 0,20

0,10 0,00 0m -0,10 -0,20 -0,30 -0,40 -0,50 -0,60

4m

8m

12 m

16 m

20 m

2D, drained, no Upd Mesh 3D, drained, trenches 3D, drained, columns

Figure 15. Development of settlement trough underneath the embankmentcomparison between 2D and 3D-modelling using trenches.
load of embankment
0% 0,00 -0,05 -0,10 20% 40% 60% 80% 100% 120%

settlement [m]

-0,15 -0,20 -0,25 -0,30 -0,35 -0,40

Using the 2D numerical model, the results from centrifuge tests are simulated effectively at working load level at specific points. Due to the geometrical distortion of the 2D model, the stress state is not reproduced exactly as it would develop in the field. For low stress levels, this calculation leads to acceptable results in a deformation analysis. The stiffness of the column and soil must be modified for 2D analysis. In order to model accurate time dependent behaviour, a smear zone has to be introduced, since smear has a dominating effect on the consolidation behaviour of stone column improved clay soil. However, under stress states approaching failure in the soft soil, results from the 2D analysis will not be correctly modelled since the failure criterion is not adapted for the 2D situation. The differences in results caused by simplifying the geometry (introducing trenches instead of columns) works well for lower load levels, but for loads approaching failure, these simplifications are no longer appropriate because the strength of the foundation will be overestimated. However it is usually not required to model the full foundation but a representative section or row will suffice. By doing so, the validity of numerical models can be significantly improved and, at the same time, the computational efforts are kept to reasonable limits. ACKNOWLEDGEMENTS The work was carried out as part of a Marie Curie Research Training Network on Advanced Modelling of Ground Improvement on Soft Soils supported by the European Community (Contract No MRTN-CT-2004512120). REFERENCES
Asaoka, A., Kodaka, T. & Nozu, M. 1994. Undrained shear strength of clay improved with sand compaction piles. Soils and Foundations. 34(4): 2332. Brinkgreve, R.B.J. 2002. Plaxis 2DVersion 8. Balkema Publishers, Lisse. Brinkgreve, R.B.J. (ed) 2007. Plaxis 3DFoundation Reference Manual Version 2. Plaxis bv.

settlement [m]

3D trenches 3D columns

Figure 16. Load settlement behaviour during embankment loading3D-simulations using trenches and columns.

increase in settlements can be observed. Figure 15 depicts these discrepancies for the fully drained analysis. The reason for this behaviour is that for the final embankment height the strength of the soil is mobilized to a high degree and therefore the assumption of a stone trench overestimates the strength. This is confirmed in Figure 16 where the load settlement curves for the two 3D models are compared. Up to about 60% of the embankment load two models show almost identical behaviour but for higher loads the

310

De Borst, R. & Vermeer, P.A. 1984. Possibilities and limitations of finite elements for limit analysis. Gotechnique. 34(2): 199210. Debats, J.-M., Guetif, Z. & Bouassida, M. 2003. Soft soil improvement due to vibro-compacted columns installation. International Workshop on Geotechnics of Soft SoilsTheory and Practice. (eds Vermeer et al.) Noordwijkerhout, Netherlands. Glckauf Essen: 551556. Desai, C.S. 1978. Effects of driving and subsequent consolidation on behaviour of driven piles. International Journal for Numerical and Analytical Methods in Geomechanics. 2: 283301. Farias, M.M., Nakai, T., Shanin, H.M., Pedroso, D.M., Passos, P.G.O. & Hinokio, M. 2005. Ground densification due to sand compaction piles. Soils and Foundations. 45(2): 167180. Hird, C.C., Pyrah, I.C. & Russell, D. 1992. Finite element modelling of vertical drains beneath embankments on soft ground. Gotechnique. 42(3): 499511. Indraratna, B. & Redana, I.W. 1997. Plane-Strain Modelling of Smear Effects Associated with Vertical Drains. Journal of Geotechnical and Geoenvironmental Engineering (ASCE). 123(5): 474478. Indraratna, B. & Redana, I.W. 2000. Numerical modelling of vertical drains with smear and well resistance installed in soft clay. Canadian Geotechnical Journal. 37(1): 132145. Kirsch, F. 2004. Experimentelle und numerische Untersuchungen zum Tragverhalten von Rttelstopfsulengruppen. Mitteilung des Instituts fr Grundbau und Bodenmechanik, Technische Universitt Braunschweig, Heft Nr. 75.

Mestat, P., Magnan, J.P. & Ddhouib, A. 2006. Results of the settlement prediction exercise of an embankment founded on soil improved by stone columns. Numerical Methods in Geotechnical EngineeringNUMGE 06. (ed. Schweiger) Graz, Austria. Taylor & Francis Group, London: 471476. Schweiger, H.F. & Gb, M. 2006. FE-Simulation von BaugrundverbesserungenMglichkeiten und Grenzen. 21. Christian Veder KolloquiumNeue Entwicklungen in der Bodenverbesserung. (eds Dietzel et al.) Graz, sterreich: 139154. Weber, T.M., Laue, J. & Springman, S.M. 2005. Modelling the inflight construction of sand compaction piles in the centrifuge. XVI International Conference on Soil Mechanics and Geotechnical Engineering. Osaka, Japan. Millpress, Rotterdam. (3): 12911294. Weber, T.M., Laue, J. & Springman, S.M. 2006. Centrifuge modelling of sand compaction piles in soft clay under embankment load. VI International Conference on Physical Modelling in Geotechnics. (eds Ng et al.) Hong Kong. Taylor & Francis Group, London. (1): 603608. Weber, T.M. 2008. Modellierung der Baugrundverbesserung mit Schottersulen. Verffentlichungen des Institutes fr Geotechnik, ETH Zrich, Band 232. Zrich: VDF-Verlag. Wehr, J. & Herle, I. 2006. Exercise on calculation of stone columnsPriebe method and FEM. Numerical Methods in Geotechnical EngineeringNUMGE 06. (ed. Schweiger) Graz, Austria. Taylor & Francis Group, London: 773776.

311

You might also like