Geo-Congres 2013 © ASCE 2013 1838

DUTCH RESEARCH ON BASAL REINFORCED PILED EMBANKMENTS

Suzanne J.M. van Eekelen1 and Adam Bezuijen2

1Senior R&D project manager, Deltares, Delft University of Technology, P.O. Box 177, 2600MHDelft, Netherlands; suzanne.vaneekelen@deltares.nl2Professor of Soil Mechanics and Geotechnics, Ghent University, Deltares, Technologiepark 905,B-9052, Gent-Zwijnaarde, Belgium; adam.bezuijen@ugent.be

ABSTRACT: Several long-term field studies, an extensive series of laboratory testsand finite element analysis are being carried out with the aim of further optimisingthe Dutch CUR 226 design guideline for piled embankments. It has been found thatthe strains measured in the geosynthetic reinforcement (GR) in the field are smallerthan the strains predicted with Zaeske’s design model, which was the model adoptedin the CUR 226. Two modifications are suggested to Zaeske’s load deflectionbehaviour for geosynthetic reinforcement: changing Zaeske’s triangular loaddistribution on the GR strip to an inverse triangular load distribution, and taking intoaccount the entire area below the GR, rather than only the area between the adjacentpiles. Adopting these two modifications leads to a reduction of 43-66% in GR strainin the case of the model experiments, and therefore to a reduction of 43-66% in therequired GR strength.

INTRODUCTION

Between 2002 and 2012, around 30 basal reinforced piled embankments wereconstructed in the Netherlands. A piled embankment of this kind consists of a field ofpiles with (or sometimes without) pile caps below a protective layer of sand or non-woven. Geosynthetic reinforcement (GR) is then put in, in one or more layers, priorto building up the embankment granular material (frictional sand or crushedaggregate). The Dutch subsoil often consists of a thick layer of soft clay or peat (withan undrained shear strength typically around 10 kPa or less). In many cases,settlement of the subsoil will exceed the deformation of the GR due to the weight ofthe working platform below the GR, leaving a gap between the mattress and thesubsoil in between the piles. The soft clay and peat layers mean that GR is needed inthe base of all Dutch piled embankments.

This paper summarises the basics of GR design, considering vertical load only, andgoes on to present the results of two field studies. The field studies led to furtherlaboratory research with model experiments. The experiments and field studies leadto suggestions for the modification of the calculation model adopted in the DutchCUR 226 and German EBGEO design guidelines for piled embankments.

SUMMARY OF DESIGN GUIDELINESThe calculation of GR strains from the vertical load usually involves two steps, as

shown in Figure 1. The first (“arching”) step divides the vertical load into two parts.

One part is transferred to the piles directly, and is called ‘A’ (kN/pile). The secondpart is the ‘remainder’ (‘B+C’ in Figure 1). Part A is relatively large due to arching.Both the German EBGEO (2010) and the Dutch CUR 226 (2010) guidelines adoptedZaeske’s arching model (Zaeske, 2001 or Kempfert et al., 2004) for this calculationstep. Filz et al. (2012) apply the load-displacement compatibility method for smalldeformations, but limit load part A with an adapted Terzaghi rule, although they statethat it is also possible to limit load part A with Zaeske’s model or any other realisticmethod.

In step 2, the “membrane” step, it is assumed that the GR strip between two piles isnormative for the GR design. In other words, it is assumed that the strains occurmainly in this strip. Assuming a load distribution on this GR strip, and support fromthe subsoil (if allowed), the GR strain e is calculated.

geometryproperties

loadstrain εstep 1

“arching”

load part A

load part B+C step 2“membrane”

GR stripB+CAA

softsubsoil

B+C

support from subsoil (C)

FIG. 1. The GR strain due to vertical load is calculated in two steps.

Step 2 implicitly results in a further division of the vertical load, as shown inFigure 5; load A goes directly to the piles (arching), load B is transferred through theGR to the piles and load C is carried by the subsoil. It should be noted that loads A,B and C are expressed in kN/pile and that A, B and C are vertical loads.

FIELD TESTS

Table 1 presents two recent Dutch field studies: a railway embankment in Houten(a) and a highway exit embankment near Woerden (b). Figures 2 and 3 compare theresulting measurements of the load distribution with EBGEO/CUR predictions. Thefigures show that the measured load part A is generally higher than the predictedload part A. Calculation step 1 may therefore underestimate arching by comparisonwith the measurements. Figure 3 shows that A increased during the dry spring of2011. The fill material may have compacted due to the combination of traffic loadand loss of moisture. Figure 3 also shows that the predictions are relatively sensitiveto the fill friction angle used in the model. The figure shows that the influence of thefriction angle on the calculations is large enough to enable a fit with themeasurements.

Table 1: Geometry and properties of two Dutch field testsHouten railwaypiled embankment

Woerden highway exitpiled embankment

in use since November 2008 June 2009soil conditions 1 m sand, 3 m soft clay,

20 m sand17 m soft clay

pile foundation cast in situ ø0.22 m, cast insitu pile caps ø0.40 m.

precast piles 0.29x0.29 m2,smoothly rounded squareprecast caps 0.75x0.75 m2

ctc distance pilesa 1.45x1.90 m2 2.25x2.22 m2

height embankmenta 2.60 m 1.53-1.89 mreinforcement across woven grid 600/50 (PET) geotextile 600/50 (PET)reinforcement along woven grid 450/50 (PVA) woven grid 600/50 (PET)references Van Duijnen et al., 2010 Van Eekelen et al., 2012c

aat monitoring location (road surface or top of rail - pile cap head)

The figures show that the measured B is always lower than the EBGEO/CURprediction, which is therefore on the safe side. As the measured B is lower than thecalculated B, it is expected that the measured GR strain will be less than thecalculated GR strain. This has been confirmed in several field tests for both thepermanent GR strain and the temporary strain due to truck passages. Two examples:(1) in Houten, less than 0.5% (permanent) GR strain was measured, while 3.5% GRstrain was predicted. (2) Figure 4 compares the calculated and measured GR strainresponse during passages of trucks in Woerden. The GR strains were measured withoptic fibres. The axle configuration of the two trucks is clear to see in themeasurements. The axle loads were measured separately. These measurements arepresented more in detail in Van Eekelen et al., 2012c.

0

40

80

120

160

200

01/06/2008

29/10/2008

28/03/2009

25/08/2009

22/01/2010

21/06/2010

18/11/2010

Load

(kN

/pile

).

passage crane (=design load)

EBGEO/CUR A no trains

EBGEO/CUR B no trains

EBGEO/CUR A design load

EBGEO/CUR B design load

measured Aat pile292

measured A at pile 285

measured B at pile 295

measured B at pile 292

in calculations: j = 43.0 o, subsoil k = 500 kN/m3

FIG. 2. Comparison of calculated and measured load distributions in Houten

0

40

80

120

160

22/01/2010

02/05/2010

10/08/2010

18/11/2010

26/02/2011

06/06/2011

14/09/2011

23/12/2011

01/04/2012

load

(arc

hing

,kN

/pile

).measured B

at pile 693

measured Bpile 692

measured Aat pile 693

measured Apile 692

EBGEO/CUR B j=37.5o

EBGEO/CUR A j=37.5o

EBGEO/CUR A j=49o

EBGEO/CUR B j=49o

FIG. 3. Comparison of calculated and measured load distributions in Woerden

0

5

10

15

20

25

30

35

13:06:06

13:06:07

13:06:09

13:06:11

13:06:13

13:06:14

arch

ing

load

part

A(k

N/p

ile)

.

0.00%

0.04%

0.08%

0.12%

0.16%

0.20%

0.24%

0.28%

GR

stra

in(%

)

measured AA CUR step 1 with phi=37.5 degA CUR step 1 with phi=49.0 degmeasured strain 10 (692)measured strain 09 (~692)strain CUR step 2 input is measured B+Cstrain inverse tr. step 2 input is measured B+C

strain 10

strain 09

FIG. 4. Comparison of calculated and measured GR strains during the passageof two trucks on 6 September 2011, Woerden

LABORATORY EXPERIMENTS

A series of nineteen piled embankment model experiments were conducted in theDeltares laboratory (Van Eekelen et al., 2012a and 2012b). The main purpose of theexperiments was to understand why the predicted GR strains are larger than the GRstrains measured in the field.

The starting point was that it had to be possible to validate calculation steps 1 and 2separately (Figure 1). It was therefore necessary to be able to measure load parts A,B, C (Figure 5) and the GR strain separately.

AA AAAA

C CC C

B B

C CC C

B B

FIG. 5. Load distribution in a piledembankment FIG. 6. Test set-up

FIG. 7. Cross section of test set-up FIG. 8. Top view of test set-up

Figures 6 to 8 show the test set-up. The insides of the box walls are located on thelines of symmetry. A foam cushion modelled the soft soil around the four piles. Thiscushion was a saturated and watertight sealed foam rubber cushion. A tap alloweddrainage of the cushion during the test to model the soft soil consolidation process.

A 1.5 to 2 cm layer of sand was applied on top of the foam cushion and the piles,followed by a stiff steel frame to which the GR was attached. The steel frame couldmove freely in a vertical direction. Since it was precisely fitted to the container, nohorizontal movement was possible. Differential settlement along the frame bars wasnot possible. It is assumed that this has a negligible influence on differentialsettlement between the piles. Several GR types were used, varying from geotextile togeogrid made from PET, PVA or PP. GR stiffness varied from test to test, and thenumber of reinforcement layers varied from one to two.

The embankment was fine sand (Itterbeck sand (125-250 mm) in two tests, andgranular fill (crushed recycled construction material 1-16 mm) in the remaining 17tests. In most tests, the embankment height was 0.42 m. The top load was appliedwith a water cushion that made it possible to apply stresses comparable with fieldstresses.

After the introduction of the fill, each test was carried out as follows: (1) onedrainage step foam cushion (subsoil consolidation), (2) first top load increment, (3)one or more drainage steps (4) second top load increment, (5) one or more drainagesteps and so on, up to the maximum top load (varying between 50 and 100 kPa) andthe subsequent drainage steps. The test concluded with the complete removal of thesubsoil support by applying vacuum to the foam cushion.

The test set-up resembled the Zaeske test set-up (2001). In our series, however, thefill in most tests was granular material instead of sand, the subsoil support wascontrolled with the foam cushion and the load distribution was measured differently.The area replacement ratio in the test series described was 2.6%, while the areareplacement ratio in Dutch practice is between 4.5 and 12%. This was a deliberatedecision in order to generate enough GR tensile forces at this smaller scale.

SUGGESTIONS FOR MODIFICATIONS TO THE ANALYTICAL MODEL

Calculation Step 1

Figure 9 compares the measured and calculated results for step 1 (load part A). Thenet load is the top load, minus the subsoil support C, minus the measured frictionbetween the fill and the box walls. The figure also shows that both load increase andconsolidation (subsoil deformation) result in a development of arching.

The arching as modelled in EBGEO/CUR does not respond to subsoilconsolidation. This is because arching (A) is assumed to be determined by the topload and the fill. Subsoil support does not have any influence. This is not confirmedby the measurements, which show that net load determines arching.

From the comparison in figure 9, it follows that EGEO/CUR always under-predictsarching A, even when full development has been reached, when subsoil support hasreached a minimum (limit state, bold dots). EBGEO/CUR therefore also yield acalculated load part B+C (and therefore GR strain) that is higher than measured inthe model tests. The improvement of step 1 is under further study, which will resultin a publication at a later date.

2550 1510 20 2550 1510 20

“arc

hing

”ste

p1:

load

part

A(k

N/p

ile)

10

5

15

0

10

5

15

0

net load on fill (kN/pile)= top load – subsoil support C

– friction between fill and box wall

measured A kN/pile

calculated A kN/pile

surchargeload step

consolidationfoam cushion(subsoil)

FIG. 9. Measured and calculated(CUR/EBGEO) result of step 1

(arching)

5

mea

sure

dst

rain

(%)t

estK

2

mea

sure

dst

rain

(%)t

estN

2

net load (kN/pile) net load (kN/pile)

straincables

0

10

100 20 30

1

0

2

201510 25

straingauges

on top of pileon and // GR stripnot on and/or across GR stripclose to and along GR strip

straingauges

FIG. 10. Measured strains occur mainly inthe tensile GR strips between adjacent piles

Calculation Step 2Calculation step 2 (the membrane step) consists of two parts. Firstly, it is assumed

that GR strains occur mainly in the GR strips as shown in Figure 1. Figure 10confirms this first assumption: all tests show that the GR strains occur mainly in thetensile strips located on top of and between adjacent piles.

In specific circumstances, such as the smooth, small-diameter pile caps in the testset-up, the largest strains are measured on top of the pile caps. This has not beenconfirmed in most field tests, in which larger, rougher pile caps have been used.Taking the strains on top of the pile caps into account would result in a lower valuefor the maximum GR strain but more deformation than when leaving these strainsout, as is the case with EBGEO/CUR 226. It was decided not to extend the step 2calculations to include strain on top of the pile cap. This is a decision on the safe sidesince it requires GR strength to be higher.

The loading and supporting of the GR strip is considered in the second part ofstep 2, as shown in Figure 1. The load distribution on the strip is directly related tothe deformed shape of the GR, as shown in Figure 11. EBGEO/CUR 226 use atriangular load distribution (blue), while the British Standard BS8006 (2010) uses theequally distributed load (red) (Van Eekelen et al., 2011). Originally, BS8006 neverallowed subsoil support to be included in the calculations, but Lawson (2012)suggests using BS8006 while applying an equally distributed net load on the GR (topload – bottom load). This strategy was also adopted by Filz et al. (2012).

x

z

cx3+dx2

bx3

ax2

GR

FIG. 11. Relationship between shape of deformed GR (left) and loaddistribution (right)

The shape of the deformed GR is measured both with a liquid levelling system andby scanning the surface of the sand layer below the GR before and after the test.From the results of these measurements (Figure 12), it can be concluded that the loaddistribution on the GR strip complies best with the green inverse triangle and that theload distribution may actually be transferred more towards the piles than reflected bythe triangular or equally distributed load. This is also confirmed by the directmeasurements of the vertical load on the GR strips between the piles, as shown inFigure 13 and by the numerical calculations described in Den Boogert et al., 2012.

Figure 14 shows the GR deformations in Woerden, as measured in a tube attachedto the GR, across piles. The second derivative of these measurements correspondsdirectly to the load distribution on the GR strips. Taking the second derivativeintroduces considerable scatter, even after some averaging, but the shape of the red

lines is much more similar to the inverse triangle than a triangular or constant loaddistribution.

If we consider subsoil support (Figure 1) for the GR strip, EBGEO/CUR assumesthat only part of the subsoil is mobilised, namely the area below the GR stripsbetween the piles. Van Eekelen et al. (2012b) suggest how, in the calculations, thesupporting subsoil area can be increased to include the entire available area belowthe GR, which is more realistic. Lodder et al. (2012) described this in more detail.Figure 15 shows the results of modifying EBGEO/CUR step 2 by improving bothsubsoil support and load distribution in the calculations. The figure shows that thesemodifications fit the laboratory measurements much better and, for this test, lead to a43-66% reduction in GR strain compared with the EBGEO/CUR assumptions.Changing these features can therefore reduce the GR strength required by 43-66%.

FIG. 12. The measured shape of the GRis closest to the third-order polynomial

Pressure on GR,measured with localtotal pressure cells

Measured total B+C(presented in inversedtriangular shape, kPa)

FIG. 13. Measured loaddistribution on the GR strip

1.5

1.0

0.5

0.0

-0.5

-1.0

-1.5

-1.12

-1.16

-1.20

-1.24

-1.28

-1.32

-1.36

19.5 20.5 21.5 22.5 23.5 24.5 25.5horizontal position across road (m)

mea

sure

dG

Rpo

sitio

n(r

elat

ive

posi

tion

NA

Pm

)

2ndde

rivat

ive

ofG

Rpo

sitio

n(1

/m)

North

South

moving average of two 2nd order derivatives

2nd derivative2nd derivative

GR positionGR position

687687686686

688688

FIG. 14. Settlement measured in Woerden (October 2011: road had been in usefor ca. 16 months)

top load75 kPa

(chronological)

0

5

4

3

2

1

0

5

4

3

2

1

measured “rest” B+C (kN/pile)

aver

age

stra

in(%

)

5 4 3 7 6 5 11 10 9 14 13 12 115 4 3 7 6 5 11 10 9 14 13 12 11

measured

top load25 kPa

top load50 kPa

top load100 kPa

EBGEO/CUREBGEO/CUREBGEO/CUR

43-66%

FIG. 15. Comparison of measurements and calculations in step 2; effect of a.modification in load distribution (from triangular shape to inverse triangular

shape) and b. modification in subsoil support (from below GR strip only tobelow entire, diamond shaped GR area)

CONCLUSIONSSeveral field tests showed that the measured GR strains in the field are

considerably smaller than the strains predicted with Zaeske’s calculation model.Model tests show that this can be explained in part by the measured increase inarching due to subsoil consolidation. Current arching models do not describe this andthe phenomenon is being subjected to further study. Furthermore, it was found thattwo modifications in the calculation model result in a considerably closer fit betweencalculations and measurements, both for static loading and for truck passages. (1)Load distribution on the GR strip should be an inverse triangle and (2) subsoilsupport should be taken into account for the entire area below the GR, not only forthe area between the adjacent piles.

Adopting these two modifications reduces GR strain by 43-66% in the case of themodel experiments, and therefore leads to a reduction of 43-66% in the GR strengthrequired.

ACKNOWLEDGMENTS

The authors are grateful for financial and practical support from Deltares, Huesker,Naue, TenCate, Tensar, the Bataafse Alliantie and Voorbij Funderingstechniek(monitoring Houten railway) and the Dutch Ministry of Public Works (Woerden).

REFERENCES

BS8006-1: 2010. Code of practice for strengthened/reinforced soils and other fills.British Standards Institution, ISBN 978-0-580-53842-1

CUR 226, 2010, Ontwerprichtlijn paalmatrassystemen (Design guideline piledembankments). ISBN 978-90-376-0518-1 (in Dutch)

EBGEO, 2010. Empfehlungen für den Entwurf und die Berechnung von Erdkörpernmit Bewehrungen aus Geokunststoffen – EBGEO, 2. Auflage, GermanGeotechnical Society, ISBN 978-3-433-02950-3 (in German).

Den Boogert, T.J.M., Van Duijnen, P.G., and Van Eekelen, S.J.M., 2012,“Numerical analysis of geosynthetic reinforced piled embankments scale modeltests.“ Plaxis Bulletin 31: 12-17

Filz, G.M., Sloan, J.A., McGuire, M.P., Collin, J.G. and Smith, M.E., 2012,“Column-Supported Embankments: Settlement and Load Transfer.” Proceedingsof Geo-Congress, Oakland, California, March; 24 p.

Kempfert, H.-G., Gobel, C., Alexiew, D. and Heitz, C. (2004). “Germanrecommendations for reinforced embankments on pile-similar elements.”EuroGeo3 - Third European Geosynthetics Conference; 279-284.

Lawson, C.R., 2012, “Role of Modelling in the Development of Design Methods forBasal Reinforced Piled Embankments. ”EuroFuge 2012, Delft, the Netherlands.

Lodder, H.J., Van Eekelen, S.J.M. and Bezuijen, A., 2012, “The influence of subsoilreaction on the geosynthetic reinforcement in piled embankments.” EuroGeo 5,Valencia in Spain.

Van Duijnen, P.G., Van Eekelen, S.J.M. and Van der Stoel, A.E.C. 2010.“Monitoring of a Railway Piled Embankment.” 9 ICG, Brazil: 1961-1964.

Van Eekelen, S.J.M., Jansen, H.L., Van Duijnen, P.G., De Kant, M., Van Dalen,J.H., Brugman, M.H.A., Van der Stoel, A.E.C., Peters, M.G.J.M., 2010. “TheDutch Design Guideline for Piled Embankments.” 9 ICG, Brazil: 1911-1916.

Van Eekelen, S.J.M.; Bezuijen, A. and Van Tol, A.F., 2011. “Analysis andmodification of the British Standard BS8006 for the design of piledembankments.”Geotextiles and Geomembranes 29: 345-359.

Van Eekelen, S.J.M., Bezuijen, A., Lodder, H.J., van Tol, A.F., 2012a. “Modelexperiments on piled embankments Part I.”Geotextiles and Geomembranes 32:69-81

Van Eekelen, S.J.M., Bezuijen, A., Lodder, H.J., van Tol, A.F., 2012b. “Modelexperiments on piled embankments. Part II.”Geotextiles and Geomembranes 32:82-94

Van Eekelen, S.J.M., Bezuijen, A., 2012c. “Does a piled embankment ‘feel’ thepassage of a heavy truck? High frequency field measurements.”EuroGeo5,Valencia, Spain.

Zaeske, D., 2001. Zur Wirkungsweise von unbewehrten und bewehrten mineralischenTragschichten über pfahlartigen Gründungselementen. SchriftenreiheGeotechnik, Uni Kassel, Heft 10, February 2001 (in German).