Sie sind auf Seite 1von 8

The 12 International Conference of International Association for Computer Methods and Advances in Geomechanics (IACMAG) 1-6 October, 2008

Goa, India

th

Influence of Anisotropy, Destructuration and Viscosity on the Behavior of an Embankment on Soft Clay
Z.Y. Yin
Department of Civil and Environmental Engineering, Helsinki University of Technology

M. Karstunen
Department of Civil Engineering, University of Strathclyde, United Kingdom Keywords: Viscosity, anisotropy, destructuration, soft clay, embankment ABSTRACT: The paper aims to investigate the influence of different features of natural clays, namely anisotropy, destructuration and viscosity, on modeling the delayed behavior of a benchmark embankment on soft clay. The elasto-viscoplastic models, which are based on elasto-plastic models (Modified Cam Clay, S-CLAY1 and SCLAY1S) and Perzynas overstress theory are coupled with Biots consolidation theory. The parameters for the models can be determined from the same set of triaxial and oedometer tests. The same set of common soil constants and state parameters is employed for all models, with additional parameters required for representing different soil features. The finite element predictions compare the results for settlement, horizontal displacement and excess pore pressure during the construction and the subsequent consolidation for all models. In addition, the stress paths under the embankment loading are also compared each other to improve the understanding of the effect of different soil features. All simulations demonstrate that all three features influence significantly the predictions. As a consequence, accounting for soil features need to be carefully considered when applied to real construction sites.

1 Introduction
The stress-strain behavior of soft soils is very complex, as different fundamental features of natural soil behavior, such as viscosity, anisotropy and destructuration, influence the soil response to foundation loading. Thus, design and construction of embankments on soft soil deposits is still a challenge. In the literature, some of the features have been considered on modeling the behavior of embankments on soft soils e.g. by Oka et al. (1991), Rowe et al. (1996), Zhu and Yin (2000), Kim and Leroueil (2001), Karstunen et al. (2005), but none of them discussed the influence of all these three features on the behavior of embankments on soft soils. A benchmark embankment on POKO clay in Finland was simulated by using different elasto-plastic models by Karstunen et al. (2006) to investigate the effect of constitutive models on the predicted response. However, these simulations did not consider the effect of soil viscosity. Yet the strain-rate effect on the undrained shear strength, the apparent preconsolidation pressure and behavior after yield for soft clay is significant. Furthermore, on natural soft soils all these features play a role during the plastic straining. The paper aims to investigate the influence of soil viscosity, anisotropy and destructuration on the delayed behavior of embankment on soft clay. Firstly an elasto-viscoplastic model based on Modified Cam Clay model (noted EVP-MCC, developed by Yin and Hicher, in press) is used for the simulations. The extended anisotropic version of the model, called EVP-SCLAY1, incorporates the effect of initial and plastic strain induced anisotropy via initial inclination and subsequent rotation of the yield surface, similarly to the S-CLAY1 model (Wheeler et al. 2003). Finally, a new EVP model EVP-SCLAY1S was used to account for all three soil features. EVP-SCLAY1S is a hierarchical elasto-viscoplastic model, and it is possible via selection of values for soil constants to switch various features on and off, reducing the model to EVP-SCLAY1 and EVP-MCC. In addition, an additional run was made to investigate the effect of viscosity on the predicted response. In order to ensure that the comparisons are realistic, the values for soil constants and state variables were determined from the same set of laboratory tests. In this paper, firstly the benchmark embankment with its foundation soil based on POKO clay is presented. Then the constitutive models are briefly described, and the derivation of values for soil constants is discussed. The values for the soil constants and state parameters of foundation soil were then determined for different models respectively from the same set of tests. Using the selected parameters, large deformation finite element (FE) analyses were carried out, using the same finite element mesh for all analyses. The model predictions are then

4728

compared to each other to investigate the effect of soil features on the embankment behavior.

Elasto-viscoplastic models
The constitutive model, called EVP-SCLAY1S accounting for soil viscosity, anisotropy and destructuration, is based on the overstress theory of Perzyna (Perzyna, 1963, 1966) and elasto-plastic model S-CLAY1S (Karstunen et al. 2005). According to Perzynas overstress theory, the total strain rate is additively composed of the elastic strain rates and viscoplastic strain rates:

& & e & vp ij = ij + ij

(1)

& where ij denotes the (i,j) component of the total strain rate tensor, and the superscripts e and vp stand,

respectively, for the elastic and the viscoplastic components. The elastic behavior in the proposed model is assumed to be isotropic similarly to the Modified Cam Clay model & vp (Roscoe & Burland, 1968). The viscoplastic strain rate ij in Eq.1 is assumed to obey an associated flow rule with respect to the dynamic loading surface following the original proposal by Perzyna (1963, 1966):

& vp ij = ( F )

f d ij

(2)

where is referred to as the fluidity parameter; (F) is the overstress function representing the difference between the dynamic loading surface and the static yield surface, and the McCauley brackets imply that <F> = F for F > 0 or <F> = 0 for F 0. The fd is the viscoplastic potential function, represented by the dynamic loading surface as explained in the following. An exponential type of scaling function is adopted:

( F ) = exp N

d pm 1 1 s pm

(3)

where N is the viscosity index (or strain rate coefficient). N and are therefore the key parameters controlling the d s viscoplastic strain-rate. The pm and pm represent the size of dynamic loading surface and the size of static yield surface respectively. Following the ideas by Wheeler et al. (2003) an elliptical surface is adopted as the static yield surface, relating to the current state of preconsolidation:

f =

3 3 s [ d p d ] : [ d p d ] M 2 [ d ] : [ d ] ( pm p) p = 0 2 2

(4)

where d is deviatoric fabric tensor, a dimensionless vector with the same form as deviatoric stress vector (Wheeler et al. 2003); M is the slope of the critical state line. For the special case of a cross-anisotropic sample, the scalar parameter = 3/ 2 ( d : d ) defines the inclination of the ellipse of the yield curve in q-p plane as illustrated in Fig.1a. The dynamic loading surface has identical shape, but a different size pmd (See Fig.1a) compared to the static yield surface. To represent the dynamic loading surface Eq. 4 can be rewritten as:

3 [ d p d ] : [ d p d ] d fd = 2 + p pm = 0 2 3 M [ d ] : [ d ] p 2

(5)

The intrinsic yield surface is assumed again to have similar shape, with size defined via pmi, which is related to the size of the static yield surface with bonding parameter (Fig. 1a). The expansion of the intrinsic yield surface, which represents the intrinsic hardening of the material, is assumed vp to be due to the inelastic volumetric strain v , similarly to the critical state models:

1+ e dpmi = pmi i

vp d v

(6)

where i is the slope of the intrinsic normal compression curve in the e-lnv plane for a stress path at constant stress ratio involving no change of anisotropy (such as isotropic loading of an isotropic sample, or oedometer test on a reconstituted sample, see Fig.1b), is the slopes of the swelling-line and e is the void ratio . The rotational hardening law, based on the formulation proposed by Wheeler et al. (2003), describes the development or erasure of anisotropy caused by viscoplastic strains. Both volumetric and deviatoric viscoplastic strains influence the rotation of the yield curve.

3 vp d d = d d d vvp + d d d d d 3 p 4 p

(7)

where the soil constant controls the rate at which the components of the deviatoric fabric tensor heads toward

4729

their current target values, which depend on the stress path, and d controls the relative effect of viscoplastic deviatoric strains in rotating the yield and loading surfaces. The amount of particle bonding is described with a scalar state variable , which is changing due to bond degradation ultimately to zero, analogously to the S-CLAY1S model (Karstunen et al. 2005):
vp d = d vvp + d d d

(8)

where the soil constant controls the absolute rate of destructuration and d controls the relative effect of viscoplastic deviatoric strains in destroying the bonds; 0 is the initial amount of bonding, which relates the sizes of the intrinsic yield surface and static yield surfaces as

pm 0 = (1 + 0 ) pmi 0

(9)
0yi0
1

q Dynamic loading surface


Mc 1

yi0

y0

ln(v)

f d ij

Static yield surface Intrinsic yield surface pmi pms pmd


1

1 Natural clay

pmi

Remoulded clay

i 1

Me

(a)

(b)

Figure 1. Definition of surfaces and parameters of EVP-SCLAY1S model Taking into account the elastic stress-strain relations, the constitutive equations of the proposed model for structured clays are derived as follows:

1 3[d pd ] [d ] :[d pd ] + [d pd ] :[d pd ] ij 2p 1 + ij (10) & & & ij = d + pij + ( F) 2G 3(1+ e0 ) p 3 2 3 M [d ] :[d ] p 2
& where d is the deviatoric stress rate tensor (d=ij-pij); p is the mean effective stress (p=kk/3); G is the

elastic shear modulus, which is related to the elastic bulk modulus (K=(1+e)p/) by assuming a constant value of Poissons ratio ' (G=3(1-2')K/2(1+')); ij is Kronecker delta with ij=1 for i=j and ij=0 for ij. The proposed model was implemented into 2D Plaxis v.8 as a user defined model. By switching on/off certain parameters, the model reduces to as EVP-SCLAY1 (0 = = d = 0) and ultimately to EVP-MCC (0 = = d = 0 and 0 = = d = 0). By using the coupled consolidation analysis based on Biots theory provided in Plaxis with the proposed model consolidation analyses of boundary value problem can be carried out.

2 Benchmark embankment 2.1 Finite element model


The geometry of the benchmark embankment is shown in Figure 2. The groundwater table is located 2m below the ground surface. Drained conditions and zero initial pore pressures are assumed above the water table, while the other boundaries were assumed to be undrained (closed). The domain to be analyzed (under plane strain conditions) has an extent of 60 m in the horizontal direction from the symmetry axis and 36 m in the vertical direction. The lateral boundaries are restrained horizontally, and the bottom boundary is restrained in both directions. Due to symmetry, only half of the embankment is represented in the finite element mesh. The construction and the subsequent consolidation of the embankment have been simulated by using twodimensional (2D) FE code Plaxis into which the proposed models were implemented as user-defined models. Plane strain condition was assumed in FE analysis. The FE mesh for the embankment on natural subsoil was constituted by 8561 nodes using 1048 15-noded triangular elements. Mesh sensitivity studies were made to check that the mesh was dense enough for the calculations to be accurate. The embankment loading was

4730

reproduced by stage construction, which means increasing the unit weight of the elements of the embankment fill, within 2 days. Because large deformations were expected, large strain analysis with updated pore water pressure was employed. The embankment, assumed to be made of granular fill, was simulated with a simple Mohr-Coulomb model 2 assuming the following material parameters: Youngs modulus E = 40000 kN/m ; Poissons ratio = 0.3; friction o o 2 3 angle = 38 ; dilatancy angle = 0 ; cohesion c = 1 kN/m and unit weight = 20 kN/m . The problem is dominated by the soft soil response and is hence rather insensitive to the embankment parameters.

Figure 2. Geometry of benchmark embankment and assumed soil profile (after Karstunen et al. 2006)

2.2 Parameters of foundation soil for constitutive models


Although the different constitutive models have different number of parameters, the values for soil constants and state parameters can be determined from the same experimental database. The common parameters of models , , e0, M were determined from triaxial and oedometer tests on Poko clay (Koskinen et al. 2002a, 2002b). The ' was assumed to be 0.2, a common value for soft soil. The soil permeability k was assumed constant during the -7 -9 consolidation, equal to be 1x10 m/s in both vertical and horizontal directions for soil crust layer, and to be 1x10 m/s in both vertical and horizontal directions for two sub-layers for all analyses. The viscosity parameters N = 10 -9 -1 and = 1x10 s obtained from long-term oedometer tests were assumed the same for all soil layers. In order to -7 -1 investigate the viscosity effect, an additional study was made with values N = 10 and = 1x10 s used for the prediction by EVP-SCLAY1S. The results for this case are referred to as low viscosity. The preconsolidation pressure p was estimated from oedometer tests. The dry crust layer was assumed overconsolidated and sub-layers normally consolidated. The Pre-Overburden Pressure (POP=p-v0) was then obtained, and can be used to calculate pm0:

pm 0 = pi +

(M

( qi 0 pi )
2

2 0 ) pi

(11)

where pi=(1+2K0)(POP+v0)/3 ; qi=(1-K0)(POP+v0). The coefficient of earth pressure at rest K0 was estimated sin c by Jakys formula K0 = 1-sinc for normally consolidated soil and K0 = (1-sinc)OCR for overconsolidated soil. The values for parameters 0 and d were estimated by equations proposed by Wheeler et al. (2003), and proposed by Leoni et al. (in press) for anisotropic models EVP-SCLAY1 and EVP-SCLAY1S, as defined below:
2 2 3 ( 4 M 2 4 K 0 3 K 0 ) M 2 K 0 0 = K 0 , d = , 2 3 8 ( K 0 + 2 K 0 M 2 )
-1

1 + e0

10 M 2 2 0d ln 2 M 2 0d

(12)

where K0=3M/(6-M), and c=sin (3M/(6+M)). For destructuration model EVP-SCLAY1S, the initial bonding 0 can be calculated from soil sensitivity:0 = St-1. The values for soils constants and d were determined from isotropic consolidation tests combined with oedometer tests. All values for soil constants and state parameters are summarized in Table 1.

4731

Table 1a. Values of common parameters for constitutive models


Depth (m) 02 26 6 - 36 3 (kN/m ) 15 15 15 e0 2.1 2.1 2.1 M 1.2 1.2 1.2 K0 1.0 0.5 0.5 0.03 0.03 0.03 0.71 0.71 0.71 POP (kPa) 15 0 0 0.2 0.2 0.2 kv (m/s) 1 x10 1 x10 1 x10
-7 -9 -9

kh (m/s) 1 x10 1 x10 1 x10


-7 -9 -9

N 10 10 10

(s1) -9 1 x10 1 x10 1 x10


-9 -9

Table 1b. Values of additional parameters for constitutive models


Depth (m) 02 26 6 - 36 EVP-SCLAY1 0 0.46 0.46 0.46 d 0.76 0.76 0.76 13 13 13 i 0.26 0.26 0.26 0 0.46 0.46 0.46 d 0.76 0.76 0.76 EVP-SCLAY1S 13 13 13 0 12 12 12 9 9 9 d 0.2 0.2 0.2

3 Influence of soil features on the embankment behavior 3.1 Settlements


Figure 3 shows the surface settlements under the centerline of the embankment predicted by models with different considerations of soil behavior. Big difference was found between the predictions by different models after 2 days of construction, especially toward the end of consolidation. Comparing the predicted settlements by EVP-SCLAY1 with that by EVP-MCC, it is clear that incorporation of soil anisotropy results in larger settlement than in the isotropic case. Accounting for initial bonding and destructuration (by EVP-SCLAY1S)increase the predicted settlement further. The predicted settlement with low viscosity model is larger than the one predicted by the standard viscosity during the early consolidation stage, but the same level for later stage. Figure 4 shows the surface settlements immediately after the construction of the embankment and at 100 years after the construction, which agree with the time-settlement curves at the centerline (Fig.3). This implies that the effects of soil features on settlements behavior for other positions underneath the embankment as the same. As for the long-term behavior, the settlements develop fast during the consolidation and become slower towards the full dissipation of excess-pore pressure (178 years after construction) for all cases. The predicted settlements continue to increase due to soil viscosity still after the full dissipation of excess pore pressures.

Full dissipation of u at 178 years

-0.5 Settlement (m) Low viscosity -1 EVP-MCC EVP-SCLAY1

-1.5

EVP-SCLAY1S

-2 0.001 0.01 0.1 1 Time (year) 10 100 1000

Figure 3. Predicted settlements by different models underneath the embankment at the centerline.

3.2 Horizontal displacements


Figure 5 presents the predicted horizontal displacements underneath the embankment toe by all models. The displacements by isotropic model (EVP-MCC) are larger than by the anisotropic models immediately after the construction but become smaller at 100 years after the construction. Slight difference can be found between the predicted displacements with and without the consideration of destructuration. The predicted displacements with lower viscosity at the end of construction are much larger than the predictions by other models corresponding to

4732

the larger settlements in Fig.4a, but at 100 years after the construction return to the same level as the ones with high viscosity model, corresponding to the same level of settlements (Fig.4b).

0.04 0.02 Surface settlement (m)

(a)

0 -0.2 0

(b) 10 20 30 40 50 60

10

20

30

40

50

60

Surface settlement (m)

0 -0.02 0 -0.04 -0.06 -0.08 -0.1 -0.12 -0.14

-0.4 -0.6 -0.8 -1 -1.2 -1.4 -1.6 EVP-MCC EVP-SCLAY1 EVP-SCLAY1S Low viscosity Distance from centreline (m)

EVP-MCC EVP-SCLAY1 EVP-SCLAY1S Low viscosity Distance from centreline (m)

Figure 4. Predicted surface settlements by different models: (a) immediately after construction, and (b) 100 years after construction.
0

-10 Depth (m)

-10 Depth (m)

-20 EVP-MCC -30 (a) -40 0 0.01 0.02 0.03 0.04 Horizontal displacement (m) 0.05 EVP-SCLAY1 EVP-SCLAY1S Low viscosity

-20

EVP-MCC EVP-SCLAY1 EVP-SCLAY1S Low viscosity

-30 (b) -40 0.05 0.1

-0.2

-0.15

-0.1 -0.05 0 Horizontal displacement (m)

Figure 5. Predicted horizontal displacements by different models: (a) immediately after construction, and (b) 100 years after construction.

3.3 Excess pore pressure


Figure 6 shows the model predictions for excess pore pressure in the foundation soil under centerline at a depth of 4.5 m. All model predictions are rather similar during the construction and the subsequent consolidation. As expected, all numerical simulations show excess pore pressures gradually dissipating with time. The predicted excess pore pressure with low viscosity is much higher during the consolidation than that with high viscosity, while the anisotropy and destructuration have only a slight influence.

3.4 Stress paths


In order to improve the understanding of the embankment behavior with different consideration of the soil features, a point at a depth of 8 m underneath the embankment centerline is selected. Figure 7(a) shows the stress paths by different models reflect the evolution of effective stresses during the construction of the embankment and subsequent consolidation. All models predict initially elastic behavior followed with a likely undrained viscoplastic behavior until a point corresponding to the end of construction. The point is different from one model to another due to different considerations of soil behavior. During the early consolidation stage, the p decreases due to the creep of soil under likely undrained condition (see Fig.6). For later consolidation stage with the dissipation of excess pore pressure, the anisotropy consideration leads the incremental stress ratio consisting with Jakys K0 = 1-sinc, while higher K0 predicted by isotropic model. Slight difference can be found for the stress paths by the two anisotropic models. Figure 7(b) shows the different stress paths predicted by EVPSCLAY1S with different viscosity. A notable difference can be found during the embankment loading corresponding to the construction, which agrees with the different response of excess pore pressure and

4733

displacements (settlements and horizontal displacements) between the predictions with low and high viscosity during the construction and subsequent early consolidation stage. This clearly indicates that viscosity is a key feature that needs to be considered when predicting the changes in undrained shear strength during construction and consolidation.

50 40 u (kPa) 30 20 10 0 0.001 0.01 0.1 1 Time (year) 10 100 1000 EVP-MCC EVP-SCLAY1 EVP-SCLAY1S Low viscosity

Figure 6. Predicted excess pore pressure in foundation soil by different models.


60 M = 1.2 50 q (kPa) EVP-SCLAY1 EVP-SCLAY1S

60 M = 1.2 50 q (kPa) High viscosity Low viscosity

40

K0 = 1-sin'c EVP-MCC (a)

40

K0 = 1-sin'c

30

30

(Predicted by EVP-SCLAY1S) (b)

20 30 40 50 p' (kPa) 60 70

20 30 40 50 p' (kPa) 60 70

Figure 7. Predicted stress paths by different models.

4 Conclusions
The behavior of a benchmark embankment on soft clay was modeled using different constitutive models coupled with Biots consolidation theory. The values for soil constants and state parameters of the models were determined from the same set of laboratory tests. The results of the finite element analyses are presented for settlements, horizontal displacements, excess pore pressures and stress paths during the construction and subsequent consolidation and time-dependent compression (creep). The comparisons between predictions by different models were used to investigate the influence of different soil features (viscosity, anisotropy and destructuration) on the embankment behavior. The soil anisotropy has slight influence on the predicted settlement at early consolidation stage but a major important effect on the later stages. For horizontal displacements, the anisotropy results in small maximum value for early consolidation stage but big value for later stage. The horizontal displacement by predicted the anisotropic models are relatively smaller than the ones predicted by the isotropic model, suggesting a lower ratio of horizontal to vertical deformations. For excess pore pressures, there is only a slight influence by the soil anisotropy on the predicted excess pore pressures during the embankment construction and its subsequent consolidation. As for soil destructuration, significant influence on settlements can be found especially during the later consolidation stage. There also seems to be a minor influence on the predicted horizontal displacements, as well as on the predicted excess pore pressures.

4734

Concerning the soil viscosity, significant influence on settlements can be seen during the early consolidation stage, but this effect vanishes with time. Similar influence can also be seen on the predicted horizontal displacements. The soil viscosity has a major influence on the generation of the excess pore pressures during the embankment construction, as well as for their dissipation during subsequent consolidation, but not for the total time of full dissipation. The stress paths with different consideration of soil behavior were also investigated to improve the understanding of the soil behavior under embankment loading. The influence of all three fundamental soil characteristics on the embankment behavior can be used as database for the guidance of the embankment construction. The soil features need to be carefully considered when simulating the soil behavior in the context of real construction sites.

5 Acknowledgement
The work by the first author was funded by the Academy of Finland (Grants 210744) and it was carried out as part of a Marie Curie Research Training Network "Advanced Modelling of Ground Improvement on Soft Soils (AMGISS)" (Contract No MRTN-CT-2004-512120) supported the European Community through the program "Human Resources and Mobility".

6 References
Karstunen M., Krenn H., Wheeler S.J., Koskinen M., Zentar R. 2005. The effect of anisotropy and destructuration on the behavior of Murro test embankment. International Journal of Geomechanics (ASCE); 5(2), 8797. Karstunen M., Wiltafsky C., Krenn H., Scharinger F., Schweiger H.F. 2006. Modelling the behaviour of an embankment on soft clay with different constitutive models. International Journal for Numerical and Analytical Methods in Geomechanics, 30(10): 953-982. Kim Y.T., Leroueil S. 2001. Modeling the viscoplastic behavior of clays during consolidation: Application to Berthierville clay in both laboratory and field conditions. Canadian Geotechnical Journal, 38(3), 484-497. Koskinen M, Karstunen M., Wheeler S.J. 2002a. Modeling destructuration and anisotropy of a natural soft clay. Proceedings of the 5th European Conference Numerical Methods in Geotechnical Engineering, Paris (France). 1120. Koskinen M., Zentar R., Karstunen M. 2002b. Anisotropy of reconstituted POKO clay. Proc., 8th International Symp. Numerical Models in Geomechanics (NUMOG), Rome (Italy), 99105. Leoni M., Karstunen M. Vermeer P.A. (in press). Anisotropic creep model for soft soils. Accepted for publication in Gotechnique. Oka F., Tavenas F., Leroueil S. 1991. Elasto-viscoplastic FEM analysis of sensitive clay foundation beneath embankment. Proceedings of the International Conference on Computer Methods and Advances in Geomechanics, 1023. Perzyna P. 1963. The constitutive equations for work-hardening and rate sensitive plastic materials. Proc., Vibration Problems, 281290. Perzyna P. 1966. Fundamental problems in viscoplasticity. Adv. Appl. Mech., 9, 244377. Roscoe K.H., Burland J.B. 1968. On the generalized stressstrain behavior of wet clay. Engineering Plasticity. Cambridge University Press: Cambridge, 553609. Rowe R. K., Gnanendran C.T., Landva A.O., Valsangkar A.J. 1996. Calculated and observed behavior of a reinforced embankment over soft compressible soil. Canadian Geotechnical Journal, 33(2), 324-338. Wheeler S.J., Ntnen A., Karstunen M., Lojander M. 2003. An anisotropic elasto-plastic model for soft clays. Canadian Geotechnical Journal, 40, 403418. Yin Z.Y. 2006. Modlisation Viscoplastique des Argiles Naturelles et Application au Calcul de Remblais sur Soils Compressibles. Thse de lEcole Centrale de Nantes et lUniversit de Nantes. Yin Z.Y., Hicher P.Y. (in press). Identifying parameters controlling soil delayed behavior from laboratory and in situ pressuremeter testing. International Journal for Numerical and Analytical Methods in Geomechanics, Zhu G., Yin J.H. 2000. Elastic visco-plastic consolidation modeling of clay foundation at Berthierville test embankment. International Journal for Numerical and Analytical Methods in Geomechanics, 24(5), 491-508.

4735

Das könnte Ihnen auch gefallen