Вы находитесь на странице: 1из 20

INTERNATIONAL JOURNAL FOR NUMERICAL AND ANALYTICAL METHODS IN GEOMECHANICS Int. J. Numer. Anal. Meth. Geomech.

, 2005; 29:577596 Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/nag.426

An elastoplastic model based on the shakedown concept for exible pavements unbound granular materials
Taha Habiballah and Cyrille Chazallonn,y
Laboratory of Mechanics and Modelling of Materials and Structures in Civil Engineering, University of Limoges, 19300 Egletons, France

SUMMARY Nowadays, the problem of rutting of exible pavements linked to permanent deformations occurring in the unbound layers is taken into account only by mechanistic empirical formulas. Finite element modelling of realistic boundary value problems with incremental rheological models will lead to unrealistic calculation time for large cycle numbers. The objective of the authors is to present a simplied model which can be used to model the exible pavements rutting with the nite elements framework. This method is based on the shakedown theory developed by Zarka which is usually associated to materials like steels. It has been adapted for granular materials by introducing a yield surface taking into account the mean stress inuence on the mechanical behaviour and a dependency of the hardening modulus with the stress state. The DruckerPrager yield surface has been used with a non-associated ow rule. Comparisons with repeated load triaxial tests carried out on a subgrade soil have been done. These comparisons underline the capabilities of the model to take into account the cyclic behaviour of unbound materials for roads. Finally, a discussion, dealing with the use of the simplied method within a nite element modelling of a full-scale experiment, is presented. Copyright # 2005 John Wiley & Sons, Ltd.
KEY WORDS:

unbound granular materials; exible pavements; rutting; shakedown theory

1. INTRODUCTION Pavements with bituminous surfacing are divided in three main types according to the main structural elements, dened as the layer or layers that provide the greatest contribution to the distribution of trac loads: bituminous, rigid and granular. The term composite is used for pavements with bituminous and a cementitious main structural element, whereas the term exible will cover pavements with unbound granular main structural element and a thickness of the bituminous layer lower than 150 mm: Recent studies, dealing with the improvement of

Correspondence to: C. Chazallon, Laboratory of Mechanics and Modelling of Materials and Structures in Civil Engineering, University of Limoges, 19300 Egletons, France. E-mail: chazallon@unilim.fr

Copyright # 2005 John Wiley & Sons, Ltd.

Received 4 April 2003 Revised 20 September 2004

578

T. HABIBALLAH AND C. CHAZALLON

design methods of exible pavements, have pointed out that in most cases, when the bearing capacity of the soil is sucient, this rutting takes place mainly in the granular base and subbase layers [1, 2]. Despite this fact, this rutting of unbound materials is still not well understood, and it is not taken into account in most pavement analysis and design methods. Despite the large annual funds utilized for construction and maintenance of exible pavements, exible pavements design and maintenance planning to date still has a strongly empirical character. To make the decision-making procedures regarding pavement design and maintenance more time and cost eective, it is necessary to thoroughly study and understand the mechanical processes that form the basis of exible pavements performance and exible pavements deterioration. These mechanical processes can be separated into two categories, namely (i) short-term mechanical processes and (ii) long-term mechanical processes. The rst category concerns the instantaneous behaviour of a exible pavement, as activated during the passage of a vehicle. The permanent pavement deformations generated during an individual vehicle axle passage are usually very small, such that the exible pavement behaviour may be viewed as reversible, and thus can be studied by means of (visco)elastic models. The second category concerns the mechanical processes characterized by a typically quasi-static time dependency, such as pavement deformations caused by ground water ow, or creep processes in clay or long-term settlements under a large number of vehicle axle passages. For all these phenomena, the generated permanent deformations may become substantial and require the use of (non-linear) plasticity-based models to study them. A separation into short-term mechanical processes and long-term mechanical processes is convenient from the mechanical point of view, though for a proper assessment of the overall exible pavement performance the interaction between these processes should be also taken into account. The overall objective of this paper is to develop an advanced model that provides detailed insight into mechanical processes in unbound granular materials of a exible pavement. Consequently, the study will be conned to the unbound granular materials of exible pavements layers. The short-term mechanical process to be studied concerns the reversible behaviour of a exible pavement under a moving vehicle axle. The hypoelastic model [3] which reproduces the resilient behaviour of unbound granular materials under laboratory repeated load triaxial tests and the reversible behaviour of a exible pavement will be used [4, 5]. The longterm mechanical process that will be studied concerns rutting as a result of a large number of vehicle axle passages. In a exible pavement, rutting is formed for the main part by the plastic deformations generated in the granular layers. A model that describes the complete response during each individual load cycle is regarded as unsuitable, since such a model requires a large amount of computational eort to simulate the structural behaviour during exible pavements rutting periods. Consequently, simplied approaches have to be developed. We will present rst a bibliographic study of elastoplastic or equivalent viscoplastic models for large cycle numbers used for unbound granular materials (UGM). Then, we will present the complete elastoplastic model based on the shakedown concept. Finally, the capabilities of the model to reproduce the long-term behaviour of the Miscillac sand (clayey sand, of 0=4 mm grading) [6], under repeated load triaxial tests will be checked. This material has been used as subgrade soil on the L.C.P.C. (Laboratoire Central des Ponts et Chauss! es) accelerated pavement testing e facility.
Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:577596

ELASTOPLASTIC MODEL FOR FLEXIBLE PAVEMENTS

579

2. UGM ELASTOPLASTIC MODEL FOR LARGE CYCLE NUMBERS The mechanical behaviour of UGM for roads is taken into account in pavement mechanics by means of two levels:
*

The rst one is related to current modelling and corresponds to pavement design methods. It is based on mechanistic pavement analysis using multilayer linear elastic model. The second one corresponds to advanced pavement design, where the cyclic triaxial test is the most widely used test to study the mechanical behaviour of UGM. The advantage of this test is the possibility to study the behaviour of the material under cyclic loadings, simulating accurately the in-situ conditions. Therefore, the short-term and the long-term mechanical behaviour are modelled separately. On one hand, the resilient behaviour of UGM for roads is modelled with non-linear elasticity [5] and on the other hand analytical models have been developed and give the vertical plastic strain function of the number of cycles [79], or a stress ratio [1012]. They are calibrated with laboratory repeated load triaxial tests with a constant or a variable conning pressure. They are, up to now, not used in pavement design or nite element modelling, though a calculation has been performed by de Buhan [13] on a railway track platform using Boussinesq solution and the analytical model of Hornych [14].

The constitutive cyclic plasticity has made great progress, and complete models have been developed which describe fairly well the monotonic, cyclic and ratchetting behaviour of materials like sands with hydro-mechanical coupling, taking into account the liquefaction [15], and the proceedings edited by Arulanandan [16, 17] for the VELACS benchmark. But these models will lead to unrealistic calculation time for large cycle numbers and for nite elements modelling of realistic boundary value problems. Nevertheless, within the plasticity framework, some authors have dened simplied approaches with accumulation laws [18, 19]. But, these last models have never been tested with UGM for roads under repeated load triaxial tests. Currently, equivalent viscoplastic models [2022] have been developed which seem to be promising. Several researchers [2326], who related the magnitude of the accumulated plastic strain to shear stress level concluded that the range of behaviours, obtained at various stress ratio for repeated load triaxial tests, can be described using the shakedown concept. But their results show that the short-term behaviour is used to determine if the shakedown state or the failure state is obtained. Consequently, no calculation of plastic strains is performed and no estimation of the rut depth can be done. Therefore, our objective will be to keep the short-term and longterm modellings, and the shakedown concept which avoids the step-by-step calculation. The plasticity framework will be used to determine the amount of plastic strains under large cycle numbers.

3. PRINCIPLES OF THE SIMPLIFIED METHOD When a structure is subjected to the action of cyclic loadings, the displacements and strains evolutions with the number of load repetitions are dened in three forms: elastic shakedown, plastic shakedown and ratchetting. When the structure is safe, the progressive strain studies require the stabilized stresses and strains knowledge. Thus, the use of step-by-step method to
Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:577596

580

T. HABIBALLAH AND C. CHAZALLON

determine the limit state is expensive. The shakedown analysis for the prediction of the stress and strain limit state can be an alternative to the step-by-step method. The present approach is based on the theoretical formulations proposed by Zarka [27] for metallic structures subjected to cyclic loadings. The main idea is to introduce a group of internal variables to characterize the local behaviour of the structure inelastic mechanisms and a group of internal structural parameters related to the rst group by non-negative symmetrical matrices. On one hand, if these coupling matrices are regular, the nature of the stabilized state is directly obtained starting from an elastic calculation. On the other hand, for a radial loading, only elastic or plastic shakedown can be obtained. Then, a local geometrical construction in the internal structural parameters plane is used to give an estimate of the stabilized structure state and the associated inelastic component (residual stresses and plastic strains elds). This inelastic analysis has been applied to an unbound granular material model with the linear kinematic hardening assumption and a non-associated ow rule. 3.1. Global evolution of a structure We consider now an elastoplastic structure with a nite volume V with boundary dV (general problem), which is subjected to:
* * * *

body forces Xjd x; t in V; surface forces Fid x; t on dFi V of dV; surface displacements Ujd x; t on dUj V of dV; initial strain eI x; t 0 in V ij

dFi V and dUj V remain constant throughout the loading history, dF V [ dU V dV and dF V \ dU V f: We shall limit ourselves to the case of small deformations and to the case of quasi-static analysis. The elasticity is linear, isotropic and elasticity coecients are not time- and temperature-dependent. Thus, the mechanical problem can be solved as follows: eij x; t Mijkl skl x; t ep x; t eI x; t ij ij ep x; t ij 1

sij x; t is the actual stress tensor, eij x; t the actual strain tensor, the plastic strain tensor, eI x; t the initial strain tensor, Mijkl the compliance elasticity matrix, where eij x; t is ij kinematically admissible with Ujd x; t on dUj V and sij x; t is statically admissible with Fid x; t on dFi V and with Xjd x; t in V: We can split this mechanical problem into an elastic problem and an inelastic problem. 3.2. Elastic problem We assume that the structure response remains virtually elastic, the elastic problem is solved as follows: eel x; t Mijkl sel x; t eI x; t ij kl ij 2

where eel x; t is kinematically admissible with Ujd x; t on dUj V and sel x; t is statically ij ij admissible with Fid x; t on dFi V and with Xjd x; t in V: Thus, the elasticity elds Uiel x; t and eel x; t can be calculated with an elastic analysis and the ij previous boundary conditions and the compliance elasticity matrix Mijkl :
Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:577596

ELASTOPLASTIC MODEL FOR FLEXIBLE PAVEMENTS

581

3.3. Inelastic problem An inelastic problem is obtained by dierence between the general problem and the elastic problem eine x; t eij x; t eel x; t ij ij The inelastic strains eld is represented as follows (Figure 1): eine x; t Mijkl rkl x; t ep x; t ij ij 4 3

eine x; t is kinematically admissible with 0 on dUj V rij x; t is the residual stress eld. It is the ij dierence between the actual stress and the elastic stress elds rij x; t sij x; t sel x; t ij 5

rij x; t is statically admissible with 0 on dFi V and with 0 in V: If the plastic strains eld is known, the inelastic elds Uiine x; t and eine x; t can be calculated ij with an elastic analysis and a null stress boundary condition, ep x; t as a xed initial strain, and ij the compliance elasticity matrix Mijkl : Then, the residual stress is calculated as follows:
1 rij x; t Mijkl eine x; t ep x; t kl kl

We can write schematically: rij x; t Zijkl ep x; t kl 7

This general formalism has been used by Zarka [27] to develop a simplied method which determines the stabilized state of metallic structures subjected to cyclic loadings. The authors have extended this simplied method to UGM with:

el

ine

el
M.

I
M .
el

Figure 1. Stressstrain relationships.


Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:577596

582

T. HABIBALLAH AND C. CHAZALLON

} the DruckerPrager yield surface has been used with a linear kinematic hardening and a non-associated ow rule. The plastic potential of von Mises has been used. q f 1 Sij yij Sij yij k aI1 sij 8 2 where I1 sij and Sij dev sij are, respectively, the rst invariant and the deviatoric part of the actual stress tensor. The kinematic hardening variable is related to the plastic strain 2H p e 9 yij 3 ij where H is the kinematic hardening modulus. q g 1 Sij yij Sij yij cste 10 2 } In most of the soil mechanics models, the hardening modulus is split in a volumetric and a deviatoric part, respectively (positive or negative hardening), which depend on the stress state history. However, in Equation (9), the hardening modulus H is a constant and we will see in Section 4.3 how its dependency with the stress state is taken into account. The elastic problem solved, the inelastic problem solution solves the general problem. Zarka [27] introduces a structural parameters eld in order to solve the inelastic problem. This structural transformed parameters eld is introduced in order to express the yield criterion with the elastic stress eld: Yij x; t yij x; t dev rij x; t 11

The yield surface expression in the deviatoric plane is f Sij yij 40: If we split the actual deviatoric stress in an elastic and inelastic part Sij x; t S el x; t dev rij x; t ij The yield surface expression becomes f S el dev rij yij 40 ij According to Equation (11), we obtain f S el Yij 40 ij 14 13 12

In the structural transformed parameters plane, the yield surface is a circle centred in S el x; t: ij Consequently, if the elastic stress eld S el x; t is known, the yield surface position is known at ij any time of the loading. The inelastic strains eld expression according to the kinematic hardening variable is expressed using Equations (4) and (9): eine x; t Mijkl rkl x; t ij
Copyright # 2005 John Wiley & Sons, Ltd.

3 yij x; t 2H

15

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:577596

ELASTOPLASTIC MODEL FOR FLEXIBLE PAVEMENTS

583

According to Equation (11), we can write 3 Yij x; t 16 2H 0 We have to dene the new compliance elasticity matrix Mijkl : According to Equation (16), the rst part of the inelastic strain formula can be developed as follows:   1 n0 n0 0 Mijkl rij 17 rij 0 trrij Iij 0 E E
0 eine x; t Mijkl rkl x; t ij

where Iij is the identity matrix and E0 EH EH   1 0 0 n n E E 2H

18

Thus, if Yij x; t is known at any point of the structure, an elastic calculation with the xed initial 3 0 strain 2H Yij x; t; a null stress boundary condition and the modied compliance matrix Mijkl ; solves the inelastic problem. Thus, the inelastic displacements and strain elds Uiine x; t and eine x; t are obtained. Then, the residual stress eld will be deduced: ij   3 01 ine Ykl x; t rij x; t Mijkl ekl x; t 19 2H We can write schematically # rij x; t Z ijkl  3 Ykl x; t 2H  20

Finally we calculate the plastic strains eld using Equation (9): 3 3 yij x; t Yij x; t dev rij x; t ep x; t 21 ij 2H 2H Consequently, a one-to-one relation exists between the kinematic hardening variables eld yij x; t and the structural transformed parameters eld Yij x; t: At any time t and for a given kinematic hardening variables eld yij x; t; there is a unique residual stress eld rij x; t; total stress eld sij x; t and structural transformed parameters eld Yij x; t: On the contrary, for a given structural transformed parameter eld Yij x; t; there is a unique residual stress eld rij x; t; total stress eld sij x; t and internal parameter eld yij x; t such as rij x; t is statically admissible with 0 in V and with 0 on dFi V and the corresponding inelastic strain eld is statically admissible with 0 on dUj V: Within this framework and for cyclic loadings, this last result will be used to build a solution satisfying all the static and kinematic conditions. 3.4. Structural transformed parameters determination for cyclic loading Under cyclic loading, the elastic stress eld can be expressed by the following expression: sel x; t 1 Ltsel min x Ltsel max x ij ij ij 22 where sel min x and sel max x are, respectively, the minimal and maximal elastic response. Lt is ij ij a monotonic periodic scalar function, which varies between 0 and 1.
Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:577596

584

T. HABIBALLAH AND C. CHAZALLON

The shakedown theory for perfectly plastic materials is based on the static theorem of Melan [28] and the Koiter [29] kinematic theorem. The Melan theorem stipulates that if there is a constant residual stress eld rij x strictly plastically admissible f sij 50 in any point of the volume V; elastic shakedown will occur. Mandel [30] generalizes the MelanKoiter theorem to materials with kinematic hardening. Then, with the concept of generalized standard materials [31], Mandel [32] describes the elastic shakedown behaviour of a structure with a kinematic hardening material, and determines the constant residual stress eld and the constant kinematic hardening variable eld which satisfy the condition f sel x; t ij rij x; yij x; t50: Nevertheless, with the new formalism (Section 3.3) the previous results are kept, and we will see in the next paragraph how a mechanical loading is described and how we can obtain the limit state nature and determine the associated stress and strain elds when a mechanical cyclic loading is performed. For a given stress sij x; t and hardening parameter yij x; t; the local stress at the level of the plastic mechanism is: * sij x; t Sij x; t yij x; t 23

with Sij x; t dev sij x; t. In the local stress plane, the plasticity convex domain C0 is a xed cone which is reduced, in the deviatoric plane, to a circle centred on the isotropic stress axis (Figure 2(a)). The normality law is written with the Moreaus notation [33]: ij * ep 2 @cC0 sij * with sij 2 C0 24

* * @cC0 sij is the subdierential to the convex C0 at sij ; where the plastic strain rate is an external normal to the convex C0 :

q
(qmax , pmax )

. ijp

. ijp

el S ij

. C (S
. ijp

el ij

y .
ij

C( y )
ij

(a)

(b)

Figure 2. Stress path in the p; q plane: (a) deviatoric plane; and (b) structural transformed parameters plane.
Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:577596

ELASTOPLASTIC MODEL FOR FLEXIBLE PAVEMENTS

585

Using Equation (11), the structural transformed parameter at the level of the inelastic mechanism is:
el el * * sij Sij Yij ) Yij sij Sij el el el Yij 2 CSij and CSij C0 Sij

25 26

el Equation (26) implies that Yij belongs the convex set CSij obtained from C0 with the el translation Sij (Figure 2(b)). The normality law is

ij ep 2 @cCSel Yij
ij

el with Yij 2 CSij

27

@cCSel Yij is the subdierential to the convex CS el at Yij x; t; where the plastic strain rate is ij ij an internal normal to the convex CS el : This convex is locally built for each plastic mechanism. ij Thus, in the structural transformed parameters plane, the yield surface is a circle centred in S el x; t: This convex plastic domain undergoes translations with the hardening since its centre is ij related to S el x; t (Figure 2(b)). ij Using Equation (9) we rewrite Equation (11) in an incremental form 2H p Yij x; t 28 e x; t dev rij x; t 3 ij The normality law is rewritten for the rst member of Equation (28): 2H p 2H @cCSel Yij x; t 29 e x; t 2 ij 3 ij 3 The second member of Equation (28) is rewritten using Equation (7): ekl dev rij x; t devZijkl p x; t Finally, we can write   ij x; t 2 2H dev Zijkl @cCS el Yij x; t Y ij 3 30

31

el Equation (31) implies that the convex CSij carries onto its boundary the Yij x; t eld. Thus, a given plastic mechanism is active only if its structural transformed parameters is on the el boundary of the convex CSij : Consequently, when the material is subjected to a mechanical cyclic loading, one can follow the plasticity convex evolution in the structural transformed parameters plane and determine the Yij x; t eld corresponding to the obtained limit state. This limit state is determined with the knowledge of the mean value Yij m x and the range DYij x: el el We denote Sij min x and Sij max x the minimal and maximal deviatoric elastic stress. The periodic elastic stress is expressed by el el el Sij x; t Sij min x LtDSij x

32

el el el with DSij x Sij max x Sij min x: In the structural transformed parameters plane and according to the loading amplitude el DSij x; two cases exist:
*

el el The extreme positions of the convex centred in Sij min x and Sij max x have a common part Cl ; in this case, elastic shakedown will occur when t becomes large. The extreme positions of the convex have no common part, in this case, plastic shakedown will occur when t becomes large.

Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:577596

586

T. HABIBALLAH AND C. CHAZALLON

Cl '
Y
ij

Cl
Y ij
el S ijmax
Y =Y

S
Y

el ijmin

ij

Y ij

Cl '
Figure 3. Structural transformed parameter determination for elastic shakedown.

3.4.1. Elastic shakedown. During a cyclic loading, the convex which characterises the behaviour of a given point of the structure, undergoes translations between the extreme elastic stresses existing at this point. If Y0 ij is the initial value of the structural transformed parameters eld, Y0 ij can be transported along the mechanical loading. Three cases can be obtained (Figure 3):
* *

Y0 ij is inside Cl ; in this case Y0 ij remains immobile and the behaviour is purely elastic. Y0 ij is such that, after the rst cycle, it reaches the boundary of Cl and remains immobile. The solution or the new position Y1 ij of the point Y0 ij is obtained by orthogonal projection on Cl : Elastic shakedown is obtained and the inelastic and plastic strains elds are obtained with Equations (16) and (21). All other initial states, where Y0 ij is transported with the convex movements and nishes on one of the two points Cl 0 : In this case, the stabilized state is reached after several cycles and elastic shakedown is obtained. The inelastic and plastic strains elds are obtained with Equations (16) and (21).

el 3.4.2. Plastic shakedown. When the loading DSij x becomes very large, Zarka [27] proves that a stationary state is reached and that the kinetics of the transformed internal parameters and the deviatoric elastic stress are the same el DSij x DYij x

33

In the structural transformed parameters plane, and with (17) we determine the structural el transformed parameters eld Yij max x which belongs to the convex CSij max x; and the el structural transformed parameters eld Yij min x which belongs to the convex CSij min x; with   3 el el # DSkl x; t Dyij x; t DSij x; t dev Z ijkl 34 2H and Dyij x; t
el Sij Yij H Dl q 3 1 el S Y S el Y 2 ij ij ij ij

35

Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:577596

ELASTOPLASTIC MODEL FOR FLEXIBLE PAVEMENTS

587

We have s    3 3 ij x; t ij x; t Dl 2 y y 2H 2H and Dl is known. Finally, with Equation (35), we obtain


el Yij max x Sij max x k aI1 sij max x

36

3 3 el el # DSij x dev Z ijkl DSkl x DlH 2H

!! 37

Yij min x with

el Sij min x

!! 3 el # ijkl 3 DSel x DSij x dev Z k aI1 sij min x kl DlH 2H

38

el 0:5 el el el # # =2H Dl 3 2 DSij x dev Zijkl 3DSkl x=2H DSij x dev Z ijkl 3DSkl x=2H

39

For an intermediate loading, the structural transformed parameters eld is determined by geometrical conditions. Its value is the distance between the extreme positions of the el el mobile convex centred in Sij min x and Sij max x in the structural transformed parameters plane (Figure 4). We obtain
el el el DYij x Yij max x Yij min x DSij x1 R1 x R2 x=DSij xDSij x=20:5 40 p p p p with R1 x 3aI1 sel min x 3k; R2 x 3aI1 sel max x 3k; R1 x and R2 x are ij ij el el the radius of the convex centred in Sij min x and Sij max x: This result has been obtained by Zarka [27] with the yield surface of von Mises and we extend it to the DruckerPrager yield surface [34]. Thus, under a stress-driven radial cyclic loading, we can determine the extreme stresses and strains of the stabilized behaviour using few calculations. This method has been implemented in the nite elements code Cast3M [35] and we are going to present now results of model parameters identication.

R2 R1
S ij min
el

Yij

moy

Sij max

el

Yij

Figure 4. Structural transformed parameters determination for plastic shakedown.


Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:577596

588

T. HABIBALLAH AND C. CHAZALLON

4. CALIBRATION OF THE SIMPLIFIED MODEL The simplied method requires the determination of elasticity parameters, DruckerPrager parameters c (the elasticity cone aperture) and pn (the apex of the DruckerPrager cone on the isotropic stress axis) and nally, the hardening modulus H: The model parameters identication procedure has required monotonic and repeated load triaxial tests performed on Missillac sand at the L.C.P.C. by Hornych [6]. The Missillac sand mechanical characteristics are shown in Table I. In order to maximize the information obtained from each permanent deformation test, a dierent test procedure has been used. This multi stage procedure, developed by Gidel [36] consists, in each permanent deformation test, to perform successively several cyclic load sequences, following the same stress path, with the same q=p ratio, but with increasing stress amplitudes. 4.1. Elasticity parameters identication The elasticity parameters are determined using repeated load triaxial test results, performed for resilient behaviour study [37]. The test procedure consists of a conditioning test, following a stress path with high stress level (q 90 kPa; q=p 2) and a series of short cyclic loadings according to 22 stress paths with dierent stress ratios q=p: The conditioning test intended to stabilize the permanent strains of the material, and to attain the resilient behaviour. The wellknown non-linear elasticity model Ky [3] is chosen for the test results adjustment [38]. Its expression is  K2 p E K1 41 pa where p is the mean stress and pa is a reference mean stress equal to 100 kPa: K1 and K2 are model parameters. The Poisson ratio is constant. With the Ky model and a given stress path, we are able to solve the elastic problem (Section 3.2) with a step-by-step calculation and to obtain the initial and the nal plasticity convex positions. Then, the inelastic problem (Section 3.3) will be solved with the following Youngs modulus expression and a constant Poisson ratio:   pmax K2 E K1 42 pa where pmax is the maximum mean stress of the stress path. A representative set of elasticity parameters is listed in Table II. 4.2. DruckerPrager parameters identication The DruckerPrager parameters c and pn ; and the characteristic line jdil [39] are determined using the results of three monotonic triaxial tests [6]. According to Equation (8), the yield

Table I. The mechanical characteristics of the Miscillac sand [6]. Material Missillac sand
Copyright # 2005 John Wiley & Sons, Ltd.

% nes 7.5

rOMP kg=m3 2040

wOMP kg=m3 9.2

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:577596

ELASTOPLASTIC MODEL FOR FLEXIBLE PAVEMENTS

589

Table II. Representative values of elasticity parameters [38]. Material Missillac sand K1 82:15 kPa K2 0.04 n 0.2

Table III. Representative values of DruckerPrager parameters. Material Missillac sand c (kPa) 12.26 j 448 c 158 pn (kPa) 12.78 jdil 398

surface is a circular cone in the p; q plane. In order to make the DruckerPrager circle coincide with the outer apices of the MohrCoulomb hexagon, the parameters k and a are expressed by 2 sin j a p 33 sin j and 6c cos j k p 33 sin j 43

where c is the cohesion and j the angle of internal friction. The apex of the DruckerPrager cone is obtained with the cohesion c and the angle of internal friction j: pn c=tg j: The parameters c is determined in order to obtain a reduced initial elastic domain. Indeed, just before the plastic ow the elastic strain must be 105 for low stress path ratio q=p: A representative set of DruckerPrager parameters for the Miscillac sand are listed in Table III.

4.3. Plasticity parameters identication The hardening modulus determination requires an adjustment on repeated load triaxial tests results performed for permanent strains study, with dierent stress ratios q=p and dierent stress levels. The loading program is shown in Figure 5, where each loading stage is carried out with 10 000 cycles and growing stress levels. The evolution of the vertical plastic strain with the number of load repetitions is shown in Figure 6. We observe that the plastic strain tends toward a stabilization for all the loading stages with the exception of the fourth loading stage of the stress path q=p 3: For this case, failure occurs in the rst thousand cycles. From the axial and the radial plastic strain obtained at the end of each loading stage, the evolution of the volumetric plastic strain with the mean stress of each loading stage is shown in Figure 7. We notice that for all stress paths with stress ratio lower than 2, material tends to compact and shakedown states are obtained. For the stress path q=p 2; a low compaction occurs for the rst loading stage, a low dilation for the second loading stage, nally dilation occurs for all the over loading stages. Notwithstanding, shakedown states are obtained for all the stages. For the stress path q=p 3; dilatation and shakedown states occur for all loading stages except the fourth stage. When dilatation is observed, failure may occur inside the material, and this loading stage will not be used in the calibration procedure of the hardening modulus H:
Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:577596

590
160 120 q (kPa) 80 40 0 0

T. HABIBALLAH AND C. CHAZALLON

q/p = 1 q/p = 1.5 q/p = 2 q/p = 3


20 40 60 80 100 120 140

p (kPa)

Figure 5. Loading program.

3.5E-02 q/p = 1 3.0E-02 2.5E-02 2.0E-02 1.5E-02 1.0E-02 5.0E-03 0.0E+00 0 10000 20000 30000 40000 Number of load repetitions 50000 60000 q/p = 1.5 q/p = 2 q/p = 3

axial plastic strain

Figure 6. Evolution of the vertical plastic strain under axi-symmetrical stress conditions. Conning pressure: 10 kPa:

Each loading stage is carried out with 10 000 cycles and this number of load repetitions may be insucient to reach the stabilised state for others UGM [34, 36]. Nevertheless the same procedure can be followed. As already stated, the simplied elastoplastic model is based on the shakedown theory and gives the stabilized plastic strains. Thus, the hardening modulus calibration has to be performed using limit state plastic strains. One of the rst interpretations applied to results of permanent deformation tests consisted in describing the variation of permanent axial deformations with the number of load cycles N; leading to relationships of the type ep f N: 1 Thus, the limit state plastic strains are calculated in agreement with the Hornych model [14]   ! N B p e1 A 1 44 100 where A and B are regression parameters. ep is the vertical plastic strain. 1
Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:577596

ELASTOPLASTIC MODEL FOR FLEXIBLE PAVEMENTS

591

1.0E-02

volumetric plastic strain

5.0E-03 0.0E+00
q/p = 1

-5.0E-03 -1.0E-02 -1.5E-02 -2.0E-02 -2.5E-02 0 20 40 60 80 100 120

q/p = 1.5 q/p = 2 q/p = 3

140

p (kPa)

Figure 7. Evolution of the volumetric plastic strain under axi-symmetrical stress conditions. Conning pressure: 10 kPa:

2.0E-02 1.8E-02
Experiment - q/p = 1 Experiment - q/p = 1.5 Experiment - q/p = 2 Experiment- q/p = 3 Hornych model - q/p = 1 Hornych model - q/p = 1.5 Hornych model - q/p = 2 Hornych model - q/p = 3

vertical plastic strain

1.6E-02 1.4E-02 1.2E-02 1.0E-02 8.0E-03 6.0E-03 4.0E-03 2.0E-03 0.0E+00 0 20 40 60 80

100

120

140

p (kPa)
Figure 8. Comparison between the plastic strain at the end of each stress stage and the calculated stabilized plastic strain.

The permanent deformation is growing up until a limit value A as N increases toward innity. B controls the shape of the plastic strains curve evolution. Consequently, Hornych model calibration on each loading stage gives an estimation of the corresponding limit state plastic strain. Figure 8 shows a comparison between the plastic strain at the end of each stress stage and the corresponding stabilized plastic strain plotted as a function of the loading stage mean stress. In Figure 8, we can observe an overall good agreement, we can state that the limit state is reached with the number of load repetitions. Finite element modelling of repeated load triaxial tests have been performed with the nite element code Cast3M [35]. The elastic stress and strain elds are determined with the Ky model
Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:577596

592
5.5 5.3 5.1 4.9 4.7 4.5

T. HABIBALLAH AND C. CHAZALLON

Log(H/L)= -1.2714 Log(p/pa) + 4.5189 q/p = 1.5 ; R2= 0.98

Log(H/L)= -0.9911 Log(p/pa) + 4.8044 q/p = 1 ; R2= 0.98

Log (H / L)

q/p = 1 4.3 4.1 3.9 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 Log(H/L)= -1.469 Log(p/pa) + 3.5832 q/p = 3 ; R2= 0.98 Log(H/L)= -1.5705 Log(p/pa) + 4.0601 q/p = 2 ; R2= 0.97 q/p = 2 q/p = 3 q/p = 1.5

Log (p / pa )
Figure 9. Hardening modulus evolution law.

and the Table II parameters. The minimal and maximal values of the elastic stress eld are used to determine the nature of the shakedown state. The hardening modulus identication is carried out using elasticity and DruckerPrager parameters (Tables II and III) and the calculated stabilized plastic strains. A hardening modulus H is determined for each loading stage. We assume that the evolution of the calculated hardening modulus divided by the stress path length is linear with the stress level in a bi-logarithm diagram (Figure 9). Thus, we can write     H p Log a Log b 45 L pa Finally, the hardening modulus expression is H 10a1Logp=pa Log p
1q=p2 b

pa

46

where pa is the atmospheric pressure, a and b are material parameters. A linear regression is considered as evolution law for the parameters a and b with the q=p ratio. The relative position of the stress path with the dilatancy line and its inuence on the amount of vertical plastic strain is taken into account in the evolution law of a and b parameters by means of a break point (Figures 10 and 11). This last result has been obtained also with two others UGM for roads [34]: a calcareous material resulting from the Sorr" ze quarry (grading: e 0=20 mm), and a micro-granite resulting from the Poulmarch quarry (grading: 0=10 mm) [36]. With the proposed evolution laws, we can determine the hardening modulus for a stress path which has not been used for the calibration. Consequently, we can have an estimate of the hardening modulus for any q=p and any mean stress pmax : This result will be very useful when the nite element modelling of a exible pavement will be performed.
Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:577596

ELASTOPLASTIC MODEL FOR FLEXIBLE PAVEMENTS

593
3 3.5

0 -0.6 -0.8

0.5

1.5

2.5

q/p = 1.6 -1 -1.2

a = -1.507 q/p + 0.8 R2=0.84 a = 0.1015 q/p - 1.7735

-1.4 -1.6 -1.8 q/p

Figure 10. Variation of a with the stress ratio q=p:

5.5 q/p = 1.6 5

4.5 b

b = -1.2816 q/p + 6.3 R2=0.83

b = -0.4769 q/p + 5.0139

3.5

3 0 0.5 1 1.5 q/p


Figure 11. Variation of b with the stress ratio q=p:

2.5

3.5

5. NUMERICAL SIMULATIONS 5.1. Comparisons experiment/model Repeated load triaxial tests have been modelled with the proposed model and the loading stress paths shown in Figure 6. The comparison between the calculated stabilized plastic strains and
Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:577596

594
2.0E-02

T. HABIBALLAH AND C. CHAZALLON

s.p.s : stabilized plastic strain


experimental data s.p.s : q/p = 1 experimental data s.p.s : q/p = 1.5 experimental data s.p.s : q/p = 2 experimental data s.p.s : q/p = 3

1.6E-02
Vertical plastic strain

1.2E-02

simulation q/p = 1 simulation q/p = 1.5 simulation q/p = 2

8.0E-03

simulation q/p = 3

4.0E-03

0.0E+00 0 20 40 60
p (kPa)

80

100

120

140

Figure 12. Comparison between the predicted plastic strains and tests results.

the model, with the proposed evolution law of the hardening modulus H; is plotted Figure 12. We can observe an overall good agreement. The model predictions capture the general trend of the calculated stabilized vertical plastic strains. For the stress paths q=p 1 and 3, modelling is very close to experiment and for the stress paths q=p 1:5; stabilized vertical plastic strains are reasonably well predicted. However, the eects of the stress ratio q=p and mean stress are captured quite accurately. Indeed, simulations show that the stabilized vertical plastic strain increases with the stress ratio q=p and/or the mean stress. 5.2. The simplied method within a nite element modelling of a full-scale experiment Currently, the nite element modelling of a exible pavement with this simplied method is in progress. It has to be performed with the following strategy:
*

Model parameters
* *

A linear elastic model is used for the bituminous concrete layer. For each UGM which constitutes an unbound layer of the exible pavement: - Elasticity parameters K1 ; K2 ; n are required. - Plasticity parameters c; pn ; c; jdil ; and H with the evolution laws: ap; q; bp; q are required.

* * *

Description of the exible pavement, mesh and boundary conditions. The inuence of the principal stresses rotation on the rut depth is not taken into account with the simplied model. Consequently, the applied load is a cyclic loading with a circular plate that we consider as equivalent to the French standard axle load (dual wheel half-axle loaded). el el Calculation of Smin x and Smax x elds with the Ky model for UGM layers. Determination of the shakedown state. Calculation of the corresponding inelastic displacement and plastic strain elds.
Int. J. Numer. Anal. Meth. Geomech. 2005; 29:577596

Copyright # 2005 John Wiley & Sons, Ltd.

ELASTOPLASTIC MODEL FOR FLEXIBLE PAVEMENTS

595

6. CONCLUSION Rutting of exible pavements which occurs in the unbound granular sublayers put in evidence the lack of precision in the design method. These empirical design methods based on elasticity cannot take into account the plastic strain occurring in the granular media. Our approach is based on the main laboratory test which is used to study the resilient behaviour and permanent deformations of UGM for roads: the repeated load triaxial test. The UGM behaviour modelling under large cycle numbers is complex. The shakedown theory, widely used for metallic structures, can be an alternative to the step-by-step methods, based on incremental rheological models, which will lead to unrealistic calculation time for large cycle numbers. The simplied method based on the shakedown theory and developed by Zarka is able to reproduce the purely elastic behaviour, the elastic shakedown behaviour and the plastic shakedown behaviour of steels. His simplied method has been modied and adapted for exible pavement modelling: the DruckerPrager yield surface and the plastic potential of von Mises, with a linear kinematic hardening have been used. The plastic potential of von Mises takes into account the rut depth instead of the rut volume, but this result is sucient since there is no test to estimate ruts volume. The model parameters identication requires three monotonic triaxial tests for the Drucker Prager parameters and characteristic line determination. One repeated load triaxial test is required for the resilient behaviour study and the calibration of the non-linear elasticity parameters. The hardening modulus determination requires only test results of three repeated load triaxial tests performed for the plastic behaviour study. Those tests have to be carried out with at least three dierent stress ratios and each test has to be performed with at least three loading stages. When the asymptotic behaviour is not obtained at the end of a stage of a multistage loading, the stabilized plastic strain has been estimated with the hyperbolic model of Hornych. Finally, a nite element modelling of a repeated load triaxial test gives the hardening modulus which is expressed as a function of the q=p ratio and the radial path length. Hardening modulus interpolation laws give very satisfactory results when the comparison has been done with multistage tests on the Miscillac sand. The eects of the stress ratio q=p and mean stress variations are captured quite accurately. The model is adapted for the structural analysis with the nite element method. Indeed, hardening modulus expressed as a function of the loading parameters can be determined at any point of the structure. Although the simplied model has been tested on various UGM, a comparison with the results of a full-scale instrumented experiment remains to be done.

REFERENCES ! 1. Brown SF. 36th Rankine lecture: soil mechanics in pavement engineering. Geotechnique 1996; 46(3):383426. 2. COURAGE. Construction with unbound road agregates in Europe. Final Report. Oce for Ocial Publications of the European Communities, 2000. 3. Hicks RG, Monismith CL. Prediction of the resilient response of pavements containing granular layers using non linear elastic theory. Proceedings of the 3rd International Symposium on Asphalt Pavement, Seattle, U.S.A., vol. 1, 1972; 410429. 4. Hornych P, Kazai A, Quibel A. Modelling of full scale experiment of two exible pavement structures. Proceedings of the 4th International Conference on Unbound Aggregates in Road Construction, Nottingham, U.K., 2000; 359367. 5. COST 337. Unbound granular materials for road pavement. Final Report. Oce for Ocial Publications of the European Communities, 2004. " " 6. Hornych P, Battard G. Etude du comportement des sols supports de chauss! ees a lessai triaxial a chargements e r! p! t! s. Rapport Interne, Laboratoire Centrale des Ponts Chauss! es, France, 2003. e ee e Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:577596

596

T. HABIBALLAH AND C. CHAZALLON

7. Barksdale RD. Laboratory evaluation of rutting in base course materials. Proceedings of the 3rd International Conference on Structural Design of Asphalt Pavement, London, U.K., 1972; 161174. 8. Paute JL, Jouve P, Martinez P, Regneau E. Mod" le de calcul pour le dimensionnement des chauss! es souples. e e Bulletin de liaison des Laboratoire des Ponts et Chauss!es 1988; 156:2136. e 9. Swere GTH. Unbound granular bases for road. Ph.D. Thesis, University of Delft, The Netherlands, 1990. 10. Pappin JW. Characteristics of a granular material for pavement analysis. Ph.D. Thesis, University of Nottingham, U.K., 1979. 11. Brown SF, Hyde AFL. Signicance of Cyclic conning stress in repeated load triaxial testing of granular materials. Transportation Research Record 1985; 537:4958. 12. Lekarp F, Dawson A. Modelling permanent deformation behaviour of unbound granular materials. Construction and Building Materials 1998; 12(1):918. 13. Abdelkrim M, De Buhan P, Bonnet G. A computational procedure for predicting the long term residual settlement of a platform induced by repeated trac loading. Computers and Geotechnics 2003; 463476. 14. Hornych P, Corte JF, Paute JL. Etude des d! formations permanentes sous chargements r! p! t! s de trois graves non e e ee ! trait! es. Bulletin de liaison des Laboratoires des Ponts et Chaussees 1993; 184:4555. e 15. Popescu R, Prevost JH. Comparison between VELACS numerical class A predictions and centrifuge experimental soil test results. International Journal of Soil Dynamics and Earthquakes Engineering 1995; 14(2):7992. 16. Arulanandan K, Scott RF. Proceeding of the International Conference on Verication of Numerical Procedures for the Analysis of Soil Liquefaction Problem, vol. 1. Balkema: Rotterdam, 1993. 17. Arulanandan K, Scott RF. Proceeding of the International Conference on Verication of Numerical Procedures for the Analysis of Soil Liquefaction Problem, vol. 2. Balkema: Rotterdam, 1994. 18. Bonaquist RF, Witczak MW. A comprehensive constitutive model for granular materials in exible pavements structures. Proceedings of the 4th International Conference on Asphalt Pavement, Seattle, U.S.A., 1997; 783802. 19. Desai CS. Mechanistic pavement analysis and design using unied material and computer models. Proceedings of the 3rd International Symposium on 3D Finite Elements for Pavement Analysis, Amsterdam, 2002; 163. 20. Suiker ASJ. The mechanical behaviour of ballasted railway tracks. Ph.D. Thesis, Delft University, The Netherlands, 2002. 21. Mayoraz F. Comportement m! canique des milieux granulaires sous sollicitations cycliques: application aux e fondations des chauss! es souples. Ph.D. Thesis, Ecole Polytechnique F! d! rale de Lausanne, Switzerland, 2002. e e e 22. Suiker ASJ, de Borst R. A numerical model for cyclic deterioration of railways tracks. International Journal for Numerical Methods in Engineering 2003; 57:441470. 23. Sharp R, Booker J. Shakedown of pavements under moving surface load. Journal of Transportation Engineering 1984; 110(1):114. 24. Yu HS, Hossain MZ. Lower bound shakedown analysis of layered pavements discontinuous stress elds. Computer Methods in Applied Mechanics and Engineering 1998; 167:209222. 25. Collins IF, Boulbibane M. Geomechanical analysis of unbound pavements based on shakedown theory. Journal of Geotechnical and Geoenvironmental Engineering 2000; 126(1):5059. 26. Arnold G, Dawson A, Hughes D, Robinson D. The application of shakedown approach to granular pavement layers. Journal of Transportation Research Board 2003; 1819(2):194200. 27. Zarka J, Casier J. Elastic plastic response of structure to cyclic loading: practical rules. Mechanics Today, Nemat-Nasser (ed.), vol. 6. Pergamon Press: Oxford, 1979; 93198. 28. Melan E. Theorie statisch unbestimmter systeme aus ideal-plastischem. Bausta. Sitzber akad Wiss Wien II2 1936; 195218. 29. Koiter WT. General theorem for elastic-plastic solids. Progress in Solid Mechanics, vol. 1. North-Holland: Amsterdam, 1960. 30. Mandel J. G! n! ralisation de la th! orie de plasticit! de W.T. Koiter. International Journal of Solids and Structures e e e e 1965; 273296. 31. Halphen B, NGuyen QS. Sur les mat! riaux standards g! n! ralis! s. Journal of Mechanic 1975; 1428. e e e e ! 32. Mandel J. Adaptation dune structure plastique ecrouissable. Mechanic Research Communication 1976; 3:483488. ! ! 33. Moreau JJ. Rae par un convexe variable. Seminaire unilaterale, Montpellier, 1971. 34. Habiballah TM. Mod! lisation des d! formations permanents des graves non trait! es: application au calcul de e e e lorni! rage des chauss! es souples. Ph.D. Thesis, University of Limoges, France, 2003. e e 35. CAST3M. see web site http://www-cast3m.cea.fr/cast3m (2004). (CAST3M is a research FEM environment; its development is sponsored by the French Atomic Energy Commission). 36. Gidel G, Hornych P, Chauvin J-J, Breysse D, Denis A. Nouvelle approche pour l! tude des d! formations e e " permanentes des graves non trait! es a lappareil triaxial a chargement r! p! t! s. Bulletin de liaison des Laboratoire des e " e ee ! Ponts et Chaussees 2001; 522. " 37. Norme francaise NF P98 235 1 Mat! riaux non trait! s. Essais a chargement r! p! t! s, 1995. e e e ee 38. Hornych P. Rapport Interne Condentiel, L.C.P.C., 2003. 39. Luong MP. Stressstrain aspects of cohesionless soils under cyclic and transient loading. Proceedings of the International Conference on Soils under Cyclic and Transient Loading, Swansea, Wales, 1980; 353376. Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:577596

Вам также может понравиться