EXPERIMENTAL ANALYSIS OF YIELDING AND STRAIN LOCALIZATION OF MODERATELY OVERCONSOLIDATED KAOLIN CLAY

Amit Prashant1 (Member, ASCE) and Dayakar Penumadu2 (Member, ASCE)

ABSTRACT Elasto-plasticity theory has been commonly used to model the mechanical behavior of clays. Yielding and normalized failure conditions are the two key elements that are used in most constitutive models developed for frictional materials. For isotropically overconsolidated clays, the stress state will be inside the elastic zone. The limit of elastic deformations in three-dimensional stress space is usually defined by the initial yield surface. In prior research, the formulation for yielding of soil is strongly linked to its defined failure state of stress. However, the localized deformations within the specimen have not been considered during the formulation of previous theories; which may have significant influence on the definition of failure condition and thus the yielding behavior of clays in three-dimensional stress space. In order to study the yield and failure behavior of Kaolin clay, a series of strain controlled true triaxial undrained tests with constant intermediate principal stress ratio (b-value) were performed on cubical specimens. The true triaxial device used in the present research has flexible boundaries and uses three-axis electro-pneumatic ProportionalIntegralDerivative (PID) based real-time feedback control. Failure occurred at peak deviator stress followed by shear band formations and localized bulging. Prior theoretical formulations of bifurcation and undrained instability support these experimental observations. Keywords: Yielding, Localization, Three-Dimensional, Shear, Pore Pressure, OCR

INTRODUCTION A significant amount of progress has been made in the field of developing three-dimensional constitutive models for modeling the mechanical behavior of frictional materials. Due to the well-defined and established concept of the classical elasto-plasticity theory, it has been a favorable choice for many researchers (e.g. Roscoe and Burland 1968, Lade 1990, Whittle and Kavvadas 1994) to represent the complex stress-strain response of geomaterials. However, these models need to be improved in many aspects to best approximate the measured mechanical response of overconsolidated (OC) clays. Lack of sufficient 3-D test data has been a major issue
1 2

Department of Civil and Environmental Engineering, University of Tennessee, Knoxville, TN 37996, USA Corresponding author:223 Perkins Hall, Department of Civil and Environmental Engineering, University of Tennessee, Knoxville, TN 37996-2010, USA, email: dpenumad@utk.edu, fax:865-974-2669

in clearly identifying the influence of various segments of the complex behavior of OC clays. In a recent experimental study (Prashant and Penumadu 2004), a series of isotropically consolidated undrained true triaxial tests were performed on remolded cubical specimens of Kaolin clay. Before shearing, the specimens were moderately overconsolidated (overconsolidation ratio, OCR = 5) by initially consolidating the specimen to an effective confining stress of 275 kPa and then unloading to 55 kPa. During the stage of undrained shearing, the specimens were subjected to various total stress paths corresponding to different intermediate principal stress ratio b-values of 0, 0.25, 0.5, 0.75, and 1.0. The b-value is defined as b = (2-3)/ (1-3), where 1 is major, 2 is intermediate, and 3 is minor principal stress. All of these tests were compressive in nature in terms of total stress applied on the specimens during shearing. This means that even for the b=1 case, the shear stress was applied by increasing the same amount of normal stress in 1 and 2 directions and maintaining a constant total stress in the 3 direction, unlike triaxial extension test. Based on the data obtained from this series of tests, the present paper focuses on the elasto-plastic yielding behavior of overconsolidated Kaolin clay and its relationship with the failure conditions. The failure condition for the deforming specimen was observed to be strongly associated with the occurrence of strain localizations. The initial yield locations were identified using three relationships that were assumed to be bilinear for a certain range of shear strains. Based on the strain information obtained from the true triaxial test data, the authors present a method of identifying the onset of localized deformations. The discussion includes a theory about the possible factors causing strain localizations within the specimens under undrained condition. TRUE TRIAXIAL TESTING True triaxial tests (TTT) were performed using a device that uses three-axis electro-pneumatic Proportional Integral Derivative (PID) closed-loop control which applies three mutually perpendicular principal stresses on a 102 mm cubical soil specimens using flexible membranes (Prashant and Penumadu 2004). Total stress values applied along each direction of cubical specimen were measured using pressure transducers. The normal stress was recorded in the three principal directions. Deformation was recorded on all six faces of the specimen using linear variable displacement transducer (LVDT), and the displacement values at opposite faces were averaged to get the strain in a given direction. Internal (inside the specimen) and external (at boundary of the specimen) pore pressures were measured using absolute-pressure-transducers. Internal pore pressure was measured using a custom hypodermic needle with a ceramic porous plug and using an absolute pressure transducer. All of the electronic transducers were connected to a personal computer for use in data acquisition and control. This computer, using custom software, automates various phases of triaxial testing such as saturation, consolidation, and application of normal stresses for predetermined stress path or strain path during shearing stage. The cubical specimens were prepared by consolidating 155% water content slurry of Kaolin clay. The slurry was consolidated under Ko condition at an applied vertical stress of 207 kPa. Three axes of TTT system were identified as x, y and z-axis, and the specimens were assembled in the system such that the 1-axis at the end of slurry consolidation coincides with z-axis. The specimens were saturated using backpressure. Full saturation was ensured by checking the Skemptons parameter B to be 0.98. The specimens were isotropically consolidated to 275 kPa resulting in normally consolidated cubical clay specimens and then unloaded to 55 kPa corresponding OCR = 5. During undrained shearing, a strain rate of 0.05 %/min along 1-axis (z-axis) was found suitable for this clay to ensure pore pressure equilibrium. Strain rate along

z-axis was maintained at a constant value using PID real-time feedback control between the applied normal stress and corresponding axial deformation. STRESS-STRAIN RELATIONSHIP AND YIELD BEHAVIOR The specimens used in true triaxial testing were anisotropically consolidated during specimen preparation, which induced a small degree of anisotropy about z-axis. For various b-values, the specimens were sheared by applying the 1 along z-axis, 2 along y-axis, and 3 along x-axis. Figure 1 shows the stress-strain relationship and excess pore pressure evolution for various b-values. In this figure, the values of q and q were calculated from the general invariant expression.
2 2 2 q = ( 1 2 ) + ( 2 3 ) + ( 3 1 ) 2

(1)

q =

1 2 2 2 2 ( 1 2 ) + ( 2 3 ) + ( 3 1 ) 3

(2)

The failure point was defined as the point of peak deviator stress just before the specimen shears to large strains with negligible change in shear stress. This definition was necessary in this series of tests, since the specimens exhibited localized deformation at and beyond peak deviator stress. Figure 1 shows an interesting observation that the shear stress-strain relationships for various b-values were identical at low to medium shear strain levels; however, the q and q values at failure were significantly different. The excess pore pressure evolution was larger for higher b-values. Most of the difference in excess pore pressure evolution was due to higher rate of mean total stress change for higher b-values.
Excess Pore Pressure, u (kPa)

180 Deviator Stress, q (kPa) 150 120 90 60

Initial Yield b=0.5 - Failure Points Uniform deformation After occurrence of strain localization b=0.75 b=1.0 b=0 b=0.25

60
b=0.75 b=1.0

After occurrence of strain localization Uniform deformation - Failure Points

30

b=0.5 b=0.25 Initial Yield b=0

30 0 0 5

0 0 5 10 15 Shear Strain, q (%) 20

10 15 Shear Strain, q (%)

20

Fig. 1. Shear stress-strain relationship and excess pore pressure response To model the OC clay behavior using the classical elasto-plasticity theory, the initial hydrostatic stress state is assumed to be inside an elastic zone, whose boundary is defined by an initial yield surface. The elasticity is mostly assumed to be linear, which defines a linear shear stress-strain relationship for loading in elastic zone. Experimental identification of initial yield point is crucial however, in most cases, difficult due to highly non linear shear stress-strain

response of many OC clays. With due consideration to classical elasto-plasticity theories, the linear elastic zone of shear deformation (inside the initial yield surface), for OC clays, may only be defined at small strain values, where the stress-strain relationship can still be assumed linear. For each test performed in this study, the observed shear stress strain relationship was non-linear from start to failure. As shown in Fig. 2, an approximate limit of the elastic zone was determined using three approaches (Tavenas et al. 1979, and Graham et al. 1983). In the first two methods, the stress-strain relationship at small shear strain levels (q < 1.5%) is assumed to be 1 relationship in Fig 2a, and q q relationship in Fig 2b. In the last approach bi-linear: 1 (Fig. 2c), at the same shear strain levels, the relationship between the length of 3-D stress vector (LSSV) as the specimen moves towards yield and the strain energy absorbed by the specimen for deformation (W) is assumed to be bilinear. The LSSV and W were calculated using expression (3) and (4).
LSSV = 12 + 2 2 + 32

(3) (4)

W = 1 1 + 2 2 + 3 3

where is the average value of the effective stress in a loading increment , and represents the corresponding strain increment. In these bilinear relationships, the two lines represent the pre-yield and post-yield behavior, and their intersection point the yield point. In these relationships, the pre-yield non-linearity suggests that, to some extent, plastic deformation has occurred even before these defined yield points. However, the behavior may be assumed to be predominantly elastic. Use of these relationships by changing the scale to higher strain levels may show higher yield locations; however, in such a case, these methods will involve large errors associated with the assumption of bi-linearity for non-linear relationships.
115 80

Major Principal Stress, 1' (kPa)

(a)
Deviatoric Stress, q (kPa)
Yield Point 95 85
b=0 b=0 b=0.25 b=0.25 b=0.5 b=0.5 b=0.75 b=0.75 b=1 b=1

(b)
Strain Energy, W (kN/m )
2

(c)
0.75 b=0 b=0 b=0.25 b=0.25 b=0.5 b=0.5 b=0.75 b=0.75 b=1 b=1

60

Yield Point

40 37 20

0.5

75

b=0 b=0 b=0.25 b=0.25 b=0.5 b=0.5 b=0.75 b=0.75 b=1 b=1

0.25 Yield Point 0 35 0 40 80

55 0.0

Major Principal Strain, 1 (%)

0.5

1.0

1.5

0.0

Shear Strain, q (%)

0.5

1.0

1.5

LSSV (kN/m2)

Yield Points q (kPa) p (kPa)

(a) 38 - 50 60 - 66

(b) 37 - 40 60 - 62

(c) 38 - 39 60 - 62

Fig. 2. Determination of yield points

Figure 2 shows that the pre-yield behavior was not significantly influenced by the change in b-values. The yield points determined for various b-values were inside a narrow range so that they were practically identical. The results obtained from three methods of yield point determination were essentially identical, which provided more confidence in the interpreted yield locations. These yield locations are mapped into the shear stress-strain relationship and excess pore pressure response shown in Fig. 1. This figure shows that the shear stress corresponding to yield location (38 kPa) was much lower than the shear stress at failure (120-150 kPa). STRAIN LOCALIZATION IN KAOLIN SPECIMENS DURING UNDRAINED SHEARING Bifurcation and Undrained Instability Vardoulakis (1982) defined bifurcation as a critical state in deformation process when deformation does not follow its straight ahead continuation but turns to an entirely different mode. Under perfect boundary conditions, a perfectly homogeneous material may undergo non-homogeneous deformation, as various bifurcation modes of deformation develop. Hill (1962) proposed that the weak stationary discontinuities of the velocity gradient during this bifurcation mode must be understood as localizations of the deformation into shear bands. Bifurcation and post-bifurcation phenomena (causing non-uniform deformations in soil elements) are not only linked to the constitutive properties of the soil but also the specimen geometry, boundary conditions, and geometrical and/or material imperfections. Vardoulakis and Sulem (1995) formulated that for plastically incompressible poro-elastoplastic material, undrained shear banding goes hand in hand with pore-water pressure shocks. Under locally undrained condition when the fluid flow is zero everywhere in the considered domain, these pressure shocks diffuse so quickly that in actual time scale of any soil testing, it can be concluded that no pore-pressure shocks can develop across shear band boundaries and therefore undrained shear banding or bifurcation mode cannot be sustained. In globally undrained condition when internal flow of pore-water can be assumed to take place, shear banding is possible, if the amount of water that has to flow across the shear band boundaries due to small shear band dilation or contraction can be accommodated by elastic deformation of the surrounding mass. This phenomenon was further investigated and found to be possible only at large deformations. Han and Vardoulakis (1991) and Vardoulakis (1996) supported this experimentally by performing undrained biaxial tests. Identification of the Onset of Strain Localization During an undrained shear test, the global volumetric strain ( sum of three principal strains) should be zero for incompressible pore fluid. In this study, the normal strains were measured at the center of specimen faces in three principal directions, and the summation of observed principal strains showed small positive values (< 1%) for all b-values. With a flexible membrane, true triaxial system, the deformation profile is not completely uniform on each face of the cube (Prashant and Penumadu 2004). In order to see the profile of deformed specimen at different stages in stress-strain curve, the b=1 test (most critical case) was repeated several times. A few tests were stopped just before the failure and others just after the failure. Fig 3(a) shows the un-deformed cubical specimen, which was later sheared to 1 = 4.2% (just past failure) and then unloaded to zero stress state with the same strain rate as in loading. Fig 3(b), 3(c) and 3(d) show pictures of the same deformed specimen. Edges and corners in Fig 3(b) confirm that their overall configuration remains intact during shear; however, Fig 3(c) shows that the center of each face

Y X (a) 51 Z Y X (d) Z (c)

(b) Defused strain localization (e) Location of deformation measurement

Fig. 3. Deformation to failure in cubical specimen, (a) undeformed specimen, (b) deformed specimen, (c) strain variation at edges and corners (d) strain localization (Shear band), (e) deformation zones in cubical specimen deforms slightly more than the edges and corners. The system used in this study calculates the strain values on each face based on the deformations measured at the center of the specimen and may be slightly different from the average strains on the entire face of the cube. Therefore, the sum of the normal strains measured in all principal directions was found to be nonzero. The edges and corners responsible for this small strain difference covered a very small area of each face. The measured strain in each of the three directions was uniform for a large central area as shown in Fig 3(e). As the specimen reached peak shear stress location, large irregularities developed on the faces of the specimen due to shear band formation or undrained instability. This resulted in rapidly increasing values of the sum of measured normal strains. This information was used to verify the onset of strain localization. A detailed description of this approach was recently published by Prashant and Penumadu (2004). In the current study, the failure location is defined by considering the onset of localization, with an understanding that the deformed specimen has significant non-uniformity beyond the onset of this bifurcation/undrained instability and therefore strain information in test data becomes unreliable after the failure point. Localized Deformations at Failure Repeated b=1 tests, stopped just before the failure, had no signs of any localization on the specimen and the deformation was relatively uniform throughout the specimen, whereas the tests stopped just after the failure showed clear localization patterns on the deformed specimen. This was observed after unloading the specimen at the same strain rate as in loading. The circled area in Fig 3(b) shows localized bulging (defused localization) on the z-face with a barely visible shear band formation, which moves the soil particles locally from the cubes z-face (1-dir) to the x-face (3-dir). The y-face (2-dir) had clearer shear band formation as shown in Fig 3(d). These shear bands were inclined at an angle of 51 (approx.) from x-axis. The specimens, extruded well after failure in the original series of tests, showed similar but more complex shear-band formations, and they also had localized bulging on x-face.

RELATIONSHIP BETWEEN INITIAL YIELDING AND FAILURE CONDITION For all the b-values, the stress-strain relationship before failure was essentially identical and so was the yield location. However, the failure location varied significantly for different b-values. Figure 4 shows a comparison between the initial yield and failure location. The information about q and W, as shown in Fig. 4a, is often used to define yield surface and hardening rule in elasto-plasticity models. Figure 4b shows a deviatoric plane containing initial yield and failure stress states. Elasto-plastic yielding is strongly linked with the constitutive properties of a soil. Figure 4 shows that size of yield surface was not significantly influenced by the change in b-value, and this behavior can be approximately represented using a yield surface that has circular shape in the deviatoric plane. The specimens sheared with b > 0.25 showed brittle failure response (a sudden change in stiffness) due to large strain localization at peak shear stress level. Strain localization may have a relationship with constitutive properties; however, it is also largely influenced by many other factors such as specimen boundary, material imperfections, etc. As a result, the failure conditions were significantly influenced by the change in b-values. Therefore, the observed initial yield and failure conditions for moderately overconsolidated Kaolin clay were associated with two separate phenomena. This experimental observation is imporant to be considered for developing a suitable constitutive model for such clay. (a)
180 150 120 90 60 30 0 0 0.5 b-value 1 0.1 0 0.5 b-value 1
x y

q (kPa)

100

W (kN-m/m ) Failure Location

(b)

z Initial Yield Localtion Failure Localtion

10 Failure Location Yield Location 1 Yield Location

Fig. 4. Comparison between initial yield and failure location: (a) deviator stress and strain Energy, (b) stress state in deviatoric plane CONCLUSIONS A series of strain controlled cubical true triaxial undrained tests were performed on moderately overconsolidated (OCR = 5) specimens of Kaolin clay. The tests were performed with different b-values (b = 0 to 1); however, the direction of 1 was kept constant (along z-axis). Three approaches were used to determine the yield location, which provided consistent results. The shear stress-strain relationship before failure and yield locations were not significantly influenced by intermediate principal stress. The failure conditions were governed by strain localizations within the specimens that were observed to occur at peak deviator stress. Experimental evidence of shear banding and localized bulging was presented and a method was developed to determine the onset of localization. Considering the factors behind occurrence of yield and failure conditions, for constitutive modeling, it was found appropriate to define the yielding/material-hardening independent of the failure conditions.

ACKNOWLEDGEMENTS Financial support from the National Science Foundation (NSF) through grants CMS-9872618 and CMS-0296111 is gratefully acknowledged. Any opinions, findings, and conclusions or recommendations expressed in this material are those of authors and do not necessarily reflect the views of NSF. REFERENCES Graham, J., Noonan, M. L., and Lew, K. V. (1983). Yield states and stress-strain relationships in a natural plastic clay. Can. Geotech. J., 20, 502-516. Han, C. and Vardoulakis, I. (1991). plain strain compression experiments on water-saturated fine-grained sand. Geotechnique, 41, 29-44. Hill, R. (1962). Acceleration waves in solids. J. Mech. and Physics of Solids, 10, 1-16 Lade, P. V. (1990). Single hardening model with application to NC clay. J. of Geotech. Eng., 116(3), 394-415. Prashant A., and Penumadu D. (2004). Effect of intermediate principal stress on overconsolidated kaolin clay. J. Geotech. and Geoenv. Engrg., ASCE, 130(3), 284-292. Roscoe, K. H. and Burland, J. B. (1968). On the generalized stress strain behaviour of wet clay. Engineering plasticity (eds J. Heyman and F. A. Leckie), pp. 535609. Cambridge: Cambridge University Press. Tavenas, F., Rosiers, Des, J. P., Leroueil, S., Rochelle and La, P., Roy, M. (1979). The use of strain energy as a yield and creep criterion for overconsolidated clays. Geotechnique, 29(3), 285-303. Vardoulakis, I. (1982). Stability and bifurcation of soil samples. In: Constitutive Relations for Soils, (Editors G. Gudehus et al.), Balkema, Rotterdam, 477-483. Vardoulakis, I. (1996). Deformation of water-saturated sand: I uniform undrained deformation and shear banding. II The effect of pore-water flow and shear banding. Geotechnique, 46(3), 441-472. Vardoulakis, I. and Sulem, J. (1995). Bifurcation analysis in geomechanics. Chapman and Hall Publication, London. Whittle, A. J., and Kavvadas, M. J. (1994). Formulation of MIT-E3 constitutive model for overconsolidated clays. J. Geotech. Eng., ASCE, 120(1), 173-198. APPENDIX I. NOTATION The following symbols were used . b = Intermediate principal stress ratio B = Skemptons pore-pressure parameter q = Shear strain in invariant form LSSV = Length of stress vector in three dimensional space q = Deviatoric stress in invariant form u = Excess pore-pressure W = Internal strain energy 1 , 2 , 3 = Major, intermediate and minor principal strains 1 , 2 , 3 = Major, intermediate and minor principal stress