Okada, N. & Nemat-Nasser, S. (1994). GPotechnique 44, No.

1, 1-19

Energy dissipation

in inelastic flow of saturated granular media

and S. NEMAT-NASSER*

cohesionless

N. OKADA*

The results of a study of energy dissipation in cohesionless granular media are presented. The relation between the excess pore water pressure, accumulated in a water-saturated granular mass, and the corresponding external work in undrained cyclic loading is studied experimentally, under displacement-controlled conditions. A micromechanical model of internal energy dissipation due to slip between contacting granules is introduced, and the results are compared with experimental measurements. The specimens are subjected to two sequences of loading with an intermediate reconsolidation to simulate reliquefaction. External work per unit volume is calculated from the experimental results, and its correlation with the excess pore water pressure is examined. In the first loading, a unique non-linear relation exists between the excess pore water pressure and the external work per unit volume which is independent of the shear strain amplitude. In the second loading this relation is a function of strain amplitude. Based on a micromechanical model, it is shown that the internal dissipation per unit volume in cohesionless granular media can be expressed in terms of the time history of the applied effective pressure and a single scalar parameter which depends on the density and strain amplitude. The model is further validated by torsion tests with random variation in the applied strain amplitude, and excellent agreement with the experimental results is obtained.

KEYWORDS: friction; laboratory tests; liquefaction; pore pressures;sands; torsion.

Larticle prCsente les rbultats obtenus au tours dune Ctude sur la dissipation de Knergie dans des milieux granulaires pulvCrulents. La relation existant entre Iexc&s de pression interstitielle deau, accumul& dans une masse granulaire saturi?e, et le travail extkrieur asscn+ apparaissant au tours de cycles de chargement non-drain&s, h d&placement contrSlt, est Ctudiee expi?rimentalement. Une mod%sation microm&anique de la dissipation dbnergie interne IiCe au glissement entre grains jointifs, est proposCe. Les rCsultats obtenus i Iaide de ce modkle sont cornpark aux mesures exp&imentales. Les izchantillons sont soumis g deux cycles de chargement, s&par& par une reconsolidation permettant de simuler la reliqkfaction. Le travail extCrieur, par uniti: de volume, est calculb g partir des ritsultats exp&rimentaux. La corr6lation pouvant exister avec lexc& de pression interstitielle est i?tudi&e.Lors du premier chargement, il nexiste quune seule relation entre Iexctis de pression interstitielle et le travail extbrieur, cette relation (?tant indbpendante de Iamplitude de la d&formation au cisaillement. Lors du second chargement, cette relation devient dbpendante de lamplitude de cette dtformation. I1 apparait, i partir du micromod6le, que la dissipation interne, par unit& de volume dun milieu granulaire p&&ulent, peut sexprimer en fonction de Ihistorique de la pression effective appliqute et dun paramktre scalaire unique dependant de la densiti! et de Iamplitude de la dbformation. Le modhle est ensuite valid& P Iaide dessais de torsion, pour des variations alCatoires de Iamplitude de deformation. La correspondance avec les rCsultats expicrimentaux est excellente.

INTRODUCTION

Liquefaction is a complex phenomenon in which fluid-saturated granular media may momentarily behave like fluids. It is an important aspect of

Manuscript received 14 January 1992; revised manuscript accepted 11 March 1993. Discussion on this Paper closes 1 July 1994; for further details see p. ii. * University of California, San Diego.

earthquake-resistant foundation design of many structures, especially those located in coastal areas, which are often built on sand with high underground water levels. As liquefaction takes place under seismic loading, saturated sand behaves more like a fluid, and therefore fails to support the applied loads of the building. Severe damage to the structure is often the result. Damage resulting from liquefaction has been observed in the aftermath of many earthquakes, including the Loma Prieta earthquake (1989), the

OKADA

AND NEMAT-NASSER

Niigata earthquake (1964) and the Alaska earthquake (1964). The mechanism of liquefaction is closely related to the dilatancy of granular media. Dilatancy was first studied by Reynolds (1885). It is defined as the rate of volume expansion in granular media per unit rate of shearing. The granules are rearranged during shear deformation, and this results in a change in the total volume. If the granular medium (e.g. sand) is water-saturated and undrained, the tendency towards densification (contractancy) results in an increase in the pore water pressure and hence a decrease in the corresponding frictional resistance of the contacting granules. In continued cyclic shearing, the pore water pressure at the termination of each cycle increases until it reaches a value close to the applied hydrostatic pressure. At this stage, the contact resistance of the granules can be regarded as essentially negligible. This process leads to a loss of the load-bearing capacity of the sandmass, which hence ceases to behave like a solid body. This is what is meant by liquefaction in this Paper; see Casagrande (1975) and Seed (1979). Liquefaction has been extensively treated experimentally (see, for example, Silver & Seed, 1971; Castro, 1975; Ishihara & Yasuda, 1975). In such experiments, parameters influencing the onset of liquefaction of the sand within a controlled volume are identified and measured. These parameters typically include overall density, initial packing conditions and granule size distribution (Seed, 1979; Miura & Toki, 1982; Tatsuoka, Muramatsu & Sasaki, 1982). In addition to these internal characteristics, the applied loading affects the onset and nature of liquefaction (Ishihara & Towhata, 1983, 1985; Symes, Gens & Hight, 1984). Previous work has in general studied liquefaction experimentally. Theoretical work has focused on phenomenological considerations rather than a micromechanical approach. A unified energy model for densification and liquefaction of cohesionless sand was proposed by Nemat-Nasser & Shokooh (1979), who compared its predictions with the experimental results of Peacock & Seed (1968), Youd (1970, 1972) and DeAlba, Seed & Chan (1976), and obtained excellent agreement. The present work correlates the results of an experimental programme with a theoretical model based on micromechanics and energy principles. Models of this kind seek to relate the overall response of granular materials to the Microresponse of their microconstituents. mechanical analyses of densification and liquefaction phenomena, which are also co-ordinated with experiments, are given by Nemat-Nasser (1980), Nemat-Nasser 8~ Tobita (1982) and Nemat-Nasser & Takahashi (1984).

EXPERIMENTAL SET-UP Motivation and background

Cohesionless granular materials support general external loads through contact friction. An experimental programme must include compression and shearing of reproducible samples in a controlled manner with reliable data. This may require complex experimental facilities, with a closed-loop feedback system to control the experiment and to monitor the specimen deformation. The specimen geometry used for the present investigation is a large hollow cylinder, 25 cm high, with inner and outer diameters of 20 cm and 25 cm respectively. This geometry is such that in torsion, the shear stress remains (approximately) homogeneous throughout the thickness of the specimen; see Hight, Gens & Symes (1983) for a detailed examination of this and related issues. The specimen is supported by a triaxial load frame (Fig. 1). The axial and torsional deformations are controlled through an MTS servohydraulic loading system. In addition, the specimen is subjected to lateral hydrostatic pressure on both its inside and outside cylindrical surfaces. In this manner, triaxial states of stress can be imposed on the material under controlled conditions with complete data acquisition capability. This load frame, to the Authors knowledge, is one of four that have been constructed to date, and is fully computer-controlled. Either the stress path or the strain path can be preprogrammed with automatic mode switching capability.

Specimen

preparation

and installation

The granular material chosen for this study is Silica 60, manufactured by U.S. Silica. This sand is chosen for its fine particle size, which is necessary in order to minimize membrane penetration phenomena that would otherwise invalidate the test results. The particle size distribution is shown in Fig. 2. The mean particle diameter is 220 urn and the specific gravity of the sand is 2.645. Depending on the packing conditions, various void ratios are obtained. For Silica 60, the minimum and maximum void ratios are 0.631 and 1.095 respectively, and are measured by the JSSMFE method (Committee of JSSMFE on the Test Method of Relative Density of Sand, 1979). Special fixtures are used to prepare hollow cylindrical sand specimens. These fixtures include inner and outer moulds to which rubber membranes are attached. The sand is initially supported on the bottom by a ring of porous metal with six evenly spaced fins, called the pedestal, which in combination with a mating top ring, called the cap, applies the torsional load to the specimen. The pedestal (with associated fins) is

INELASTIC

FLOW

OF GRANULAR

MEDIA
Adjustment bar

Top plate Metal bands

Electrical

dial gauge

Potentiometer

Plexiglas

chamber Cap

Torque

load cell Supporltng Tie bars Counter Specimen balance bars

Inner membrane Outer membrane

Ram of trlaxlal

apparatus Bearings

Porous Bottom

metal plate

Pedestal

Clamp

Vertical

clamps .Horizontal

-Wheels

Fixed counter

frame

Fig. 1. Triaxial load frame

attached to the bottom support plate. The outer membrane is then slid over the inner membrane and fixed to the pedestal with o-rings. The outer mould is bolted in place, and the top of the outer membrane is draped over the outer mould and held in place by o-rings. A separate fixture is installed on top of the outer mould to prevent sand spillage on the rest of the triaxial load frame. This fixture also allows for an overfill amount of sand so that a desired packing condition can be obtained. The excess sand is removed later. It is well known that the initial packing condition of the sand has a noticeable effect on the material response of the specimen (Arthur & Menzies, 1972; Oda, 1972a; Miura & Toki, 1982). The specimen preparation method must therefore achieve a consistent initial packing condition so that experiments are repeatable. To this end, a

160 Sieve sizes: urn

Fig. 2. Particle size distribution curve

OKADA

AND NEMAT-NASSER

technique has been adopted that is known as the rodding method. This method consists of pouring an approximately 2 cm deep layer of sand into the mould and then inserting a rod approximately l-l.5 cm into the latest layer. The rod is moved around the circumference of the sand in an up-and-down motion for 2-3 revolutions. This procedure is continued until the mould is filled. Ten layers were used to obtain loose packing conditions, and 14 layers for dense packing conditions. Experiments performed under loose conditions use the sand in a wet form, where the sand has been air dried and then mixed with 8 wt% water before pouring into the mould. The water is needed to prevent non-homogeneous initial packing conditions in the loose form. The void ratio for this condition varies between 0.865 and 0.874. Experiments performed with dense conditions use only air-dried sand with a void ratio varying between 0.708 and 0.725. The fixture that was attached to the top of the outer mould is removed and the amount of overfilled sand is cut away. The cap is then installed; it consists of the same porous metal as the pedestal and also has six fins. A second vacuum system is connected to the cap and pedestal. The purpose of this vacuum system is to make the under atmospheric pressure. specimen rigid The vacuum level is maintained at 29.4 kN/m2. The first vacuum system that keeps the outer membrane fixed to the outer mould is then released. The outer mould is removed, followed by the inner mould. A torque load cell unit is first bolted onto the ram of the triaxial load frame (Fig. l), and then bolted onto the cap. Next, a potentiometer is attached to the load frame (Fig. 1). The potentiometer measures the twist angle during the experiment. A Plexiglas chamber with steel bands is installed over the entire specimen, and a top plate is installed. The top plate is affixed to the bottom plate by stainless steel tie bars, which hold the chamber firmly in place. The chamber has three purposes: it provides confinement of the experiment if the sand mould loses integrity, it holds the water that is used to apply hydrostatic pressure to the specimen, and it is used as a viewport to observe the progress of the experiment. The specimen assembly is now complete. The assembly is raised to the level of the MTS load frame by a forklift. A special work frame has been built onto the MTS load frame: this allows attachment of all connections to the specimen assembly, and provides a railway for installation and removal of the specimen into and from the MTS load frame. The MTS load frame used for this experiment has an axial capability of 89 kN and a torsional capability of 565 Nm, which can be used inde-

pendently. The system uses a Digital Equipment Corporation PDP-11 computer to control the servohydraulic actuators. The system is closedloop, so that feedback from any selected transducer can be used to control the test. Once the triaxial load frame has been rolled into place over the ram of the MTS load frame, it is secured in place by both vertical and horizontal clamps (Fig. 1). The hydraulics for the MTS system are turned on, and the MTS ram is raised to the level of the universal joint by use of displacement control. An air clamp that is fixed to the top of the MTS ram is then actuated and grips the universal joint on the bottom of the ram of the triaxial load frame. The universal joint is required to accommodate any misalignment between the ram of the MTS load frame and the ram of the triaxial load frame. The first step in the experimental procedure is to fill the Plexiglas chamber with water until the specimen is completely submerged. The remaining space above the specimen is pressurized with air to 29.4 kN/m, the same value as the vacuum inside the specimen. During this operation the vacuum in the specimen is released and water pressurized in such a manner as to keep the effective pressure in the specimen constant at 29.4 kN/m2. The specimen is then water saturated as follows. To attain full saturation, the specimen is first saturated with CO, gas through the porous metal in the pedestal and cap. The flow of gas is continued until all air is removed from the specimen. CO, gas is used because of its high solubility in water. A fixed amount (41) of de-aired water is used to saturate the specimen. As much as possible of the small amount of air and CO, gas remaining in the specimen must then be removed. The pore water pressure is increased to 196 kN/m as back pressure, using a burette system, while at the same time the external hydrostatic pressure is increased to 225.4 kN/m2, in order to keep the effective pressure constant (29.4 kN/m2) during this procedure. The volume of the excess gas in the specimen is hence reduced due to the relatively high pore water pressure. For experiments of this type, the specimen is required to be highly saturated. The degree of saturation is measured by the B value. For the B value to be measured, the specimen must be in the undrained condition. This condition is met by closing the valve to the burette, ensuring that the specimen remains at a fixed volume. The specimen is said to be perfectly saturated (B = 1) if an increase in the external hydrostatic pressure has the effect of increasing the pore water pressure in the specimen by an identical amount. The B value is defined as the ratio of the incremental increase of pore water pressure to the incremental increase

INELASTIC

FLOW

OF GRANULAR

MEDIA

of hydrostatic pressure. The value for all the present experiments is higher than 0.99. The last step of specimen preparation is to increase the effective pressure to 196 kN/m* by reopening the valve to the burette, allowing water to drain from the specimen. The external hydrostatic pressure is thereby increased to 392 kN/m, where pore water pressure is 196 kN/m. Finally, the specimen is left undisturbed in this condition to consolidate isotropically for a period of 3 h.

Experimental procedure and data acquisition The MTS load frame has a computer-operated controller system. The computer operates three independent controllers. Each controller has three independent feedback channels. Controller 1 is associated with the vertical movement of the MTS/triaxial load frame ram assembly; channel 1 is used to monitor the load from the torque load cell and channel 2 is used to monitor the vertical displacement of the specimen. Channel 3 is not used with any controller. Controller 2 is associated with the pressure; channel 1 is used to monitor the chamber pressure P, and channel 2 is used to monitor the pore water pressure Pi. Controller 3 is associated with the twist of the ram assembly; channel 1 monitors the torque from the torque load cell and channel 2 monitors the angle of twist from the potentiometer. The experiment is conducted by use of two closed-loop feedback systems. The first feedback system uses channel 1 of controller 1 in load control to keep the specimen in a state of hydrostatic compression in accordance with the external pressure P,. The second feedback system uses channel 2 of controller 3 in displacement control to twist the specimen cyclically to desired shear strain amplitudes at desired shear strain rates.

The imposed cyclic angular displacement that produces the applied shear strain has a triangular time variation with constant strain rate, f%/min over each quarter cycle. Shear strain amplitudes are 0.2%, 0.5% and 1.0% for both loose and dense specimens. Tests at 0.4% and 2.0% shear strain amplitudes are performed on dense specimens. All tests are conducted in an undrained condition, and continue until the excess pore water pressure reaches 95% of the initial effective pressure, i.e. 186.2 kN/m. The tests are actually stopped at the end of the cycle after which transchannel 2 of controller 2 (pore pressure transducer) reaches a value of 382.2 kN/m*. This entire process is defined as the first loading. The valve to the burette is then opened, and Pi is reduced to its initial value of 196 kN/m2. The specimen is not disturbed for 3 h, for reconsolidation purposes. The valve to the burette is closed, and the exact procedure for the first loading is repeated. This is called the second loading. After the second loading, the experiment is disassembled. Care is taken to remove the sand from the specimen and place it in an oven for drying. The sand is dried for 24 h and then weighed.

EXPERIMENTAL RESULTS First loading All the experiments described were performed under strain-controlled conditions, in contrast to the work of most other researchers, who conducted the undrained cyclic shear tests under stress-controlled conditions (see, for example, Ishihara & Yasuda, 1975; Seed, 1979; Tatsuoka et al., 1982). Fig. 3 shows the relation of the shear strain and the effective pressure for loose samples, for two strain amplitudes, 0.2% and 1.0%. The effective pressure decreases during each cycle; the

Effective pressure: kN/m

Fig. 3. Relation between shear strain and effective pressure in first loading of loose specimens; strain amplitudes are 0.2% and l-O%

OKADA AND NEMAT-NASSER

Number of cycles 27 2

75---,

No. 40 4, ,--_

Amplitude 0.2% 1 .O% --. .

Void ratio 0.874 0.871

-25-

-50-I 0

50

100 Effective pressure:

150 kN/m*

200

t 250

Fig. 4. Relation between shear stress and effective pressure in first loading of loose specimens; strain amplitudes are @2% and 19% reduction after the first cycle is especially large. The number of cycles required for the excess pore water pressure to attain 95% of the initial effective pressure depends on the employed strain amplitude: 27 cycles are needed for 0.2%, and only two cycles for 1.0% strain amplitude. Fig. 4 shows the relation of the shear stress and the effective pressure; Fig. 5 plots the corresponding shear stress against shear strain. The peak shear stress (and the secant modulus) decreases after each cycle. The energy supplied through external work is mainly consumed by the frictional loss at contacting granules, resulting in a change of the microstructure in the granular mass. Therefore, the external work can be used to measure the history of fabric change in a granular mass.
Number of cycles 27 2

The rate of external work per unit volume (tir) can be evaluated in terms of the applied boundary tractions z and the boundary velocity field i

where a dot denotes the inner product and aD is the boundary of the sample domain D. If it is assumed that the boundary tractions are uniform, then equation (1) can be expressed in terms of the overall stresses and strain rates, as (Hill, 1963, 1967) <GE) = (~,,>(~,,> + (a,,)<&) + (~,,>(3,e> (2)

s
L?D

t.tidS

+ (QX&Je>

50

No. 40 41

Amplitude 0.2% 1 .O%

Void ratio 0.874 0.871

I -0.5 0.0 Shear strain: 0.5 % 1.0

I 1.5

Fig. 5. Relation between shear stress and shear strain in first loading of loose specimens; strain amplitudes are @2% and lQ%

INELASTIC

FLOW

OF GRANULAR

MEDIA

Since the specimen is isotropically pressurized by the external pressure P, throughout the experiment, all three normal stresses equal P,. If p/V is the average volumetric strain rate, equation (2) can be rewritten as (GE) = PJ(&,) + (&,> + (&,)I + (c,&jlrCJ (3)

= P0CQ/V + (e,,>(3,e>

Since, for the pressure levels used here, sand particles and water can be assumed to be incompressible, v/V is zero during the undrained experiment if the sample is completely saturated and the change in the elastic rebound of the membrane due to the reduction of the effective lateral stress is neglected. The rate of external work per unit volume then becomes (&) = (a,,>(?.s> (4)

The pressure term does not contribute to the rate of external work for incompressible materials, as is evident at the outset. The external work per unit volume (wr) up to time t can then be evaluated by the time-integration of equation (4) (WE) = (e,e>(L> s f0 dr (5)

where t, is the time at which the experiment is started. The external work is calculated from the experimental results, and correlated with the accumulated pore water pressure for both loose

and dense specimens. The relation of the external work per unit volume and the excess pore water pressure for the loose specimens is shown in Fig. 6. Three strain amplitudes, 0.2%, 0.5% and l.O%, are used here. The data at the end of each cycle in each experiment are plotted in Fig. 6, which shows a unique non-linear relation of the external work and the accumulated pore water pressure; this relation is independent of the shear-strain amplitude employed, but the number of cycles to 95% of the initial effective pressure does depend on the strain amplitude. The relation of external work per unit volume and excess pore water pressure for dense specimens is also shown in Fig. 6. Five strain amplitudes, 0.2%, 0.4%, 0.5%, 1.0% and 2.0%, are employed. Again, there is a unique non-linear relation of the external work and the accumulated pore water pressure. The effect of specimen density on the relation discussed above is now considered. Both loose and dense specimens during the first loading show a unique relation of the external work and the corresponding excess pore water pressure. These essentially coincide up to 130 kN/m pore water pressure, i.e. 65% of the initial effective pressure. A significant difference appears thereafter, with the loose samples developing higher pore pressure, as expected. To explore this interesting phenomenon further, randomly varying shear-strain amplitudes up to 1.0% are applied to a dense specimen. The

No. 0 40

Type L0CSe

Amplitude 0.20%

Void ratlo O-674 O-665 0.871 0.722 o-719 0.706 0.725 0.716 0.719

NO of cycles 27 2 50 14 9 3

0
l

37 34 32 36 38 39

Dense Dense Dense Dense Dense Dense

0.20% 0.40% 0.50% 1 .OO% 2.00% Random

Y 0 9 +

1 :o External

1:5 work per unit volume:

2:o kJ/m3

215

$0

Fig. 6. Relation between external work per unit volume and excess pore water pressure in first loading of both loose and dense specimens

OKADA

AND

NEMAT-NASSER

Speclmen no. 39 Void ratio 0.719

-,
0

i0

160 Effectwe pressure

160 kN/m

260

2;o

Fig. 7. Relation between shear stress and effective pressure in first loading of dense specimen subjected to random torsional loading

relation of the shear strain and the effective pressure is shown in Fig. 7. The relation of the external work and the excess pore water pressure for this random loading is shown in Fig. 6, where the accumulated pore pressure at zero shear strain is plotted against the corresponding external work. The randomness in loading does not affect the unique non-linear relation of these two quantities. Second loading Experimental results for the second loading for both loose and dense specimens are presented below. After the first loading, specimens are reconsolidated under the same initial effective pressure, 196 kN/m*, as in the first loading. The second loading is then applied to the specimens.

The results where the strain amplitude in the second loading is the same as that in the first loading, and then the results where the strain amplitude in the second loading is different from that in the first loading, are examined. The shear deformation characteristics in the second loading are compared with the results of the first loading. Fig. 8 shows the first two cycles of the relation between the shear strain and the effective pressure in both the first and second loading for a dense specimen deformed at a strain amplitude of 0.5%. The excess pore water pressure accumulated during the second loading is much lower than in the first loading, and the number of cycles required to reach 95% of the initial effective pressure in the second loading is much greater than in the first loading.
Strain ampktude First ----Second 0.50% Number of cycles 9 36

Void rat10 0.706 0.696

No. 32, dense 0.5-

speamen

-0.5

I 0

50

100 Effectw pressure:

150 kN/m

I 200

I 250

Fig. 8. Relation between shear strain and effective pressure in first and second loading
of dense specimen; strain amplitude is 05%

INELASTIC Fmt load,ng 0 og.. 00 oOp go: . . + + ++,++++* +

FLOW

OF GRANULAR

MEDIA

Second loading +++++

+++++++++++++++

l+

Strain amplttude NO. 33 (FL) 33 (SL) 32 (FL) 32 (SL)

0.50% Number of cycles 6 11 9 36

0 0 0 +

Void ratio 0.865 0.646 0.706 0.696

lb Number

i0 of cycles

i0

4b

Fig. 9. Relation between number of cycles and excess pore water pressure in first and second loadings of both loose and dense specimens; strain amplitude is 05% Figure 9 shows a direct comparison of the pore water pressure variation in loose and dense specimens, deformed at a strain amplitude of 0.5%. It takes a greater number of cycles for the pore water pressure to reach a specified level in the second loading than in the first loading, for both

cases. This is due to the ordered arrangement of the granules, attained on completion of the first loading. Specifically, the contact normals after the first loading tend to be oriented such that the specimen is better able to resist a similar shearing (see Oda, 1972b; Konishi, Oda & Nemat-Nasser,

200
O-5% (D) **x ,*** 2.0% (D)

150 3 Y hi ; I a, 1oc a a, kz I z 6 50

D = dense

* .& 0 0 l w 0 9

NO. 37 34 41 37 34 32 36

Type LOWe Loose Loose Dense Dense Dense Dense

Amphtude 0.20% 0.50% 1 .OO% 0.20% O-40% 0.50% 1 .OO% 2.00%

Void Ratva O-657 0.646 0.652 O-710 0.706 0.696 0.711 0.700

Number

of cycles (26) 11 (749, 61 36 9 4

38

Dense

2 External

1
kJlm3

work per unit volume.

Fig. 10. Relation between external work per unit volume and excess pore water pressure in second loading of both loose and dense specimens; strain amplitude used in second loading is the same as in first loading

10

OKADA

AND

NEMAT-NASSER

1982; Subhash, Nemat-Nasser, Mehrabadi & Shodja, 1991). Another notable trend is that the pore water pressure builds up faster during the first loading of the dense specimen than during the second loading of the loose specimen. These trends are seen in all results of the present study when the behaviour of loose and dense specimens deformed at a constant strain amplitude is compared. However, in a strain-controlled test the deformation of the specimen is limited by the prescribed strain amplitude. This prevents extensive particle rearrangement, which often occurs in stress-controlled tests, once sufficiently high pore pressures are attained (see Nemat-Nasser & Tobita, 1982). The relation of the external work per unit volume and the excess pore water pressure in the second loading is now presented. Fig. 10 shows this for loose and dense specimens. There are no significant differences up to half the initial effective pressure, but after that the relation in the second loading seems to depend on the strain amplitude employed, although the same relation in the first loading does not. This is because, for the same density, the sample packing is the same for all samples at the start of the first loading, whereas at the start of the second loading each sample has experienced a different stress and strain history during its first loading, with a strain amplitude different from the other samples. It is still not clear whether or not the strain amplitude used in the first loading affects the relation of the

external work and the excess pore water pressure in the second loading. In order to investigate the true effect of the strain amplitude during the second loading, the specimens are subjected to cyclic loading at several different strain amplitudes, and then the second loading is performed for each specimen at a common strain amplitude. The results are shown in Fig. 11. It is clear that the stress and strain history in the first loading does not affect the relation of the external work and the excess pore water pressure in the second loading, i.e. this relation in the second loading depends on the strain amplitude employed during the second loading, and not on that employed during the first loading.

THEORETICAL MODELLING Energy dissipation in granular media In the present context, the external work that is supplied at constant temperature to a material sample is either dissipated through friction or stored in the material as strain energy. The relation of the external work and the internal dissipation provides a basic constitutive constraint for the flow of granular media (Rowe, 1962). For the stress levels considered in the present experiments, essentially the entire external work is dissipated by slip between contacting granules, and hence a negligible amount is stored in the granules and the fluid as strain energy (Schofield & Wroth, 1968). Energy dissipation in granular

0.50% *A L 150. 3 Y hi 5 I g 100, E B a z G 50

I (Loose)
0.20%

(Loose) 1 0-f. *o *

1 .OO%

(Dense)

+@DbQ) o
0.20% (Dense)

Amplitude No. * 0 A 0 * t 0 9 + 0 40 57 33 37 54 56 48 36 49 TYPO LOOS? Loose Loose Dense Loose Dense Dense Dense Dense 1st 0.20 1 .oo 0.50 0.20 1.00 0.20 1.00 2.00 2nd 0.20 0.20 0.50 0.20 0.50 0.20 1 .oo 1 .oo 1 .oo Void ratio Number of cycles (28) (23) 11 (::, (79) 7 9 8 0.857 0.853 0.846 0.710 0.846 0.706 0.708 0.711 0.705

3 External work

4 per unit volume:

5 kJ/m3

between external work and exea pore water pressure in second loading of both loose and dense specimens; strain amplitude used in second loading is different from that used in first loading

Fig. 11. Relation

INELASTIC

FLOW

OF GRANULAR

MEDIA

11

media is micromechanically modelled below, and is related to the pore water pressure built up in cyclic shearing of saturated undrained samples. The low-strain-rate shear loading of a granular mass that occupies spatial region D of volume V is considered. The region R within D is occupied by the granules, whereas D - R is occupied by water (the specimen is saturated). The rate of external work per unit volume is expressed in terms of the boundary tractions z and the boundary velocity field i by <%(Q) = ; The tractions n on aD by 7(x, n, t) . ri(x, t) dS(x) s aD r relate to the exterior (6)

unit normal

7(x, n, t) = a(x, t) . n(x, t)

(7)

where Q is some stress field in equilibrium with the applied tractions. Substitution of equation (7) into equation (6) gives ($(t)) = $ (a(~, t) . n(x, 1)). i(x, t) dS(x) (8)

The rate of external work can be related to the rate of internal dissipation by the application of the Gauss theorem (see Malvern, 1969; NematNasser & Hori, 1993). In the application of the Gauss theorem for granular materials, slip at contacting granules renders the velocity field discontinuous there. Since the Gauss theorem requires that both stress and velocity fields be differentiable, it cannot be applied directly to the whole body. However, after the body is divided into the necessary number of sub-regions by the introduction of suitable surfaces, the Gauss theorem can be applied to each sub-region in which the derivatives are continuous. Suppose that at time t there are N slip surfaces in D at contacting granules (Fig. 12(a)). Let s be the area on which the ath slip occurs. Then, the velocity field is discontinuous on the collection of all slip surfaces, i.e. on If= 1 s. Let 5 be a typical point on this slip area. Associated with the velocity field i(x, t) now is the velocity jump [ri(x, t)] defined at 5 by [fi(& t)] = lim ri(x+, t) - lim i(x-, x--t x+-t; t) (9)

(b)

(d) Fig. 12. Surface S and slips 9 in the considered region D

12

OKADA AND NEMAT-NASSER

Consider a surface S (Fig. 12(b)-(d)) which includes the slip surfaces cf=r sa in D at time t, and which divides region D into M sub-regions, within each of which the velocity field is continuously differentiable. The Gauss theorem is applied to each sub-region, to give

=-

1 V {S BD

t.idS

- .f,
=-

1;.[4x+) - 4x-11ds}
(18) (13) reduces to

1 r.idS-irI;.[ti,dS) V (1 aD

and equation

=tjlI8
V.a=O and the symmetry o = 07 give, from equation

[(V.a).ri+o:Vi]dl

(10)

where aDB is the boundary of subregion DB. The equilibrium equation with no body and inertia forces (11) of the stress tensor (12) (10)

s
aD

t.idS

o:kdV

+il

j--.[i,

dS)

(19)

The first and second terms on the right-hand side of equation (19) are the strain energy and frictional energy terms respectively. The strain energy term in equation (19) can be decomposed to the strain energy for granules and for water, leading to 1 v
s aD

1 =-Ja:idV VD where i: = i[Vi + (Vri)T]

(13) which is the energy dissipation rated granular media. equation for satu-

(14) side of

is the strain-rate tensor. The left-hand equation (13) is expressed as

+I;.idS+ir.idS) where S+ and S- are opposite surface S, i.e. s+(x, n, t) = s-(x, --n, t) Since the surface tractions
5(X, n, t) = -t(X, --n, t)

(15) faces of the same

Formulation offrictional energy loss The frictional energy term in equation (20) is formulated below as a linear function of the effective pressure P. The unit contact normal to a slip plane and the unit vector in the slip direction are denoted by n and s respectively. The traction r on the slip surfaces If= 1 soI D is decomposed as in T(X,n, t) = F(x, t)s(x, t) + tyx, t)n(x, t)

(21)

(16)

z satisfy (Fig. 12(d)) (17)

the second and third terms in equation (15) cancel out for any x on (S - I:= 1 f) where the velocity field is continuous. Equation (15) then becomes

where T' and T" are the normal and shear components of the tractions respectively. The velocity jump [i](x, t) is assumed to have only a shear component [ti](x, t). The frictional energy term in equation (20) then becomes

21
s-1

T(X, . CW, t) dS(4 t)

S(X*
II,

I)

T(X,t)[d(-% t) dS(x) (22)

=1 V [I
aD

t.idS

The frictional coefficient for If= r sil is denoted by ~(x, t). Then the shear component T' is expressed in terms of the normal component T" as
T(X, t) = p(X, t)T(X, t)

(23)

INELASTIC

FLOW

OF

GRANULAR

MEDIA

13 in

Since the granules carry the effective pressure through intergranular frictional contacts, it is reasonable to assume that the normal tractions r at a slipping contact in I:= 1 sa are linearly dependent on the effective pressure P 51(x, t) = W(w, t)P(t) (24)

volume under these conditions is expressed terms of the average stresses and strains as 1 v Equation r(x, n, t) . Yx, t) dS = (c,e(t)><d,s(r)) (29) yields = <W)>W) + (&)>

(30)

where Y is a scalar-valued function defined at points where slip occurs. In general, Y depends on the size of the granules, the packing, the loading condition and other relevant factors. Substitution from equations (23) and (24) into equation (22) yields r(x, t) .

<~&)><3&)>

(31) of

Consider the energy per cycle. Integration equation (3 1) over the nth cycle gives <e&)X3,&)> s *.-1 = dt

CW, t) dW

,: (k(t))P(t) sn L

dt +

fn (i(t)) s *.-1

dt

(32)

x WCW,
Since the effective pressure ofx

t) W-4

(25)

where t, is the time at which the nth cycle is completed. If equation (32) is decomposed by taking P outside the integral and accounting for the resulting error by the addition of a term E,, then (e,s(t)>(3,s(t)) s r.- 1 fn = I. (i(t)) i I_I dr dt + E, + (e(t)) dt s fn-L pressure (33) in the

P(t) is not a function

4 . Cdb, t) dS(4
= <W>P(t) (26)

where

where P,, is the average nth cycle, given by 1 p, = t - t,-1 1.

effective

P(t) dt

s &,

x W, t)Cirl(x, dS(x) 4

(27)

Since a negligibly small part of the rate of external work is stored in the granules and the water at the stress levels considered in the experiments, the strain energy terms for granules and water in equation (20) are negligibly small compared to the frictional energy term, therefore (~~=d(S,.:idV+S,_~:PdV) The energy becomes 1 dissipation in granular media (28) then

If the efective pressure changes only slightly in the nth cycle, the error E, is small. Summation of equation (33) over all cycles up to the kth cycle gives

(~,s(t)><hs@)> dt

s
7.
a0

ri dS

Finally, equation = (E)P + (i)

(35) is written as dt

<~,&)>(L+(t)> s 10

Cyclic torsional loading

The energy dissipation for a cyclic torsion test performed on a hollow cylindrical specimen is now considered. Experiments are performed under undrained conditions. The shear strain is cyclically applied to the isotropically consolidated specimen, as already described in the experimental results. The rate of external work per unit

n=l to P = Ck P(t) dt s LO

&,+

*(c(t)) dt

(36)

(37)

14
0.3 l-00%

OKADA

AND

NEMAT-NASSER

Ftrst loading Second loadmg

O-50%

0
Number

4b of cycles

80

&I

Fig. 13. Variation of C, with number of cycles in first and second loading of loose specimens

where c, = t, 1 f:^G) dt s

Equation (37) represents the energy dissipation up to the kth cycle. C, is called the energy dissipation coefficient, and has dimensions of l/time. It is regarded as representing the microstructural arrangement of the granular mass at each instant. The coefficient C, in equation (37) can be estimated experimentally. Its variation is plotted against the number of cycles in Figs 13 and 14 for
0.5-

loose and dense specimens respectively, and against the excess pore water pressure in Figs 15 and 16. These results show that C, is nearly constant throughout the experiment, when a constant strain amplitude is cyclically applied to the specimen. There is only a small difference in the C, value between the first and second loadings, although there is a large difference between the corresponding numbers of cycles. These results suggest that C, does not depend on the strain history, but does depend on the strain amplitude. The relation of C, and the strain amplitude for both loose and dense specimens is shown in Fig. 17, which indicates that C, depends on the strain amplitude and the density, but not on the strain history.

2.00% P 0.41 .OO% .O

0
l

First loadmg Second loading

o! 0

20
Number

40 of cycles

60

80

Fig. 14. Variation of C, with number of cycles in first and second loading of dense specimens

INELASTIC

FLOW

OF GRANULAR

MEDIA
1 .OO% 0

15
l

P*

0
l

First loading Second loading 0.50%

l l

Of

eO*e

0.20%
l

to*

l P

l+?**Q*

nv

40

80 Excess pore pressure

120 kN/m*

160

200

Fig. 15. Variation of C, with excess pore water pressure in first and second loading of loose specimens

From these results it seems reasonable to assume that at constant density and strain amplitude C, is a constant, say C, related to the internal work per unit volume of a given granular mass. It can be evaluated experimentally, provided the same strain amplitude is cyclically applied to the specimen. The C values shown in Fig. 17 are used to calculate the internal work per unit volume at the end of each cycle; these results are plotted in Figs 18-21, with the corresponding external work. Clearly, the data points match closely in each case. These results show that energy dissipation in granular media for cyclic torsional loading can be

expressed in terms of the time history of the effective pressure, together with a constant C value that depends on the strain amplitude and density only.

Random torsional loading Energy dissipation for a torsional loading test in which the shear strain is applied randomly (rather than cyclically) was studied as an application of the above results. The experiments were performed using dense specimens. Shear strains were applied to a specimen at randomly varying strain amplitudes but at a constant shear strain rate,

o-5-

0 o-4
l

First loading Second loadfng

2.00%
l

l 0**0

03 z 0 Y x 0 0.2

0.1

0, 0 4b .CiO Excess pore pressure: li0 kN/m 160 260

Fig. 16. Variation of C, with excess pore water pressure in first and second loading of dense specimens

16

OKADA AND NEMAT-NASSER

Stratnamplitude.%

Fig. 17. C values obtained experimentally

10

Numbercycles of

20

30

Fig. 18. Relation between external work and calculated internal work in first loading of loose specimens

10 Number

15
of cycles

20

25

Fig. 19. Relation between external work and calculated internal work in second loading of loose specimens

INELASTIC
z 4] / / 2.00% (0.421 1.00% x 1Om4) (0.335 O-50% x 10-T (0.267

FLOW

OF GRANULAR

MEDIA
Stratn amplitude 0 t (C value)

17

External work Predtctlon

x 1O-4) 0.20% (0.155 x 10-e)

40 Number of cycles

60

60

Fig. 20. Relation between external work and calculated internal work in first loading of dense specimens

(0.421 (D * Y E : 4.F zl z a s = 2-

x 10m4)

1.00%

Strain amplitude 0

(C value)

External work l Predicton

1

20 Number

40 of cycles

60

60

Fig. 21. Relation between external loading of dense specimens

work and calculated

internal work in second

Specimen no. 39 Void ratlo 0.719 1.0 I 0 I 150 pressure: kN/m I 200 I 250

50

100 Effective

Fig. 22. Relation between shear stress and effective pressure in second loading of dense specimen subjected to random torsional loading

18

OKADA

AND

NEMAT-NASSER

i0

4b Number of zero strain

6b crossqs

Fig. 23. Relation between external work and calculated internal work in first and second loading of dense specimen subjected to random torsional loading

over each quarter cycle. The relation of the shear strain and the effective pressure in the first loading is shown in Fig. 7 and examined in the experimental results for first loading. The same relation in the second loading is shown in Fig. 22. Since the C values are essentially independent of the strain history, as discussed above, the energy balance for a random torsional loading gives, modification of equation (37) dt = f Ci
i=l

depends on the strain amplitude and the density, but is essentially independent of the stress or strain history. However, the effective pressure clearly depends on the strain history, as shown by the large difference in the number of cycles required to attain the same pore water pressure in the first and second loading. Therefore, the righthand side of equation (37) comprises a strainhistory-dependent part s& P dt and a strain-history-independent part C.

P dt s ti- 1

(39) ACKNOWLEDGEMENTS The Authors wish to express their appreciation to Professor Muneo Hori of Tokyo University for his critical suggestions and to Mr Ryuichi Sugimae for his assistance in carrying out the experiments. This work has been supported by the US Air Force Office of Scientific Research Grants AFOSR 87-0079 and AFOSR F49620-92J-01 17 to the University of California, San Diego.

where ti is the time of the ith zero shear strain and Ci is the ith C value. The value of Ci is determined as follows. First, the maximum shear strain between L,_~ and ti is found. Then, the C value corresponding to the shear strain is obtained from Fig. 17, which shows the relation of the C value and the shear strain amplitude (broken line). The internal work per unit volume is calculated from the right-hand side of equation (39), using the experimental results, and is shown in Fig. 23 together with the external work per unit volume. These two quantities are in excellent agreement; this supports the validity of equation (39).

NOTATION
energy dissipation coefficient

region considered flth sub-region in D

volume average of the rate of strain energy number of sub-regions DB at time t exterior unit normal vector number of slips in D at time t effective pressure pore water pressure average effective presure in nth cycle external pressure unit vector in the slip direction ccth slip surface union of all surfaces containing slip surface instant when the nth cycle is completed

CONCLUSIONS Energy dissipation in the flow of cohesionless granular media has been considered. A theoretical formulation is proposed, based on a simple micromechanical model. The internal work for cyclic torsional loading is shown to depend on the time-history of the effective pressure and an experimentally obtainable parameter C. The results of a series of experiments show that C

INELASTIC

FLOW

OF GRANULAR

MEDIA

19

i
[i] [ti] V (ws) (tis) dD aDB i: n 5 o r z r5 Y R

velocity velocity jump shear component of velocity jump volume of D volume average of the external work volume average of the rate of external work boundary of D boundary of Ds strain rate friction coefficient point on slip surface stress tensor tractions normal component of tractions acting on slip surface shear component of tractions acting on slip surface operator connecting effective pressure P to normal traction component T region occupied by granules in D

REFERENCES Arthur, J. R. F. & Menzies, B. K. (1972). Inherent anisotropy in a sand. Geotechnique 22, No. 1, 1155 128. Casagrande, A. (1975). Liquefaction and cyclic deformation of sands-a critical review. Proc. 5th Pan Am. Conf on Soil Mech., Buenos Aires. Castro, C. (1975). Liquefaction and cyclic mobility of saturated sands. J. Geotech. Engng Div. Am. Sot. Civ. Engrs 101, GT6, 551-569. Committee of JSSMFE on the Test Method of Relative Density of Sand (1979). Maximum-minimum density test method of sand. Procedure for testing soils, 2nd revised edn, Part 2, pp. 172-188 (in Japanese). Tokyo: Japanese Society for Soil Mechanics and Foundation Engineering. DeAlba, P., Seed, H. B. & Chan, C. K. (1976). Sand liquefaction in large-scale simple shear tests. J. Geotech. Engng Div. Am. Sot. Civ. Engrs 102, GT9, 909-927. Hight, D. W., Gens, A. & Symes, M. J. (1983). The development of a new hollow cylinder apparatus for investigating the effects of principal stress rotation in soils. Geotechnique 33, No. 4, 355-383. Hill, R. (1963). Elastic properties of reinforced solids: some theoretical principles. J. Mech. Phys. Solids 11, 357-372. Hill, R. (1967). The essential structure of constitutive laws for metal composites and polycrystals. J. Mech. Phys. Solids 15, 79-95. Ishihara, K. & Towhata, I. (1983). Sand response to cyclic rotation of principal stress directions as induced by wave loads. Soils Fdns 23, No. 4, 1 l-26. Ishihara, K. & Towhata, I. (1985). Sand response to cyclic rotation of principal stress directions as induced by wave loads (closure). Soils Fdns 25, No. 1, 117-120. Ishihara, K. & Yasuda, S. (1975). Sand liquefaction in hollow cylinder torsion under irregular excitation. Soils Fdns 15, No. 1, 45559. Konishi, J., Oda, M., & Nemat-Nasser, S. (1982). Inherent anisotropy and shear strength of assembly of oval cross-section rods. Proc. IUTAM Symp. on

Deformation and Failure of Granular Materials (eds P. A. Vermeer and H. J. Luger), 403-412. Malvern, L. E. (1969). Introduction ta the mechanics of a continuous medium. New Jersey: Prentice-Hail. Miura, S. & Toki, S. (1982). A sample preparation method and its effect on static and cyclic deformation-strength properties of sand. Soils Fdns 22, No. 1, 61-77. Nemat-Nasser, S. (1980). On behavior of granular materials in simple shear. Soils Fdns 20, No. 3, 59-73. Nemat-Nasser, S. & Hori, M. (1993). Micromechanics; overall properties of heterogeneous solids. Amsterdam: Elsevier. Nemat-Nasser, S. & Shokooh, A. (1979). A unified approach to densification and liquefaction of cohesionless sand in cyclic shearing. Can. Geotech. J. 16, 659-678. Nemat-Nasser, S. & Takahashi, K. (1984). Liquefaction and fabric of sand. J. Geotech. Engng Dio. Am. Sot. Civ. Engrs 110, No. 9, 1291-1306. Nemat-Nasser, S. & Tobita, Y. (1982). Influence of fabric on liquefaction and densification potential of cohesionless sand. Mech. Mater. 1,43-62. Oda, M. (1972a). Initial fabric and their relations to mechanical properties of granular materials. Soils Fdns 12, No. 1, 17-36. Oda, M. (1972b). Deformation mechanism of sand in triaxial compression tests. Soils Fdns 12, No. 4, 455 63. Peacock, W. H. & Seed, H. B. (1968). Sand liquefaction under cyclic loading simple shear conditions. J. Geotech. Engng Div. Am. Sot. Civ. Engrs 94, SM3, 689-708. Reynolds, 0. (1885). The dilatancy of media composed of rigid particles in contact. Phil. Mag. 20, 469-481. Rowe, P. W. (1962). The stressdilatancy relation for static equilibrium of an assembly of particles in contact. Proc. R. Sot. 269,50&527. Schofield, A. N. & Wroth, C. P. (1968). Critical state soil mechanics. New York: McGraw-Hill. Seed, H. B. (1979). Soil liquefaction and cyclic mobility evaluation for level ground during earthquakes. J. Geotech. Engng Div. Am. Sot. Civ. Engrs 105, GT2, 201-255. Silver, M. L. & Seed, H. B. (1971). Deformation characteristics of sands under cyclic loading. J. Soil Mech. Fdn Engng Div. Am. Sot. Civ. Engrs 97, SMS, 10811098. Subhash, G., Nemat-Nasser, S., Mehrabadi, M. M. & Shodja, H. M. (1991). Experimental investigation of fabric-stress relations in granular materials. Mech. Mater. 11, 87-106. Symes, M. J. P. R., Gens, A. & Hight, D. W. (1984). Undrained anisotropy and principal stress rotation in saturated sand. Geotechnique 34, No. 1, 11-27. Tatsuoka, F., Muramatsu, M. & Sasaki, T. (1982). Cyclic undrained stress-strain behavior of dense sands by torsional simple shear test. Soils Fdns 22, No. 2, 55570. Youd, T. L. (1970). Densification and shear of sand during vibration. J. Soil Mech. Fdn Engng Div. Am. Sot. Civ. Engrs 96, SM3, 863-880. Youd, T. L. (1972). Compaction of sands by repeated shear straining. J. Soil Mech. Fdn Div. Am. Sot. Civ. Engrs 98, SM7,709-725.