Professional Documents
Culture Documents
Chemoplasticity I
Chemoplasticity I
ABSTRACT: Employing a thermodynamic framework, chemomechanical couplings for shotcrete are treated in
this paper. A new material model based on multisurface chemoplasticity is presented. It accounts for hydration
kinetics and chemomechanical couplings related to the strength growth and the evolution of stiffness properties
as well as for autogeneous shrinkage in early-age shotcrete. The underlying intrinsic material functions, which
are independent of field and boundary conditions, are determined from standard material tests. As for the
numerical treatment of the constitutive equations of the material model, an extended form of the return mapping
algorithm is proposed. The constitutive equations are applied to a two-surface chemoplastic model, consisting
of a Drucker-Prager loading surface and a tension cut-off. In a companion paper, the proposed material model
is employed for 2D structural analyses of tunnels driven according to the New Austrian Tunneling Method.
=
⭸⌿
⭸ε
⭸
= ⫺ p,
⭸
⭸U
=⫺ ,
⭸
Am = ⫺
⭸⌿
⭸m
(4a–c) dp = 冘
␣僆Jact
d␣
⭸f␣
⭸
, d = 冘
␣僆Jact
d␣
⭸f␣
⭸
(11a,b)
In (4), denotes the hardening force; and Am = chemical af- where d␣ = plastic multipliers; Jact denotes the set of n active
finity. Am roughly represents the difference of chemical poten- yield surfaces, n ⱕ N. It is defined by Jact := {␣ 僆 1, 2,
tials of the reactant (water, cement) and the product phase . . . , N兩 f␣(, ) = 0}. All yield surfaces obey the Kuhn-Tucker
(hydrates). The chemical affinity is the driving force of the loading/unloading (complementarity) conditions
hydration reaction. Autogeneous shrinkage, aging elasticity,
and strength growth represent (thermodynamic) cross-effects f␣ ⱕ 0, d␣ ⱖ 0, f␣d␣ = 0 (12)
between chemical, elastic, and plastic hardening evolutions. along with the consistency requirement
These cross-effects result from Maxwell symmetries on the
free energy ⌿. To obtain these couplings, it is useful to intro- d␣df␣ = 0 (13)
duce the Legendre-Fenchel transform of , defined as
For all active yield surfaces, the consistency requirement, (13),
* = : ( ⫺ p) ⫺ , * = *(, m) (5) together with the hardening rule, (11b), results in
yielding
⭸* ⭸(* ⫺ U)
df␣ = 0 ⇔
⭸f␣
⭸
: d ⫹
⭸f␣ ⭸
⭈
⭸ ⭸m
⭸f␣ ⭸
dm = ⫺ ⭈ ⭈
⭸ ⭸ 僆Jact
d
⭸f
⭸
冘 (14)
⫺ p = , Am = (6)
⭸ ⭸m Eq. (14) constitutes a system of n equations for the n unknown
plastic multipliers d. In matrix notation, these unknowns can
The (infinitesimal) increment of ⫺ pl is obtained as be determined from
⭸2* (d) = [H]⫺1(df ⫹ dmf ) (15)
d ⫺ dp = ⌳ : d ⫹ bdm, b= (7)
⭸⭸m
where the n components of vectors (df ) and (dmf ) are given
where ⌳ = C⫺1 denotes the elastic compliance tensor; and C by
= elastic stiffness tensor. As suggested by experiments by By-
fors (1980) and Boumiz et al. (1996), C depends on m (aging ⭸f␣ ⭸f␣ ⭸
d f␣ = : d; dm f␣ = ⭈ dm (16a,b)
elasticity). The second-order coupling tensor b relates the in- ⭸ ⭸ ⭸m
crease of hydrate mass, dm, to the strain increment of chemical [H] is a quadratic matrix with n ⫻ n coefficients:
origin, dshr. In a stress-free experiment, b reduces to an iso-
tropic tensor
᭙( = 0) : dshr = b1dm (8)
H␣ = m 冉 ⭸f␣
⭸
嘸
⭸f ⭸2U0()
⭸
:
⭸2 冊 (17)
function of the hydration products (Mindess et al. 1978), with the hardening modulus H given as
冉冊
which may also affect the hardening/softening properties of 2
the material. With regards to constitutive modeling, this im- ⭸f ⭸2U0()
H=m (19)
plies the dependence of the hardening force on both the ⭸ ⭸2
plastic hardening variable and the hydrate mass m. Assum-
If kinematic hardening/softening is accounted for, it is nec-
ing, for the sake of simplicity, a linear chemoplastic coupling,
essary to introduce the hardening force as a second-order
U(, m) = mU0(), the (infinitesimal) incremental state equa-
tensor, together with a corresponding second-order tensor for
tion resulting from (4b) reads
the internal variable, . In this case, the flow rule and the
⭸2U0() hardening rule are still given by (11). The expression for the
d = ⫺0 dm ⫺ m ⭈ d (9) plastic multipliers remains to be (15), together with (16a)
⭸2
whereas the components of the vector (dmf ) are now given as
where the first term represents the strength growth because of
chemical hardening (with 0 = ⭸2U(, m)/⭸⭸m), whereas the ⭸f␣ ⭸
dm f␣ = : dm (20)
second term refers to plastic hardening/softening. The che- ⭸ ⭸m
H␣ = m 冉⭸f␣
⭸
嘸
⭸f
⭸
::冊⭸2U0()
⭸2
(21)
The described chemoplastic material model involves three
material functions depending on the hydration degree. As
shown in the following, they can be determined uniquely from
Next, a suitable evolution law for the hydration kinetics standard material tests, exploring the Maxwell symmetries of
must be designed, linking the rate ṁ to its driving force, i.e., the material model. The experimental data used hereafter are
to the chemical affinity Am. The following expression may be taken from Huber (1991). The experimentally investigated
used [e.g., Ulm and Coussy (1995)]: shotcrete is characterized by a water-cement ratio of 0.45. It
冉 冊
is composed of 350 kg cement per m3 of shotcrete and of 1,800
Ea kg aggregate per m3 of shotcrete, with the size of the aggregate
Am = ṁ exp (22)
RT ranging from 0 to 12 mm. The rapid hardening of this shot-
where the Arrhenius factor exp(Ea/RT) accounts for the ther- crete results from the admixture of an alkali-based accelerator.
mal activation of the hydration reaction. Ea denotes the acti- [For more details concerning the test device, sample prepara-
vation energy of the shotcrete; and R = universal constant of tion, etc., see Huber (1991).]
ideal gas. The viscosity = (m) increases with increasing
hydrate mass m, since the free water has to bypass the growing Hydration Kinetics
layers of hydrates. This leads to an increase of the microdif- The prediction of the hydration degree (and, thus, of its
fusion time of the free water to reach the unhydrated cement. influence on the growth of strain and strength) requires knowl-
It is noted that the couplings [(6)] a priori imply the depen- edge of the function Ã(). On the macroscopic level of a con-
dence of the chemical affinity Am on the loading history and crete sample, the hydration kinetics can be followed indirectly
the hardening history (i.e., on microcracking) of the material through the cross-effects between the hydration and the evo-
[i.e., Am = Am(, , m)]. Fortunately, these couplings can be lution of strength and autogeneous shrinkage. Thus, if the kind
considered as weak. Hence, it can be assumed that the affinity of the coupling is known, the Maxwell symmetries of the
Am depends mainly on m (Ulm and Coussy 1995), resulting in model provide direct macroscopic access to the hydration de-
the partial decoupling hypothesis gree and, thus, to the kinetics of hydration. In particular, the
Ã() = ˙ exp 冉 冊
Ea
RT
(25)
Combining fc(t) [Fig. 1(a)] and fc() (Fig. 2) yields the evo-
lution of the hydration degree, described by (t):
(t) fc(t) ⫺ fc,0 fc(t)
(t) = = = 0 ⫹ (1 ⫺ 0) (29)
⬁ fc,⬁ ⫺ fc,0 fc,⬁
where fc,0 (ⱕ0) is a reference value at = 0. Substituting (29)
into (25) yields a parameter form of the kinetics function Ã
with t as the parameter, namely
Ã(t) =
dfc(t)/dt
fc,⬁ ⫺ fc,0
exp 冉 冊
Ea
RT(t)
(30)
Aging Elasticity
Similarly, the treatment of aging elasticity only requires ex-
perimental determination of the elastic properties and of the
hydration degree as functions of the material age [i.e., C =
C(t)]. Assuming an isotropic elastic law, two material param- FIG. 5. Chemomechanical Couplings for Shotcrete: (a) Aging
Elasticity E(); (b) Dependence of Chemical Shrinkage Strains
eters, namely Young’s modulus E and Poisson’s ratio , are on Degree of Hydration
involved. The elasticity tensor is given by
E E (1993). Fig. 4(a) shows the evolution of E(t) according to the
C = 1 嘸 1K ⫹ 2GIdev, K= , G= (31) test series performed by Huber (1991). Then, using (29), E =
3(1 ⫺ 2) 2(1 ⫹ )
E() can be determined, as shown in Fig. 5(a). Since is a
with the bulk modulus K and the shear modulus G. 1 is the constant, C = C() is known. The points shown in Fig. 5(a)
second-order unit tensor; and Idev = I ⫺ 1/3 ⭈ 1 嘸 1 denotes are mean values of a series of tests. It is noted that the evo-
the deviatoric operator, with I standing for the fourth-order lution of the modulus of elasticity of shotcrete as a function
unit tensor. At a very early age of shotcrete, may be consid- of the hydration degree is quasi-linear. Most probably, this is
ered a constant; see, by analogy, Byfors (1980) and Laplante a consequence of the rather small aggregate, with diameters
JOURNAL OF ENGINEERING MECHANICS / JUNE 1999 / 695
not exceeding 12 mm. In fact, similar results were found for and
mortar and cement pastes (Boumiz et al. 1996). In contrast to
the macroscopic determination of material characteristics cho- f␣,n⫹1 ⱕ 0, ⌬␣,n⫹1 ⱖ 0, f␣,n⫹1⌬␣,n⫹1 = 0, ⌬␣,n⫹1⌬f␣,n⫹1 = 0
sen for the present research, in Boumiz et al. (1996), the elastic (37)
properties and the degree of hydration were determined by
Jact denotes the set of active yield surfaces at tn⫹1
ultrasonic measurements and by calorimetric experiments, re-
spectively. The similarity of the relations between E and , in Jact = {␣ 僆 1, 2, . . . , N兩 f␣(n⫹1, n⫹1) = 0} (38)
the sense of their linearity, demonstrates the intrinsic nature of
the sought material function E = E(). In order to be able to use the standard return mapping scheme,
at first, the trial state needs to be defined. It is attained by
Autogeneous Shrinkage means of freezing the plastic flow (i.e., ⌬ p,trial
n⫹1 = 0, ⌬ n⫹1 =
trial
n⫹1 = 冏
⭸⌿
⭸ n⫹1
= C(n⫹1) : (n⫹1 ⫺ pn⫹1 ⫺ 1shr(n⫹1)) (34)
viewpoint, it follows that
⌬n⫹1 = ⌬ trial
n⫹1 (45)
where refers to the material function illustrated in Fig.
shr
where ⌬n⫹1 = increment of the hydration degree. Conse-
5(b). quently, ⌬n⫹1 is obtained from the ordinary differential equa-
Evaluating the state equation = ⭸⌿/⭸ [(4)] at t = tn⫹1 and tion (25). The solution of (25) is uncoupled from the mechan-
accounting for the linear chemomechanical coupling shown in ical problem. For example, applying a first-order integration
Fig. 2 yields scheme with ˙ = ⌬n⫹1/⌬tn⫹1, the following relation is ob-
n⫹1 =
⭸⌿
冏 =⫺
⭸U
冏 =⫺
具n⫹1 ⫺ 0典 ⭸U⬁
冏 tained:
冉 冊
(35)
⭸ n⫹1 ⭸ n⫹1 1 ⫺ 0 ⭸ n⫹1 EA
⌬n⫹1 ⫺ ⌬tn⫹1 exp ⫺ ˜ n⫹1) = Ᏻ(⌬n⫹1) = 0
A( (46)
where 具x典 = (x ⫹ 兩x兩)/2. RT
The evolution equations of plasticity (11) as well as the
Kuhn-Tucker conditions (12), along with the consistency re- where Ã() denotes the normalized chemical affinity, shown in
quirement (13), are integrated by means of a Euler backward Fig. 3.
scheme for the interval [tn, tn⫹1], resulting in As far as the solution of the nonlinear equation (46) is con-
冘 冏
cerned, the Newton algorithm is not unconditionally stable,
⭸f␣ because Ᏻ(⌬n⫹1) is not necessarily a convex function [Fig.
⌬ pn⫹1 = ⌬␣,n⫹1 (36a)
␣僆Jact ⭸ n⫹1 6(a)].
⌬n⫹1 = 冘
␣僆Jact
⌬␣,n⫹1
⭸
冏
⭸f␣
n⫹1
(36b)
In order to find an unconditionally stable algorithm for the
solution of (46), the following characteristics of the material
function Ã() are noted:
696 / JOURNAL OF ENGINEERING MECHANICS / JUNE 1999
APPLICATION: DRUCKER-PRAGER CRITERION WITH
TENSION CUT-OFF
In this section, the preceding developments will be applied
to a plasticity model suitable for shotcrete. A Drucker-Prager
yield surface and a tension cut-off are chosen to describe the
behavior of shotcrete under primarily uniaxial and biaxial
stress states.
The Drucker-Prager loading surface is given by
f1 = ␣I1 ⫹ 兩s兩 ⫺ kDP(1, ) = 0 (51)
FIG. 6. Characteristic Shapes of Function Ᏻ(⌬n⫹1): (a) With
One Point of Inflection; (b) Without Point of Inflection The tension cut-off reads
f2 = I1 ⫺ TC(2, ) = 0 (52)
I1 = tr denotes the first invariant of the stress tensor; s =
⫺ (I11)/3 denotes the stress deviator; and ␣, kDP, and TC =
material parameters. Fig. 7 shows the loading surfaces in the
I1-兩s兩-stress space.
冉 冊 冉 冊
d2 = d23 (59)
⫺⌬ p,k ⭸ f ( kn⫹1, ( kn⫹1, n⫹1))
R kn⫹1 := n⫹1
⫹ ⌬ kn⫹1 (49) Hence
⫺⌬ n⫹1
k
⭸ f ( n⫹1
k
, ( n⫹1
k
, n⫹1))
with k indicating the number of the iteration step. For multi- d1 ⫹ d2 = d13␣ ⫹ d23 (60)
surface chemoplasticity, the plastic flow and hardening resid- Comparison of (60) and (57) yields
uals are given as
冉 冊 冘 冉 冊
d1 ⫹ d2 = tr dp (61)
⫺⌬ p,k ⭸ f␣( kn⫹1, ( kn⫹1, n⫹1))
R k
:= n⫹1
⫹ ⌬ k
␣,n⫹1
n⫹1
⫺⌬ n⫹1
k
␣僆Jact
⭸ f␣( n⫹1
k
, ( n⫹1
k
, n⫹1)) resulting in (53), q.e.d. Note that, depending on the number
(50) of active yield surfaces, both, one, or none of the plastic mul-
tipliers d1 and d2 in (57)–(60) are equal to zero [Kuhn-
For strategies to obtain the actual set of active yield surfaces Tucker conditions; (12)].
Jact during the Newton iteration scheme, one may refer to Simo Next, some comments on the structure of the matrix of hard-
and Hughes (1998). ening moduli, [H], defined by (17), for the chosen yield sur-
In the same manner, the consistent tangent operators for the faces are made. Because
global Newton-Raphson equilibrium iteration (Simo and
Hughes 1998) are defined for purely plastic evolutions with ⭸fi
=0 for i ≠ j, i, j = 1, 2 (62)
initially updated aging elasticity and chemical hardening. ⭸j
␣= 冑 2 ⫺1
3 2 ⫺ 1
(65)
Next, the general return mapping algorithm [(32)–(50)] will
be specialized for the chosen yield surfaces. Because of the
comparatively simple mathematical structure of these yield
冉 冊
surfaces, this specialization results in a significant simplifica-
1 2 ⫺ 1
1(1, ) = 1⫺ Rc(1, ) (66) tion of the general formulation.
3 ⫺1 First, the increment of the degree of hydration, ⌬n⫹1 =
where Rc = uniaxial chemoplastic compressive threshold. Em- ⌬ trial
n⫹1, is computed by solving (46). Then the trial state is
ploying a quadratic plastic hardening law and the linear determined. It is defined by (39), whereas the trial hardening
strength evolution described by fc() (Fig. 2), Rc can be ex- forces for the chosen yield surfaces read as follows [see (66)
pressed as and (70)]:
1
具 ⫺ 0典 trial
2,n⫹1 = ⫺ ft(n⫹1) (72)
Rc() = fc() = fc,⬁ (68) 3
1 ⫺ 0
The trial values of the yield surfaces are obtained [(54) and
for 1 ⱖ 1,u (Fig. 8). (55)] as
The first term in (67) accounts for pure chemical hardening,
1,n⫹1 = ␣(I 1,n⫹1 ⫹ 3 1,n⫹1) ⫹ 兩s n⫹1兩
f trial trial trial trial
whereas the second term is associated with plastic hardening. (73)
1,u denotes the value of the internal variable 1 when fc is f trial
=I trial
⫹ 3 trial
(74)
2,n⫹1 1,n⫹1 2,n⫹1
reached. As a consequence of the assumption of linear che-
moplastic coupling, i.e., of U(, m) = mU0(), [see (9)], 1,u Afterwards, the active yield surfaces are determined, and
can be determined from a uniaxial compression test at com- the trial state is projected back onto the yield surface.
1,n⫹1 < 0 and f 2,n⫹1 < 0, the (n ⫹ 1)th load step is
If f trial trial
plete hydration
(chemo)elastic. In the stress space, the trial stress state is sit-
9␣兹2(E⬁ε uc,⬁ ⫺ fc,⬁) uated within the elastic domain. It coincides with the actual
1,u = ⫺ (69)
E⬁(3␣兹2 ⫺ 2兹3) stress state.
If f trial trial
1,n⫹1 > 0 and f 2,n⫹1 < 0, only the Drucker-Prager yield
Thus, 1,u is independent of the degree of hydration, as illus-
surface is active. If f trial trial
1,n⫹1 < 0 and f 2,n⫹1 > 0, only the tension
trated in Fig. 8.
cut-off is active. Projection of the trial state from outside the
Concerning the tension cut-off, perfect chemoplasticity
elastic domain onto the yield surface allows determination of
(purely chemical hardening) is assumed, i.e.
the actual stress state. The respective projection algorithms are
1 1 具 ⫺ 0典 described in Appendix I.
2 = ⫺ ft() = ⫺ ft,⬁ (70) In the case of f trial trial
1,n⫹1 > 0 and f 2,n⫹1 > 0, either both yield
3 3 1 ⫺ 0
surfaces or one of them are active. First, both yield surfaces
Thus, the stress relaxation resulting from plastic softening is are assumed to be active. If the projection algorithm for the
neglected. Nevertheless, since experiments for shotcrete under corner region (see Appendix I) yields ⌬1,n⫹1 > 0 and ⌬2,n⫹1
tension are lacking, purely chemical hardening in the tensile > 0, this assumption is correct. If, however, either ⌬1,n⫹1 or
regime is considered to be a reasonable assumption. ⌬2,n⫹1 is negative, only the second and the first yield surface,
698 / JOURNAL OF ENGINEERING MECHANICS / JUNE 1999
respectively, are active. In this case, the respective projection ⌬1,n⫹1 = ⌬1,n⫹13␣ (76)
algorithm (see Appendix I) has to be employed.
Appendix II contains the derivation of the consistent tangent Inserting (75) into (34), making use of
moduli for the chosen yield surfaces. C(n⫹1) = 2G(n⫹1)Idev ⫹ K(n⫹1)1 嘸 1 (77)
with G and K according to (31), and recalling the definition
SUMMARY AND CONCLUSIONS of the trial stress state (39), yields
This paper has dealt, in a thermodynamic framework, with n⫹1 = trial
n⫹1 ⫺ ⌬1,n⫹1[3K(n⫹1)␣1 ⫹ 2G(n⫹1)nn⫹1] (78)
chemomechanical couplings for shotcrete. The main items of
The stress deviator can be written as follows:
the paper may be summarized as follows:
sn⫹1
sn⫹1 = n⫹1 : Idev = s trial
n⫹1 ⫺ ⌬1,n⫹12G(n⫹1) (79)
• For constitutive modeling, the chemoplastic model pro- 兩sn⫹1兩
posed by Ulm and Coussy (1996) was extended to mul- Solving (79) for s trial
n⫹1 gives
冉 冊
tisurface chemoplasticity. It is noteworthy that the as-
sumption of the partial decoupling hypothesis, stating the ⌬1,n⫹1
s trial
n⫹1 = sn⫹1 1 ⫹ 2G(n⫹1) (80)
independence of hydration kinetics from the loading and 兩sn⫹1兩
hardening history, leads to the introduction of the degree
of hydration as in cement chemistry. It follows from (80) that (Simo and Hughes 1998)
• For the experimental determination of the involved intrin- sn⫹1 s trial
n⫹1
sic material functions, the normalized chemical affinity, nn⫹1 = = trial = n trial (81)
兩sn⫹1兩 兩s n⫹1兩
n⫹1
constituting the driving force of the hydration reaction,
was determined on the basis of the time-dependent In other words, for a Drucker-Prager yield surface, the direc-
strength evolution, taking both the linear chemoplastic tion of the plastic flow at the time instant t = tn⫹1, ␣1 ⫹ nn⫹1,
coupling between the degree of hydration and strength is known as soon as the trial state has been evaluated. This
and the current temperature of the specimen into account. allows reduction of the residuals (49) to a single nonlinear
Knowledge of the hydration kinetics and the time-depen- equation.
dent evolutions of stiffness and autogeneous shrinkage al- If the Drucker-Prager loading surface is active, the consis-
lowed the determination of the intrinsic material functions tency condition
for stiffness and autogeneous shrinkage.
• For the numerical treatment of the constitutive equations f1(n⫹1, 1,n⫹1) = 0 (82)
of the material model, the return mapping algorithm must be fulfilled [see (37)]. Inserting (78) and (66) into (82),
(Simo and Hughes 1998) was extended to multisurface and accounting for (76) and (81), one gets
chemoplasticity. In contrast to classical multisurface plas-
ticity, the trial hardening force is not equal to the hard- f1,n⫹1 = ␣I trial
1,n⫹1 ⫹ 兩sn⫹1兩 ⫺ ⌬1,n⫹1[9K(n⫹1)␣ ⫹ 2G(n⫹1)]
trial 2
冉 冊
ening force at the end of the preceding load increment.
However, because of the partial decoupling hypothesis, 2 ⫺ 1
⫹␣ 1⫺ Rc(1,n ⫹ ⌬1,n⫹13␣, n⫹1) = 0
the degree of hydration can be updated independently ⫺1 (83)
from the mechanical state variables. Hence, the standard Eq. (83) represents a nonlinear equation for determination of
return mapping algorithm was performed with an initially the plastic multiplier ⌬1,n⫹1. Using the threshold Rc according
updated chemical hardening. In the same manner, the con- to (67), (83) can be solved in an unconditionally stable man-
sistent tangent moduli were derived with an initially up- ner, using Newton’s method. Then, (75) yields the plastic
dated aging elasticity and chemical hardening. strains, whereas (78) gives the actual stress state.
• For the application of the material model, the constitutive
equations were elaborated for a Drucker-Prager loading Projection Algorithm for Tension Cut-Off
surface with a tension cut-off. The Drucker-Prager loading
surface represents the behavior of shotcrete under uniaxial For the tension cut-off, the discretized form of the flow rule
and biaxial compression, whereas the tension cut-off ac- [(36a)] reads
counts for the material behavior in tension.
⌬ pn⫹1 = ⌬2,n⫹11 (84)
Inserting (84) into (34), making use of (77) and of (39), gives
In the companion paper, the presented material model is
applied to 2D structural analyses of tunnels driven according n⫹1 = trial
n⫹1 ⫺ ⌬2,n⫹13K(n⫹1)1 (85)
to the NATM.
If the tension cut-off is active, the consistency condition
⌬ pn⫹1 = ⌬1,n⫹1 冉
␣1 ⫹
sn⫹1
冊
兩sn⫹1兩
= ⌬1,n⫹1(␣1 ⫹ nn⫹1) (75) ⌬2,n⫹1 =
1
9K(n⫹1)
1,n⫹1 ⫺ ft(n⫹1)]
[I trial (88)
with n = s/兩s兩. The discretized form of the hardening rule, With the help of (88), the plastic strains are obtained from
(36b), reads (84) and the actual stress state from (85).
JOURNAL OF ENGINEERING MECHANICS / JUNE 1999 / 699
Projection Algorithm for Corner Region df1 df1
df1 = 0 = : d ⇔ =0 (97)
If both loading surfaces are active, the discretized form of d d
the flow rule [(36a)] reads permits computation of the derivative of the plastic multiplier
⌬ p
= ⌬1,n⫹1(␣1 ⫹ nn⫹1) ⫹ ⌬2,n⫹11 (89) with respect to :
冏 冉 冏冊
n⫹1
⫺1
In this case, the discretized form of the hardening rule [(36b)] d⌬1 d1
= 9K(n⫹1)␣2 ⫹ 2G(n⫹1) ⫺ 3␣
is obtained as d n⫹1 d1 n⫹1
f1(n⫹1, 1,n⫹1) = 0 and f2(n⫹1, 2,n⫹1) = 0 (92) Considering (91), evaluation of d/d at t = tn⫹1 gives
must be fulfilled [see (37)]. Inserting (91), (66), and (70) into
(92) and accounting for (81) and (90) gives
Cep =
d
d 冏
n⫹1
= C(n⫹1) ⫺ [3K(n⫹1)␣1 ⫹ 2G(n⫹1)nn⫹1]
f1,n⫹1 = ␣I trial
1,n⫹1 ⫹ 兩s n⫹1兩 ⫺ ⌬1,n⫹1(9K(n⫹1)␣ ⫹ 2G(n⫹1))
冏
trial 2
d⌬1 1
嘸 ⫺ 4G2(n⫹1)⌬1,n⫹1 (Idev ⫺ nn⫹1 嘸 nn⫹1)
冉 2 ⫺ 1
冊 d n⫹1 兩s trial
n⫹1兩
冏
⫹␣ 1⫺ Rc(1,n ⫹ ⌬1,n⫹13␣, n⫹1)
⫺1 d⌬2
⫺ 3K(n⫹1)1 嘸
⫺ ⌬2,n⫹19K(n⫹1)␣ = 0 (93) d n⫹1 (100)
The consistency conditions [see (13)]:
1,n⫹1 ⫺ ⌬1,n⫹19K(n⫹1)␣ ⫺ ⌬2,n⫹19K(n⫹1)
f2,n⫹1 = I trial
df1 df2
⫺ ft(n⫹1) = 0 (94) =0 and =0 (101)
d d
Eqs. (93) and (94) constitute a system of two equations for form a system of two linear equations for the two unknowns
determination of the two unknown plastic multipliers ⌬1,n⫹1 d⌬1/d兩n⫹1 and d⌬2/d兩n⫹1, resulting in
and ⌬2,n⫹1. This system of equations is linear with respect to
⌬2,n⫹1. Therefore, it can be reduced to one nonlinear equation d⌬1
冏 2G(n⫹1)n trial
冏
n⫹1
= (102)
for ⌬1,n⫹1, namely d d1
冉 冊
n⫹1
2G(n⫹1) ⫺ 3␣
2 ⫺ 1 d1 n⫹1
兩s trial
n⫹1兩 ⫺ ⌬1,n⫹12G(n⫹1) ⫹ ␣ 1⫺
⫺1 and
⭈ Rc(1,n ⫹ ⌬1,n⫹13␣, n⫹1) ⫹ ␣ft(n⫹1) = 0 (95) d⌬2
冏 1 2G(n⫹1)␣n trial
冏
1⫺
n⫹1
= (103)
Using the threshold Rc according to (67), (95) can be solved d n⫹1 3 d1
in an unconditionally stable manner, using Newton’s method. 2G(n⫹1) ⫺ 3␣
d1
Then, ⌬2,n⫹1 can be determined from (94), the increment of
n⫹1
plastic strain from (89), and the actual stress state from (91). ACKNOWLEDGMENTS
APPENDIX II. CONSISTENT TANGENTS This research work was done through close cooperation between the
Vienna University of Technology and the Laboratoire Central des Ponts
Differentiation of (78), (85), and (91) with respect to re- et Chaussées, Paris. The writers would like to express their gratitude to
sults in the consistent tangents, which are a necessary prereq- these institutions for having made this scientific exchange possible.
uisite for a quadratic rate of convergence of the global New-
ton-Raphson equilibrium iteration. APPENDIX III. REFERENCES
Acker, P. (1988). ‘‘Comportement mécanique du béton: apports de
Consistent Tangent for Drucker-Prager Loading l’approche physico-chimique [mechanical behavior of concrete: a phys-
Surface ico-chemical approach].’’ Tech. Res. Rep. LCPC 152, Laboratoires des
Ponts et Chaussées, Paris (in French).
Considering (78), evaluation of d/d at t = tn⫹1 gives Boumiz, A., Vernet, C., and Cohen Tenoudij, F. (1996). Mechanical prop-
冏
erties of cement pastes and mortars at early age.’’ Advanced Cement
d Based Mat., 3, 94–106.
C ep
n⫹1 = = C(n⫹1) ⫺ [3K(n⫹1)␣1 ⫹ 2G(n⫹1)nn⫹1] Byfors, J. (1980). ‘‘Plain concrete at early ages.’’ Tech. Rep., Swedish
d n⫹1
Cement and Concrete Research Institute, Stockholm, Sweden.
嘸
d⌬1
d 冏n⫹1
⫺ 4G2(n⫹1)⌬1,n⫹1
1
兩s trial
n⫹1兩
(Idev ⫺ nn⫹1 嘸 nn⫹1)
Catharin, P. (1978). ‘‘Hydratationswärme und Festigkeitsentwicklung [hy-
dration heat and strength evolution].’’ Tech. Rep. 31, Forschungsinstitut
des Vereins der österreichischen Zementfabrikanten, Vienna (in
(96) German).
Coussy, O. (1995). Mechanics of porous continua. Wiley, Chichester,
The consistency condition [see (13)]: U.K.