A thermodynamically consistent formulation of the

Johnson–Cook model
Charles Mareau

To cite this version:

Charles Mareau. A thermodynamically consistent formulation of the Johnson–Cook model. Mechanics
of Materials, 2020, 143, pp.103340. �10.1016/j.mechmat.2020.103340�. �hal-04279018�

HAL Id: hal-04279018

https://hal.science/hal-04279018
Submitted on 10 Nov 2023

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
Highlights
A thermodynamically consistent formulation of the Johnson-Cook
model
Charles Mareau

• The model of Johnson and Cook is revisited from the view point of
thermodynamics

• The resulting analysis shows that the original model displays some
limitations

• Such limitations are circumvented with an alternative but nevertheless

similar model

• Some applications of both the original and alternative models are pre-
sented
 A thermodynamically consistent formulation of the
Johnson-Cook model
Charles Mareau
Arts et Métiers Paris Tech - LAMPA
Arts et Métiers ParisTech, Campus d’Angers, LAMPA, 2 bd du Ronceray, 49035 Angers
Cedex 1, France

Abstract
The model of Johnson and Cook, which includes a viscoplastic flow rule
and a damage criterion, is widely used to describe the mechanical behaviour
of metallic materials subjected to severe loading conditions, such as those
encountered during fabrication operations or impact. This model has been
built on empirical, rather than physical, grounds. The present paper there-
fore aims at revisiting the model of Johnson and Cook from the view point
of thermodynamics with internal variables. The interest of this approach is
twofold. First, it provides a guide for the construction of a complete thermo-
mechanical constitutive model, with some constitutive relations not only for
the stress tensor but also specific internal energy, specific entropy and heat
flux vector. Second, it allows highlighting some possible limitations of the
original model of Johnson and Cook. Such limitations can be circumvented
with an alternative model, which is described in the present work. For il-
lustration purpose, some applications of both the original and alternative
models are presented in the final section.
Keywords: Johnson-Cook, Viscoplasticity, Damage, Thermodynamics,
Finite strain

1. Introduction
During fabrication operations or impacts, materials often experience a
wide range of strain rates and temperatures. Different models have therefore
been developed to include the impact of strain rate and temperature on
the mechanical behaviour of solid materials. Relying on either empirical or

Preprint submitted to Mechanics of Materials January 21, 2020

physical arguments, these models provide some relations to (i) connect the
flow stress to the equivalent plastic strain, the equivalent plastic strain rate
and the temperature and (ii) determine whether the conditions for fracture
are met or not. The former set of relations is usually referred to as the flow
rule while the latter is a fracture criterion.
The most common model for metallic materials is likely the one of John-
son and Cook (1983, 1985), which includes a thermo-viscoplastic flow rule
and a ductile damage criterion. This model is largely used for the numerical
simulation of fabrication operations, with numerous applications to machin-
ing (Umbrello et al., 2007; Shrot and Bäker, 2012; Hor et al., 2013; Agmell et
al., 2014) as well as forging (Yoo and Yang, 1997; Djavanroodi et al., 2019).
It has been built upon an experimental basis, with some observations of both
the deformation behaviour and fracture behaviour of different metallic mate-
rials over a wide range of strain rates and temperatures. The resulting flow
rule is based on a multiplicative decomposition of the flow stress into strain
hardening, strain-rate hardening and thermal softening contributions. To
extend the range of applicability of the Johnson-Cook model, some modified
versions of this model have emerged. For instance, many options have been
explored to consider the coupled effects of temperature and strain rate (He
et al., 2013; Chen et al., 2010; Wang et al., 2017). A modified version has
also been proposed by Couque et al. (2006) to account for the dislocation
drag regime that governs the deformation behaviour at very high strain rates
(≥ 105 s−1 ). Camacho and Ortiz (1997) modified the strain-rate sensitivity
term of the viscoplastic flow rule to describe the behaviour at low strain
rates. A strain-rate softening contribution, which is ignored in the original
model, has been included in the model developed by Zhao et al. (2017).
Though the model of Johnson and Cook (1983, 1985) is often used for
thermomechanical simulations, it has not been developed in a consistent ther-
modynamic framework. The Johnson-Cook model therefore has two major
drawbacks. First, it does not provide any constitutive relation for thermo-
dynamic quantities such as internal energy, heat flux vector or entropy. The
direct consequence is that it is not possible to identity the expressions of
the different heat sources contributing to the heat diffusion equation. Sec-
ond, the compatibility of constitutive relations with the fundamental laws of
thermodynamics cannot be checked so that violations of these laws cannot
be excluded. Specifically, the non-negativity of the heat dissipation source,
which is implied by the second of law of thermodynamics, is not ensured.
The present work therefore aims at reformulating the Johnson-Cook model

2
in a consistent thermodynamic framework. For this purpose, the concept of
internal state variables (Maugin and Muschik, 1994) is used to consider the
microstructural transformations (e.g. strain hardening, ductile damage) that
have been included in the original Johnson-Cook model. This concept pro-
vides a way of revisiting the Johnson-Cook model with an alternative point of
view. This allows deriving a complete thermomechanical finite strain consti-
tutive model, with some constitutive relations not only for the stress tensor
but also specific internal energy, specific entropy and heat flux vector. The
present paper is organized as follows. The first section is dedicated to the de-
scription of constitutive relations. Because the original Johnson-Cook model
displays some limitations, different options for considering thermal softening
and strain rate effects are explored. For illustration purpose, some appli-
cations, which allow comparing these options, are presented in the second
section.

2. Constitutive relations
For the purpose of describing the evolution of a deformable body, the
mass, momentum and energy conservation equations must be supplemented
with a constitutive model. Within a thermomechanical context, a constitu-
tive model consists of some relations for the stress state, the specific entropy,
the specific internal energy and the heat flux vector. Some additional evolu-
tion relations can also be included to consider the microstructural transfor-
mations (e.g. hardening, damage) having an impact on the thermomechanical
behaviour of a material point. In the present section, a set of constitutive
relations, which includes the Johnson-Cook viscoplastic flow rule and the
Johnson-Cook fracture criterion, is detailed.

2.1. Kinematics
For the presentation of the constitutive model, a single material point is
considered. For such a material point, the multiplicative decomposition of
the deformation gradient tensor F suggested by Lee (1969) is adopted so
that:
F = Fθ · Fp (1)
where F θ and F p are respectively the thermoelastic and plastic contributions
to the deformation gradient tensor. The concept of the isoclinic relaxed
intermediate configuration proposed by Mandel (1973) is used here as it
provides a convenient way of formulating constitutive models.

3
 Because we restrict our attention to the specific, but rather common, case
where the plastic flow is incompressible, the initial and current mass densities
ρ0 and ρ are connected to each other with:

ρ0 = Jρ (2)

with J = detF = detF θ .

In the following, the Green-Lagrange strain tensor E is used as a strain
measure. According to the above multiplicative decomposition, the Green-
Lagrange strain tensor E is separated into thermoelastic and plastic contri-
butions as follows:
1
F T · F − 1 = F Tp · Ẽ θ · F p + E p


E= (3)
2
with:
1
F Tθ · F θ − 1


Ẽ θ = (4)
2
1
F Tp · F p − 1


Ep = (5)
2
While E and E p are associated with the initial configuration, the thermoe-
lastic strain tensor Ẽ θ is attached to the isoclinic intermediate configuration.
In a similar fashion, the velocity gradient tensor L is decomposed accord-
ing to:
L = Ḟ · F −1 = Lθ + F θ · L̃p · F −1
θ (6)
with:

Lθ = Ḟ θ · F −1
θ (7)
L̃p = Ḟ p · F −1
p (8)

The above relation indicates that L̃p is associated with the intermediate
configuration while both L and Lθ are attached to the current configuration.

2.2. Thermodynamics
For a closed system, according to the first law of thermodynamics (i.e.
energy conservation), the evolution of the specific internal energy u for a
material point is given by:
σ :L ∇·q
u̇ = − +r (9)
ρ ρ

4
where σ is the Cauchy stress tensor, q is the heat flux density vector and r is
the specific heat source. From the decomposition of the velocity gradient ten-
sor L, the above energy conservation equation is conveniently reformulated
on the intermediate configuration as follows:
˙
S̃ : Ẽ Σ̃ : L̃p ∇˜ · Q̃
θ
u̇ = + − +r (10)
ρ0 ρ0 ρ0

where S̃ is second Piola-Kirchoff stress tensor associated with the intermedi-

ate configuration and Σ̃ is the Mandel stress tensor (Mandel, 1973). These
stress measures are connected to the Cauchy stress tensor σ with:

S̃ = JF −1 −T
θ · σ · Fθ (11)
Σ̃ = JF Tθ ·σ· F −T
θ (12)

˜ operator, which appear in equation

The heat flux density vector Q̃ and the ∇
(10), are defined according to:

Q̃ = JF −1
θ ·q (13)
˜ = ∇ · Fθ
∇ (14)

According to the second law of thermodynamics, the rate of entropy pro-

duction per unit mass η must be non-negative. For a closed system, the
corresponding inequality is:
r 1 q
η = ṡ − + ∇ · ≥0 (15)
T ρ T
where s is the specific entropy and T is the absolute temperature. The
expression of the specific dissipation source d is obtained from:

d = ηT ≥ 0 (16)
1 1
= ṡT − r + ∇ · q − q · ∇T ≥ 0 (17)
ρ ρT
Using the local form (9) of the first law of thermodynamics, one obtains the
following expression of the specific dissipation source d:
σ:L 1
d= − u̇ + ṡT − q · ∇T ≥ 0 (18)
ρ ρT

5
The above dissipation inequality can alternatively be formulated on the in-
termediate configuration with (10), that is:
˙
S̃ : Ẽ Σ̃ : L̃p 1
θ ˜ ≥0
d= + − u̇ + ṡT − Q̃ · ∇T (19)
ρ0 ρ0 ρ0 T
2.3. Specific free energy
The specific Helmholtz free energy a = u − sT is a thermodynamic poten-
tial that can be used to obtain the different state equations. In the following,
the list of state variables, on which the specific free energy a depends, includes
the thermoelastic strain tensor Ẽ θ , the absolute temperature T , a scalar
isotropic hardening variable P , which is the accumulated plastic strain, a
tensorial kinematic hardening variable Ãk , which is a Green-Lagrange strain-
like variable, and a damage variable D (with D ∈ [0, 1]). In the present work,
the following decomposition of the specific free energy is assumed:
h i h i h i
a Ẽ θ , Ãk , P, D, T = aθ Ẽ θ , D, T +ath [T ]+ak Ãk , D, T +ai [P, T ] (20)

The thermoelastic contribution aθ is such that:

1 1
aθ = Ẽ θ : L : Ẽ θ − Ẽ θ : L : α (T − T0 )
2ρ0 ρ0
(21)
1
+ α : (L − Lu ) : α (T − T0 )2
2ρ0
The above expression uses the current and initial fourth-rank stiffness ten-
sors L and Lu , the second-rank thermal expansion tensor α and the refer-
ence temperature T0 . The current fourth-rank stiffness tensor L depends on
the damage variable D, the thermoelastic strain tensor Ẽ θ and the absolute
temperature T . To consider the stiffness reduction associated with the devel-
opment of damage, the following strategy is adopted. To account for possible
closure effects, two different situations, which depend on the elastic strain
tensor Ẽ e = Ẽ θ − α (T − T0 ), are considered:
(
Lt , Ẽ e : 1 ≥ 0
L= (22)
Lc , Ẽ e : 1 < 0

In the above equation, Lt (respectively Lc ) is the stiffness tensor correspond-

ing to a positive (respectively negative) spherical elastic strain tensor. At a

6
given time t, the stiffness tensors Lt and Lc are calculated from the damage
variable D and the initial stiffness tensor Lu with:

Lt = (1 − D) Lu (23)
Lc = Lt + Ps : (Lu − Lt ) : Ps (24)

Equation (24) uses the fourth-rank spherical projection tensor Ps . This tensor
is defined by:
1
Ps = 1 ⊗ 1 (25)
3
According to equation (24), the stiffness tensor Lc is quite similar to Lt ,
except that the spherical contribution is recovered because of closure effects.
The impact of temperature on the undamaged stiffness tensor Lu , which
corresponds to the situation where D = 0, is described with:

Lu = L0 (1 − β (T − T0 )) (26)

where L0 is the initial stiffness tensor at T = T0 and β is a material parameter.

To consider the evolution of thermal expansion properties with respect to
temperature, the following relation is used:

α = α0 (1 + ζ (T − T0 )) (27)

where α0 is the thermal expansion tensor at T = T0 and ζ is a constant that

allows considering the impact of temperature on thermal expansion proper-
ties.
The thermal contribution to free energy ath , which solely depends on the
absolute temperature T , is:
  
T 1
ath = (c0 − γT0 ) T − T0 − T ln − γ (T − T0 )2 (28)
T0 2
As we shall see later, the evolution of the specific heat capacity as a function
of temperature is controlled with the material parameters c0 and γ.
The contribution of isotropic hardening to free energy is given by:
1 B
ai = P n+1 (29)
ρ0 n + 1
In the above equation, n is the strain hardening exponent and B is the
temperature-dependent isotropic hardening modulus. To consider the yield

7
stress reduction caused by a temperature increase, the hardening modulus B
must depend on the absolute temperature T . In the following, two different
options are explored for the consideration of thermal softening:
  m 
T − T0
B = B0 1 − (30a)
TM − T0
T m − T0m
 
B = B0 1 − m (30b)
TM − T0m
where B0 and m are material parameters and TM is the absolute melting
temperature. The first option (30a) corresponds to the original proposition
of Johnson and Cook (1983). Except from the specific case where m = 1,
this option has two major drawbacks. First, it is valid only for temperatures
comprised between T0 and TM . Second, as will be discussed later, this option
does not allow determining the specific heat capacity at T = T0 when m < 2.
The aforementioned drawbacks are avoided with the second option (30b),
which is nevertheless similar to the original proposition of Johnson and Cook
(1983) in the sense that the conditions B[T0 ] = B0 and B[TM ] = 0 are both
met.
A quadratic form is assumed for the contribution ak of the kinematic
hardening variable Ãk to free energy so that:
1
ak = Ãk : K : Ãk (31)
2ρ0
where K is the kinematic hardening moduli tensor. The kinematic hardening
moduli tensor is related to the damage variable in a manner similar to that
used for the stiffness tensor. Depending on the spherical part of the kinematic
hardening variable Ãk , the following situations are therefore considered:
(
Kt , Ãk : 1 ≥ 0
K= (32)
Kc , Ãk : 1 < 0
The kinematic hardening moduli tensors Kt and Kc are given by:
Kt = (1 − D) Ku (33)
Kc = Kt + Ps : (Ku − Kt ) : Ps (34)
The initial kinematic hardening moduli tensor Ku depends on the absolute
temperature T according to:
Ku = K0 (1 − β (T − T0 )) (35)

8
The kinematic hardening moduli tensor K0 should be defined in agreement
with the possible constraints due to material symmetry.

2.4. State equations

The specific free energy a allows determining the state equations connect-
ing the state variables to the corresponding thermodynamic forces. First,
with the above definition of the free energy, the second Piola-Kirchoff stress
tensor S̃ associated with the intermediate configuration is:
∂a
S̃ = ρ0 (36)
∂ Ẽ θ
 
= L : Ẽ θ − α (T − T0 ) = L : Ẽ e (37)

The dual variable of the accumulated plastic strain P is the scalar stress
R, which represents the yield stress increase associated with strain hardening.
The stress R is connected to the accumulated plastic strain P with:
∂a
R = ρ0 (38)
∂P
= BP n (39)

The relation between the kinematic stress tensor X̃, which defines the
center of the yield surface in the stress space, and the kinematic hardening
variable Ãk is:
∂a
X̃ = ρ0 (40)
∂ Ãk
= K : Ãk (41)

It is emphasized that, because of the definition of the kinematic hardening

variable, the kinematic stress tensor X̃ behaves similarly to the second Piola-
Kirchoff stress tensor S̃.
The thermodynamic force driving the development of damage is denoted
by −Y . This thermodynamic force is given by:
∂a
−Y = ρ0 (42)
∂D
1 ∂L 1 ∂K
= Ẽ e : : Ẽ e + Ãk : : Ãk (43)
2 ∂D 2 ∂D

9
Because of closure effects, the following cases have to be considered when
evaluating the derivative of the stiffness tensor with respect to the damage
variable: (
∂L −Lu , Ẽ e : 1 ≥ 0
= (44)
∂D −Lu + Ps : Lu : Ps , Ẽ e : 1 < 0
In a similar fashion, the derivative of the kinematic moduli tensor K with
respect to the damage variable is given by:
(
∂K −Ku , Ãk : 1 ≥ 0
= (45)
∂D −Ku + Ps : Ku : Ps , Ãk : 1 < 0

Finally, the specific entropy s, which is the dual variable of temperature

T , is given by:

∂a
s=− (46)
∂T  
1 ∂L ∂α
= Ẽ θ : : α (T − T0 ) + L : (T − T0 ) + L : α
ρ0 ∂T ∂T
 
1 ∂L T
− Ẽ θ : : Ẽ θ + (c0 − γT0 ) ln + γ (T − T0 )
2ρ0 ∂T T0
  (47)
1 ∂λu ∂λ 1
+ − (T − T0 )2 + (λu − λ) (T − T0 )
2ρ0 ∂T ∂T ρ0
1 ∂B n+1 1 ∂K
− P − Ãk : : Ãk
ρ0 (n + 1) ∂T 2ρ0 ∂T
For the purpose of conciseness, the above state equation for specific entropy
uses the variables λ and λu that are defined according to:

λu = α : Lu : α (48)
λ=α:L:α (49)

The list of state equations for the proposed model consists of equations
(37), (39), (41), (43) and (47).

2.5. Specific dissipation source

For the evolution of a material point to be compatible with the second law
of thermodynamics, the specific dissipation source d must be non-negative.

10
Using the relation u = a + sT , the dissipation inequality (19) becomes:
˙
S̃ : Ẽ Σ̃ : L̃p 1
θ ˜ ≥0
d= + − ȧ − sṪ − Q̃ · ∇T (50)
ρ0 ρ0 ρ0 T
In the following, the distinction is made between the thermal dissipation
source dth and the intrinsic dissipation source din . These two dissipation
sources are assumed to be separated in the sense that each of them is required
to be non-negative:
˙
S̃ : Ẽ Σ̃ : L̃p
θ
din = + − ȧ − sṪ ≥ 0 (51)
ρ0 ρ0
1 ˜ ≥0
dth = − Q̃ · ∇T (52)
ρ0 T
Using the definitions (36), (38), (40), (42) and (46) of the thermodynamic
forces, the time derivative of the specific free energy is expressed by:
1 ˙ + 1 X̃ : Ã
˙ + 1 RṖ − 1 Y Ḋ − sṪ
ȧ = S̃ : Ẽ θ k (53)
ρ0 ρ0 ρ0 ρ0
The above expression allows reformulating the specific intrinsic dissipation
din source as follows:
1 1 ˙ − 1 RṖ + 1 Y Ḋ
din = Σ̃ : L̃p − X̃ : Ã k (54)
ρ0 ρ0 ρ0 ρ0
For the construction of a yield criterion, it is necessary to define a kine-
matic stress tensor that behaves like Σ̃. For this purpose, it is worth mention-
ing that the Mandel stress tensor Σ̃ is connected to the second Piola-Kirchoff
stress tensor S̃ with:
Σ̃ = F Tθ · F θ · S̃ (55)
Guided by the above relation, the following Mandel stress-like tensor χ̃ is
introduced:
χ̃ = F Tθ · F θ · X̃ (56)
This definition of the kinematic stress tensor χ̃ leads to the following expres-
sion of the intrinsic dissipation source:
1 1 1 1
din = Σ̃ : L̃p − χ̃ : L̃k − RṖ + Y Ḋ (57)
ρ0 ρ0 ρ0 ρ0

11
with:
L̃k = F −1 −T ˙
θ · F θ · Ãk (58)
Expressions (52) and (57) emphasize that, for the constitutive model to be
complete, some evolution rules for L̃p , L̃k , Ṗ , Ḋ and Q̃ must be proposed,
with the restriction that the specific intrinsic and thermal dissipation sources
remain non-negative.

2.6. Evolution equations

Fourier’s law of heat conduction. According to Fourier’s law of heat conduc-
tion, the constitutive relation for the heat flux density vector Q̃ is:
˜
Q̃ = −κ · ∇T (59)

where κ is the thermal conductivity tensor, which has to be positive-definite

for the thermal dissipation source dth to be non-negative. The influence
of temperature and damage on the thermal conductivity tensor is modelled
with:
κ = κ0 (1 + ϑ (T − T0 )) (1 − D) (60)
where κ0 is the thermal conductivity tensor at the reference temperature and
ϑ is a material parameter.

Viscoplastic flow rule. For the construction of the plastic flow rule, it is con-
venient to introduce a scalar function φ that allows determining an equivalent
stress Σeq such that: h i
Σeq = φ Σ̃d − χ̃d (61)

Because plastic flow is assumed to be incompressible (i.e. L̃p : 1 = 0), the

equivalent stress solely depends on the deviatoric parts Σ̃d and χ̃d of the
applied and kinematic stress tensors Σ̃ and χ̃. The deviatoric stress tensors
Σ̃d and χ̃d are given by:

Σ̃d = Σ̃ − Ps : Σ̃ = Pd : Σ̃ (62)
χ̃d = χ̃ − Ps : χ̃ = Pd : χ̃ (63)

where Pd = I − Ps is the deviatoric projection tensor.

12
 If the normality rule is adopted, the evolution laws for L̃p and L̃k have
the following form:
∂φ ∂φ
L̃p = Ṗ = Ṗ : Pd (64)
∂ Σ̃ ∂ Σ̃d
∂φ ∂φ
L̃k = −Ṗ = −Ṗ : Pd (65)
∂ χ̃ ∂ χ̃d

Since the equivalent stress depends only on the difference between Σ̃d and
χ̃d , we have1 :
∂φ ∂φ
=− =N (66)
∂ Σ̃ ∂ χ̃
where N is the plastic flow direction. Using the above definition of the plastic
flow direction, one obtains:

L̃k = L̃p = Ṗ N (67)

Also, because φ is a homogeneous function of degree one, the application of

Euler’s homogeneous function theorem leads to:
 
Σ̃ − χ̃ : N = Σeq (68)

Using the above property, the expression of the intrinsic dissipation source
din reduces to:
1 1
din = (Σeq − R) Ṗ + Y Ḋ (69)
ρ0 ρ0
Because the equivalent plastic strain rate Ṗ is non-negative, the following
conditions must hold for the intrinsic dissipation source to be non-negative:
(
Ṗ = 0, Σeq < R
(70)
Ṗ ≥ 0, Σeq ≥ R

1
For isotropic materials, the Mandel stress tensor Σ̃ is symmetric (Gurtin et al., 2010).
For this specific case, the plastic flow is irrotational in the sense that, due to the adoption of
the normality rule, the plastic contribution to the velocity gradient tensor L̃p is symmetric
as well.

13
 Prior to discussing the possible choices for the evolution equation for Ṗ ,
usually known as the viscoplastic flow rule, the yield stress Σy of a material
point is defined according to:
Σy = A + R (71)
where A represents the initial yield stress, i.e. when no plastic strain has
been accumulated yet. The initial yield stress A is usually very dependent
on the absolute temperature. To include this dependence of A with respect
to temperature, a similar relation to that of B is used. As a result, the
following options for the initial yield stress are considered:
  m 
T − T0
A = A0 1 − (72a)
TM − T0
T m − T0m
 
A = A0 1 − m (72b)
TM − T0m
where A0 is the yield stress at the reference temperature. According to the
state equation (39), the yield stress is therefore given by either one or the
other following relations:
  m 
n T − T0
Σy = (A0 + B0 P ) 1 − (73a)
TM − T0
T − T0m
m
 
n
Σy = (A0 + B0 P ) 1 − m (73b)
TM − T0m

A stronger requirement than (70) consists in assuming that no plastic

flow can occur when the equivalent stress Σeq is inferior to the yield stress
Σy , that is: (
Ṗ = 0, Σeq < Σy
(74)
Ṗ ≥ 0, Σeq ≥ Σy
In the present work, two different possibilities are considered for the con-
struction of the viscoplastic flow rule:
(
0,  Σeq < Σy
Ṗ = 
Σeq 1 (75a)
Ṗ0 exp CΣy − C , Σeq ≥ Σy
(
0, Σeq < Σy
Ṗ = 
Σeq
 (75b)
Ṗ0 exp CΣ y
− C1 − Ṗ0 , Σeq ≥ Σy

14
 where C is a material parameter and Ṗ0 is a reference strain rate. The
flow rule (75a) corresponds to the original proposition of Johnson and Cook
(1983). It is clear that, when the conditions for plastic flow are met (i.e.
Σeq ≥ Σy ), the minimum value of the equivalent plastic strain rate is Ṗ0 .
The Johnson-Cook model is therefore not applicable for low strain rates (i.e.
Ṗ < Ṗ0 ). Also, the equivalent plastic strain rate is a discontinuous function
of Σeq /Σy , which may give rise to a non-smooth elastic-plastic transition.
To circumvent these limitations, an alternative flow rule, given by equation
(75b), is proposed. As illustrated by Figure 1, for large strain rates (i.e.
when Ṗ /Ṗ0  1), this alternative flow rule is quasi-identical to the original
Johnson-Cook model. For low strain rates, the modified Johnson-Cook model
of Camacho and Ortiz (1997) is retrieved.

Figure 1: Comparison between the original and alternative viscoplastic flow rules. For the
present comparison, the C parameter is set to 0.01.

The classical presentation of the Johnson-Cook constitutive model is ob-

tained from the definition of the yield stress (73) and the flow rule (75).
Depending on the chosen flow rule and yield stress definition, one has:
!!   m 
n Ṗ T − T0
Σeq = (A0 + B0 P ) 1 + C ln 1− , Ṗ ≥ Ṗ0 (76a)
Ṗ0 TM − T0
!! 
T m − T0m

n Ṗ
Σeq = (A0 + B0 P ) 1 + C ln 1 + 1− m , Ṗ > 0 (76b)
Ṗ0 TM − T0m
The original Johnson-Cook model, which results from the combination of
equations (73a) and (75a), is given by (76a). An alternative Johnson-Cook

15
model, which has been obtained from equations (73b) and (75b), is provided
by equation (76b). These models will be compared with each other in the
next section.

Damage rule. According to the Johnson and Cook (1985) ductile fracture
criterion, fracture occurs when the equivalent plastic strain P reaches a criti-
cal value Pf . The critical value Pf , which can be interpreted as an equivalent
fracture strain, depends on the equivalent strain rate Ṗ , the temperature T
and the triaxiality ratio τ :
!!  
Ṗ T − T0
Pf = (D1 + D2 exp (D3 τ )) 1 + D4 ln 1 + D5 (77)
Ṗ0 TM − T0

where D1 , D2 , D3 ,D4 and D5 are material parameters. The triaxiality ratio

τ is given by the ratio between the hydrostatic stress and the von Mises
equivalent stress:
1 Σ:1
τ= q (78)
3 3Σ : Σ
2 d d

In the general case, the fracture strain Pf is not constant during a defor-
mation process. It is however possible to determine whether the conditions
for ductile fracture are met or not by introducing a damage variable H. As-
suming a linear cumulative damage rule, the evolution of the damage variable
H is given by:
Ṗ
Ḣ = (79)
Pf
It should be noticed that, in contrast with D, the damage variable H has
no influence on material properties. To account for the stiffness degradation
caused by the development of ductile damage, the following damage rule is
used: (
0, H < 1 or Y < 0
Ḋ = H−1 z

(80)
G
(1 − D) , H ≥ 1 and Y ≥ 0
According to the above equation, the damage variable D increases if and only
if the conditions for ductile fracture are met (i.e. H ≥ 1). Also, as shown by
equation (57), the progression of damage contributes to the intrinsic dissipa-
tion source. For this contribution to be non-negative, damage is allowed to

16
grow if and only if the corresponding driving force Y is non-negative2 . For a
given material, the progression of the damage variable can be adjusted with
the G and z material parameters.

2.7. Heat diffusion equation

The proposed Johnson-Cook model consists of the state equations (37),
(39), (41), (43) and (47), the equivalent stress definition (61) and the evolu-
tion equations (59), (64), (65), (75) and (80). To understand the implications
of the thermomechanical couplings that are considered in the constitutive
model, it is instructive to derive the heat diffusion equation. According the
first law of thermodynamics, we have:
1
u̇ = ȧ + ṡT + sṪ = σ : L + r − ∇ · q (81)
ρ
From the definition (51) of the specific intrinsic dissipation source din , one
obtains:
ṡT = din + r − ∇ · q (82)
Using the state equation (47) for the specific entropy s, the left-hand term
of the above equation becomes:
∂s ˙ + T ∂s : Ã
˙ + T ∂s Ṗ + T ∂s Ḋ + T ∂s Ṫ
ṡT = T : Ẽ θ k (83)
∂ Ẽ θ ∂ Ãk ∂P ∂D ∂T

Combining equations (82) and (83) leads to the following form of the heat
diffusion equation:

cṪ = din + ϕθ + ϕic + r − ∇ · q (84)

According to the above equation, different heat sources control the temper-
ature evolution of a material point. The thermoelastic heat source ϕθ is:
∂s ˙
ϕθ = −T : Ẽ θ (85)
∂ Ẽ θ
   
1 ∂L ∂L ∂α ˙
= T : Ẽ θ − :α+L: (T − T0 ) − L : α : Ẽ θ (86)
ρ0 ∂T ∂T ∂T

2
For isotropic and cubic materials, Y is necessarily non-negative, in which case the
evolution law for the damage variable D solely depends on H.

17
 The specific heat source resulting from the couplings between internal
(i.e. hardening and damage) variables and temperature ϕic is given by:
∂s ˙ − T ∂s Ṗ − T ∂s Ḋ
ϕic = − T : Ã k (87)
∂ Ãk ∂P ∂D
1 ∂K ˙ 1 ∂B n
= T Ãk : : Ãk + T P Ṗ
ρ0 ∂T ρ0 ∂T
∂ 2L ∂ 2K
 
1
+ T Ẽ e : : Ẽ e + Ãk : : Ãk Ḋ (88)
2ρ0 ∂T ∂D ∂T ∂D
 
1 ∂L ∂α
− T Ẽ e : : (T − T0 ) + α Ḋ
ρ0 ∂D ∂T
Finally, the specific heat capacity c is given by:
∂s
c =T (89)
∂T  
1 ∂L ∂α ∂L ∂α
= 2T Ẽ θ : :α+L: + : (T − T0 )
ρ0 ∂T ∂T ∂T ∂T
 2
∂ 2λ
  
1 ∂ λu 2 2 ∂λu ∂λ
+ T − (T − T0 ) + T − (T − T0 ) (90)
2ρ0 ∂T 2 ∂T 2 ρ0 ∂T ∂T
1 1 ∂ 2 B P n+1
+ T (λu − λ) + c0 + γ (T − T0 ) − T
ρ0 ρ0 ∂T 2 n + 1
It is emphasized that, due to the impact of temperature on the stiffness
and thermal expansion properties and the isotropic hardening modulus, the
specific heat capacity c depends on both the thermoelastic strain tensor Ẽ θ
and the equivalent plastic strain P . Also, depending on the retained op-
tion, it is worth mentioning that the second derivative of B with respect to
temperature is:
∂ 2B
= −B0 m (m − 1) (T − T0 )m−2 (TM − T0 )−m (91a)
∂T 2
∂ 2B
2
= −B0 m (m − 1) T m−2 (TMm
− T0m )−1 (91b)
∂T
According to equation (91a), this derivative does not necessarily exist for T =
T0 when m < 2 and m 6= 1, in which case the specific heat capacity cannot
be evaluated. The classical formulation of the Johnson-Cook model should
therefore be used with caution for coupled thermomechanical problems.

18
 As shown by equation (84), the construction of the heat diffusion equation
does not require the introduction of the so-called Taylor-Quinney coefficient
(Taylor and Quinney, 1934). Indeed, this coefficient, which measures the
fraction of plastic work rate being dissipated into heat, is entirely determined
by constitutive relations. In constrast with some studies considering this
coefficient as a constant material parameter (Børvik et al., 2001; Su and
Stainier, 2015), the present approach is consistent with the experimental
observations showing that it depends on both the loading conditions and
deformation history (Rittel, 1999).

3. Illustrative examples
In this section, the proposed set of constitutive equations is used to model
the behaviour of a 4340 steel. The corresponding material parameters are
listed in Table 1. Both the mechanical and thermophysical properties are
assumed to be isotropic. The parameters of the Johnson-Cook model, for
both the viscoplastic flow rule and the damage rule, have been obtained by
Johnson and Cook (1985). The reference equivalent plastic strain rate Ṗ0
and temperature T0 are respectively fixed to 1 s−1 and 293 K. For simplicity,
the variation of the thermal expansion coefficient, thermal conductivity and
heat capacity with respect to temperature are not considered. The von Mises
yield criterion is used for all the following examples. Except for the specific
case of a Bauschinger test (see below), kinematic hardening is not considered.

3.1. Isothermal tension tests

To compare the original and alternative formulations of the Johnson-Cook
model, the specific case of uniaxial tension is first considered. The loading
conditions are such that both the temperature T and the axial logarithmic
strain rate ˙ are constant. The axial Cauchy stress σ is plotted as a func-
tion of the axial logarithmic strain  for different strain rates and different
temperatures in Figure 2. According to the results, only minor differences
between both formulations are observed, even for low strain rates (i.e. when
Ṗ ≈ Ṗ0 ). This indicates that the original and alternative formulations of
the Johnson-Cook model are very similar, but only the alternative formu-
lation allows extrapolating to either low strain rates (i.e. Ṗ < Ṗ0 ) or low
temperatures (i.e. T < T0 ).

19
 Parameters
Physical ρ0 c0 γ TM
7830 kg/m3 477 J/kg/K 0 K−1 1793 K
Heat conduction κ0 θ
37 W/m/K 0 K−1
Thermoelastic E0 ν α0 β ζ
200 GPa 0.29 11 MK−1 0.00047 K−1 0 K−1
Viscoplastic A0 B0 C n m
792 MPa 510 MPa 0.014 0.26 1.03
Damage D1 D2 D3 D4 D5
0.05 3.44 -2.12 0.002 0.61
G z
0.01 MPa 10
Table 1: Material parameters used for the application of the Johnson-Cook constitutive
model.

3.2. Adiabatic tension tests

To evaluate the relative importance of the different heat sources, the case
of an adiabatic uniaxial tension test is now examined. The initial tempera-
ture T is set to 293 K and the axial logarithmic strain rate ˙ is equal to 1
s−1 . The evolutions of the axial stress σ and the absolute temperature T as
a function of the axial logarithmic strain , which have been obtained from
the alternative Johnson-Cook model, are displayed in Figure 3. The contri-
butions of the different heat sources to the temperature evolution are plotted
in Figure 4. According to the results, except for the final stage of damage
growth, the contribution of the thermoelastic source ϕθ is relatively small.
It can thus be neglected for most practical applications. Nevertheless, prior
to total failure, the thermoelastic contribution varies significantly. Indeed,
during the final deformation stage, the applied stress is too low for plastic
deformation to occur. As a result, since further plastic deformation is almost
impossible, the thermoelastic strain rate increases when damage becomes
significant. As indicated by Equation (86), this causes a rapid variation of
the thermoelastic source. Also, though the intrinsic dissipation source din
provides the most important contribution, the temperature evolution is sig-
nificantly impacted by the internal coupling source ϕic . The importance of
the internal coupling source, which should therefore be considered for ther-
momechanical problems, has also been highlighted by Ranc and Chrysochoos

20
Figure 2: Evolution of the axial Cauchy stress σ as a function of the axial logarithmic
strain  during an isothermal uniaxial tension tests for different strain rates: 1 s−1 (top),
10 s−1 (middle) and 100 s−1 (bottom).

21
(2013).

Figure 3: Evolution of the axial Cauchy stress σ and absolute temperature T as a function
of the axial logarithmic strain  during an adiabatic uniaxial tension test for a strain rate
of 1 s−1 and an initial temperature of 293 K.

Figure 4: Evolution of the different heat sources as a function of the axial logarithmic
strain  during an adiabatic uniaxial tension test for a strain rate of 1 s−1 and an initial
temperature of 293 K.

3.3. Bauschinger tests

For the purpose of evaluating the impact of kinematic hardening on the
flow behaviour, the alternative Johnson-Cook model is used to evaluate the
response of a material point during a Bauschinger test. In the present case,

22
the Bauschinger test consists in first performing an isothermal shear test, up
to a shear strain γ of 10%, with a shear strain rate γ̇ of 1 s−1 . The loading
direction is then reversed (i.e. γ̇ = −1 s−1 ) until the shear strain γ vanishes.
To limit the number of material parameters, the initial kinematic harden-
ing moduli tensor K0 is assumed to be equal to some fraction ξ of the initial
elastic stiffness tensor L0 , that is:

K0 = ξ L0 (92)

The absence of kinematic hardening corresponds to the situation where ξ is

equal to zero. The evolution of the shear stress τ as a function of the shear
strain γ is presented in Figure 5 for different values of ξ. The results show
how the ξ parameter allows controlling the translation of yield surface in the
stress space.

Figure 5: Evolution of the shear stress τ as a function of the logarithmic shear strain γ
during an isothermal Bauschinger test for a shear strain rate of ±1 s−1 and a temperature
of 293 K.

3.4. Charpy impact test

For the next application, a notched specimen, whose geometry is pre-
sented in Figure 6, is subjected to an adiabatic impact test. The velocity of
the striker is set to 5 m·s−1 . For the resolution of the field equations resulting
from equilibrium and compatibility conditions, the ABAQUS explicit finite
element solver is used. Both the striker and the support are modelled as an-
alytical rigid surfaces while the specimen is meshed with C3D8R elements.
The friction coefficient between the specimen and both the striker and the

23
support is fixed to 0.3. Elements for which the damage variable D of the
corresponding integration points has reached a unit value are deleted.

Figure 6: Geometries of the notched specimen, striker and support used for the simulation
of a Charpy Impact test.

The evolution of the equivalent plastic strain field P is presented in Figure

7. As experimentally observed, the crack initiates from the notch and propa-
gates toward the upper surface of the specimen. Also, because the proposed
model has been developed within a consistent thermodynamic framework,
the partition of energy during a thermomechanical process can be investi-
gated. The partition of the total work into dissipated and internal energies
during impact is presented in Figure 8. According to the results, most of the
total work developed by external forces is dissipated into heat, essentially as
a result of intrinsic dissipation. Nevertheless, the increase of total internal
energy indicates that an important part of the total work is stored within
the specimen because of the microstructural transformations (e.g. hardening)
occuring during the deformation process.

24
3.5. Quenching
The last application aims at demonstrating the interest of the proposed
model for thermomechanical problems. For this purpose, the impact of a
thermal loading on the formation of residual stresses is investigated. Specif-
ically, the alternative Johnson-Cook model is used to determine the residual
stress field resulting from the quenching of a cylindrical specimen. The di-
mensions of the cylindrical specimen, together with the boundary conditions,
are given in Figure 9. For the application of the thermal load, the temper-
ature history is prescribed on the external surface. It consists of cooling
the specimen from 893 K to 293 K with a cooling rate of 100 K·s−1 . The
temperature is then held constant until thermal equilibrium is reached. As
before, the ABAQUS explicit finite element solver is used for the resolution
of field equations. Because of symmetry planes, only one eighth of the speci-
men is considered for the finite element model. The specimen is meshed with
C3D8T elements, for which the nodal variables are the three components of
the displacement vector and the temperature.
As shown in Figure 10, important temperature gradients due to the high
cooling rate produce an internal stress field. The internal stress field allows
the accumulation of plastic strains near the external surface. Though the
magnitude of plastic strains is rather small (about 3×10−3 ), the plastic strain
field is responsible for the formation of important residual stresses, which can
be observed once thermal equilibrium has been reached (at t = 600 s).

4. Conclusion
In this work, the Johnson-Cook model, which is widely used for the de-
scription of the mechanical behavior under severe loading conditions, has
been revisited from the point of view of thermodynamics with internal vari-
ables. Specifically, to construct a thermomechanical constitutive model, some
internal state variables have been introduced to consider both isotropic hard-
ening, kinematic hardening and ductile damage. These internal state vari-
ables have then been used to propose a definition of the specific free energy,
whose derivation has lead to the state equations governing the behaviour of
a material point. From the expression of the specific dissipation source, the
dissipative forces controlling the evolution of the internal variables have been
identified. Some evolution equations, which include the viscoplastic flow rule
of Johnson and Cook (1983), have been detailed. Such evolution equations,
when associated with state equations, form a complete thermomechanical

25
constitutive model. Finally, the constitutive relations have been used to ob-
tain the expressions of the different heat sources involved in the heat diffusion
equation.
When constructing the constitutive model, some alternative options have
been explored for the definition of the thermal softening function and the
viscoplastic flow rule. As shown by some illustrative examples, these alter-
native options provide a description of the mechanical behaviour that is quite
similar to the original Johnson-Cook model. Nevertheless, these alternative
options allow circumventing the difficulties associated with the definition of
the specific heat capacity and the validity of the flow rule at low strain rates
and low temperatures.

References

Agmell, M., Ahadi, A., Ståhl, J.E., 2014. Identification of plasticity constants
from orthogonal cutting and inverse analysis. Mechanics of Materials 77,
43–51. https://doi.org/10.1016/j.mechmat.2014.07.005.

Børvik, T., Hopperstad, O.S., Berstad, T., Langseth, M. 2001. A computa-

tional model of viscoplasticity and ductile damage for impact and pene-
tration. Eur. J. Mech. A/Solids 20, 685–712.

Camacho, G.T., Ortiz, M., 1997. Adaptive Lagrangian modelling of ballistic

penetration of metallic targets. Int. J. Comp. Meth. Appl. Mech. Engng.
142, 269–301.

Chen, G., Ren, C., Ke, Z., Li, J., Yang, X., 2016. Modeling of flow behav-
ior for 7050-T7451 aluminum alloy considering microstructural evolution
over a wide range of strain rates. Mechanics of Materials 95, 146–157.
https://doi.org/10.1016/j.mechmat.2016.01.006.

Couque, H., Boulanger, R., Bornet, F., 2006. A modified Johnson-Cook

model for strain rates ranging from 10−3 to 105 s−1 . J. Phys. IV France
134, 87–93.

Djavanroodi, F., Hussain, Z., Irfan, O.M., Al-Mufadi, F., 2019. Strain Behav-
ior of Nickel Alloy 200 during Multiaxial Forging through Finite Element
Modeling. Metals 9, 1–12. https://doi.org/10.3390/met9020132.

26
Gurtin, M.E., Fried, E., Anand, L., 2010. The Mechanics and Thermody-
namics of Continua. Cambridge University Press, New York.
Hor, A., Morel, F., Lebrun, J.L., Germain, G., 2013. Modelling, identifica-
tion and application of phenomenological constitutive laws over a large
strain rate and temperature range. Mechanics of Materials 64, 91–110.
https://doi.org/10.1016/j.mechmat.2013.05.002.
He, A., Xie, G., Zhang, H., Wang, K., 2013. A comparative study on John-
sonCook, modified JohnsonCook and Arrhenius-type constitutive models
to predict the high temperature flow stress in 20CrMo alloy steel. Materials
and Design 52, 677–685. https://doi.org/10.1016/j.matdes.2013.06.010.
Johnson, G., Cook, W., 1983. A constitutive model and data for metals
subjected to large strains, high strain rates, and high temperatures. In:
Proc. 7th Int. Symposium on Ballistics, page 541.
Johnson, G., Cook, W., 1985. Fracture characteristics of three metals sub-
jected to various strains, strain rates, temperatures, and pressures. Engi-
neering Fracture Mechanics 21, 31–48.
Lee, E.H., 1969. Elastic Plastic Deformations at Finite Strains. J. appl. Mech.
36.
Mandel, J., 1973. Thermodynamics and plasticity. In: Delgado Domingas,
J. J., Nina, M. N.R., Whitelaw, J.H., (Eds.), Proceedings of The Interna-
tional Symposium on Foundations of Continuum Thermodynamics, Hal-
sted Press, New York, 283–304.
Maugin, G.A., Muschik, W., 1994. Thermodynamics with Internal Variables.
Part I. General Concepts. J. Non-Equilib. Thermodyn. 19, 217–249.
Ranc, N., Chrysochoos, A., 2013. Calorimetric consequences of thermal
softening in JohnsonCooks model. Mechanics of Materials 65, 44–55.
https://doi.org/10.1016/j.mechmat.2013.05.007.
Rittel, D., 1999. On the conversion of plastic work to heat during high strain
rate deformation of glassy polymers. Mechanics of Materials 31, 131–139.
Shrot, A., Bäker, M., 2012. Determination of Johnson-Cook parameters
from machining simulations. Computational Materials Science 52, 298–
304. https://doi.org/10.1016/j.commatsci.2011.07.035.

27
Su, S., Stainier, L., 2015. Energy-based variational modeling of adiabatic
shear bands structure evolution. Mechanics of Materials 80, 219–233.
https://doi.org/10.1016/j.mechmat.2014.04.013.

Taylor, G.I., Quinney, H., 1934. The latent energy remaining in a metal after
cold working. Proc. R. Soc. A143, 307–326.

Umbrello, D., M’Saoubi, R., Outeiro, J.C., 2007. The influence of Johnson-
Cook material constants on finite element simulation of machining of AISI
316L steel. International Journal of Machine Tools and Manufacture 47,
462–470. https://doi.org/10.1016/j.ijmachtools.2006.06.006.

Wang, S., Huang Y., Xiao, Z., Liu, Y., Liu, H., 2017. A Modified Johnson-
Cook Model for Hot Deformation Behavior of 35CrMo Steel. Metals 7,
1–10. https://doi.org/10.3390/met7090337.

Yoo, Y.H., Yang, D.Y., 1997. Finite element modelling of the high-
velocity impact forging process by the explicit time integration
method. Journal of Materials Processing Technology 63, 718–723.
https://doi.org/10.1016/S0924-0136(96)02713-6.

Zhao, Y., Sun, J., Li, J., Yan, Y., Wang, P., 2017. A comparative
study on Johnson-Cook and modified Johnson-Cook constitutive ma-
terial model to predict the dynamic behavior laser additive manufac-
turing FeCr alloy. Journal of Alloys and Compounds 723, 179–187.
http://dx.doi.org/10.1016/j.jallcom.2017.06.251

28
Figure 7: Evolution of the equivalent plastic strain field P during a Charpy impact test:
t = 0.001 s (a), t = 0.002 s (b), t = 0.003 s (c) and t = 0.006 s (d).

29
Figure 8: Partition of the total work developed by external forces during a Charpy impact
test. The fraction of the total work that has been converted to kinetic energy is negligible,
hence not visible.

Figure 9: Geometry of the cylindrical specimen and boundary conditions for the simulation
of a quenching operation.

30
 Figure 10: Distribution of the temperature (top), accumulated plastic strain (middle) and
equivalent stress (bottom) along a radial path from the center to the external surface of
the cylindrical specimen. The time t = 6 s corresponds to the end of the cooling stage
while the end of the holding stage is at t = 600 s.

31

View publication stats