Escolar Documentos
Profissional Documentos
Cultura Documentos
4
PLASTICITY
From the early plasticity scientific work of Tresca in 1864, plastic theory has
developed greatly, especially with the use of high capacity computers.
Some aspects of modern plasticity theory were applied from the very beginning
of research on continuum mechanics. Take for instance, the well-known
Galileo’s experiment from his “Dialogues concerning two new sciences” (1638)
with a built-in cantilever beam (Figure 4.1.a) written almost 40 years before
Hooke´s linear constitutive equation.
1
assumed a limit load R at the centroid of the fixed section AB (Figure 4.1.b) and
taking moments about B, he obtained the collapse load as
R = 2(L/h) P
2
It is possible to see (ref. 3) that the combination of a hypothetical collapse
mechanism along with the establishment of its global equilibrium, leads to a
bound to the collapse load. This is considered one of the bound theorems for
elastoplastic bodies accepted today for limit design.
It is also interesting that Coulomb choose this approach in his Memoir to the
French Academy (1773) when trying to obtain the design formula for collapse
loads of retaining walls, arches or columns. (For a lucid and instructive
description refer to J. Heyman (ref.4)).
3
Figure 4.3 is a copy of Plate I in Coulomb´s Memoir. It includes, in addition to
of the description of a testing device (Coulomb´s fig.1) and equilibrium
considerations (fig. 2, 3 and 4), the famous idea of a plane of rupture (fig.5) in
a column and the surface of rupture for retaining walls (figs. 7 and 8) where a
force of cohesion (stress concept was not yet discovered) acted simultaneously
with other due to friction. Equilibrium was written as a function of a parameter
(for instance the angle of the assumed failure plane), and a rationally based
minimum or maximum condition allowed the estimation of the collapse load. He
obtained the formula that is still taught and applied in engineering practice
today. In fact this was the first application of bounding plasticity ideas to soil
mechanics as shown in Heyman´s essay (ref. 4).
In the XIX century, with the advent of the elasticity theory, the research
emphasis was mainly on the linear elastic behaviour, but some substantial
progress occurred thanks to theoreticians like Saint Venant and Poncelet, or
bright experimentalists as Tresca and Bauschinger.
4
Figure 4.4 Tresca’s apparatus used in his
flow experiments (1872)
At the same time Bauschinger (1875-1886) conducted and published many
results from his experiments on steel, iron, wood, etc. in which he discovered
the lowering of the yield stress in a second loading when the stress was of
opposite sign from that of the first loading, which is part of the experimental
justification of the today´s kinematic strain hardening concept.
In spite of those early attempts, when Hill (ref. 1) published his epoch-making
book in 1950, he wrote “significant advances did not follow (to the ones above
described) until after 1920. The theory of plasticity still remains a young science,
with few serious students”.
In those years of the first half of the twentieth century several points became
clear: the importance of the load trajectory in the stress space and the
impossibility of obtaining a direct stress-strain relationship. This lead to the
Saint Venant procedure of an incremental constitutive equation that later
received the name of “flow theory” as opposed to the “deformation theory”
proposed by Hencky in 1924 to simplify the resolution of practical problems.
The last one was applied intensively by URSS researchers (see Katchanov
ref.6). The development of “flow theory” applied to metals is mainly due to
5
German researches between the two World Wars. Some of them have been
already mentioned in Chapter 1.
a. Under very general conditions, the solids can reach a “limit” or “collapse”
load whose estimation is a very important objective when we need to
calculate the ultimate strength of the structure to quantify its safety.
b. In ductile materials, the collapse load is reached after important
displacements. This capacity to make excursions on the plastic range
without losing strength is one of the most appreciated virtues of ductile
structures, especially in seismic engineering where energy dissipation
capacity guaranteed by structural ductility is as important as stiffness or
strength.
c. It is possible to obtain bounds of the collapse load using equilibrium
conditions. This is achieved either postulating a failure mechanism, as
Coulomb’s did, or searching for a statically admissible stress field along with
a limiting flow condition.
6
4.1 Uniaxial behavior
Figure 4.5, shows that the behavior is linear elastic up to point A. In this range
O − A, the behavior of the material is independent from the load path.
The stress at point A (σy ) is called yield stress or initial yield point.
From point A and beyond, the material is yielding, i.e., there is permanent strain
when the stress goes back to zero. We observe two clearly defined regions: a
flat region A − B where strain increases at constant stress (perfect plasticity)
and a region B − U where an increase in stress produces an increase in strain
(plasticity with hardening).
7
If unloading occurs in any point C of the range A − U, the only part of the strain
that is recovered is DE. The other part (OD) remains even without the load. This
irreversible strain which is associated to energy dissipation is the plastic strain.
Therefore, the strain at an arbitrary point C in the plastic region is the sum of
the plastic strain OD (εp ) and the elastic strain DE (εe ):
ε = εe + εp (1)
This separation of the strain into an elastic or reversible part and a plastic or
irreversible part is justified by the physics of the elastic and plastic phenomena
and is only valid for infinitesimal deformations.
Yield without hardening corresponds to the case when the stress remains
constant during yield (fixed yield stress). If there is hardening, yield only occurs
when the stress increases and consequently the yield stress also increases
(variable yield point). This is shown in Figure 4.5: after unloading, CD, if the
material is loaded again yielding would not occur until point C.
Consider, for instance, that we reach point C (Figure 4.5) in a tensile test on a
thin-wall tube that, afterwards, is unloaded until point D. Now we reload again
but subjecting the tube to a torsional moment until point C is reached again with
a stress equivalent to the shear stress caused by the torsion moment (recall
Chapter 1). We shall have the same plastic strain, but the stress state will be
totally different!
Therefore, it is very important to know the loading trajectory and, if the yield limit
is reached, the only possibility we have is to establish a relation between the
infinitesimal increments of stress σ̇ and the infinitesimal increments of plastic
strain ε̇ p . This approach is called “flow theory” and was started by Barré de Saint
Venant in the XIX century.
8
4.1.3 Loading and unloading criteria
Since the constitutive relationships of a plastic material are different for the case
in which there is plastic loading, elastic loading or unloading, it is necessary to
identify if an increment of stress or strain causes plastic loading or elastic
unloading.
On the other hand, the end of the structure serviceability, the so-called
“collapse load”, will occur when all the three members reach their plastic load
Np .
9
Let us analyze closely this particular example. Consider that the three members
are made of the same material: Young modulus E, yield stress σy and cross
section A.
The collapse load Fp is obtained with the vertical equilibrium equation and
plugging in the values of the limit loads N3 and N1 previously obtained
Now let us see how we can follow the progression of plasticity from the elastic
situation.
N1 = N2
F = N3 + 2 N1 cos α (4)
Elastic solution
Constitutive equation
Nj Lj EA
∆lj = ; Nj = ∆l (6)
EA Lj j
10
1 3 2
L
L/ α α
co
sα
(a)
σ N Fp
Fe ?
σy Np
ε Δl v
Np = σy A
σy ? Fe
11
Elastic limit load
When the stress in the middle bar is the yield stress σ3 = σy , bar 2 starts
yielding and, from there on, the linear behavior of the truss is characterized by
the constant slope (Figure 4.6.g)
F EA
= (1 + 2 cos 3 α) (8)
v L
stops to be valid. Then, since
N3 = A σ y
(9)
[Fe = N3 (1 + 2 cos 3 α) = A σy (1 + 2 cos 3 α)]
Elastoplastic solution
From this moment on, the central member has lost its stiffness but the other two
maintain their loading capacity because only portion (cos α)σy has been used
and two symmetric members constitute a statically determinate (isostatic) truss
that can resist increments of load that will produce additional vertical
displacements. These displacements will be passively followed by the central
member that, after yielding, cannot share any more of the load.
12
Setting
∆N3 = 0 (13)
The equations now are
∆ F = 2(cos α) ∆N1
∆I3 = ∆v (14)
∆I1 = ∆I2 = ∆v cos α
N3 N1=N2
A σy
ΔN1 ΔN2
A σy/4
ve v ve v Δv; ΔF
EA EA
∆N1 = ∆N2 = ∆ l1 = (cos2 α) ∆v (15)
L/ cos α L
and
∆v
∆F = 2 ∆N1 cos α = 2(cos 3 α) EA (16)
L
The truss stiffness is
∆F EA EA
k ep = = 2 cos 3 α ( ) < (1 + 2 cos 3 α) = ke (17)
∆v L L
For instance, if α = π/3
1 3
k ep 2 ( ) 1/4 1
= 2 = = (18)
ke 1 3 1 + 1/4 5
1 + 2( )
2
13
and we can continue to plot the (F, v) diagram as a line, until member 1 and 2
start to yield and the structure reaches the so-called plastic limit collapse load
Fp .
F
Fp=Fe+ΔF
kep
Fe
A σy/4 ke
ve ve+Δve v
Figure 4.8 Force-displacement plot
14
Fe 1 5 ve
=1+ = σy = 1
Aσy 4 4 L( )
𝐸
∆F 3 ∆v
= σy = 3 (23)
Aσy 4 L( )
𝐸
Fp 8 v
= =2 σy = 4
Aσy 4 L( )
𝐸
F
pa
Ay ra
lle
l
2
5/4
v
1 4 σ
L y
E
Here we can see the ductility of the structure, where it is capable of having
large excursions in the plastic domain without losing strength.
We can plot also the elastic domain and the yield surface of the structure by
representing the stress-state in a multidimensional space where each axis
corresponds to every member. For instance, in the truss model problem, we
only need two axes because due to the symmetry the inclined members present
the same behavior.
15
σ2=N2/A
σy
σ1=N1/A=σ3
-σy σy
-σy
Figure 4.10
Something similar can be done with the strains (see Self Evaluation Exercise
1.).
σ
|σ| < σy ⇒ ε = εe =
E
σ (σ − σy ) (24)
|σ| > σy e p
⇒ ε=ε +ε = +
E Et
dσ
where the tangential modulus Et = is assumed to be constant.
dε
16
B
B
A
F
B A
O
O C
C B
D E
D
C
b) Ramberg-Osgood model
The hardening rule is used to represent the evolution of the yield stress σy
throughout the loading process. That evolution is established through a
hardening parameter α which is related to the plastic deformation.
17
4.2 Plasticity theory. General formulation
ε̇ = G(σ̇ ) (27)
t′ = k t
∂ ∂ (28)
=k
∂t ∂t′
Then:
ε̇ t = k ε̇ t′ σ̇ t = k σ̇ t′ (29)
Using the stress strain relationship
ε̇ t = G(σ̇ t )
(30)
k ε̇ t′ = G(k σ̇ t′ )
If G is homogeneous and first order function of the stress rate
ε̇ t′ = G(σ̇ t′ ) (32)
which shows there is no dependence on the time reference as mentioned
previously. The dot over the symbols that is reserved for time derivate in
viscoelasticity and dynamics, here it only has meaning of increment or rate. An
incremental relationship linear with respect to stress rate is then proposed:
ε̇ = C e σ̇ (33)
18
where C e is the material elastic (or compliance) tensor, or expressed in terms
of the stiffness matrix De
σ̇ = De ε̇ (34)
According to the constitutive equation, by extending the concepts from the one-
dimensional case, the plasticity theory is based on four fundamental aspects:
c) A flow rule that allows the establishment of the constitutive rules for
plastic strain increments.
These four aspects establish a general theory of plasticity and will be developed
in the following sections.
F(σij ) = 0 (35)
For points inside that surface, the material behaves elastically. For points
located on that surface, the material begins to yield.
The stress space is a space with six stress components which is hard to
visualize. If the body is isotropic, the yield does not depend on the orientation
of axis but only on the values of principal stresses:
19
F(σ1 , σ2 , σ3 ) = 0 (36)
or the stress invariants I1 , J2 , J3
F(I1 , J2 , J3 ) = 0 (37)
which makes representing it easier. (Review Appendix Chapter 1)
In this criterion, yield occurs when the elastic distortion energy reaches a critical
value (Recall Appendix Chapter 1). The yield surface is defined:
F(σ) = J2 − k 2 = 0 (38)
R 2k
σ3
σ2
σ1
20
Figure 4.13 Von Mises section with the deviatoric 𝛑 plane
(isometric projection)
The cross-section of the yield surface with a deviatoric plane is a circle of radius
R = √2 k (Figure 4.13).
If one of the principal stresses is zero (σ3 = 0), then the yield surface is an
ellipse in the σ1 − σ2 plane (Figure 4.14)
21
4.2.1.2 Tresca’s yield surface
In this criterion, yield is not related to energy but to the maximum shear stress.
The yield surface is defined in terms of principal stresses by the six planes:
1
|σ − σ2 | = k
2 1
1
|σ − σ3 | = k (40)
2 2
1
|σ − σ1 | = k
2 3
Or more compactly written
1
F(σ) = Sup(|σi − σj |) − k = 0 (41)
2 i≠j
which represents a prism that has the same axes as the von Mises yield surface
and it is contained within it. In fact, von Mises’ idea came when he tried to
smooth the corners of the Tresca yield surface.
In the deviatoric plane, the yield surface is a regular hexagon with six corners
(Figure 4.15).
σ3
2 3 1
1 2 k
2 1
1
2 3 k
2 2
1 1
1
3 1 k
2 3
σ1 3 2 σ2
22
If one of the stresses is zero (σ3 = 0), the yield surface becomes an irregular
hexagon (Figure 4.16). For example, in the quarter σ1 ≥ 0, σ2 ≥ 0 , then:
σ1 − σ2 = 2 k (42)
For a perfectly plastic material, the yield surface F(σij ) = 0 is a fixed surface
in the stress space. Plastic strain occurs if the current stress point is on the
surface. In a loading case, this point stays there after an increment of load dσij .
On the other hand, when the load is removed, the stress point would move
inside the surface.
23
For a hardening material, the yield surface changes when the stress point
reaches the yield surface2.
The existence of the yield surface F that borders the elastic domain E (Figure
4.19) allows the definition of a relationship between the plastic strain increment
and the stress increment compatible with F.
2 Recall that the components of gradient of surface F(σ) is defined by vector 𝜕F/ ∂σ that points in the
direction of the normal n to the surface which can be, then, defined as n = 𝜕F/ ∂σ
24
Consider point A (Figure 4.20) which represents the stress state on the
boundary ∂E of the yield surface F . At that point, the normal vector is n .
Consider three different stress increments σ̇ 1 , σ̇ 2 , σ̇ 3 . The first increment points
towards the inside of F and according to the previous section, this implies
elastic behavior:
ε̇ = C e σ̇ (43)
where C e is the material’s elastic tensor or elastic compliance tensor.
n 𝜕𝐹
𝜕σ1
n1
𝜕𝐹
n = [n2 ] =
n3 𝜕σ2
𝜕𝐹
[𝜕σ3 ]
σ1
σ = [σ2 ]
σ3
ε1
ε = [ε2 ]
ε3
Figure 4.20 Plastic load
On the other hand σ̇ 3 produces and increment of load and therefore
ε̇ = C ep σ̇ (44)
where C ep is the elastoplastic tensor for the incremental behavior.
This implies:
25
a) Continuity through the yield surface
C e σ̇ t = C ep σ̇ t (45)
b) Additive decomposition in the elastic and plastic components
ε̇ = C ep σ̇ n + Ce σ̇ t (47)
Adding and subtracting C e σ̇ n to the right-hand-side leads to:
Ε̇ = (C ep − C e ) σ̇ n + C e (σ̇ n + σ̇ t ) (48)
C e (σ̇ n + σ̇ t ) = C e σ̇ = ε̇ e (49)
Therefore, it follows the additive decomposition
ε̇ = ε̇ e + ε̇ p (50)
where:
26
ε̇ p = C p σ̇ n
(51)
C p = C ep − C e
c) Flow rule
ε̇ p = C p σ̇ n = C p n (nt σ̇ ) (52)
In general it is possible to write:
1
Cp n = m (53)
h
where h is a parameter that depends on the material and m is the orientation
vector of the strain increment.
Therefore:
1 t
ε̇ p = (n σ̇ ) m (54)
h
A particular case, called Associative Plasticity occurs when m = n. In this
case:
1 t
ε̇ p = (n σ̇ ) n (55)
h
Since:
∂F
∂σ
n=
(56)
∂F t ∂F
√( ) ·
∂σ ∂σ
∂F
Understanding that the vector expressed in components is
∂σ
∂F ∂F ∂F ∂F T
=[ ] (57)
∂σ ∂σ1 ∂σ2 ∂σ3
Then the plastic strain rate can be written as
27
∂F t
1 (∂σ) σ̇ ∂F
ε̇ p = t (58)
h ∂F ∂F ∂σ
( ) ·
[ ∂σ ∂σ]
The definition of h is found by multiplying ε̇ p by itself:
t 1 2
(ε̇ P ) ε̇ P = 2
(nt σ̇ )
h
(59)
P 2
1 2
|ε̇ | = 2 |σ̇ n |
h
And since
|ε̇ P | = |dεp |
(60)
|σ̇ n | = |dσn |
We obtain,
|dσn |
h= (61)
|dεp |
ε̇ e = ε̇ − ε̇ p
C e σ̇ = ε̇ − ε̇ p (62)
σ̇ = De (ε̇ − ε̇ p )
Using the definition of the plastic strain rate in terms of h leads to
1 t
σ̇ = De [ε̇ − (n σ̇ ) n] (63)
h
Multiplying both terms by n and rearranging leads to:
28
t
nt De n
(n σ̇ ) [1 + ] = nt De ε̇
h
(64)
1 t nt De ε̇
(n σ̇ ) =
h h + nt De n
Therefore the tress rate is
e
De nnt De
σ̇ = [D − ] ε̇ (65)
h + nt De n
The elastoplastic stiffness matrix is:
ep
De nnt De
e
D =D − (66)
h + nt De n
Sometimes, the plasticity problem is expressed with the strain vector dεPij
(corresponding to a certain stress σij ) being normal to an assumed Plastic
Potential Function G(σij ):
p ∂G
dεij = dλ (67)
∂σij
In the case of Associative Plasticity, the plastic potential and the yield surface
coincide (G = F), while if G ≠ F the flow rule would be called Non-Associative.
The plastic multiplier dλ is a positive scalar than is only non-zero when there is
plastic strain. If dλ is negative there is unloading. Since elastic unloading is the
only case possible, a negative dλ should be replaced by a zero value.
To obtain yielding, the stresses have to remain on the yield surface, i.e.: at any
instant the yield surface passes through the stress point and
Then, for a case of perfect plasticity with the associative rule the consistency
condition is:
∂F ∂F ∂F
dF = σ̇ = 0 ⇒ (Dijkl dεkl − dλ Dijkl )=0 (69)
∂σij ij ∂σij ∂σkl
29
And solving for the plastic consistency parameter:
∂F
D
∂σij ijkl
dλ = dεkl (70)
∂F ∂F
D
∂σpq pqrs ∂σrs
∂F ∂F 𝑒
D𝑒ijkl D
ep e
∂σkl ∂σij ijkl
Dijkl = [Dijkl − ] dεkl (72)
∂F 𝑒 ∂F
D
∂σpq pqrs ∂σrs
This is the same matrix seen previously in eq. (66) but with h = 0 (perfectly
plastic case).
4.2.4 Hardening
When a material is loading, there are changes in the limit values, either as an
increase (hardening materials) or a decrease (softening materials). In these
cases, as the plastic strain appears, the yield surface is transformed into a new
loading surface. Since the changes depend on the plastic deformation, the
constitutive laws depend on the loading path.
The changes that occur in the yield surface are the main difference between
plasticity with hardening and perfect plasticity. The changes are controlled
through two hardening parameters: αij and χ
30
F(σij , αij , χ) = 0 (73)
For elastoplastic materials, the loading criteria was seen in section 4.2.2 and
the flow rule does not change. Here we are going to describe three possible
hardening rules, depending on the nature of the parameter αij : isotropic
hardening, kinematic hardening and mixed hardening.
The isotropic hardening rule assumes that the yield surface expands uniformly.
The size of the subsequent loading surfaces is controlled by a hardening
parameter χ that affects k 2 :
σ3 σ3 2 k(1 )
σ1=σ2=σ3
2 k(2 )
2 k(2 )
2 k(1 )
σ2
σ1 σ2 σ1
31
which simplifies to:
1
J2 = σ2e (77)
3
To verify the condition in the yield surface:
σe = σy
| 1 (78)
J2 = k 2 (χ) = σ2y
3
Therefore, for the general case, σe can be interpreted as an effective stress.
σe = √3 J2 (79)
|s| = √2 J2
t (82)
|ε̇ p | = √(ε̇ p ) ε̇ p
So, substituting (79) on the left-hand-side of (80) and eq. (81) & (82) on the
right-hand-side
p t
√3 J2 ε̇ e = √2 J2 √(ε̇ p ) ε̇ p (83)
p 2 t
ε̇ e = √ (ε̇ p ) ε̇ p (84)
3
32
The effective plastic strain represents a measure of the amount of plastic strain
p p
accumulated during the process. In simple tension σe = σy and ε̇ e = ε̇ y so the
relationship between the effective stress and the effective plastic strain can be
obtained from the one-dimensional plastic stress-strain relationship.
Two currently used alternatives for χ are based on the total amount of effective
plastic strain eq. (84).
t
p
χ1 = ∫ ε̇ e dt (85)
0
Several rules have been proposed to describe this type of hardening. One of
them is Prager’s rule which postulates describe that the yield surface does not
change its size or its shape but translates in stress space in the direction of its
normal
33
p
dα = c dε (89)
Due to some inconsistencies in Prager’s rule for a general stress space, Ziegler
proposed that the translations occurs in a radial direction determined by the
vector σ − α:
dα = (σ − α)dμ (90)
σ1=σ2=σ3
Figure 4.24 represents both hardening rules for the two-dimensional von Mises
criterion.
34
P
P: Praguer
Z: Ziegler
The mixed hardening rule combines the behaviors of isotropic and kinematic
hardenings so the yield surface can both translate and expand uniformly without
a change in orientation.
In this case, there are two hardening parameters, a “kinematic vector” α that
represents the translation, and an isotropic parameter, that represents the
expansion:
35
4.2.4.4 Experimental determination of the hardening parameters
σe = √3 J2 (94)
√J 2 = k
σe σ2e (95)
2
σe = √3 k k= k =
√3 3
Or, using eq. (90) and generalizing the concept of equivalent unidimensional
stress for mixed von Mises Hardening
1
3 T 2
(96)
(s − α)e = [ (s − α) (s − α)]
2
and, the yield criteria can be expressed as
that can be identified in a (σ, ε) diagram of the tensile test (Figure 4.26), where
R(χ) is variable.
From those tests some researches (see ref. 17) infer that:
a) The kinematic stress α increases, in matrix form, with the plastic strain and
tends to saturate at a limit value α∞ .
36
εp (Figure 4.26) makes it possible to identify α∞ and calculate γ on a semi-log
plot or with least squares.
where R ∞ and b are parameters that depend on the material and temperature
and can be compared to α∞ and γ, respectively.
37
4.3 Plasticity: Analysis with the finite element method
In a classical step by step analysis by the finite element method, the equilibrium
equation is weakly satisfied, resulting in a system of equations:
Where:
and the sub index n + 1 represents the values in the step t n+1 . Since:
In a nonlinear analysis, such as plastic analysis, the internal forces vector Fint
is a non-linear function in u, so the calculation for un+1 can only be performed
iteratively with a numerical method such as Newton-Raphson.
The well-known relation that defines the Newton-Raphson method to obtain the
solution for the equation r(un+1 ) = 0 is:
r(ukn+1 )
uk+1
n+1 = ukn+1 − (105)
r ′ (ukn+1 )
38
where k represents each iteration, uk+1 k
n+1 is the value we are looking for, r(un+1 )
is the residual at the previous iteration, and r ′ (ukn+1 ) is the derivative of the
residual.
Figure 4.27 shows this method graphically for a general function f. (See Self
Evaluation Exercise 13).
The residual vector contains both the external force and the internal force
evaluated at each iteration. Matrix A is called tangent operator and changes for
each iteration. In a modified Newton-Raphson method, matrix A is fixed for a
certain number of iterations, reducing the solution cost, but also convergence.
In the following section, it will be discussed how to evaluate the residual vector,
b, and the tangent operator, A, in an elastoplastic problem.
39
4.3.3 Evaluation of the residual vector: Integration of the elastoplastic
constitutive equations
For each step, the external force vector is constant. The problem, however,
consists of calculating the stress at each iteration σkn+1 so that the plastic
consistency condition is satisfied. The way to do this is by using robust
algorithms for integrating the plastic constitutive rules.
40
MAIN PROGRAM
Data input
k 0, unk 1 un
Start……………….
MATERIAL SUBROUTINE
given n and kn 1 obtain…..
kn 1
b=r unk 1
A=F'INT unk 1
Assembly…….....……y……………....
YES Check Convergence
Print results
r k r 0 ???
NO
A u=b
Solve………….
Check Convergence
unk 11 unk 1 u
k k 1
41
4.3.3.1 Generalized trapezoidal rule
p
Starting from the known values of strain (εn , εn ) and stress (σn ), in a step t n , it
is necessary to obtain those values in a next step t n+1 .
p p ∂G ∂G
∆εp = εn+1 − εn = ∆λ [(1 − α) +α ] (109)
∂σn ∂σn+1
where the integration parameter,α varies from 0 to 1 and the plastic multiplier
is obtained by consistency, as shown in 4.2.3.
The stresses σn+1 are calculated and therefore the internal force vector.
Separating the stress, σn+1 , into an elastic part, σtest
n+1 , and a plastic part,
(D ∶ ∆εp ), it can be shown that the stress value at t n+1 is obtained in two
phases. The first phase, which corresponds to the elastic stress-strain
relationships, transforms the initial stress state σn into an elastic predictive
stress state, σtest
n+1 . In the second phase, this predictive state returns to the yield
surface, F(σn+1 ) = 0, thereby restoring the plastic condition. The plastic return
can also be divided into two steps: in the first step the stress is projected onto
the direction of the initial yield ∂G/ ∂σn; the stresses obtained are then projected
onto the yield surface in the direction of the final yield ∂G/ ∂σn+1 (Figure 4.29).
42
ntest
1
F(n ) 0
G
(1 )
n
n1
G
n
F(n1 ) 0
According to this interpretation, the variable parameter α would give the relative
weight assigned to the directions of the initial and final yield and the integration
would only be exact in the case of radial loading where the normal does not
vary from t n to t n+1 . If α = 0, the algorithm is termed explicit and called forward
Euler algorithm. For α > 0 the algorithm is implicit, and in particular for α = 1
and associative plasticity is called implicit backward Euler method. This choice
of α was initially proposed as a generalization of the radial return algorithm for
plasticity models other than von Mises.
̂ k B dΩ
A = Fint (ukn+1 ) = ∫ B t D (110)
Ω
where
∂σ(εkn+1 )
̂k =
D (111)
∂εkn+1
represents the consistent tangent modulus, which is strongly influenced by the
constitutive laws and the algorithms used for integration.
43
Let us calculate the expression of the tangent modulus consistent with the von
Mises plastic perfect model with the radial return algorithm.
According to the Self Evaluation Exercise 14, using the radial return, the
following expression is obtained:
test
R sn+1
σn+1 = +pI
test test
(112)
√sn+1 : sn+1
where
1
p = tr(σtest
n+1 )
test
sn+1 = σtest
n+1 − p I σtest
n+1 = σn + D: ∆ε (113)
3
In initial form and without writing the sub index n + 1:
R sijtest
σij = + p δij
test s test
(114)
√spq ij
σtest
ij = σn,ij + Dijkl (εkl − εn,kl ) (116)
where σn,ij and εn,kl represent the values of stress and stress for t n .
∂σtest
ij 2
= Dijkl = G(δik δjl + δil δjk ) + (K − G) δij δkl (117)
∂εkl 3
44
∂p 1 ∂σtest
ii
= = K δkl
∂εkl 3 ∂εkl
∂sijtest ∂σtest
ij ∂p 2
= − δij = G (δik δjl + δil δjk − δij δkl )
∂εkl ∂εkl ∂εkl 3
̂ ijkl leads to:
Substituting these derivatives in D
Where:
1
Iijkl = (δik δjl + δil δjk ) (119)
2
45
SELF EVALUATION EXERCISES
EXERCISE 1
The three-bar truss shown below is loaded by a vertical force P. All bars have
the same cross-sectional area, A. The bars are made of an elastic-perfectly
plastic material with elastic modulus E and yield stress σ0 . The load P is first
increased to the plastic limit Pp and then is unloaded to zero. Afterward, P is
increased in the reversed direction until all three bars yield again in
compression and then is unloaded again to zero.
a) Determine the elastic limit load Pe+ and the plastic limit load Pp+ in the
initial loading.
b) Determine the elastic limit load Pe− and the plastic limit load Pp− in the
reversed loading.
c) Determine the residual stresses and strains of the bars and the residual
horizontal and vertical displacements at load point C at the end of the
loading history.
Figure 4.30
Solution:
46
P
σ1 = σ2 σ1 + σ3 =
A
ε1 = ε2 4 · ε1 = ε3
Using the elastic constitutive laws, we obtain the elastic solution of the problem.
In this case, the results are in the plane (σ1 , σ2 ). The elastic domain has a
square shape bounded by the lines σ1 = σ0 , σ1 = −σ0 , σ2 = σ0 , σ2 = −σ0
(Figure 4.31).
2
0
0
1
0
0
a) Determine the elastic limit load and the plastic limit load
Since the stress in bar 3 is the largest, this bar yields first. The elastic limit load
is obtained by equating σ3 to σ0 .
Pe+ 5
= σ0
A 4
at P = Pe+ the stress, strain and displacements are
1 1
σ1 = σ2 = σ0 σ3 = σ0 ε1 = ε2 = ε0 ε3 = ε0
4 4
u=0 v = L ε0
47
where u and v are the horizontal and vertical displacements respectively at the
point of application of the load.
For P > Pe+ , ∆σ3 = 0 since the material behaves as a perfectly plastic material.
From the equilibrium equations:
P
σ1 + σ0 = σ1 = σ2
A
The plastic limit load is reached when:
σ1 = σ2 = σ0
so
Pp+
= 2 σ0
A
In Pp+ the stress, strain and displacements are:
σ1 = σ2 = σ3 = σ0 ε1 = ε2 = ε0 ε3 = 4 ε0
u=0 v = 4 L ε0
Unloading
When the load P decreases from Pp+ to zero, the material is unloaded
elastically. The incremental stresses are related to the load increments ∆P
through the elastic solution:
1 ∆P 4 ∆P
∆σ1 = ∆σ2 = ∆σ3 =
5 A 5 A
This leads to:
σ1 = σ2 = σ0 + ∆σ1 σ3 = σ0 + ∆σ3
48
∆P = −Pp+ = − 2 σ0 A the residual stresses are obtained as:
3 3
σ1 = σ2 = σ0 σ3 = − σ0
5 5
The residual strains are:
∆σ1 2 3
ε1 = ε2 = ε0 + = ε0 − ε0 = ε0
E 5 5
∆σ3 8 12
ε3 = 4 ε0 + = 4 ε0 − ε0 = ε
E 5 5 0
12
u=0 v= L ε0
5
b) Determine the compression elastic load and plastic limit loads
Pp−
= −2σ0
A
And
σ1 = σ2 = σ3 = −σ0 ε1 = ε2 = −ε0 ε3 = −4 ε0
u=0 v = −4 L ε0
49
c) Residual values
When unloading occurs ∆P = − Pp− , the residual stresses and strains are:
3 3 3 12
σ1 = σ2 = − σ0 σ3 = σ0 ε1 = ε2 = − ε0 ε3 = − ε
5 5 5 5 0
12
u=0 v=− L ε0
5
Figure 4.32 represents a plot in the load-vertical displacement P − v plane with
the load path including loading, unloading and residual values.
50
EXERCISE 2
Plot the load path in the stress-strain plane from the load in Example 1.
Solution:
Using the elastic domain in stresses (Figure 4.33), during the elastic path the
load follows the straight line OP. At P, σ3 reaches the yield stress σ0 and due to
the perfect plasticity the stress in bar 3 is constant.
During the elastoplastic phase, the load follows path PQ until Q, where yielding
of all the structure is reached.
Unloading occurs along path TU and point U represents the residual final
stresses (σ1 = −3/5 σ0 ; σ3 = 3/5 σ0 )..
51
Figure 4.34 Loading path in the strain space
During the elastic phase the path is along line OA. Past this point, the strains
are plastic and continue along the same line until point B:
ε1 = ε0 ε3 = 4 ε0
Upon unloading, the path reaches point C which corresponds to the residual
strains (ε1 = 3/5 ε0 ; ε3 = 12/5 ε0 ).
The following final reload and unload steps follow path AD and DE respectively.
It can be observed that, as expected, the strain path follows along a straight line
corresponding to the compatibility equation of the structure 4 ε1 = ε3 .
52
EXERCISE 3
Solve the structure from Exercise 1 using the finite element method.
Solution:
The stiffness matrix in the global coordinate system of a truss structure is:
c2 sc −c 2 −sc
EA s2 −cs −s 2 )
K= (
L c2 sc
sim. s2
where c = cos α and s = sin α.
3 √3 3 √3
− −
4 4 4 4
1 √3 1 0 0 0 0
EA − EA 1 0 −1
K1 = 4 4 4 K2 = ( )
2L 3 √3 L 0 0
− sim. 1
4 4
1
(sim. 4 )
3 √3 3 √3
− −
4 4 4 4
1 √3 1
EA − −
K3 = 4 4 4
2L 3 √3
4 4
1
(sim.
4 )
Assembling the overall system of equations and applying the boundary
conditions, the equilibrium equation at node C (load node) is obtained.
53
3
EA 4 0 uC
0
( )= ( ) (v )
−P L 1 C
0 1+
4
Assuming unit values for E and A, the solution is:
uC = 0
4
vC = − PL
5
Starting from these global values, the stresses and strains in each bar can be
calculated in their local coordinates as
1 uj − ui
εij = (cos α sin α) ( v − v )
L j i
and then:
2
ε1 = ε2 = P = σ1 = σ2
5
4
ε3 = P = σ3
5
In the numerical solution process, the load is applied incrementally and the
elastic stresses are calculated until the stress in any bar reaches over the yield
stress σ0 . From that point, iterations are made until equilibrium is achieved.
σ1 = 0.4 σ0 σ3 = 0.8 σ0
In this case, the yield stress has not been reached. Increasing the load by ∆P =
0.5 σ0, the total applied load is P = 1.5 σ0 and
σ1 = 0.6 σ0 σ3 = 1.2 σ0
The stress in bar 3 now exceeds the yield stress. Since the material behavior is
perfectly plastic, if the force σ3 = σ0, equilibrium is not achieved:
3
0.6 σ0 + σ0 ≠ σ0
2
54
To achieve equilibrium, subtract the excess stress value in bar 3, 0.2 σ0 . This
leads to P = 1.3 σ0 which produces the following stresses:
σ1 = 0.52 σ0 σ3 = 1.04 σ0
Repeating the same iterative process with the excess values (0.04 σ0 ), the
solution is reached when the difference between the calculated stress σ3 and
σ0 is less than a specified residual value.
EXERCISE 4
σ1 = −σ0 ; σ2 = −σ0 ; σ3 = 0
Using the von Mises and Tresca criteria, find the stresses in tension and pure
shear as well as the yield load Py , for the following load path σ1 = P, σ2 =
2P, σ3 = 3P.
Solution:
F(σ) = J2 − k 2 = 0
55
1
J2 = σ20 = k 2
3
σ0
k=
√3
In a simple tensile test there is only one component of stress, σ1
σ1 = σ1 σ2 = 0 σ3 = 0
σ1 = σ0
σ1 = τ1 σ2 = 0 σ3 = −τ1
1 2
σ20 σ0
(1 + 1)σ1 = ⇒ σ1 = τ1 =
2 3 √3
For the given load path, plugging in the yield criterion leads to a yield load Py
1 σ20 σ0
(1 + 1)Py2 = ⇒ Py =
2 3 √3
Tresca criterion
1
max ( |σi − σj |) = k
i≠j 2
- k calculation
1
σ =k
2 0
-Tensile strength
56
σ1 = σ1 , σ2 = 0, σ3 = 0
1 1
σ1 = σ0 ⇒ σ1 = σ0
2 2
- Shear strength
σ1 = τ1 , σ2 = 0, σ3 = −τ1
1 1 1
2τ1 = σ0 ⇒ τ1 = σ0
2 2 2
-Yield load Py
1 σ0 σ0
(3 Py − Py ) = → Py =
2 2 2
EXERCISE 5
Determine h for a von Mises yield surface F and the stress-strain (σ − ε) rule
from Figure 4.35
dσn
H=
dεp
(a) (b)
Figure 4.35
Solution:
57
In the uniaxial tensile test
σ̇ 2
σ̇
3 3
σ̇ σ̇ σ̇
σ̇ = ( 0 )= + − = σ̇ oct + ṡ
3 3
0 σ̇ σ̇
( ( − )
3) 3
The projection of σ̇ on n gives the deviatoric stress tensor ṡ since σ̇ oct is
perpendicular to n. (See Figure 1.18 Appendix in Chapter 1).
4 2 1 2 2
|σ̇ n | = nt σ̇ = √ |σ̇ | + 2 |σ̇ | = |σ̇ |√
9 9 3
|dσn | 2 dσ 2
h= =√ = √ H
|dεp | 3 dεp 3
58
EXERCISE 6
Solution:
As shown in section 4.2.1.1 the von Mises yield surface can be written as:
F(σ) = J2 − k 2 = 0
i.e., the direction of the plastic strain increment is parallel to the deviatoric stress
tensor (Figure 4.36).
59
EXERCISE 7
Solution:
F(σ) = max(|σi − σj |) − 2 k = 0
i≠j
Similar to the von Mises criterion, the plastic strain increment is purely deviatoric
since the walls of the yield surface are parallel to the hydrostatic axis (σ1 = σ2 =
σ3 ).
In the deviatoric plane, the directions of dεp have to be determined for the six
planes and six corners of the Tresca regular hexagon (Figure 4.37).
F(σ) = σ1 − σ3 − 2 k = 0
p p p
dεp = (dε1 dε2 dε3 ) = dλ (1 0 −1)
For the corners, the direction of the plastic flow is undefined and lies between
the cone formed between the normal vectors of neighboring faces.
60
EXERCISE 8
Obtain the expression for the plastic multiplier dλ for a perfectly plastic material
using the associative flow rule for the von Mises criterion. Assume that in the
elastic domain the behavior is linear and isotropic.
Solution:
For a linear elastic and isotropic material the elastic tensor Dijkl is
2
Dijkl = (K − G) δij δkl + G(δik δjl + δil δjk )
3
Therefore:
∂F 2 ∂F ∂F
Dijkl = (K − G) ( δij ) δkl + 2 G
∂σij 3 ∂σij ∂σkl
∂F ∂F 2 ∂F ∂F ∂F ∂F
Dijkl = (K − G) ( δij ) ( δkl ) + 2 G =
∂σij ∂σkl 3 ∂σij ∂σkl ∂σkl ∂σkl
2
2 ∂F ∂F ∂F
= (K − G) ( δij ) + 2 G
3 ∂σij ∂σij ∂σij
∂F 2 ∂F ∂F
Dijkl dεkl = (K − G) ( δij ) dεkk + 2 G dε =
∂σij 3 ∂σij ∂σkl kl
61
2 ∂F ∂F
= (K − G) ( δij ) dεkk + 2 G dε
3 ∂σij ∂σij ij
Therefore
∂F
δ = sij δij = sii = 0
∂σij ij
EXERCISE 93
62
Figure 4.38
Solution:
63
Limit pressure 𝐩𝐲 , and plastic strain increment analysis
p1 D 2 p1 D 2 2
p1 D 2
( ) + 4( ) (1 − r) − 2 ( ) (1 − r) = 3 k 2 = σ2e
4t 4t 4t
and plugging in the value of k and p1 = py leads to:
4t 3
p1 = py = k√ 2
D 4r −6r+3
4t
If r = 0 ⇒ p0 = k
D
and then
3
p1 = py = √ 2 p
4r −6r+3 0
1 2
− σa + σc 1 3
3 3 2σ − σ 2(1 − r) − − 2r
2 1 1 c a 1 p1 D 2 1 p1 D 2
s= σ − σ = [ 2σa − σc ] = 1−1+r = r
3 a 3 c 3 −σ − σ 3 2t 3 6 t 3
a c
−
1
σ −
1
σ [ −
2
+ r ] [ 2 + r]
−
[ 3 a 3 c]
3 − 4r
p ̇
ε̇ = λ [ 2r ]
−3 + 2r
- Tresca criterion
64
ε̇ p = [ε̇ pc p
ε̇ a
p
ε̇ r ]
1
(1) 0≤r≤
2
σ1 = σc σ2 = σa σ3 = 0
σc − σr = σy
p
dεij = (dεpc dεa
p p
dεr ) = (1 0 −1) dλ
1
(2) ≤r≤1
2
σ1 = σa σ2 = σc σ3 = 0
Tresca’s criterion
1 D D
k = σa ⇒ p1 = p0
2 8t 4t
Limit pressure
p1 = py = 2 p0
σa − σr = σy
p
dεij = (dεpc dεa
p p
dεr ) = (0 1 −1) dλ
65
(3) r≥1
σ1 = σa σ2 = 0 σ3 = σc < 0
Tresca’s criterion
1 D D D
k = (σa − σc ) ⇒ p1 − (1 − r) p1 = p0
2 8t 4t 4t
Limit pressure
p0
p1 = py =
1
r−
2
Plastic strain increment
σa − σc = σy
p
dεij = (dεpc dεa
p p
dεr ) = (−1 1 0) dλ
66
EXERCISE 10
Obtain the expression for the plastic strain increment for the von Mises loading
surface considering associative plasticity and an isotropic hardening rule with
the slope of the effective stress-effective plastic strain curve H
dσe
H= p
dεe
Solution:
F(σij , χ) = J2 − k 2 (χ) = 0
Using the effective stress concept, σe , from section 4.2.4.1, the criterion can be
written as:
1
F(σij , χ) = σ2e − k 2 (χ) = 0
3
which leads to:
σe
k(χ) =
√3
Therefore, the von Mises criterion in terms of the effective stress is:
1 1
F(σij , χ) = sij sij − σ2e (χ)
2 3
Applying the consistency condition leads to:
∂F 2
dF = 0 ⇒ dσij − σe dσe = 0
∂σij 3
67
p 2 p
sij Dijkl (dεkl − dεkl ) = σe H ε̇ e
3
But:
p ∂F
dεkl = dλ = dλ skl
∂σij
2
sij Dijkl dεkl = sij [(K − G) δij δkl + G(δik δjl + δil δjk )] dεkl = 2 G sij dεij
3
p 2
sij Dijkl dεkl = sij Dijkl skl dλ = 2 G sij sij = 2 G σ2e
3
p 2 ∂F ∂F 2 2
ε̇ e = dλ√ = dλ√ sij sij = dλ σe
3 ∂σij ∂σij 3 3
p ∂F 2G
dεij = dλ = sij skl dekl
∂σij 4
(3G + H) σ2e
9
This expression could have been obtained as a function of the stress increment
as:
∂F p 4 sij dσij
dσij = dλ H σ2e ⇒ dλ
∂σij 9 4
H σ2e
9
Then:
p ∂F 1 1 ∂J2 1
dεij = dλ = sij skl dσkl = sij dσkl = sij dJ2
∂σij 4 2 4 2 ∂σkl 4 2
H σe H σe H σe
9 9 9
68
EXERCISE 114
Start with the same vessel as Exercise 9, and load it with the same pressure.
The ends of the tube are closed. Assume a von Mises loading function with
isotropic hardening rule and the relationship εep = a σ3e , where a is a constant.
p p
Determine the plastic strains (εa , εc ) at the end of the following two loading
paths:
Solution:
69
p 1
dεij = sij dJ2
4
H σ2e
9
and for this problem:
dσe 1 4 4
H= p = ⇒ H σ2e =
dεe 3 a σ2e 9 27a
Then:
p 9a
dεij = [(2σa − σc ) dσa + (2σc − σa ) dσc ]sij
4
Then:
p 9 27a
dεij = 6 a σa dσa sij = σ dσ s
4 2 a a ij
which in component form is:
p 27a 2 27a D 3 2
dεa = σ dσ = ( ) p1 dp1
2 a a 2 4t
p 27a 2 27a D 3 2
dεc =− σ dσ = − ( ) p1 dp1
2 a a 2 4t
Integrate the above expressions to obtain the plastic strains at the end of
the load path as:
p 27a D 3 p1 2 9a D p1 3
εa = ( ) ∫ p1 dp1 = ( )
2 4t 0 2 4t
70
p 27a D 3 p1 2 9a D p1 3
εc =− ( ) ∫ p1 dp1 = − ( )
2 4t 0 2 4t
p 9a 9a
dεij = 2 σc dσc sij = σc dσc sij
4 2
In component form, this is:
p 9a 1 3a D 3 2
dεa = σc dσc (− σc ) = ( ) p2 dp2
2 3 2 2t
p 9a 2 D 3 2
dεc = σc dσc ( σc ) = −3a ( ) p2 dp2
2 3 2t
At the end of path p1 = 0, p2 = R p1 , the equations become:
p 3a D 3 Rp1 2 a 3D p1 3
εa = ( ) ∫ p2 dp2 = ( )
2 2t 0 2 4t
p D 3 Rp1 2 3D p1 3
εc = −3a ( ) ∫ p2 dp2 = −a ( )
2t 0 4t
and
1 2 3 D 2 2
J2 = σc = ( ) p1
3 4 2t
In the second part, p1 increases from 0 to p1 and p2 is constant and equal to
R p1 . The stresses are:
71
D D D D
σa = p σc = p1 − R p1 = 2 σa − R p1
4t 1 2t 2t 2t
When the value of σc decreases, it is necessary to check the loading criteria. If
J2 is equal at the beginning of the step than the current J2 , we can plug in the
values of stresses σa and σc from above into the expression for J2 from the
beginning of the example:
2
1 2 D D 3 D 2 2
[σ + (2σa − Rp1 ) − σa (2σa − Rp1 )] = ( ) p1
3 a 2t 2t 4 2t
Or:
D
σ2a − σa R p1 = 0
2t
Solving for σa , leads to:
D 3D D
σa = 0 or σa = R p1 = p1 > p1
2t 4t 4t
This result implies that the tube is either in an unloading state or in an elastic
loading state and there is no new plastic strain. Therefore, at the end of this
path, the strains are:
p 27a D p1 3 p D p1 3
εa = ( ) εc = −27a ( )
2 4t 4t
72
EXERCISE 12
Solution:
1. k = 0 ukn+1 = u0
2. Calculate r(ukn+1 )
4. Calculate r′(ukn+1 )
k
r(ukn+1 )
5. Obtain ∆u =
r′(ukn+1 )
6. Evaluate uk+1 k
n+1 = un+1 − ∆u
k
7. Next k k = k+1
8. Go to 2
73
EXERCISE 13
Solution:
74
EXERCISE 14
Develop the algorithm to calculate the stresses at time step t n+1 , by applying
the minimum distance point integration scheme for a von Mises yield surface
without hardening with an associative flow rule.
Solution:
Since von Mises criterion is purely deviatoric, the direction of the plastic return
according to an associative flow rule in σn+1 , is normal to the hydrostatic axis
and therefore radial in the deviatoric plane. (Figure 4.39)
radial return
75
This algorithm, known as the radial return algorithm, follows these steps:
1. Calculate σtest
n+1 = σn + D: ∆ε
test test
4. Calculate A = sn+1 : sn+1
R
If A > R2 → Plastic process σn+1 = s test +pI → END
√A n+1
76
SUMMARY OF MAIN IDEAS
77
7. All possible stress combinations that maintain the elastic response of the
body define an “Elastic domain” E in the stress space where the material
behaviour follows the Linear Elasticity equations.
8. The elastic domain is bounded by a “Yield Surface” 𝐅(σ) that separates
the elastic from the elastoplastic stress states. Since the XIX century, the
definition of that surface has been one of the main research objectives in
plasticity theory. Von Mises criterion is a well-established alternative for
metals but in concrete or soil modelling, where dilatancy is coupled with
shear strains, more appropriated Yield Locus are being used.
9. When the point representing the stress-state has a trajectory that reaches
or crosses the yield surface, plastic strains appear. Under very reasonable
hypothesis of continuity of response and linear relation between stress and
strain rates, it is possible to establish a “Flow Rule” that depends on the
existence of a new surface 𝐆(𝛔), called “Plastic Potential”, whose normal
vector defines the direction of the plastic strain increment.
12. The key to numerical models is the development of criteria to follow the
nonlinear path and the definition of a stiffness matrix that depends on all
the concepts defined above.
78
REFERENCES
10. W.F. Chen & A.F. Saleeb “Constitutive equations for engineering
Materials vol1.” Wiley 1982
13. W.F. Chen, X.L. Liu “Limit analysis in soil mechanics” Elsevier 1990
14. D. Muir Wood “Soil behaviour and critical state soil mechanics”
Cambridge UP 1990
15. W.F. Chen & .Zhong “Structural Plasticity Theory, problems and CAE
software” Springer 1991
79
18. A. Muttoni, J. Schwartz, B. Thürlimann “Design of Concrete Structures
with Stress Fields” Birkhäuser 1997
20. P. Nielsen “Limit Analysis and concrete plasticity” 2nd ed. CRC press
1999
80