Professional Documents
Culture Documents
Yue Qi
Philip Eisenlohr
1
contact
• eisenlohr@egr.msu.edu
• +1 517 432-4506
• room 3520
2
introduction
3
introduction
deformation in manufacturing
4
introduction
deformation in manufacturing
http://www.autoform.com/en/glossary/springback/
deep drawing
5
introduction
deformation in manufacturing
spinning
6
introduction
deformation in manufacturing
roll forming
7
introduction
deformation in service
8
introduction
deformation in service
9
introduction
• ductility
• toughness
• fatigue resistance
10
introduction
11
introduction
033905-5 Diffraction contrast tomography Rev. Sci. Instrum. 80, 033905 !2009"
FIG. 7. Image processing steps for the completion of the grain boundaries
not visible in PCT. "a! Shows the original PCT image after median filtering
and contrast enhancement. One grain boundary is not visible and shall be
completed. First, the grain boundaries are labeled semiautomatically as far
as visible "b!. The overlay of the slimmed labeled grain boundaries with the
DCT image "c! shows that some DCT grains extend over the grain bound-
FIG. 5. "Color online! Rendition of the 3D grain structure in a cylindrical
10.1063/1.3100200 aries. Those parts have to be removed "d! before the DCT image can be used
beta-Ti specimen containing 1008 grains, as obtained by the DCT process- as marker image for the watershed algorithm. "e! The missing grain bound-
ing route described in this paper. ary was completed. Note that the originally labeled grain boundaries remain
unaffected by the algorithm. The accuracy of DCT was determined from the
completed network of grain boundaries and the DCT data "f!. "The average
error for the grain shown in this figure is 4.1 voxels or 2.3 "m!.
monochromator with a relative bandwidth of 10−3!. The
grain map was reconstructed using the data processing route 12
described in this paper. A 3D rendered view of the resulting B. Image processing and segmentation
introduction
• ferritic–martensitic steel
• Al-4%Cu
• Mg-5%Zn
• Ni-base alloy
13
introduction
(a)
the challenges: deformation mechanisms
(a)
(b)
Fi
si
(b
do
to
d
an
10.1016/j.msea.2010.02.041
o
st
(b)
re
st
p
(c) re
re
T
F i g. 1.6. (a) A TEM picture of dislocation structure in pureFig.single crystal BCC
5. EBSD map of steel LG (average grain size 50 !m) tensile deformed to 0.3 al
molybdenum deformed at temperature 278 K (courtesylogarithmicof L.strain. L. Diffraction
Hsiung).
axis; IPF: inverse pole figure).
pattern quality map (a), TA-IPF map (b) (TA: tensile
14 re
M
(b) Dislocations formed bundles (braids) in single crystal copper deformed at
introduction
hierarchical modeling
15
introduction
hierarchical modeling
gime extrac
10.1115/1.3183783
Preußner e
which allow
alloys by m
tems, with a
Their mode
related to
Ohashi et a
model the
ties of meta
hardening l
according t
tion loop in
!97". Using
able to repr
function of
Fig. 1 Hierarchical multiscale modeling of the plasticity in- crystal, Gro
volving three characteristic length scales model
16 of K
introduction
the plan…
reading
Tuesday Thursday Friday Lab homework
assignments
Elasto-plastic
decomposition, lattice orientation
Polycrystal plasticity, single crystals and bi-
Apr 14 configurations,
grain interactions crystals under tension
evolution in bicrystal
numerical solution deformation
strategies
17
introduction
the plan…
reading
Tuesday Thursday Friday Lab homework
assignments
Dependence of
Single-point, mean- Grain aggregate
Homogenization strength on
Apr 21 schemes
field, interaction, full- deformation with
composition in two-
field orientation shuffling
phase polycrystal
Performance
Spectral methods for Spectral methods for
comparison:
Bonus strong property strong property
spectral and finite
contrast contrast
element method
Finals to be decided…
18
THE CONCEPT OF CRYSTAL DISLOCAT
discrete dislocation
(a) dynamics (b) (c)
b
b
• Burgers circuit
dislocation
line sense
21
discrete dislocation dynamics
!1 P
C b1 q
23
discrete dislocation dynamics
dislocation character
• edge type
• mixed type
• screw type
dislocation
line sense
Burgers
vector
direction
24
LIVINGSTOK: THE DEiYSlTY AND DISTRIBUTION OF DISLOCATIONS 231
discrete dislocation dynamics
dislocations with densities between 7 x lo3 and
5 x 104/cm2, so that the total initial dislocation
densities ranged from lo4 to about 105/cm2.
The dislocation densities observed in the various
deformed crystals are plotted against shear strain and
basic idea of DDD shear stress in Pigs. 3 and 4. To calculate shear strain
and shear stress, lattice rotation w-as assumed only
for crystals of orientation 1. Included are data for
deformed polycrystalline copper obtained by Baileyol)
compute and follow the dynamics
by transmission electron microscopy. His tensile
‘O”/
• relationship between
/
dislocation structure and
strength(BAILEY1
10.1016/0001-6160(62)90120-7
10.1007/s11661-009-9863-8
RESOLVED SHEAR STRESS-KG / MM2
• spatial discretization
• driving forces
• mobility
• topological changes
26
discrete dislocation dynamics
spatial discretization
• eigenstrain field
• front tracking
dislocation
27
discrete dislocation dynamics
• assumption of straight
dislocation lines —>
points on plane
• positive / negative
Burgers vector or line
sense
• multiple slip
b1 b2
b3
28
discrete dislocation dynamics
advantages drawbacks
• no curvature effects
• dislocation sources?
29
discrete dislocation dynamics
30
discrete dislocation dynamics
198 L I N E D I S L O C AT I O N DY NA M I C S
• degrees of freedom
• positions of nodes
• Burgers vectors (connectivity) at
{ri , bij }
node
node
segment
32
discrete dislocation dynamics
advantages drawbacks
33
discrete dislocation dynamics
exemplary simulation of
dislocation penetration
through low angle grain
boundary
(i.e. dislocation mesh)
34
discrete dislocation dynamics
driving force
35
discrete dislocation dynamics
driving force
36
∂ 2 R/∂xi ∂xj . b and b ′ are the Burgers vectors at locations x and x′ , respe
discrete dislocation dynamics
Each line integral C is taken over all segments in the dislocation network
!
once. The direction of each line segment is arbitrary but the Burgers vector
be defined with respect to the chosen line direction.
driving force It is easy to see that both the integrand and the resulting integral in eq.
1
are infinite, because the derivatives ∂i ∂j R diverge as ∥x − x′ ∥ → 0. This is
energy ofceptable in anetwork
dislocation numerical implementation. A common approach to get rid
0.5
unwanted singularity is to truncate the integrals in eq. (10.4) so that the int
is set to zero whenever R becomes smaller than some 0cut-off radius1 rc . A
Etot (C) = Eel (C, rc ) + Ecore (C, rc ) 0.5
native"approach, to be followed here, is to replace every R inside eq. (10.4
Ra ≡ R 2 + rc2 [101], i.e.
# #
µ
Eel (C, rc ) = bi bj′ ∂k ∂k Ra dxi dxj′
16π C C
# #
µ
− ϵij q ϵmnq bi bj′ ∂k ∂k Ra dxm dxn′
8π C C
# #
µ ′
+ ϵikl ϵj mn bk bm ∂i ∂j Ra dxl dxn′ .
8π (1 − ν) C C
Obviously, the modified energy reduces to the original see
oneHirthin&the (1982) rc → 0
Lothelimit
“Theory of dislocations”
are two advantages to this approach. First, it completelyforremoves
details the sing
because the derivatives ∂i ∂j Ra become finite and smooth everywhere. Seco
derivatives of Ra can be integrated analytically over straight segments. Thi
37
to closed-form expressions that look very similar to and are as simple as th
discrete dislocation dynamics
driving force
@Eel
fiel =
@ri
38
elasticity
discrete theory alone.
dislocation dynamics
Suppose that the total energy Etot (C) of an arbitrary dislocation netw
could be obtained from an atomistic model that, in the present context, is re
asdriving Given that the elastic energy Eel (C, rc ) does not account for the th
exact. force
linear interactions in the core, let us define the core energy Ecore (C, rc )
difference between E
energy of dislocation tot (C) and Eel (C, rc ). Thus, the core energy is introdu
network
account for whatever is left unaccounted for in the continuum theory. Fortu
this contribution
Etot (C) = Eel is(C,
usually
rc ) + small andrclocal
Ecore (C, ) (or short-range), and can be wri
a single integral along the dislocation network:
!
Ecore (C, rc ) = Ec (x; rc ) dL(x), (
C
where •Ecdepends
(x; rc ) isonthe energy
local per unit length of the line at point x on the ne
line direction
By comparison, the effect of the elastic interaction is non-local (or long-r
since E•el (C,
from atomistic energy minus elastic
rc ) involves a double integral, as in eq. (10.5). The locality of th
• tension to reduce line length
39
energy means also that the integrand in eq. (10.15) depends only on
point x on the network.4 Usually, the line direction is spe
direction at dynamics
discrete dislocation
orientation angles θ and φ.
If we accept that eq. (10.15) holds for an arbitrary dislocation s
drivingthe
force
easiest way to compute the core energy function Ec (θ, φ ; rc )
the atomistic and elastic energies for a set of parallel, infinitely lon
energydislocations
of dislocation of different orientations. This is precisely what we did i
network
Suppose that the core energy function Ec (θ, φ ; rc ) has been cons
on(C)
Etot such=Ecalculations.
el (C, rc ) + EThen the
core (C, rc )core energy for a dislocation consisti
segments is given by the following:
!
Ecore (C, rc ) = Ec (θi−j , φi−j ; rc ) ∥ri − rj ∥,
(i−j )
driving force
41
C
only when x lies on a segment connected to node i. On a given segment
(10.7)
nd andprovide
discrete
(10.8) (10.8) provide
dislocation
analytic analytic
dynamics
∥x − r
expressions ∥ expressionsfor the for the Peach–Koehler
Peach–Koehler
Ni (x) decrease linearly from
= one at nodej i to zero at node j , i.e. (10.10)
to the dislocation
dislocation lines themselves.
N i lines themselves. The physical meaning
(x)
The
,
physical meaning of the of the
∥ri − rj ∥
ehler
ce can beforce can
explainedbe explained by considering
∥x − r ∥ the following thought pro-
driving forceby N considering
(x) = the following
j
, thought pro- (10.10)
nine that the shape Now of the
he shape of the dislocation linea changes
Fig. 10.2(b). dislocation
consider
i
∥riline
virtual changes
− rdisplacement
jby∥ δr(x), byi.e. of
δr(x),
every i.e.point
node every
i by point
δr i
slocation
ne moves its by
shape
δr(x). by In
N ithe
(x)δr limit
n moves by δr(x). In the limit of small δr(x), the corresponding The
to change i , of
as small
shown in
δr(x), Fig.the corresponding
10.2(a).
the
ated
changeelastic
in
ic energy isFig.
in energy
10.2(b).
change
the is Now
in elastic
elastic consider
energy
energy a virtual
from Peach–Koehler
is displacement
force of node i by δri
he line to change its shape
! by Ni (x)δri , as shown in Fig. 10.2(a). The
202 L I N E D I S L O C AT I
!
PK f energy PK
nding change in the elastic is
!
=δE − el = PK−
f · (x) · δr(x) dL(x).
dL(x), (10.11) (10.9)
δEel =
δE el − f (x)C · δr(x)i dL(x).
(x) N (x)δr i (10.9)
C C !
PK
teforce
the PK on
K force to energy
force
node δE to
elisforce
= −
force on node i,C let us
i on f
node (x)i, let
· N us remember
(x)δr
remember that in our model the
i i dL(x), that in our
dr(x) model the
(10.11)
dr2
displacement
re constrained variation to remain straight between every pair of connected nodes.
rained to remainδE straight202 between every
! variation
pair
LINE of connected
DISLOCATION DYNAMICSnodes.
r 1 r2
ne
lastic a shape
el
force
fi = N function
on node
−i (x) for
el Ni (x) for
is
= every PK every node i in such a way that
f node dL(x). N (x) is
ape function i (x) N i iin (x)such a way that Ni (x) 2(10.12) N is
(x) i
nly when x liesδr oni a segment C ! connected to node i. On ra0 given segment r3
n x lies on a segmentδEconnected to node i. On a given segment r0
(x) decreasef linearly el
= − from el one
= dr(x)at PK
f node dr2
i to zero at
dL(x). node j , i.e. (a) (10.12)
ease linearly
stic part of• the from one at node to zero
(x)
i nodal force is equal to the PK force integrated with
i Nati node
(x) j , i.e.
introduce shape δri function r1 C r2 F i g. 10.2. The force on node 2 is com
over all segments connected to nodej i. ∥x − r ∥ Taken together, eqs. (10.12),
N ∥x(x) −=rj ∥ , Peach–Koehler
r1 r2 (10.10)
force on segments 1–2
he
.8)elastic
furnishpart Niof (x)the
expressions = nodal
i
for force
computing
∥r ,
i −is requal
j ∥ the to the PK
elastic force
contribution ofintegrated
ment(10.10) node 2to bythe with
δr2 . As a result of th
Ni.i (x) over all segments ∥r i
r−
connected
0 r j ∥ to node
r
i.3
Takenr0
together,
1–2 and 2–3eqs. r3
(10.12),
is displaced by δr(x) = N2
shown in (b), is equal to one at node
ed
d in
(10.8)Fig. 10.2(b).
furnish Now
expressions consider for a (a)virtual displacement
computing the elastic of
(b) node
contribution i by1 δr to i the
g. 10.2(b). Now consider a virtual displacement of node by
neighboring
i δr nodes
i
and 3. The work
e line
node i. to change its shape
Fig. 10.2. by The
N i force on, node
(x)δr i as shown in
2 is computed Fig.as a10.2(a).
integrating weighted
the PK The
average
force 42ofthe
over theswep
ivative of this
forcediscrete
on node work 2 iswith
dislocation respect
computeddynamics toasδra2 . weighted average of the
ler force on segments 1–2Nand O DA2–3.L RE (a)P RConsider
E S E N TATaIvirtual
O N A Ndisplace-
D FORCES
e that
2 by the
δr2 .internal
As a result stress field,
of this eq. (10.8), is
displacement, eachan point
integral over the entire
on segments
nisnetwork.where
displaced This
drivingσ (x)
by is =
integral
force
δr(x) theNcan local be
(x)δr stress
written
, whereat point
as
thethe x sum
on the
shape of dislocation
line
function integrals
N (x), line,
over b is the Burg
2 2 2
vector,
t, segments
is equal and atξ up
tomaking
one (x)
node is 2the
the local
network,
and tangentlinearly
i.e.
decreases vector to of zero
the dislocation
at two line. Within
nodes non-singular
1 and 3. The continuum
work theory of
performed
stress from the dislocation network
! by dislocations,
the stress is the internal
obtained by stress field genera
σαβ (x) =
by the dislocation σαβ (x; k −
lines themselves canl).be written down explicitly, (10.13)
he PK force over the swept area (shaded). The force on node 2 is
! (k−l)
e of this work with respect µ to δr2 . " #
σαβ (x) = ∂i ∂p ∂p Ra bm ϵimα dxβ′ + bm ϵimβ dxα′
(x; k − l) is the stress8πfield C at point x produced by the straight segment
ghenodeinternal
k andstress
node field,
l and the eq. sum
µ
(10.8), is is an over
! taken integral
all over
segments the entire
in the dislo- % ′
work.
work.This integral can
+ be written for σαβas(x; the sum of
∂i ∂line integrals
− δαβ ∂over a dxk . (10
$
Analytic expressions bkm ϵ− l) have
imk α∂been
βR a derived i[101]
∂p ∂p Rbut
ments making up the 4π(1 −i.e.
network, ν) C
d here to save space.
see Hirth & Lothe (1982)
r to arrive at the final
Equations (10.7)expressions
! and (10.8) for the nodal
provide force,
analytic the line integral
expressions for the
“Theory in Peach–Koeh
of dislocations”
nodal mobility
• typically over-damped
• strongly anisotropic
• depends on direction of motion
• depends on line direction ✓
44
discrete dislocation dynamics
nodal mobility
vi = M({fj })
45
discrete dislocation dynamics
nodal mobility
drag
f (x) = B(⇠(x))v(x)
per line length drag tensor
drive
f (x) = B(⇠(x))v(x)
46
discrete dislocation dynamics
nodal mobility
I
Ni (x)[f drag (x) + f drive (x)] dL(x) = 0
IC
Ni (x)B(⇠(x)) v(x) dL(x) = fi
C
I zX }| {
Ni (x)B(⇠(x)) vj Nj (x) dL(x) = fi
C j
XI
Ni (x)B(⇠(x))Nj (x) dL(x) vj = fi
j |C {z }
X
Bij v j = fi
j
47
discrete dislocation dynamics
edge
Be,glide
✓
screw
b
Bs,glide
2 2
Bs,glide cos ✓ + Be,glide sin ✓
48
discrete dislocation dynamics
edge
Be,climb
✓
screw
b
Bs,glide
49
discrete dislocation dynamics
{1 1 1}
50
discrete dislocation dynamics
• no out-of-plane
component
vi · nij = 0
0 1
• remove normal
X
components vi glide
= @I nij ⌦ nij A vi
j
51
discrete dislocation dynamics
B(⇠) / I ⇠⌦⇠
52
discrete dislocation dynamics
2 2
B(⇠) = (Bs,glide cos ✓ + Be,glide sin ✓)(I ⇠ ⌦ ⇠)
B(⇠) = B(I ⇠ ⌦ ⇠)
53
discrete dislocation dynamics
• pure B(⇠) = Bs (I ⇠ ⌦ ⇠)
screw
2 2 2 2 2
Bglide = Be,glide |b ⇥ ⇠| + Bs (b · ⇠)
2 2 2 2 2
Bclimb = Be,climb |b ⇥ ⇠| + Bs (b · ⇠)
54
discrete dislocation dynamics
topological changes
possible reasons
• rediscretization (numerical)
55
discrete dislocation dynamics
junction formation in a
low-angle grain boundary
56
bEBdynamics
discrete dislocation := bAB .
3. Deallocate old Burgers vectors bAB and bBA .
topological changes: rediscretization
Conversely, the following algorithm deletes node D between no
elementary topological operators
B D C
A
E
delete node
• add add node
bFig.
AE := 10.5.
bAB andTo improve
bEA := bBA representation of dislocation lines such
bebnecessary
bBE := to add a new node E between A and B or to delet
BA and bEB := bAB
B and C.
• delete
bBC := bBD and bCB := bCD
57
This is because in the regions of high local curvature, it is necessary to discretize
the lines into fine segments, whereas the same is not required in the regions of low
discrete dislocation dynamics
curvature. Let us now consider a slightly more sophisticated approach and demand
that, for any directization node (a node with only two neighbors), the orientations
of two sequential segments connected to the node should not differ too much.
The following algorithm
topological enforcesrediscretization
changes: this requirement by monitoring the area of the
triangle formed by node (N0 ) and its two neighbors (N1 and N2 ). The r1 inputs for r0
this algorithm are lmin , lmax , Amin , and Amax .
algorithm r2
Algorithm 10.5 (remesh)
1. If node N0 has more than two arms, go to 7. Otherwise, find nodes N1 and
N2 connected to node N0 . Obtain positions r0 , r1 , r2 and velocities v0 , v1 ,
v2 of nodes N0 , N1 , N2 .
2. Compute the area of the triangle formed by nodes N0 , N1 and N2 ,
A := ∥(r1 − r0 ) × (r2 − r0 )∥. Compute the rate of change of the triangle
area, dA/dt.
3. If ∥r2 − r1 ∥ < lmax , A < Amin and dA/dt < 0, use Algorithm 10.4 to delete
node N0 , exit.
4. If ∥r1 − r0 ∥ < lmin or ∥r2 − r0 ∥ < lmin , delete node N0 , exit.
5. If A > Amax and ∥r1 − r0 ∥ ≥ lmin , use Algorithm 10.3 to add a new node
between N0 and N1 .
6. If A > Amax and ∥r2 − r0 ∥ ≥ lmin , add a new node between N0 and N2 .
7. If any arm of node N0 is longer than lmax , add a new node in the middle of
the arm.
58
node 0 has n arms. Suppose the decision is made to split node 0 into
discrete dislocation
and dynamics
a. Assume also that, in this split, node a is entitled to keep arms
the new node 0 retains arms s + 1, . . . , n (Fig. 10.6(b)). The followi
describes how this can be accomplished. For the case s = 1, the spli
topological changes:
equivalentjunctions
to adding a new node on segment 0–1.
Fig. 10.6. Split node 0 into two nodes 0 and a. The new node a inheri
bai := b0i andwhile
bia :=node
bi0 0i keeps . , s remaining arms s + 1 . . . n. Depending o
= 1, . .the
of the sBurgers vectors, new nodes 0 and a may or may not be con
X
ba0 = b0a = b= bai if b 6= 0
i=1
59
Algorithm 10.7 (merge nodes)
discrete dislocation dynamics
1. Deallocate Burgers vectors b and ba0 if they exist.
0a
2. Allocate new Burgers vectors b0i := bai , bi0 := bia , i = 1, . . . , s.
3. Deallocate
topological changes: Burgers vectors bai , bia , i = 1, . . . , s. Deallocate nod
junctions
4. Double-loop through all segments connected to node 0. If tw
segments
elementary topological i and j are found to connect node 0 to the same neig
operators
a
• merge b b b
0 0 0
F i g. 10.7. (a) Before merge, nodes 0 and a may have a common neigh
will result in a double connection
b0i := bai and bi0 := bia i = 1, . . . , s between nodes 0 and
check for b after merg
double
double connection is replaced by a single connection or completely
connectivity
depending on the sum of the Burgers vectors of two overlapping seg
60
Let us define the two segments to be in contact if d < ra , where ra is a col
distancedynamics
discrete dislocation parameter (subscript a stands for annihilation).
Let us now see what has to happen in the line DD model in order to ena
possible reaction between two segments 1–2 and 3–4. The first step is to add
two nodes at points P and Q on the two colliding segments.13 Because the
topological changes: junctions
nodes just introduced are within contact radius ra of each other, the merge is c
upon to merge them into new node P ′ , as shown in Fig. 10.8(b). To comple
when to merge
2 2 2
4 4 4
P d Q!
• segment P!
Q P!
collision
1 3 1 3 1 3
d < ra
(a) (b) (c)
Fig. 10.8. (a) The minimum distance d between two unconnected segment
and 3–4 is reached at points P and Q. Two segments are considered to
contact if d < ra . New nodes are introduced at points P split
and Q. (b) Nod
• minimum nodal
and Q are merged into a single node P ′ . (c) Node P ′ splits into two new n
separation P ′ and Q′ , possibly leading to a topology different from that in (a).
d < ra
13 This is assuming that points P and Q do not coincide with the end nodes of two co
segments. If P or Q does coincide with one of the end nodes, the appropriate end node sho
used in place of P or Q in the following topological rearrangements. 61
Let us define the two segments to be in contact if d < ra , where ra is a col
distancedynamics
discrete dislocation parameter (subscript a stands for annihilation).
Let us now see what has to happen in the line DD model in order to ena
possible reaction between two segments 1–2 and 3–4. The first step is to add
two nodes at points P and Q on the two colliding segments.13 Because the
topological changes: junctions
nodes just introduced are within contact radius ra of each other, the merge is c
upon to merge them into new node P ′ , as shown in Fig. 10.8(b). To comple
when to split
2 2 2
4 4 4
• multi-arm node with
P d P! Q!
more than 3 segments P!
Q
1 3 1 3 1 3
(a) (b) (c)
Fig. 10.8. (a) The minimum distance d between two unconnected segment
and 3–4 is reached at points P and Q. Two segments are considered to
contact if d < ra . New nodes are introduced at points P and Q. (b) Nod
• select maximum
and among
dissipation rate Q are merged intoQ̇Pa = fP · vnode
single P <PfP′0. ·(c) 0 + fQ0 ·′ vQ0 = Q̇P 0 Q0
vPNode P splits into two new n
P ′ and Q′ , possibly leading to a topology different from that in (a).
possible splits
13 This is assuming that points P and Q do not coincide with the end nodes of two co
segments. If P or Q does coincide with one of the end nodes, the appropriate end node sho
used in place of P or Q in the following topological rearrangements. 62
discrete dislocation dynamics
summary of 3D line DD
63
continuum mechanics
$SZTUBM1MBTUJDJUZ'JOJUF
• course follows
&MFNFOU.FUIPET
JO.BUFSJBMT4DJFODFBOE&OHJOFFSJOH
Roters, Eisenlohr, Bieler, & Raabe: “Crystal Plasticity
Finite Element Methods in Materials Science and
Engineering”, chapter 3
Wiley-VCH, 2010
64
continuum mechanics
content
• kinematics:
study of (typically position-dependent) displacements and, if considering time-dependence,
motions of a material body without explicitly asking about the forces that are causing them
• mechanical equilibrium:
conditions for forces acting on the body of material and causing above kinematic reactions
65
continuum mechanics: kinematics
B0
x y
reference current
y(x) : x 2 B0 7 ! y 2 B
66
continuum mechanics: kinematics
deformation gradient
u(x+dx)
dx dy
u(x)
x y
reference current
@y
y + dy = y(x) + dx + O(dx2 )
@x
@y
dy = dx = F dx
@x
67
continuum mechanics: kinematics
deformation gradient
u(x+dx)
dx dy
u(x)
x y
reference current
@y
F= = Grad y or
@x
Fij = @yi /@xj in Cartesian components.
68
continuum mechanics: kinematics
deformation gradient
u(x+dx)
dx dy
u(x)
x y
reference current
1 @x
F = = grad x or
@y
1
Fij = @xi /@yj in Cartesian components.
69
continuum mechanics: kinematics
deformation gradient
1
• push forward:
A = F A0 F
acting in reference configuration
1
• pull back
A0 = F AF
acting in current configuration
70
continuum mechanics: kinematics
change in volume
dV ⇢0
• Jacobian J = det F = =
dV0 ⇢
71
continuum mechanics: kinematics
change in area
T
• directed surface dn = JF dn0
72
continuum mechanics: kinematics
reference current
1 1
• stretch dy · dy = F dx · F dx dx · dx = F dy · F dy
✓ ◆2 ✓ ◆2
dl dl0 1 1
=Fa·Fa =F b·F b
dl0 dl
= a · FTF a =b·F T
F 1
b
= a · Ca T 1
= b · (FF ) b
2
= =b·B 1
b
2
=
right Cauchy–Green tensor
left Cauchy–Green tensor
73
continuum mechanics: kinematics
change in angle
dy1
dx1 dx2 dy2
✓
y
x
1 2 cos ✓ = a1 · C a2
74
continuum mechanics: kinematics
T
• right Cauchy–Green C:=F F
tensor ⇣ ⌘T ⇣ ⌘T
= FT FT = FTF
= CT
• left Cauchy–Green B : = FF T
tensor ⇣ ⌘T ⇣ ⌘T
= F T F T = FF T
= BT
75
continuum mechanics: kinematics
3
X
• C is symmetric and positive-definite C= µi n i ⌦ n i
i=1
• orthogonal eigenvectors n1 , n2 , n3
76
continuum mechanics: kinematics
2
j = n j · C nj
3
!
X
= nj · µi n i ⌦ n i nj
i=1
3
!
X
= nj · µi ni (ni · nj )
i=1
3
!
X
= nj · µi n i ij
i=1
= µj n j · n j
= µj
77
continuum mechanics: kinematics
3
X
2
• right Cauchy–Green C= i ni ⌦ ni
tensor i=1
UU = U2 = C
3
X
• right stretch tensor U= i ni ⌦ ni
i=1
78
continuum mechanics: kinematics
• invertible tensor F F = RU = VR
• R is proper orthogonal
(rotation)
• U, V are positive-definite
and symmetric
79
continuum mechanics: kinematics
F = RU = VR
C = FTF B = FF T
T T
= (RU) (RU) = (VR) (VR)
= UT RT RU = VRRT VT
= UT U = VVT
= U2 = V2
80
continuum mechanics: kinematics
u(x)
x y
connection to strains and rotations
from infinitesimal framework o
reference current
@y @(x + u)
F= =
@x @x
@u
=I+
@x"
✓ ◆T # " ✓ ◆T #
1 @u @u 1 @u @u
=I+ + +
2 @x @x 2 @x @x
✓ ◆ ✓ ◆
@u @u
=I+ +
@x sym @x skew
=I+"+!
81
continuum mechanics: kinematics
u(x)
x y
reference current
✓ ◆T ✓ ◆
T @u @u
C=F F = I+ I+
@x @x
✓ ◆T ✓ ◆T ✓ ◆
@u @u @u @u
=I+ + +
@x @x @x @x
⇡ I + 2"
U = C1/2 ⇡ I + "
82
continuum mechanics: kinematics
strain measures
• Biot E(1) = U I
(2) 1
• Green’s Lagrangian E=E = U2 I
2
1 1⇣ T ⌘
= (C I) = F F I
2 2
(m) 1
• Doyle–Ericksen E = (Um I)
m
83
continuum mechanics: kinematics
u(x)
x y
time-dependent displacement, motion y(t)
o
reference current
d
• velocity field v= u = u̇ = ẏ
dt
@v
• spatial velocity gradient
L= = grad v
@y
(relative velocity between two
points in current configuration)
84
continuum mechanics: kinematics
velocity gradient
@v d d 1
L dy = dy ⌘ dy = F dx = Ḟ dx = ḞF dy
@y dt dt
1
L = ḞF
Ḟ = LF
85
continuum mechanics: kinematics
velocity gradient
86
continuum mechanics: kinematics
velocity gradient
additive decomposition
1 T 1
L = Lsym + Lskew = L+L + L LT = D + W
2 2
87
continuum mechanics: kinematics
elasto-plastic decomposition
• crystalline solid
• elastic and/or
plastic shape
change
• pure plastic is
lattice-invariant
and stress-free
88
continuum mechanics: kinematics
elasto-plastic decomposition
intermediate n
α
(or relaxed) α
m
reference current
Fp Fe
X
x y(x)
F(x)
infinitesimal
neighborhood of x
89
continuum mechanics: kinematics
elasto-plastic decomposition
F = Fe Fp
intermediate n
α
(or relaxed) α
m
reference current
Fp Fe
X
x y(x)
F(x)
infinitesimal
neighborhood of x
90
continuum mechanics: kinematics
elasto-plastic decomposition
1
L = ḞF
1 1
= (Fe Fp )˙Fp Fe
= Ḟe Fp Fp 1
Fe 1
+ Fe F˙p Fp 1
Fe 1
= Ḟe Fe 1
+ Fe F˙p Fp 1
Fe 1
1
= Le + Fe Lp Fe
acts in
acts in current
intermediate
configuration
configuration
91
continuum mechanics: mechanical equilibrium
B0
x y
reference current
• body force field g
Z Z
• surface tractions k= t0 dS0 = t dS
S0 S
Z Z
= P dn0 = dn
S0 S
first Piola–
Kirchhoff Cauchy stress
(nominal) stress
92
continuum mechanics: mechanical equilibrium
B0
x y
reference current
• body force field
Z Z
• surface tractions k= P dn0 = dn
S0 S
T
dn = JF dn0
T
P=J F
93
continuum mechanics: mechanical equilibrium
reference
Z Z Z
f= ⇢0 v̇ dV0 = ⇢0 g dV0 + t0 dS0
B0 B0 S0
94
continuum mechanics: mechanical equilibrium
reference
Z ✓ ◆
@P
0= ⇢0 g + ⇢0 v̇ dV0
B0 @x
Z
= (Div P + ⇢0 (g v̇)) dV0
B0
95
continuum mechanics: mechanical equilibrium
B0
x y
reference current
PF T = FPT () = T
96
continuum modeling of microstructured solids
overview
B
B0
x y
reference current
• modeling objective
• material point
97
continuum modeling of microstructured solids
• anisotropy
• physically-
based model
mechanical
response ?
solve partial differential
• traction,
equation system under displacement
time-varying boundary • temperature
conditions
• other fields…
Div P = 0
Curl F = 0
98
continuum modeling of microstructured solids
solution methods
• finite differences
• spectral collocation /
Green’s function
• finite elements
99
continuum modeling of microstructured solids
material point
what is there?
P(F) ?
100
continuum modeling of microstructured solids
material point
101
continuum modeling of microstructured solids
http://axisorigin.com/hkane/wp-content/uploads/Stent_01.jpg
10.1016/j.ijfatigue.2012.04.022
F=F
P = P(F)
102
continuum modeling of microstructured solids
problems
103
continuum modeling of microstructured solids
F1 P1
F2 P2
F .. .. P
. .
FN PN
104
continuum modeling of microstructured solids
elasto-plasticity
105
continuum modeling of microstructured solids
system boundary
response condition
P(F(t))
time- instantaneous
dependent elasticity
plasticity
106
continuum modeling of microstructured solids
elasto-plastic decomposition
intermediate n
α
(or relaxed) α
m
reference current
Fp Fe
X
x y(x)
F(x)
infinitesimal
neighborhood of x
107
continuum modeling of microstructured solids
elasto-plastic consistency
plastic
deformation
rate
S Lp
stress
(in intermediate
sp itu l
re nst eria
se e
on tiv
configuration)
t
co ma
Ḟp = Lp Fp
s
ic
at
m
ne
Fe ki Fp
elastic F = Fe Fp plastic
deformation deformation
108
continuum modeling of microstructured solids
109
• elasticity
• plasticity
constitutive law
dx
physics of deformation
Fig. 11—Mg2Si precipitates in cell boundary regions of as-cast AS21
x
reference
stress
deformation
S
lished in the as-cast structure and the $ phase is supersaturated
Lp
with solutes. The larger supersaturation in the grain bound-
Fe
(b) ary region compared to the grain interiors observed by Ref-
oundary area of AZ91 erence 5 explains the contrast difference in Figure 6. During
a: (a) bright-field TEM annealing at elevated temperature, precipitation of solute ele-
from matrix (circles)
ments occurs[5] to establish thermodynamic equilibrium. Con-
current creep promotes precipitation by stimulating nucleation.
o
crystallite
Similar coarsening of the Mg2Si precipitates in AS21
(thermo-)
has not been observed for T % 423 K. However, the pre-
P = P (F , Ḟ )
&
aggregate
continuum
solver for
• equilibrium
P̄
• compatibility
def. partitioning
homogenization
110
continuum modeling of microstructured solids
upcoming lectures
111
continuum modeling of microstructured solids
homogenization
scheme <homogenization>
<phase>
<microstructure>
[Taylor]
[Aluminum_J2isotropic]
[Aluminum_j2]
type
elasticity
crystallite isostrain
hooke 1
microstructure Ngrains j2 1
(constituent)
plasticity phase 1 texture 1 fraction 1.0
[GrainCluster]
[DP_Steel]
(output) flowstress
(output) strainrate
type
crystallite RGC
1
phase texture
Ngrains
(constituent)
c11 8
phase
110.9e92 texture 1 fraction 0.82
phase texture clustersize
(constituent)
c12 222 3
phase
58.34e9 texture 1 fraction 0.18
…
taylorfactor 3
phase texture tau0 31e6
…
112
constitutive description of crystal plasticity
overview
• experimental background
• phenomenological descriptions
113
constitutive description of crystal plasticity
resolved stress
n
slipped area
m
• work done by
dislocation motion
W = dA ⌧ b
⌧ = ( n) · m
= · (m ⌦ n)
114
constitutive description of crystal plasticity
m n
• rotation of slip
direction and slip
plane normal
115
constitutive description of crystal plasticity
n n
λ
B 0 C
D λ
λ
s s
E
l cos = l0 cos 0
n
y l0
l
κ0
x l sin = l0 sin 0
κ
s
A
!
cos κ0 sin2 λ0 116
τ = σ cos κ cos λ = σ · · 1−
0
constitutive description of crystal plasticity
λ
ε
lattice slip kinematics (tension)
initial slip
direction
primäre Gleitrichtung
1 01
1 11
P''
l cos = l0 cos 0
1 12
P'
l sin = l0 sin 0
1 11 P
tensile axis
@u
h=b
@y
@u
hn = bm
@y
@u b
= m⌦n
@y h
@u bA
= m⌦n= m⌦n
@y V
@v b Ȧ
= m⌦n= ˙ m⌦n
@y V | {z }
slip system
geometry
118
constitutive description of crystal plasticity
X
Lp = ˙ ↵ m↵ ⌦ n ↵
↵
119
constitutive description of crystal plasticity
experimental background
shear 50
2
stress 1
40
Schubspannung τ [MPa]
4 5
30 III
20 6
1
II
3
10
5
6 4
I 2
0
0.1 0.2 0.3 0.4
Abgleitung γ shear
120
constitutive description of crystal plasticity
experimental background
Schubspannung τ
Bereich I Bereich II Bereich III
dislocation deposition on
2nd systems
τIII
121
constitutive description of crystal plasticity
phenomenological description
• slip kinetics
122
constitutive description of crystal plasticity
phenomenological description
• resistance to deformation on
each slip system
g↵
positive and
negative sense
123
constitutive description of crystal plasticity
phenomenological description
ġ ↵ = h↵ ˙
✓ ◆ a
g g
= q ↵ h0 sgn 1 1 ˙
g1 g1
124
constitutive description of crystal plasticity
phenomenological description
slip kinetics
stress sensitivity or
rate insensitivity
125
constitutive description of crystal plasticity
phenomenological description
advantages drawbacks
• no clear
mechanism(s) for
internal state
evolution
126
constitutive description of crystal plasticity
• slip kinetics
127
constitutive description of crystal plasticity
128
constitutive description of crystal plasticity
↵
• dislocation production %̇+
129
constitutive description of crystal plasticity
• geometric parameter:
slipped area per deposited ⇤
dislocation length
↵
• rate of dislocation density V d
b⇤ = ↵
increase V d%
130
constitutive description of crystal plasticity
0.5
• geometric parameter is structure ⇤/%
dependent
p ↵
↵ % ↵
d% = |d |
b c⇤
p ↵
↵ % ↵
%̇ = |˙ |
b c⇤
131
constitutive description of crystal plasticity
dA+ ?
dlnV = 2d
V
d
= 2d?
b
132
constitutive description of crystal plasticity
• …
133
constitutive description of crystal plasticity
slip kinetics
↵ ↵ ↵
• Orowan equation ˙ = % bv
velocity of
dislocation motion
134
constitutive description of crystal plasticity
• dislocation structure
• Peierls barrier
135
constitutive description of crystal plasticity
q
↵
• slip system interaction ⌧cr = µ b a↵ %
136
presume no dislocation activity. Above, only an effective resolved
constitutive description of crystal plasticity
shear stress teff that is reduced by the critical resolved shear stress
is acting as driving force for dislocation motion.22 22
A. Se
8 penden
< (|t | t ) sign t if |t | > t and of
cr cr Crysta
teff = (3.38)
:0 if |t | t ries 7, 4
dislocation density-based description cr 10.1080
The flowstress tcr accounts for the different strength of the reaction
products of the interacting dislocations depending on the slip sys-
slip system interactions (fcc) tems involved. In the case of fcc crystals, six distinct reactions
x
1 2 3 4 5 6 7 8 9 10 11 12 Table
z betwe
1 s cp cp h l g cl g g h g l x in a
2 cp s cp l h g g h l g cl g s: self
• self
3 cp cp s g g cl g l h l g h nar in
intera
• coplanar 4 h l g s cp cp h g l cl g g g: gli
locks.
5 l h g cp s cp g cl g g h l
• collinear in tab
6 g g cl cp cp s l g h g l h
• Hirth lock 7 cl g g h g l s cp cp h l g
8 g h l g cl g cp s cp l h g
• Lomer lock 9 g l h l g h cp cp s g g cl
10 h g l cl g g h l g s cp cp
• glissile junction
11 g cl g g h l l h g cp s cp
12 l g h g l h g g cl cp cp s
23
L. Ku
with characteristic strength can be identified (see table 3.3 for a T. Hoc
age rat
correlation between slip systems and interaction types). Following face-ce
the work of Kubin et al.,23 the critical resolved shear stress of a slip Mater.,
system x reads 137 10.1016
constitutive description of crystal plasticity
• solid solution
MOTION OF A CRYSTAL DISLOCATION 21
atoms t= 0 Eb = Ep
Eb
• Peierls barrier
Kinks
0 < τ < τp 0 < Eb < Ep
Eb
Peierls valley
τ = τp Eb = 0
(a) (b)
Fig. 1.12. (a) A schematic of the periodic Peierls potential for dislocation motion
through a discrete lattice, viewed along the dislocation line (the positions of the
lines are marked by the small filled circles). Eb is the energy barrier that the
dislocation sees when it moves from left to right. The energy barrier decreases
with the increasing stress τ and disappears completely when the stress reaches
the Peierls threshold τ = τP . (b) A three-dimensional sketch of the Peierls
potential. When τ < τP , the dislocation moves by nucleating a kink pair 139 and
constitutive description of crystal plasticity
• solid solution
atoms
• Peierls barrier
activation energy Q = ⌧ˆ A b = ⌧ˆ l s b
✓ ✓ ✓ ◆ p ◆q ◆
• probability to Q |⌧e↵ |
overcome obstacle P = exp 1
kB T ⌧ˆ
140
constitutive description of crystal plasticity
• solid solution b
lS = p sS = dobst
atoms cat
activation energy Q = ⌧ˆ A b = ⌧ˆ l s b
✓ ✓ ✓ ◆ p ◆q ◆
• probability to Q |⌧e↵ |
overcome obstacle P = exp 1
kB T ⌧ˆ
141
constitutive description of crystal plasticity
1
• waiting time at obstacle t = (⌫a P )
142
constitutive description of crystal plasticity
✓ ◆ 1
t 1
v= = +
t+ vT
vT
t tP tS
= + + ...
P S
143
demonstration
144
homogenization schemes
problem
B
B0
x y
reference current
F P
known unknown
P F
145
homogenization schemes
solutions
• mean-field methods
• computational homogenization
146
homogenization schemes
mean-field methods
• no rigorous solution of
microscopic boundary value
problem
1 X (i) (i)
P= V0 P = hP(i) i
V0
147
homogenization schemes
mean-field methods
iso-strain
(i)
F =F
(i)
P = hP i
(1) (2)
F F
148
homogenization schemes
mean-field methods
iso-stress
(i)
P =P
(i)
F = hF i
149
homogenization schemes
P(i) = (i)
P⇤
(i)
F = hF i
Sachs, G. (1928). Zur Ableitung einer Fließbedingung.
Z. Ver. Deutsch. Ing., 72, 734–736.
150
homogenization schemes
mean-field methods
iso-work rate
˙ X
(i) (i)
• volume average F= ⌫ Ḟ
X
• constraint ⌫ (i) (i)
=1
151
homogenization schemes
mean-field methods
iso-work rate
• non-linear equation h i
system for free (i)
P(i) ( (i)
) (N )
P(N ) ( (N )
) · F˙ = 0
parameters
152
homogenization schemes
mean-field methods
(elastic) composites
0 = C0 ✏ 0
• strain ✏1 = A✏
localization
tensors ✏1 = B✏0
1 1 = C1 ✏ 1
A = B : [⌫0 I + ⌫1 B]
= ⌫0 0 + ⌫1 1
✏ = ⌫ 0 ✏0 + ⌫ 1 ✏1
153
homogenization schemes
mean-field methods
(elastic) composites
0 = C0 ✏ 0
• macro
stiffness
= C✏
1 = C1 ✏ 1
C = [⌫0 C0 + ⌫1 C1 : B] :
1
[⌫0 I + ⌫1 B] = ⌫0 0 + ⌫1 1
✏ = ⌫ 0 ✏0 + ⌫ 1 ✏1
154
homogenization schemes
mean-field methods
(elastic) composites
• macro C = [⌫0 C0 + ⌫1 C1 : B] :
stiffness
1
[⌫0 I + ⌫1 B]
• iso-strain B=I
1
• iso-stress B = C1 : C0
155
homogenization schemes
mean-field methods
Eshelby formalism
156
homogenization schemes
mean-field methods
Eshelby formalism
• cut out
157
homogenization schemes
mean-field methods
Eshelby formalism
• cut out
• stress-free
eigenstrain ✏⇤
158
homogenization schemes
mean-field methods
Eshelby formalism
• cut out
• stress-free
eigenstrain ✏⇤
• apply stress
⇤
(or traction)
159
homogenization schemes
mean-field methods
Eshelby formalism
• cut out
• stress-free
eigenstrain
✏⇤
• apply stress
⇤
(or traction)
• put back
160
homogenization schemes
mean-field methods
Eshelby formalism
• cut out
• stress-free
eigenstrain
✏⇤
• apply stress
⇤
(or traction)
• put back
161
homogenization schemes
mean-field methods
Eshelby formalism
• cut out
• stress-free
eigenstrain ✏⇤
• apply stress
⇤
(or traction)
• release Green’s function for displacement resulting from (point) body forces
c ⇤
) ✏ = S ✏ ⌘ const
within ellipsoidal inclusions
162
homogenization schemes
mean-field methods
✏
isolated inclusion
C0
✏1 = H(I, C0 , C1 ) ✏ C1
| {z }
1
[I+S(I,C0 ):C0 1 :(C
1 C0 ) ]
163
homogenization schemes
mean-field methods
isolated inclusion
unknown composite modulus
• self-
consistent
✏1 = H(I, C, C1 )✏
inclusion experiences
matrix strain as far-field
strain
164
homogenization schemes
mean-field methods
summary
165
homogenization schemes
strategy
166
homogenization schemes
• stack experiences
plastic velocity
gradient
0 1 0 1
2
X 0 0 1 0 0 0
1 @
• each grain may Lp = r
˙ rlx Krrlx = ˙ rlx 0 0 0A + ˙ rlx
2 @
0 0 1A
deviate by two r=1 0 0 0 0 0 0
0 1
shear relaxation 0 0 1
˙ rlx
modes = @0 0 2 A
˙ rlx
0 0 0
167
homogenization schemes
Grai
nb e2
e1
a
N
X
La = Lp + Lp = (m↵ ⌦ n↵ ) ˙ ↵
• symmetric distribution
↵=1
(valid for equal grain volume)
b
N
X
Lb = Lp Lp = m ⌦n ˙ ,
=1
• usually slip system
activity differs in both
grains
168
homogenization schemes
Grai
nb e2
e1
• solution minimizes
plastic dissipation rate
a b
N
X N
X X
P= ⌧↵ ˙ ↵ + ⌧ ˙ + r
⌧rlx r
| ˙ rlx | = min
↵=1 =1 r=1
169
homogenization schemes
LAMEL advancements
170
homogenization schemes
strategy
• extension of mono-directional
to tri-directional “stack”
e3
e1 e2
Cluster axes
171
homogenization schemes
• relax all grain pairs stacked along the shortest and second
shortest dimension as in LAMEL
• Wagner, P. (1994). Zusammenhange zwischen mikro- und makroskopischen Verformungsinhomogenitäten und der Textur. Ph.D.
thesis, RWTH Aachen
• Crumbach, M., Pomana, G., Wagner, P., & Gottstein, G. (2001). A Taylor type deformation texture model considering grain
interaction and material properties. Part I – Fundamentals. In G. Gottstein, & D. A. Molodov (Eds.), Recrystallisation and Grain
Growth, Proceedings of the First Joint Conference (pp. 1053–1060). Berlin: Springer
172
homogenization schemes
• generalization of GIA
• finite strain
• arbitrary relaxations
173
homogenization schemes
ng–3
174
homogenization schemes
Grain g2 Grain g1
• relaxation vector
displaces mutual
interface
ag–12 = ag11
• compatible relaxation
ng–12
ng11
e3
e1 e2
175
homogenization schemes
Grain g1
• relaxation vector
displaces mutual
interface
ag22
• compatible relaxation ag21 ng21
ng22
• incompatible relaxation
e3
Grain g2 e1 e2
176
homogenization schemes
3
X
• individual grain g 1
deformation gradient
F = F̄ + (ag↵ ⌦ ng↵ )
d
±↵=1 ↵
1⇣ g ⌘T
• mismatch (surface
gT
dislocation tensor) Mg↵ = n↵ ⇥ F↵
across interface 2
magnitude determines
penalty energy density
177
homogenization schemes
computational homogenization
strategy
x y
178
homogenization schemes
computational homogenization
reference coordinates
e
y = Fx + w
deformed coordinates
@y e
@w
=F +
@x @x
e
F =F +F
179
homogenization schemes
computational homogenization
Z
F= F dV0
B0
Z
= e dV0
F+F
B0
Z
=F+ e dV0
F
B0
| {z }
!
=0
180
homogenization schemes
computational homogenization
boundary conditions
S+
Z
0= e dV0
F
B0
S–
Z
= e ⌦ dn0
w
S0
Z Z
= e ⌦ dn0 +
w e + ⌦ dn0 +
w
S0 S0+
181
homogenization schemes
computational homogenization
examples
!"#$%&'$
67%& /(/)4 5%&'"&&)& 3-5)##%08 -2 $7) 9)7(.%-"4 -2 &$4"'$"4)& 4)%02-4')5 91 #-08 :94) ;%<=6% '-3/-&%$) 3($)4%(# >%$7
( /)4%-5%' 3%'4-&$4"'$"4)? @ 0)> 3"#$%&'(#) 9)7(.%-"4 3-5)# 9(&)5 -0 ( 3"#$%#).)# :0%$) )#)3)0$ A*+, B (//4-('7 %&
"&)5 $- $(C) %0$- (''-"0$ 7)$)4-8)0)%$%)& %0 $7) 9)7(.%-"4 9)$>))0 $7) :94) (05 3($4%D? E$ %& &7->0 $7($ '-39%0%08 $7%&
3-5)# >%$7 /(4(##)# '-3/"$($%-0 $)'70%F")& 0-> 3(C)& %$ /-&&%9#) $- '-0&%5)4 4)(#%&$%' '-3/-&%$) &$4"'$"4(# '-3/"G
$($%-0& 1%)#5%08 ( 5)$(%#)5 8)-3)$4%' 5)&'4%/$%-0 (05 '-0&$%$"$%.) )F"($%-0& 8%.%08 ('')&& $- 3%'4-&$4"'$"4(# 5($(H %0G
&$)(5 -2 -0#1 $- /7)0-3)0-#-8%'(# 3('4-&'-/%' 5($( 5%I'"#$ $- '-44)#($) >%$7 $7) #-'(# 3)'7(0%'(# &$($)? ! JKKK
9/$-%'I Z,'0/2) E--&I Z,*0I "34.4I CG[ J#QQQK [QF\[[Q
+#&).%)4 ;'%)0') L?M? @## 4%87$& 4)&)4.)5? HHHI,&),5(,.I*/$:&/*+',:*$+
$"%.&).'#++6 .7 5$#'" 0%&3 *(" ,"(#-.&/% &0 &7" "+"3"7*#%6 $#**"%7 =" " # ":/#*.&7B8
% " .)2*'? $2*6(26*%3 ! !
C* '#7 ," )"3&75*%#*") *(#* *(" 5'#+"5 #%" 5"$#%#*" .0 *(" 7/3,"% &0 $#**"%75 .5 .717.*"8 9(" 3.'%&5'&$.'
" )//*#)("%$ )*% 6$%, )2 JKLMH NO;PQR; +62 "%*% &% #-1= ,'$(6$$ 2"%
).5$+#'"3"7* '#7$%(#-,3
," E%.**"7 #5 *(" 5/3 &0 # $"%.&).' 1"+) #7) # 3#'%&5'&$.' 1"+) =" )"7&*"5 # +&'#+ %"0>
T
@%(2'#- T; &% >'7% ) .%2"#,#1#>'()1 ,%$(*'/2'#- #8 2"% EL "%"7'" &7 # '"++ #7) ! #U;3#'%&5'&$.'
.#,%13 S- @%(2'#- &% ,'$(6$$ 5*%#.7
'2$ '.<*"75&%B8
-2)2'#- )-, 2"% *%56'*%.%-2 8#* 6$'-> /)*)11%1 (#./62'-> 2%("-'($3 @%(2'#- ! V >'7%$ %?)./1%$ #8
E("%" $ .5 # $"%.&).' 1"+) E.*( *(" $"%.&) &0 *(" ,#5.' '"++8 9(" +&'#+.5#*.&7 #7) (&3&;"7.5#*.&7 ":/#*.&75
1
!
!"#$%%&' "( )*+,%- ./ #+(/- (*$( (*/ !0 )+%,("+# #%" &,*#.7") ,6 #-"%#;.7;D
%&'()*)+),-&. !
2"3/) $ )$(")4$5(+6& examples: $776+8"9$("+#
turbine rotor 4+6 $ :.6/ -";
'! ( %
F
! )#!
"
$9/(/6 +4 (*/ )$9/ 9$2#"(,-/ $) (*/ 3+%,9/ "
$ -%& 2=/% #8 .#,%1 8)11$ &'2"'- 2"% >%-%*)1 ()2%>#*= #8 .612'$()1% .#,%1$3 !"% (#-$2'262'7% %56)2'#-$
! % '"%56)2'#-$
(!
'22%- #-1= #- .'(*#$(#/'( $()1% )-, "#.#>%-'$)2'#- )-, 1#()1'$)2'#-
1 ! !
)*% 6$%, 2# (#./62%
$6+,#- $ <$,))
)(*#$(#/'( $2*)'-$
#
7+"#(
! )-, $2*%$$%$ +4
! D-#&'->
#
(*/ !0 9$56+)5+7"5
2"% .%(")-'()1 $2)2%
! %)2 .'(*#$(#/'( 1%7%13 J-% #8 2"%
'# (" #A$
)1 /*#+1%.$ &'2" 2"'$ 2=/% #8 )//*#)(" '$ 2# ("##$% 2"% .'(*#$(#/'( $()1% 2# +% (#-$',%*%,3 H12"#6>"
! !
9/)*= >( *$) .//# 3/6":/- (*$( (*") 9/)*&%"))$$6.%,
*%$ 2"%.$%17%$ )*% .612'1)=%* (#./#$'2% .)2%*')1$ A$%% E'>3 PB;
#+( (++ 2"%.*E&
2# +% "#.#>%-%#6$
9("%" #%" *("%"0&%" 0&/%*(>&%)"% *"75&%5 % #7) & E(.'( "75/%" +&'#+.5#*.&7 &0 *(" 5*%"55"5 #7)
$+$6)/
$26,= )-,$#- -+/)
("#$% 2"% +)$'( #+(
8'+*% ""#(6+-,5("+# $#& %"5$"'*.-"+6D
.)2*'? /)22%*- )$ .'(*#$(#/'(
5*%#.75 !0$()1%3-")5;
W#&%7%*; 2"'$ "=/#2"%$'$ '$ ) )
# *%$2*'(2'7%; $'-(% .'(*#$2*6(26*)1 #+$%*7)2'#-$ ")7% $"#&- 2")2 2"% F+*% '$ #-1= $1'>"21= ,%8#*.%,
/(")$("+#
.)'-$ %1)$2'(3 H/66+6= " % %!! #8 %)(" 1)=%*; &"'(" )*% -#2
.#*% ,%2)'1%, )-)1=$'$ =, '-7#17% 2"% (")*)(2%*'$2'($
! )!
+1%3 # % &! " #G$
'1 -#&; 2"% +%")7'#6* .#,%1 #8 2"% .'(*#$(#/'( $()1% ")$ *%.)'-%,
! ) ! /"%-#.%-#1#>'()1 A/#1=(*=$<
$ -%& 2=/% #8 .#,%1 8)11$ &'2"'- 2"% >%-%*)1 ()2%>#*= #8 .612'$()1% .#,%1$3 !"% (#-$2'262'7% %56)2'#-$
'22%- #-1= #- .'(*#$(#/'( $()1% )-, "#.#>%-'$)2'#- )-, 1#()1'$)2'#- %56)2'#-$ )*% 6$%, 2# (#./62%
)(*#$(#/'( $2*)'-$ ! )-, $2*%$$%$ ! D-#&'-> 2"% .%(")-'()1 $2)2% )2 .'(*#$(#/'( 1%7%13 J-% #8 2"%
# #
)1 /*#+1%.$ &'2" 2"'$ 2=/% #8 )//*#)(" '$ 2# ("##$% 2"% .'(*#$(#/'( $()1% 2# +% (#-$',%*%,3 H12"#6>"
*%$ 2"%.$%17%$ )*% .612'1)=%* (#./#$'2% .)2%*')1$ A$%% E'>3 PB; &% )$$6.%, 2"%. 2# +% "#.#>%-%#6$
$ $26,= )-, ("#$% 2"% +)$'( 8'+*% " .)2*'? /)22%*- )$ .'(*#$(#/'( $()1%3 W#&%7%*; 2"'$ "=/#2"%$'$ '$
# *%$2*'(2'7%; $'-(% .'(*#$2*6(26*)1 #+$%*7)2'#-$ ")7% $"#&- 2")2 2"% F+*% '$ #-1= $1'>"21= ,%8#*.%,
.)'-$ %1)$2'(3 H .#*% ,%2)'1%, )-)1=$'$ =, '-7#17% 2"% (")*)(2%*'$2'($ #8 %)(" 1)=%*; &"'(" )*% -#2
+1%3
'1 -#&; 2"% +%")7'#6* .#,%1 #8 2"% .'(*#$(#/'( $()1% ")$ *%.)'-%, /"%-#.%-#1#>'()1 A/#1=(*=$<
.#,%1B3 !"% %56)2'#-$ 6$%, 2# ># 8*#. #-% $()1% 2# )-#2"%* &%*% +)$%, #- 2"% $%18<(#-$'$2%-2 $("%.%
#8 '2$ ,%*'7)2'7%$3 !"%$% .#,%1$ )*% $)2'$8)(2#*= &"%- )//1'%, 2# .)2%*')1$ &'2" ) *)-,#. .'(*#<
6*% $6(" )$ .%2)1 /#1=(*=$2)1$; +62 )*% 6-$6'2)+1% 8#* #6* &#*D; +%()6$% #8 2"% (#./1%2%1= #*,%*%,
2*6(26*%; &"'(" '$ -#2 (#-$'$2%-2 &'2" 2"% $%18<(#-$'$2%-2 8*).%&#*D3
184
homogenization schemes
W computational homogenization
8" 80)0(9 :";<" =/'>,4/0 ? =,672." @0./,A* %77(" @04/" B&C1C" #DE F$GGGH EGI3EEG
6)1 ?
2#*,#/0+ /: # *../23#0$ ,/ ,41 )/.*) 81+4
X1+4 H examples:
? turbine rotor
A B
+, +-*-'. /,01, 23(4%$3- 5 2%6&7), 8-)3%9# :&&', 8-$3, ;<=>=, ?@A BCDDDE ADFGAAD
#$% &'% ()*+,#-.*,#/0 #+ ./0.10,2*,13 #0 * ,4#0 21$#/0 *2/503 ,41 -62178*,2#9 #0,12:*.1 *03 /52 -2+, 81+4 #+
/ ./*2+1 #0 ,4#+ *21* !* +#0$)1 1)1810,' ,/ /6,*#0 * +5;.#10,)< ,4#0 =)*+,#. 21$#/0 +5.4 *+ ,4*, /6+12>13 /0 @,I6 O6 H)' 5*41 8*#$8 ;'()'( 4('.6
1+41+ ?@ A *03 B% C4#+ 19=)*#0+ D4< ,41 >#+./=)*+,#. +,2*#0 #+ 503121+,#8*,13 *03 ,41 +,21++1+ />121+,#E
*,13% 185
F+ 19=1.,13@ ,41 -012 ,41 )/.*) 81+4@ ,41 61,,12 ,41 21+5),+@ :/2 6/,4 ,41 1)*+,#. *03 >#+./=)*+,#. *+=1.,+%
homogenization schemes
W computational homogenization
8" 80)0(9 :";<" =/'>,4/0 ? =,672." @0./,A* %77(" @04/" B&C1C" #DE F$GGGH EGI3EEG
6)1 ?
2#*,#/0+ /: # *../23#0$ ,/ ,41 )/.*) 81+4
X1+4 H examples:
? turbine rotor
A B
!" !#$#%& '"()" *+,-./+# 0 *.1234" 5#4+.67 822%" 5#/+" 9:;<;" =>? @ABBBC ?BDE??B
#$% &'% ()*+,#-.*,#/0 #+ ./0.10,2*,13 #0 * ,4#0 21$#/0 *2/503 ,41 -62178*,2#9 #0,12:*.1 *03 /52 -2+, 81+4 #+
/ ./*2+1 #0 ,4#+ *21* !* +#0$)1 1)1810,' ,/ /6,*#0 * +5;.#10,)< ,4#0 =)*+,#. 21$#/0 +5.4 *+ ,4*, /6+12>13 /0 5+:8 I8 J$,E&'-#$ ,1 %-2-'&(#0 E+$%,4'&$(+% $()&+* 1,) ("# 1,-) 2#$"#$8
1+41+ ?@ A *03 B% C4#+ 19=)*#0+ D4< ,41 >#+./=)*+,#. +,2*#0 #+ 503121+,#8*,13 *03 ,41 +,21++1+ />121+,#E
*,13% 186
F+ 19=1.,13@ ,41 -012 ,41 )/.*) 81+4@ ,41 61,,12 ,41 21+5),+@ :/2 6/,4 ,41 1)*+,#. *03 >#+./=)*+,#. *+=1.,+%
1#"3&-$5 ?(- 0"$1&-,- )*$4&%,*"! &%),-. ;@ 4)-#
homogenization schemes ("4#)5
?*/ ./$9 ./#-"#2 5+97,($("+# ") $ @@(/8(; !"2= FG= K5*/9$("5 -"$26$9 +4 6/"#4+65/- (,6."#/ 6+(+6=
.++ABB 5$)/ )*+C"#2 (*/ 4/$)"."%"(& +4 (*/ !01
9/(*+-= ># (*") )/5("+#' C/ -")5,)) $ 6/$% )(6,5(,6/D
?*") 2/+9/(6& ") 6/%$("3/%& $6."(6$6&' ./5$,)/ +4 $
$ (,6."#/ 6+(+6 ),.E/5(/- (+ 5/#(6"4,2$% %+$-"#2=
%$5A +4 -$($ +# (*/ -"9/#)"+#) +4 $ 6/$% 6+(+6
0#2"#/ 9$#,4$5(,6/6) $6/ 5+#)"-/6"#2 (*/ 6/;
9$-/ (*") C$&=
7%$5/9/#( +4 9/($%%"5 (,6."#/ 6+(+6) .& .%$-/-
J+$-"#2 ") 5/#(6"4,2$%= ?*/ )7//- +4 6+($("+#
"#2) C*+)/ .+6/ ") 6/"#4+65/- .& deformation
$ 5+97+)"(/'
5x $)
exaggerated 3$6"/) %"#/$6%& +3/6 ("9/= ?*/ *+9+2/#/+,) 7$6( +4
%%,)(6$(/- "# !"2= FG=
(*/ 6+(+6 ") 9$-/ +4 ("($#",9 $#- (*/ 6/"#4+65/;
?*") 6$")/) (*/ 76+.%/9 +4 *+C (+ 5+97,(/ (*")
9/#( ") 9$-/ +4 K"LM?" 5+97+)"(/ N(*/ 5+97+)"(/
(6,5(,6/' ),.E/5(/- (+ 5/#(6"4,2$% %+$-"#2' ($A"#2
9$(6"8 ") 9$-/ +4 (*/ )$9/ 9$(/6"$% $) (*/ 6/)( +4
*/ 6/"#4+65/9/#( 9"56+)(6,5(,6/ "#(+ $55+,#( "#
(*/ 6+(+6O= ?*/ 9/)* +4 (*/ ,#-/6%&"#2 9"56+;
-/($"%= ?*/ !01 9+-/% ") /)7/5"$%%& C/%% ),"(/- 4+6
)(6,5(,6/' )("%%!;>(*/
<*+5 ;C5 J,#-)) KLM%N %, ")
)$9/' ,(-2"3/#
-!. "2
"# &"%.*!+/
(*/ )$9/-!&%#+-.
:2,6/=O
187
*$(= ,*$-)5
homogenization schemes
computational homogenization
examples
!"#$%&'$ ! "#$%&'"($%& )*%(*+,- ).#*(/0+ 1&% "&2+0#3, ?4+ )#"50+)* "+*4&2 0+(2#3, *& *4+ 4&"&,+3#<+2 "&'
*4+ "+$4(3#$(0 %+)5&3)+ &1 4+*+%&,+3+&.) "(*+%#(0) (* 2.0# &1 ( 4+*+%&,+3+&.) "(*+%#(0 (550#+) *4+ %.0+ &1 "#:'
0(%,+ 2+1&%"(*#&3) (32 3&3'0#3+(% 4#)*&%- 2+5+32+3* *.%+)7 ?4+ &6+%(00 5%&5+%*- #) *4+3 $(0$.0(*+2 () (3 (6+%(,+
"(*+%#(0 /+4(6#&.% #) 5%+)+3*+27 84+3 .)#3, *4#) "#$%&' &6+% *4+ %+)5+$*#6+ 5%&5+%*#+) &1 *4+ $&3)*#*.+3*)> 9+#,4*+2
"($%& (55%&($4 9#*4#3 *4+ $&3*+:* &1 ;3#*+ +0+"+3* #"' 9#*4 *4+#% 6&0."+ 1%($*#&3)7 ?4#) (55%&($4 *(B+) #3*&
188
50+"+3*(*#&3 *4+%+ #) 3& 3++2 *& )5+$#1- *4+ 4&"&,+3#<+2 $&3)#2+%(*#&3 &30- &3+ "#$%&)*%.$*.%(0 $4(%($*+%#)*#$> *4+
0(*&(3%. (> 5!% A$".',#"+%'"# L -(.% &#"-% 3+'",- '%.$0%.
homogenization schemes ,-+%4'"+,(- %#%*%-+31 F9% $-.%>('*%. "-. .%>('*%. 4%6
(*%+',%3 (> +9% *"0'(*%39 "'% 39(J- ,- S,41 V1 M+ +9% #%>+
3,.% +9% 3+',& ,3 5=%. ,- "=,"# B9(',N(-+"#C .,'%0+,(-D +9%
.,3&#"0%*%-+ ,- +'"-3!%'3% B!%'+,0"#C .,'%0+,(- ,3 #%>+ >'%%1
M+ +9% ',49+ 3,.% +9% '(+"+,(- (> +9% 0'(33 3%0+,(- ,3 &'%6
30',)%.1 M3 &$'% )%-.,-4 ,3 0(-3,.%'%. +9% )%9"!,($' (>
+9% 3+',& ,3 $-,>('* ,- "=,"# .,'%0+,(- "-.D +9%'%>('%D "
3,-4#% #"E%' (> %#%*%-+3 (- +9% *"0'(#%!%# 3$>50%3 +(
3,*$#"+% +9% 3,+$"+,(-1
computational homogenization !"
I- +9,3 %="*&#% +J( 9%+%'(4%-%($3 *,0'(3+'$0+$'%3
0(-3,3+,-4 (> " 9(*(4%-%($3 *"+',= *"+%',"# J,+9 ?2 "-.
VUW !(#$*% >'"0+,(- (> !(,.3 "'% 3+$.,%.1 F( 4%-%'"+% "
'"-.(* .,3+',)$+,(- (> 0"!,+,%3 ,- +9% *"+',= J,+9 " &'%6
30',)%. !(#$*% >'"0+,(-D *"=,*$* .,"*%+%' (> 9(#%3 "-.
*,-,*$* .,3+"-0% )%+J%%- +J( -%,49)(',-4 9(#%3D >(' "
examples: porous aluminum +J( .,*%-3,(-"# :;<D +9% &'(0%.$'% >'(* X"## B?RR?C "-.
/*,+ B?RRLC 9"3 )%%- ".(&+%.1 F9% :;<3 $3%. ,- +9%
0"#0$#"+,(-3 "'% &'%3%-+%. ,- S,41 @1
computational homogenization
($ (+$ ,$)-")%*&.$ "- (+$ %$(+"# -") (+$ ,-+%4'"+,(- %#%*%-+31 F9% $-.%>('*%. "-. .%>('*%. 4%6 M+ +9% ',49+ 3,.% +9% '(+"+,(- (> +9% 0'(33 3%0+,(- ,3 &'%6
30',)%.1 M3 &$'% )%-.,-4 ,3 0(-3,.%'%. +9% )%9"!,($' (>
+9% 3+',& ,3 $-,>('* ,- "=,"# .,'%0+,(- "-.D +9%'%>('%D "
3,-4#% #"E%' (> %#%*%-+3 (- +9% *"0'(#%!%# 3$>50%3 +(
(*%+',%3 (> +9% *"0'(*%39 "'% 39(J- ,- S,41 V1 M+ +9% #%>+ 3,*$#"+% +9% 3,+$"+,(-1
I- +9,3 %="*&#% +J( 9%+%'(4%-%($3 *,0'(3+'$0+$'%3
3,.% +9% 3+',& ,3 5=%. ,- "=,"# B9(',N(-+"#C .,'%0+,(-D +9% '"-.(* .,3+',)$+,(- (> 0"!,+,%3 ,- +9% *"+',= J,+9 " &'%6
30',)%. !(#$*% >'"0+,(-D *"=,*$* .,"*%+%' (> 9(#%3 "-.
*,-,*$* .,3+"-0% )%+J%%- +J( -%,49)(',-4 9(#%3D >(' "
%*()16 %*($)1*0 +*' 3$$& #$'.)13$# 32 (+$ .,3&#"0%*%-+ ,- +'"-3!%'3% B!%'+,0"#C .,'%0+,(- ,3 #%>+ >'%%1
+J( .,*%-3,(-"# :;<D +9% &'(0%.$'% >'(* X"## B?RR?C "-.
/*,+ B?RRLC 9"3 )%%- ".(&+%.1 F9% :;<3 $3%. ,- +9%
0"#0$#"+,(-3 "'% &'%3%-+%. ,- S,41 @1
*'(1. %"#$0 1&1(1*002 #$4$0",$# 32 7"#&$) M+ +9% ',49+ 3,.% +9% '(+"+,(- (> +9% 0'(33 3%0+,(- ,3 &'%6
#$%& '() *& O-.%>('*%. ( "-. .%>('*%. * 0(-54$'"+,(-3 (> +9%
*"0'(*%39
:;<=>? !+$2 ,)","'$# * '()$'' #$,$&#$&( 30',)%.1 M3 &$'% )%-.,-4 ,3 0(-3,.%'%. +9% )%9"!,($' (>
-,+% %#%*%-+ .,30'%+,3"+,(-1
71 8(*&$+% +9% :;< !%'+%= >('0%3 ! ! ! ! ! ?! 2! @ "-. 3$)6
3+,+$+% +9%3% ,-+( <A1 B?7CD E,%#.,-4 +9% :;< "!%'"4%.
3+'%33 !:;< 1 F9,3 3+'%33 ,3 '%+$'-%. +( +9% *"0'(30(&,0
" #$'.)13$ (+$ '()*1& )*($ #$,$&#$&( 21$0# (> +9% *"0'(30(&,0 -(."# .,3&#"0%*%-+31
*0? 9:;;;*> +*4$ $%3$##$# (+1' 7"#&$)E 0(.% ,- QMFPMH B?RRGC1 S(' +9% *,0'(30(&,0 &"'+ (> +9%
309%*% +9% 0(**%'0,"##E "!",#")#% 5-,+% %#%*%-+ &"0T"4%
VUW !(#$*% >'"0+,(- (> !(,.3 "'% 3+$.,%.1 F( 4%-%'"+% " !"
QM:8 B?RRKC ,3 $3%.1
30$ 4$)'1"& "- (+$ %"#$0 $'(*301'+$# 32 30',)%. !(#$*% >'"0+,(-D *"=,*$* .,"*%+%' (> 9(#%3 "-.
%="*,-%.1 H(+9 +9% #%-4+9 "-. +9% 9%,49+ (> +9% 3"*&#% #$%& +() *& :;<3 $3%. ,- +9% 0"#0$#"+,(-3 J,+9 ?2W !(,.3 ( "-.
%A$"# U12 *D +9% +9,0T-%33 ,3 +"T%- ? *1 F9% *"0'(*%39 ,3 VUW !(,.3 *
@ ,)","'$# 32 7**1I$&' 9:;;:>? *,-,*$* .,3+"-0% )%+J%%- +J( -%,49)(',-4 9(#%3D >(' "
+J( .,*%-3,(-"# :;<D +9% &'(0%.$'% >'(* X"## B?RR?C "-.
$0 (+$ ("(*0 #$-")%*(1"& 1' #$."%,"'$# 1& *& /*,+ B?RRLC 9"3 )%%- ".(&+%.1 F9% :;<3 $3%. ,- +9%
0*'(1. ."&()135(1"&? !+$ $0*'(1. #$-")%*(1"& 0"#0$#"+,(-3 "'% &'%3%-+%. ,- S,41 @1
+$ '()$'' *..")#1&C (" *& 1'"()",1. &$"/
1"&'+1, 9*0'" '$$ D$.(? L?:>? !+$ #1''1,*(14$
#$%& '() *& O-.%>('*%. ( "-. .%>('*%. * 0(-54$'"+,(-3 (> +9%
*(1"& )*($ ($&'") !! 1' )$0*($# (" (+$ #$41*(")
*"0'(*%39
'()$'' ($&'") !" *..")#1&C (" (+$ C$&$)*01F$# "#$% &% U"%$&(E.5)4*(5)$ #1*C)*% )$'50(1&C -)"% (+$
A )50$ %1.)"/%*.)"
!"#$% &'()#%* ($+#,-%. ,- /%0+1 212 $3,-4 " 3%&"'"+% .*0.50*(1"&'
56 190
? !+$ %*($)1*0 ,*)*%$($)' "R *&# $ )$O$.( (+$ 4"1#' *( * .5)4*(5)$ "- :?P= %&: *&# *& *,,01$# %"%$&(
(+&$'' homogenization schemes
"- (+$ $0*'(1./("/,0*'(1. ()*&'1(1"& *&# (+$ $T5*0 (" H"V ' :R= N % 1& (+$ #$-")%$# %*.)"'()5.(5)$ *&#
& )*($ '$&'1(141(2@ )$',$.(14$02? !+$ '(*($ 4*)1*30$ # 1& (+)$$ #$-")%$#@ 1&1(1*002 1#$&(1.*0 Z[\' *( #1--$)$&(
"0' (+$ +*)#$&1&C *' (+$ )$'1'(*&.$ (" ,0*'(1. O"A@ 0".*(1"&' 1& (+$ %*.)"'()5.(5)$? \*.+ +"0$ *.(' *' * ,0*'(1.
+ 1' #$Q&$# 32 (+$ -"00"A1&C $4"05(1"& $T5*(1"& '()*1& ."&.$&()*(") *&# .*5'$' +1C+$) '()*1&' 1& (+$ Z[\
&%"!
#: % "#R & #computational
: #$ homogenization
"PV# (+*& "..5))1&C 1& (+$ +"%"C$&1F$# %*.)"'()5.(5)$? J& (+$
,)$'$&( .*0.50*(1"&' (+$ %*61%5% $--$.(14$ ,0*'(1. '()*1&
$ (+$ ."&'(*&(' # R *&# # : #$&"($ (+$ 0"A$) *&# 5,,$) 1& (+$ %*.)"'()5.(5)$ 1' *3"5( P=^@ A+$)$*' *( Z[\ 0$4$0
#' "- #@ )$',$.(14$02@ *&# % 1' * %*($)1*0 ."&'(*&(
examples: porous aluminum
"001&C (+$ )*($ "- A")B +*)#$&1&C? !+$ 1&($)&*0
(+1' '()*1& )$*.+$' VR^? J( 1' "341"5' -)"% (+$ #$-")%$#
C$"%$()2 "- (+$ +"0$' 1& ]1C? H (+*( (+$ Z[\ 1& (+$ 5,,$)
30$ "! )$,)$'$&(' (+$ $--$.(14$ ,0*'(1. '()*1&@ A+1.+ ,*)( "- (+$ 3$&#$# '()1, 1' '53I$.($# (" ($&'1"& *&# (+$
Q&$# 32 (+$ -"00"A1&C
&!!!!!!!!!!!!!!!!!!!!!!!!!
! $4"05(1"& $T5*(1"&
P!S"!! ! !! #? 0(*&(3%. (> 5!% A$".',#"+%'"# L -(.% &#"-% 3+'",- '%.$0%.
5$' ."&.$)&1&C (+$ &5%$)1.*0 (1%$ 1&($C)*(1"& "- (+1' ,-+%4'"+,(- %#%*%-+31 F9% $-.%>('*%. "-. .%>('*%. 4%6
1(5(14$ %"#$0 *)$ #1'.5''$# 1& #$(*10 1& D%1( $( *0? (*%+',%3 (> +9% *"0'(*%39 "'% 39(J- ,- S,41 V1 M+ +9% #%>+
*>? 3,.% +9% 3+',& ,3 5=%. ,- "=,"# B9(',N(-+"#C .,'%0+,(-D +9%
.,3&#"0%*%-+ ,- +'"-3!%'3% B!%'+,0"#C .,'%0+,(- ,3 #%>+ >'%%1
(+$ ,)$'$&( .*0.50*(1"&' (+$ %*($)1*0 ,*)*%$($)' -") M+ +9% ',49+ 3,.% +9% '(+"+,(- (> +9% 0'(33 3%0+,(- ,3 &'%6
*0$# *05%1&5% WW :R=R #$($)%1&$# 32 4*& #$) W* 30',)%.1 M3 &$'% )%-.,-4 ,3 0(-3,.%'%. +9% )%9"!,($' (>
> +*4$ 3$$& 5'$#X $0*'(1. ,*)*%$($)'Y '+$*) %"#505' +9% 3+',& ,3 $-,>('* ,- "=,"# .,'%0+,(- "-.D +9%'%>('%D "
P"H ' :RL U8*@ 350B %"#505' ' ! <"V ' :RL U8* *&# 3,-4#% #"E%' (> %#%*%-+3 (- +9% *"0'(#%!%# 3$>50%3 +(
3,*$#"+% +9% 3,+$"+,(-1
'1(2 ,*)*%$($)'Y "R ! :RV '&P @ % ! :S"V@ $ ! S"L@ I- +9,3 %="*&#% +J( 9%+%'(4%-%($3 *,0'(3+'$0+$'%3
!"
V:"L U8*@ #: ! :<R U8*? 0(-3,3+,-4 (> " 9(*(4%-%($3 *"+',= *"+%',"# J,+9 ?2 "-.
VUW !(#$*% >'"0+,(- (> !(,.3 "'% 3+$.,%.1 F( 4%-%'"+% "
'"-.(* .,3+',)$+,(- (> 0"!,+,%3 ,- +9% *"+',= J,+9 " &'%6
(' 30',)%. !(#$*% >'"0+,(-D *"=,*$* .,"*%+%' (> 9(#%3 "-.
*,-,*$* .,3+"-0% )%+J%%- +J( -%,49)(',-4 9(#%3D >(' "
"/%*.)" .*0.50*(1"&' -") * +$($)"C$&$"5' '()5.(5)$@ +J( .,*%-3,(-"# :;<D +9% &'(0%.$'% >'(* X"## B?RR?C "-.
'$&($# 32 (+$ Z[\' '+"A& 1& ]1C? L +*4$ 3$$& .*))1$# /*,+ B?RRLC 9"3 )%%- ".(&+%.1 F9% :;<3 $3%. ,- +9%
1%50*(1&C ,5)$ 3$&C *( * ,)$'.)13$# %"%$&( )*($ 0"#0$#"+,(-3 "'% &'%3%-+%. ,- S,41 @1
(" = ' :R= N % '&: ? J& ]1C? = (+$ %"%$&(E.5)4*(5)$ "#$% '% M"&("5) ,0"(' "- (+$ $--$.(14$ ,0*'(1. '()*1& 1& (+$
*% )$'50(1&C -)"% (+$ %1.)"/%*.)" *,,)"*.+ 1' #$-")%$# %*.)"'()5.(5)$ *&# 1& (+)$$ #$-")%$# Z[\'@
#$%& '() *& O-.%>('*%. ( "-. .%>('*%. * 0(-54$'"+,(-3 (> +9%
&($#? !" C14$ *&*"0'(*%39
1%,)$''1"& "- (+$ 1&O5$&.$ "- (+$ ."))$',"&C (" #1--$)$&( ,"1&(' "- (+$ %*.)"'()5.(5)$
• random porosity
!"#$ %&' ($ !"# +$&' *%( '*)( &2 ,(-,(1(%&$%= . ,(=6)., 1&,6/&6,(2
.%8 FL ,.%8*0 !"#1 (
!"#$ %&' ($ !"# +$&' *%( '*)( &2 ,(-,(1(%&$%= . ,(=6)., 1&,6/&6,(2
.%8 FL ,.%8*0 !"#1 (
192
#(2 E*'%(, 4%/)0/%2* $#/)'%7#$ %( ?#)29 K %/ . "#+3&1+ 200+9
($ %(2'%(/%) /0&2#(%(,
homogenization
%/ 3.3#' 20 #(2#' %( schemes
L/ .( #A.-3+# 2"# %(R1#()# 0& 2"# /3.2%.+ .''.(,#-#(2
0& 2"# -%)'0/2'1)21'# 0( 2"# 04#'.++ -0-#(2J)1'4.21'#
%/ .$$'#//#$ 20 2"# '#/30(/# 0& 2"# 40%$#$ -.2#'%.+ 1($#' 31'# 7#($%(, %/ !"# $% &'( )*+(, -.,& &* /*0-,(11$*%2 +'$)( &'( !"# $% /*%&.$%$%= *%( 1$%=)( '*)( BJ$=: K.D: ;* .3*$8 . 8$1/611$*%
&'( 3$/$%$&4 *5 &'( %(6&,.) .7$1 $1 )*.8(8 /*%1$8(,.9)4 *% &'( 0$%$060 ,(I6$,(8 1$>( B$%/)68$%= &'( %609(, *5
2 #2 .+9 6:@@@.<9 /21$%#$9 !"# 7#".4%01' 0& 2"# -.2'%A -.2#'%.+ %/ $#/)'%7#$
0$)8(, &'.% &'( *&'(, !"#1: ;'$1 /*%<,01 &'( /*%/)61$*%
&'.& &'( 0(&'*8 ,(.)$1&$/.))4 8(1/,$9(1 &'( (3*)6&$*% *5 &'(
0$/,*1&,6/&6,(: ;'( 0(&'*8 0.4 9( .--)$(8 %*& *%)4 .1 .%
.)&(,%.&$3( 5*, &'( '*0*=(%$>.&$*% *5 &'( 0.&(,$.) 9('.?
3*$81D *5 . ,.%8*0 !"#2 FL 8$55(,(%& ,.%8*0 !"#1 '.3(
9((% =(%(,.&(8 BJ$=: K9D: ;'( .3(,.=(8 9('.3$*6, *5 &'(1(
FL !"#1 $1 (7-(/&(8 &* 9( ,(-,(1(%&.&$3( 5*, &'( ,(.) ,.%8*0
1&,6/&6,( +$&' . =$3(% 3*)60( 5,./&$*% *5 '(&(,*=(%($&$(1:
&0' 1(%.A%.+ 2#(/%0( 0& 7* 2"# #+./20B4%/)0B3+./2%) -0$#+ >%2" ".'$#(%(, 012+%(#$ 3$*6,2 96& .)1* &* 8(1/,$9( 0(/'.%$/.) &,.%15*,0.&$*%1 *5
&'( 0$/,*1&,6/&6,(:
@% &'( 1691(I6(%& 1(/&$*%1 &'( /*0-.,$1*% $1 -(,5*,0(8
5*, &',(( 8$55(,(%& /*%1&$&6&$3( 0*8()1 *5 &'( 0.&,$7
!:
! .%8 ().1&*?3$1/*?-).1&$/ +$&' $%&,$%1$/ 1*5&(%$%=:
#2#'/ .'# 2"# /.-# ./ 2"#/# 1/#$ %( ?#)29 Q9 0./,*1/*-$/ 9('.3$*6,2 $% &'$1 1(/&$*% &'( *3(,.)) ,(?
1-*%1( *5 . -(,5(/&)4 ,(=6)., 1&,6/&6,( B/69$/ 1&./C$%= *5
,.%8*0 .%8 &'( ,(=6)., 0$/,*1&,6/&6,(1 .,( /*0-.,(8:
examples: porous
#/30(/# 0& 2"# '#,1+.' &'0- 2"# &1++ -%)'0B-.)'0 .(.+*/%/ 0& 31'# 7#($%(, 0& 2"#
(,(8: ;'( ,(=6)., 1&,6/&6,( $1 0*8()(8 94 . 1I6.,( !"# &'( ,(=6)., .%8 ,.%8*0 1&,6/&6,(1 $1 /.,,$(8 *6& 5*, &'(
!!
.%8 FL ,.%8*0 !"#1 (
!"#$ %&' ($ !"# +$&' *%( '*)( &2 ,(-,(1(%&$%= . ,(=6)., 1&,6/&6,(2
,.%8*0 1&,6/&6,(:
'(&(,*=(%($&$(1D $1 /*0-.,(8 +$&' &'( *3(,.)) ,(1-*%1( *5 .
1-*%1( *5 . -(,5(/&)4 ,(=6)., 1&,6/&6,( B/69$/ 1&./C$%= *5
0./,*1/*-$/ 9('.3$*6,2 $% &'$1 1(/&$*% &'( *3(,.)) ,(?
A6(%/( *5 &'( ,.%8*0%(11 *5 &'( 0$/,*1&,6/&6,( *% &'(
0.&(,$.)1: ;* /*%&,$96&( &* &'( 8$1/611$*% .9*6& &'( $%?
$0-./& *% &'( 0./,*1/*-$/ ,(1-*%1( *5 '(&(,*=(%(*61
.,,.%=(0(%& *5 &'( 0$/,*1&,6/&6,( 0.4 '.3( .% $0-*,&.%&
@% &'( $%&,*86/&$*% $& '.1 9((% -*$%&(8 *6& &'.& &'( 1-.&$.)
"#$%&'( )#(*%* ('+,-. (#/(#*#+0'01-+
E 0.&(,$.) +$&' FGH 3*)60( 5,./&$*% *5 3*$81 $1 /*%1$8?
!"#$ %&' ($ !"# +$&' *%( '*)( &2 ,(-,(1(%&$%= . ,(=6)., 1&,6/&6,(2
.%8 FL ,.%8*0 !"#1 (
0 1 0 1
X x
• rotation matrix relating @ Y A = Q@ y A
reference to crystal frame
Z z
infinitesimal volume
of orientation space
194
homogenization schemes
orientation space
1
dQ = 2
d'1 dcos d'2
8⇡ '1
sin '2
= 2
d'1 d d'2
8⇡
195
homogenization schemes
1
dQ = 2
d'1 dcos d'2
8⇡ '1
sin '2
= 2
d'1 d d'2
8⇡
.Z
discrete volume of
orientation space
f i = ⌫i dQ
boxi
196
homogenization schemes
i i
i f sin
p(Q ) =
max(f sin )
197
homogenization schemes
i=1
n*
n* == 010 1
2
0 1
3 1
3 2 2 0
3 1
1 0
Eisenlohr, P., & Roters, F. (2008). Selecting sets of discrete orientations for accurate
texture reconstruction. Computational Materials Science, 42(4), 670–678.
198
homogenization schemes
• systematic over-prediction
at (relatively) low numbers
of discrete orientations
199
homogenization schemes
(! )*+
of discrete orientations hybrid
!
• blend of probabilistic and 10-1
deterministic schemes
(“hybrid”)
10-4 10-2 100 102
N*/N
200
finite element method
purpose
• an (irregular) domain
201
finite element method
classification of PDEs
independent dependent
variables variable
202
finite element method
203
finite element method
governing PDE
u satisfies PDE
everywhere
• strong form F =0
u satisfies PDE on
average in domain
Z
• weak form ⌘F = 0
⌦
function space for u has
infinite dimension…
204
finite element method
• “triangulation” of domain
• low-order polynomial in
each element
p = a1 node
+ a2 x + a3 y element
+ a4 xx + a5 xy + a6 yy
+ ...
205
finite element method
nodal
value function
• function determined by
nodal values
node
element
206
finite element method
N
X
• globally continuous u(x) = 10 ui Ni (x)
14
5
i
11 15
6
2
16
12 18
• uniquely defined by set of nodal 3 7
values 1
{ui }
8 17
4
13
Figure 1.8: The graph of a nodal basis function: it looks like a207
camping tent.
finite element method
Z
• nodal basis functions
as test function
Ni F = 0
system of N
equations to solve
for unknown nodal
degrees of freedom
208
finite element method
1.1 illustrates an idealization of a taut wire. The wire is under prestress by the fo
d to be uniform along the length of the wire. The left-hand end is immovably fixed, wh
and end is held in a fixture which can slide perpendicularly to the axis of the wire (occas
example problem: taut wire
to as a “roller”). A transverse force FL is applied at the movable end. In addition
some distributed force q (in physical units of force per unit length) acting along the
tance gravity). The transverse displacement is a function of both the axial coordinate
e t, w = w(x, t) . The transverse displacement is assumed to be very small compared
of the wire. The deformation in Figure 1.1 is highly magnified in order to be apparen
alance equation
P. Krysl, Thermal and stress analysis with the finite element method, Pressure Cooker Press, 2010
e a segment of length ∆x of the wire (see Fig. 1.2). The forces acting on the 209
segment
may be some distributed force q (in physical units of force per unit length
finite element method
(for instance gravity). The transverse displacement is a function of both t
the time t, w = w(x, t) . The transverse displacement is assumed to be ve
length of the wire. The deformation in Figure 1.1 is highly magnified in o
2 Thermal and Stress Analysis with the FEM
governing equation
′′ ∂ 2 w(x)
w (x) = ,
∂x2
and we equate the resultant of the vertical forces to the inertial force (Newton’s la
a balance equation for the taut wire
P w′′ + q = µẅ ,
∂2w
Fig. 1.1. Schematic of=taut
where ẅ
∂t2
wire
is the acceleration.
P w00 + q = µẅ
1.2 Balance equation
We take a segment of length ∆x of the wire (see Fig. 1.2). The forces acti
Fig. 1.2. The forces acting on a segment of the taut wire
prestressing forces in either cross-section and the resultant of the distribu
the deflection is very small compared to the span, w ≪ L, and we also ass
wire is very small, w′ = ∂w/∂x ≪ 1. These geometrical features are introdu
1.3 horizontal
the forces. In the Boundary
P. Krysl,
conditions
Thermal and stressdirection weelement
analysis with the finite havemethod,
justPressure
the Cooker
twoPress,
prestressing
2010 fo
and hence theyThe cancel.
functionIn w the verticalthedirection
that describes we add takes
transverse deflection up the components
two arguments, x, an
210
on a rectangular domain shown in Fig. 1.3: 0 ≤ x ≤ L, and 0 ≤ t ≤ t̄. The deflectio
may be some distributed
ment, force q (in physical units of force per unit length
finite element method
(for instance gravity). The transverse displacement is a function of both t
w(0, t) = w̄0 (t) .
the time t, w = w(x, t) . The transverse displacement is assumed to be ve
As we shall find out, there is a good reason why this kind of conditio
length of the wire. The deformation
essential boundaryin Figure 1.1 is highly magnified in o
condition.
boundary condition At the other end the boundary condition is of a different nature. It is
as we have to derive it. Again, we take a short section of the wire of leng
time there are terms that are multiplied by ∆x, but there are also other
latter survive when we make ∆x go to zero
∂w
−P (L, t) + FL (t) = 0 .
∂x
w(0, t) = w̄This
(t)boundary condition is simply the balance of forces at the end of the
0 this kind are called natural boundary conditions.
of
Fig. 1.1. Schematic of taut wire
“essential”
boundary condition
initial condition
212
finite element method
Z L
!
• minimizing the residual rB (x) dx = 0
0
Z L
• weighting the residual by test !
⌘(x) rB (x) dx = 0
function 0
Z L
• use multiple test functions with !
⌘j (x) rB (x) dx = 0
limited support 0
213
finite element method
Substituting for the balance residual, we get two terms
2.5 Shifting
! L derivatives ! L ! L
shifting
2.5ηjShiftingderivatives
(x)r (x) dx =for piecewise
derivatives η (x)P w linear
′′
(x) dxapproximation
+ ηj (x)q(x) dx .
B j
0
Substituting for the balance 0residual, we get two terms 0
! LSubstituting
Integration by parts for the
may ! balance
Lbe appliedresidual,to the wefirst
get
! L
two terms
term on the right-hand sid
! L′B (x) dx =
ηj (x)r η j !
(x)P L w ′′
(x) dx + η j ! L
(x)q(x) dx .
′ ′ ′ ′′
0(ηj P w ) η=(x)r ηj P w (x) +dx
0 ηj=P w η (x)P w′′ (x) dx 0 + ηj (x)q(x) dx .
j B j
0 0 0
Integration
and integratingby parts frommay 0 beto Lapplied
we obtain to the forfirst
the term
left-handon the sideright-hand side
Integration by parts may be applied to the first term on the right-hand
! ′ ′ ′
(ηj P w ) = ηj ′P w + ηj P w L
L ′ ′′
! Land integrating
Therefore we see that fromwe 0canto replace
L we obtain the term for the
with left-hand
the second side derivatives
! ′ ′ ′ L ′ ′
(η
! Lj L P w ) dx = [η j P w ] 0 = η j (L)P w (L) − η j ! L (0)
(0)P w
′ ′ ′ L ′ ′
0 (η′′j P w ) dx = [ηj ′P w ]0 = ηj (L)P′ w (L) − ηj (0)P ′ w
′ (0)
η P w dx = η (L)P w (L) − η (0)P w (0) −
j0 j j j η P w dx .
Therefore
0 we see that we can replace the term with the0 second derivatives as
! Therefore we see that we can replace the
ThisL does the trick: the second derivatives are gone term
! with the second derivat
L from the trial functi
ηj P! wL′′ dx =and
first derivatives, ηj (L)P w′ (L) − ηjlinear
the piecewise ′
(0)P wfunctions
(0) − ηj ′!Pwe
that ′
L spoke
w dx . about ar
0 ηj P w′′ dx = ηj (L)P w′ (L) − ηj (0)P w′ (0)0 − ηj ′ P w′ 214
dx .
finite element method
215
finite element method will need to put up with an imbalance (residual) r
F
is applied
natural boundary condition
rF = −P w′ (L) + FL .
ηjWe
(L)rmay
F = take
η j
a list
(L)(−P from
w ′
(L) the
+ Fbook
L) = 0
written for t
function and integrate. In this case integration
Therewhere
are two
rF ways in which
“lives”. Since we thishave
could happen:
taken ηj (x) eit
for
we said above
here, is in general
evaluated on the not going to
boundary. happen),the
Therefore or
however:
mayallbeitwritten
means is asthat the force FL does not pl
not what we want. So we cannot keep the natural b
′
η (L)r
own. Overj the years,
F = η (L)(−P w
jthe following (L)clever
+ FL )manipulat
216
finite element method
Z L
• combine nj rB dx + nj (L)rF = nj (L)P w0 (L) nj (0)P w0 (0)
both 0
Z L Z L
residuals
n0j P w0 dx + nj q dx
0 0
+ nj (L)( P w0 (L) + FL )
= nj (0)P w0 (0)
Z L Z L
n0j P w0 dx + nj q dx
0 0
force to zero
+ nj (L)FL
=0
217
i=1 i
w(x) =also called
Ni (x)w ,
i where
degrees of freedom.
the wi ’s are the Thecoefficients
Ni (x)’s areofknown (suitably
the linear combic
finite element method where the the wi ’snumber
i=1 are theof coefficients
alsotermscalled inofthe
N degrees thetrial linear
of freedom.combination
function The Ni(real
matches numbers)
the number
(x)’s are known of
also called degreesthe of freedom.
the number of Theof Ni (x)’sN are known be (suitably chosen) bas
where the because
wi ’s are thenumber coefficients of terms
unknowns
the linear in combination
needs theto trial function
matched to
(real matches
the numb
numbers) th
the number ofSubstituting
terms N because in the intotrialthe
(2.14),function
number ofmatches
unknowns the needs
number to of
be the test f
matched
also called degrees of freedom. The Nwe i
obtain
(x)’s are known (suitably chosen) bas
because the number of unknowns needs to be matched to the number of equa
the number of terms N in"Substituting the trial Nfunction into (2.14),
matches
"
wethe
obtain
number of the test f
stiffness matrix and Substituting
loadthe vector into (2.14), Lwe obtain ! L
because number
ηj (L)Fof unknowns ηj ′ P needs NiL′to
" wibe dxmatched ηj qtodxthe="number
0,L j = of 1, equ
L− !+
N ..., N
Substituting" into (2.14),
! N 0η we obtain− " Lηj ′ P 0 ′ w dx +
L j (L)FLi=1 N i i ηj q dx = 0,
′ ′
ηj (L)FL − " ηj P Ni wi dx + 0" ηj q i=1 dx = 0, j = 1,0..., N ,
N
which0may L
′
be! simplified
i=1 ′
to 0 L
• residuals ηj (L)FL − ηj Pwhich N# imaywi dxbe +simplified ηj q$dxto = 0, j = 1, ..., N ,
0
which may be simplified !
to
i=1 N " L 0 " L
′ N # ′ " $ "
ηj (L)F # L − η P
j !i
$ N dx L w i + η j q dx = 0,L j = 1,
which may be simplified
! N " L to i=1 ηj (L)F 0 L− " L ηj ′ P Ni ′0 dx wi + ηj q dx =
′ ′
ηj (L)FL − # " η j P Ni dx $ w i=1i + " 00 ηj q dx = 0, j = 1,0..., N
! N η0jL(x = 0) = 0, ηj ∈ C0 L, j = 1, ..., N , w(x = 0) = w̄
i=1 0
ηj (L)FL − ηj ′ P Ni ′ dx ηj (xw=i + 0) = 0,ηj qηdx j ∈= C0, , j = 1, ..., N N, w
With 0
=the
ηj (x i=1 0) definitions
=0 0, ηj ∈ C , j = 1,0..., N , w(x = 0) = w̄0 , w ∈ C
" LWith the definitions 0
η (x
With the definitions
j = 0) = 0, ′ η j ∈ ′ " ,
C j = 1, ..., N , w(x = 0) = w̄0 , w ∈ C
• stiffness matrix Kji = ηj P Ni dxL,
" L 0 K = η ′
P N ′
dx ,
With the definitions ji j i
Kji = "andηj ′ P Ni ′ dx , 0
0L
Kji = ηj ′ P Ni ′ dx and, " L
and 0 L = η (L)F + ηj q dx " L
j j L
• load vector " L Lj = 0ηj (L)FL + ηj q dx
and 0
Lj = ηjwe(L)F L +
may write η j q dx
"0 L (2.18) in the form
Lj = ηj (L)F + weη may
q dx write (2.18) in the form
N L j
we may write ! (2.18) in
0 the form
Kji wi = L ! N, j = 1, ..., N
XN j
we may
!N write i=1
(2.18) in the form Kji wi = Lj , j = 1, ..., N
• algebraic system K
Kji wi ji
w = L
= Lij , j j= 1,
, i=1
j
..., N
= 1, . . . , N
!N
i=1 i=1
Kji wi = Lj , j = 1, ..., N
i=1 218
spectral method
outline
• background and
opportunities
• methodology
• exemplary results
• limitations
219
spectral method
background
macromechanics
• complex geometry
220
spectral method
background
micromechanics
• periodic geometry
• shape-conforming mesh
• regular grid
221
spectral method
opportunities
222
spectral method
methodology
historical development
• Moulinec, H., & Suquet, P. (1994).
A fast numerical method for computing the linear and nonlinear properties of composites.
C. R. Acad. Sci. Paris Ser. II, 318, 1417–1423.
• Moulinec, H., & Suquet, P. (1998).
A numerical method for computing the overall response of nonlinear composites with complex
microstructure.
Computer Methods in Applied Mechanics and Engineering, 157(1-2), 69–94.
• Lebensohn, R. A. (2001).
N-site modeling of a 3D viscoplastic polycrystal using Fast Fourier Transform.
Acta Materialia, 49(14), 2723–2737.
• Lahellec, N., Michel, J. C., Moulinec, H., & Suquet, P. (2001).
Analysis of inhomogeneous materials at large strains using fast Fourier transforms.
In C. Miehe (Ed.), Proc. IUTAM Symposium (pp. 247–258). Solid Mechanics and its Applications, 108. Dordrecht, The
Netherlands: Kluwer Academic Publishers.
• Lebensohn, R. A., Kanjarla, A. K., & Eisenlohr, P. (2012).
An elasto-viscoplastic formulation based on fast Fourier transforms for the prediction of
micromechanical fields in polycrystalline materials.
Int. J. Plast., 32-33, 59–69.
• Eisenlohr, P., Diehl, M., Lebensohn, R. A., & Roters, F. (2013).
A spectral method solution to crystal elasto-viscoplasticity at finite strains.
Int. J. Plast., 46, 37–53.
223
spectral method
methodology
reference coordinates
e
y = Fx + w
deformed coordinates
@y e
@w
=F +
@x @x
e
F =F +F
224
spectral method
methodology
polarization
field
225
spectral method
methodology
static equilibrium
Z n o
1 i k·x
• transformation 0= 3
e A [ (k) ⌦ i k] + ⌧ (k) i k dk
(2⇡)
into Fourier space ⇣n o ⌘
= F 1 A [ (k) ⌦ i k] + ⌧ (k) i k
• equilibrium n o
0 = A [ (k) ⌦ i k] + ⌧ (k) i k
fulfilled if
226
spectral method
methodology
static equilibrium
n o
• equilibrium 0 = A [ (k) ⌦ i k] + ⌧ (k) i k
condition
227
spectral method
methodology
(
1
A(k) ⌧ (k) i k if k 6= 0
(k) =
(0) if k = 0
228
spectral method
methodology
(
1
A(k) ⌧ (k) i k if k 6= 0
(k) =
(0) if k = 0
(
1
A(k) ⌧ (k) (k ⌦ k)
F(k) =
F(0)
(
(k) ⌧ (k) if k 6= 0
=
F if k = 0
229
spectral method
methodology
iterative algorithm
⌧ =P AF
(
(k) ⌧ (k) if k 6= 0
F(k) =
F if k = 0
h i
e
F(x) = (x) ⇤ P(F(x)) AF(x)
= (x) ⇤ P(F(x)) + (x) ⇤ [AF(x)]
| {z }
= (x) ⇤ P(F(x)) + e
F(x)
230
spectral method
methodology
iterative algorithm
• fix-point iteration e
F(x) = e
(x) ⇤ P(F(x)) + F(x)
( !
1
(k) {P(k)}n if k 6= 0
{F(x)}n+1 = {F(x)}n F
F n
{F BC }n+1 if k = 0
⇢
@F
• boundary {F BC }n+1 = F 0
+ ḞBC t P n
PBC
@P
condition n
231
spectral method
exemplary results
• load case
232
spectral method
exemplary results
55
• mesh 163
convergence of 323
643
volume-average 163
50
response
Pyz / MPa
323–2563
c t r a l
FEM spe d
et h o
m
45
40
0 0.05 0.1 0.15 0.2
Fyz
233
spectral method
Author's personal copy
comparison between
FEM and spectral method
163 crystallites
323 2563
643 1283
(a) Finite element method simulation (b) Spectral method simulation
Fig. 6. Local shear deformation gradient F yz at average shear deformation of F yz ¼ 0:2 mapped onto deformed configuration. Light arrow: intense she
emanating from close to a horizontal grain boundary into neighboring grains. Dark arrow: geometry evolution in FEM in contrast to essentially consta
geometry for spectral method.
234
3 3
spectral method
Author's personal copy
comparison between
FEM and spectral method
163 crystallites
323 2563
643 1283
(a) Finite element method simulation (b) Spectral method simulation
Fig. 7. Local first PIOLA–KIRCHHOFF shear stress P yz at average shear deformation of F yz ¼ 0:2 mapped onto the deformed configuration. Arrows highlight tw
examples of low-stress grains within the aggregate that are captured at all resolutions for the spectral method but emerge for the FEM only at high
resolution.
235
spectral method
exemplary results
resolution
10 100
comparison between 108
cpu time / s
r
ea
lin
104
spectral
method
102
104 106 108
elements or grid points
236
spectral method
exemplary results
a r t i c l e i n f o a b s t r a c t
Article history: A viscoplastic approach using the Fast Fourier Transform (FFT) method for obtaining local
Received 9 October 2009 mechanical response is utilized to study microstructure–property relationships in compos-
Received in final revised form 31 August 2010 ite materials. Specifically, three-dimensional, two-phase digital materials containing iso-
Available online 15 September 2010
tropically coarsened particles surrounded by a matrix phase, generated through a Kinetic
Monte Carlo Potts model for Ostwald ripening, are used as instantiations in order to calcu-
Keywords: late the stress and strain-rate fields under uniaxial tension. The effects of the morphology
Micromechanical modeling
of the matrix phase, the volume fraction and the contiguity of particles, and the polycrys-
Composite materials
Stress and strain-rate fields
tallinity of matrix phase, on the stress and strain-rate fields under uniaxial tension are
Viscoplasticity examined. It is found that the first moments of the stress and strain-rate fields have a dif- 237
ferent dependence on the particle volume fraction and the particle contiguity from their
spectral method
exemplary results
• percolating vs
granular matrix
• stress
• strain rate
Fig. 3. Two contrasting microstructures used for instantiation of the simulations: (a) a microstructure from the coarsening simulation with particle volume
fraction of !0.6; (b) a modified single-phase polycrystal (‘‘disordered” microstructure) wherein about 40 vol% of grains have the properties of the matrix
phase. (c) and (d) are the corresponding stress fields, and (e) and (f) are the strain-rate fields from the simulation of uniaxial tension in the coarsened and
modified polycrystal microstructures, respectively (each of stress and strain-rate scale bars is shown only once). Note the drastic difference in morphologies
of two microstructures and its effect on stress and strain-rate fields. The arrows indicate the direction of applied uniaxial tension.
238
spectral method
718 S.-B. Lee et al. / International Journal of Plasticity 27 (2011) 707–727
exemplary results
• different
contiguity
• stress
• strain rate
Fig. 7. Cross-sections of microstructures having (a) high particle contiguity and (b) low particle contiguity in Fig. 6 under uniaxial tension and the
corresponding stress fields ((c) and (d)), and strain-rate fields ((e) and (f)) (each of stress and strain-rate scale bars is shown only once).
239
in the BCC particles with increasing particle volume fraction. At high volume fractions, the distribution tends to be towards
that calculated for the BCC polycrystal. As the particle volume fraction increases, the average stress of the particles increases
spectral method
as noticed before, which results in the shift of the distribution curve to the right. At the microstructural scale, as the particle
contiguity increases with increasing volume fraction, particles in soft orientations are no longer shielded by the softer matrix
phase and load is transmitted to particles in harder orientations (i.e., at higher stresses). This provides a simple explanation
for the changes in skewness from left to right as the particle volume fraction increases.
However, the distribution of Taylor factors from the 643 isolated BCC voxels is quite different from that of the FFT sim-
ulation on the same voxels in a 64 " 64 " 64 simulation domain. This is because of the effect from the neighbors on the
stress and strain-rate state for each voxel during the FFT simulation. Note that, as n increases, the results from the BCC poly-
exemplary results
crystal case evolve toward to match the distribution of Taylor factors when those 643 voxels are isolated with no neighboring
interactions, such that the frequencies over the bins with both large and small stress values become higher while those over
the intermediate bins get lower.
Fig. 10. Average strain-rate and its standard deviation of both particles and matrix phase in the microstructures from coarsening simulation with particle
volume fraction of !0.8 as a function of contiguity of particles. Note that the standard deviation of strain-rate decreases strongly in a linear fashion in the
matrix with increasing particle contiguity.
240
spectral method
limitations solution
241
announcements
homework
242
announcements
file transfer
• copy
• *.geom files of 20 and 40 grain structure
• material.config
• loadcase file
• *.spectralOut binary result file
243
announcements
file transfer
• copy
• *.geom files of equiaxed and sliced structures
• material.config
• loadcase file
• *.spectralOut files of all five simulations
244
spectral method for large property contrast
• rate of convergence
proportional to contrast
factor
245
spectral method for large property contrast
microstructure
246
spectral method for large property contrast
247
spectral method for large property contrast
248
kinetics
strain energy
density
functional
W
P(x) = = f (x, F, Ḟ, ⇠)
F(x)
249
static equilibrium
inv. Fourier
transform
1
min W =) Div P(x) = F [P(k) i k] = 0
c[ (k)] := P(k) i k = 0
F
residual
body force
field
250
static equilibrium
251
static equilibrium
reference
stiffness
252
static equilibrium
h i
c 1 c
P F [ (k)] ⌦ i k = A(k)
1
P(k) i k ⌦ i k = 0 8 k 6= 0
deformation
deformation map
in
gradient
reference
in
reference
material
material
cdirect [F(k)] := (k) P(k) = 0 8 k 6= 0
F
253
static equilibrium
Lagrange
L[F, , ⇤] = W multiplier
Z
+ F(x)⇤(x) · [Grad (x) F(x)] dx
B0
Z
1
+ [Grad (x) F(x)] · A [Grad (x) F(x)] dx
2 B0
penalty
254
static equilibrium
L n o
= P(x) F(x)⇤(x) + A F(x) Grad (x) = 0,
F(x)
L h n oi
= Div F(x)⇤(x) A F(x) Grad (x) = 0,
(x)
L
= Grad (x) F(x) = 0.
⇤(x)
255
static equilibrium
n o
P(k) ⇤R (k) + A F(k) (k) ⌦ i k = 0
n o
1
(k) = A(k) AF(k) ⇤R (k) i k
(k) ⌦ i k F(k) = 0
256
static equilibrium
saddle point
8 n n oo9
>
<P(k) ⇤R (k) + A F(k) (k) AF(k) ⇤R (k) >=
cmixed [F(k), ⇤(k)] :=
F n o
>
: >
F(k) (k) AF(k) ⇤R (k) ;
=0
257
numerical solution
• non-linear GMRES
• inexact Newton-GMRES
258
spectral solver — versatility
• isotropic elasticity
• 2048 x 2048
• uniaxial tension
103 Lagrange
Basic Lagrange
multiplier Polarisation
multiplier
Polarisation
Basic (filtered)
evaluations
259
solutions
http://www.mathpages.com/home/kmath330/kmath330.htm
260