Sei sulla pagina 1di 41

Fracture analysis with ADINA

Theodore Sussman, Ph.D.

© ADINA R&D, Inc, 2016 1


3D linear elastic fracture mechanics
• Stress intensity factors overview
• Energy release rate overview
• Virtual crack extensions
• Nodal virtual shift (NVS) method
• Station virtual shift (SVS) method
• Crack definition using the SVS method
• Modeling hints
• Example

• Note: this lecture is only a brief introduction to fracture mechanics, with


emphasis on the features of ADINA. For a more in-depth description of
fracture mechanics, see the literature, e.g. Anderson, Fracture
Mechanics, Fundamentals and Applications, 3rd ed, Taylor and Francis,
2005.

© ADINA R&D, Inc, 2016 2


Goal of linear elastic fracture mechanics
• Given a crack in a stressed body, determine if the crack will propagate.

• Calculate the stress intensity factors and/or the energy release rate.

© ADINA R&D, Inc, 2016 3


Stress intensity factors
• In a planar crack, stress analysis very close to the crack tip finds:

KI 5  1 3  K II  5  3 3 
 rr   4 cos  cos     sin  sin 
2 r 2 4 2 2 r  4 2 4 2

KI 3  1 3  K II  3  3 3 
    4 cos  cos   sin  sin 
2 r 2 4 2  2 r  4 2 4 2

 zz    rr     , plane strain assumed

KI  1  1 3  K II  1  3 3 
 r  sin  sin  cos  cos
2 r  4 2 4 2  2 r  4 2 4 2 
K III  K III 
 rz  sin  z  cos
2 r 2 2 r 2

• The stresses have a 1/ r singularity at the crack tip. K I , K II , K III


are the stress intensity factors.
© ADINA R&D, Inc, 2016 4
Stress intensity factors
• The dimensions of stress intensity factor are [stress] [length] , for
example MPa m

• The three stress intensity factors correspond to the three basic modes of
deformation:

• In 3D, K I , K II , K III vary from point to point along the crack front.

• The stress intensity factor concept is applicable only for linear elastic
materials (brittle fracture).

© ADINA R&D, Inc, 2016 5


Relationship between stress intensity factors and
Ritchie-Knott-Rice model
• In the Ritchie-Knott-Rice model, failure is assumed to occur when the
fracture stress is exceeded over a characteristic distance.

• Suppose that this characteristic distance is known from experiment.


Then, under the assumption of linear elastic fracture mechanics, the
fracture stress at this characteristic distance can be calculated from the
stress intensity factors.
KI  5  1 3 
 rr  cos  cos
2 r  4 2 4 2 
K  5  3 3 
 II   sin  sin  , etc
2 r  4 2 4 2

For example, the maximum principal stress at the critical over-stressed


depth d can be computed from the stress intensity factors.
© ADINA R&D, Inc, 2016 6
Energy release rate
• Another way to characterize the severity of a crack is with the energy
release rate.

Potential energy   U   Ru
 d
G  lim  
A 0 A dA
© ADINA R&D, Inc, 2016 7
Energy release rate
d
• The formula G   works for both load-controlled conditions and
dA
displacement-controlled conditions.

[force][length] N-m
• The dimensions of G are [length]2 , for example
m2

• Conceptually G can be calculated as follows:


– Compute  for an analysis in which the crack has area A
– Perturb the geometry slightly so that the crack has area A  A ,
then rerun the analysis and compute   

– Compute G  
A

• However we will use more efficient methods to compute G.

© ADINA R&D, Inc, 2016 8


Energy release rate
• In 3D, G varies from point to point along the crack front.

• G and the stress intensity factors are related by


K I2 K II2 K III2 E
G   E  for plane strain   shear modulus
E  E  2 1 v 2

• The energy release concept can be applied in inelastic analysis, under


various assumptions (see the literature).

© ADINA R&D, Inc, 2016 9


Energy release rate and the J-integral
• The energy release rate is closely related to the J-integral.

ui
J  G   Wdx2   ij n j ds

x1

W  strain energy density

• J is path-independent, so long as the path surrounds the crack tip.

• This is why, in the literature, critical values of the energy release rate are
referred to using the symbol J, e.g. J1c

© ADINA R&D, Inc, 2016 10


The energy release rate and the virtual shift field
• The material near the crack tip is subjected to a “virtual shift”.

X
• The virtual shift of each
material particle is
denoted X

© ADINA R&D, Inc, 2016 11


The energy release rate and the virtual shift field
• Using the concept of the virtual shift field, the energy release rate can be
calculated as

1  u j  X k
G
Ac V  ij xk
  W d ik 
 xi
dV

• This integral is path-independent, so that different virtual shift fields


yield the same numerical value.
X for first virtual shift field
X for second virtual shift field

© ADINA R&D, Inc, 2016 12


The virtual shift field in 3D
• In 3D, the virtual shift field is nonzero in a tube that surrounds the crack
front.

X

The virtual shift field is zero on the tube boundaries, both on the radial
boundary and on the axial boundary.

© ADINA R&D, Inc, 2016 13


Calculating the energy release rate
• The energy release rate can be evaluated, given a finite element
solution, and also the virtual shift field:

1  u j  X k
G
Ac V  ij xk
  W d ik 
 xi
dV

• Therefore only one finite element solution needs to be calculated.

• The virtual shift field is also used to calculate Ac . So the energy
release rate calculation is unaffected by the magnitude of the virtual
shift field.

• When there are pressure loads acting on the crack faces, a correction
needs to be added to the above formula. For brevity, we will not
include this correction in the following equations.

© ADINA R&D, Inc, 2016 14


Nodal virtual shifts
• One way to describe the virtual shift field is in terms of nodal virtual
shifts. The virtual shift at each node is specified.

Example in 2D A nodal virtual shift Another nodal


virtual shift
• In ADINA, this technique is termed the "nodal virtual shift" technique
(NVS).

© ADINA R&D, Inc, 2016 15


Nodal virtual shifts in 3D

Example in 3D A nodal virtual shift

Another nodal
virtual shift
© ADINA R&D, Inc, 2016 16
Nodal virtual shifts and mesh layout
• The nodal virtual shift technique requires a structured mesh near the
crack front.

• This means that specialized meshing techniques need to be used near


the crack front.

© ADINA R&D, Inc, 2016 17


Example meshing layout using CRACK-M
• The CRACK-M commands can be used to create a structured mesh near
the crack front, and a transition region to the rest of the mesh.

© ADINA R&D, Inc, 2016 18


Nodal virtual shift implementation
• Using the nodal virtual shift technique, the energy release rate is
calculated at each corner node on the crack front.

• More than one nodal virtual shift field can be included for each corner
node on the crack front. This allows for the path-independence (virtual
shift field independence) of the energy release rate to be checked.

• Only one crack is allowed in the model.

• Due to the above-mentioned limitations, the station virtual shift method


(described next) is preferred and should be used whenever possible.

© ADINA R&D, Inc, 2016 19


Station virtual shift method
• In ADINA 9.1 and above, the station virtual shift method (SVS method)
is introduced.

• The major advantages of the SVS method are


– More than one crack allowed per model
– Simplified specification of cracks in model definition
– Unstructured meshes allowed near crack front
– Calculation of stress intensity factors in addition to energy release
rate (especially useful in mixed-mode fracture)
– Stations along crack front (where the results are obtained) can be
evenly spaced and need not correspond to nodes.

• Using the SVS method and the available meshing tools in the AUI,
fracture problems can be rapidly set up and solved.

© ADINA R&D, Inc, 2016 20


Station virtual shift method
• In the SVS method, the virtual shift field is set up without explicit
reference to the underlying mesh.

© ADINA R&D, Inc, 2016 21


Station virtual shift method
• Each crack advance station undergoes a virtual shift.

© ADINA R&D, Inc, 2016 22


Station virtual shift method
• The crack advance stations need not be at node points. Typically, the
crack advance stations are evenly spaced along the crack front.

• The tube boundaries need not correspond to element boundaries.

• Different tube radii (radial domains) can be used within the same crack,
to check the path-independence of the results.

• The crack front can be open or closed.

• Because the crack specification is independent of the mesh, specialized


meshes need not be used near the crack front. The model can be set up
with whichever tools are most convenient.

© ADINA R&D, Inc, 2016 23


Calculation of stress intensity factors
• The energy release rate integral can be viewed as an "interaction
integral":

1  1  A u Bj u Aj   X k
G (u , u )  V  2  ij xk   ij xk   W d ik  xi dV
A B B A, B

Ac    

u A  displacement field A  strain field A, stress field A


u B  displacement field B  strain field B, stress field B

1  u j  X k
G (u, u) 
Ac V  ij xk  W d ik  xi dV  energy release rate
u  displacement field from FE analysis

© ADINA R&D, Inc, 2016 24


Calculation of stress intensity factors
• We now introduce
u I  displacement field for mode I with K I  1
u II  displacement field for mode II with K II  1
u III  displacement field for mode III with K III  1

• It turns out that, analytically,


 G (u I , u I ) G (u I , u II ) G (u I , u III )  1/ E  0 0 
 III   0 
 G (u I
, u II
) G (u II
, u II
) G (u II
, u ) 
  1/ E  0 
G (u , u ) G (u , u ) G (u , u )   0
I III II III III III
0 1/  
 

• Therefore
 G (u I , u I ) G (u I , u II ) G (u I , u III )   K I   G (u, u I ) 
 I II II II II III      G (u, u II ) 
 G (u , u ) G (u , u ) G (u , u ) K
  II   
G (u , u ) G (u , u ) G (u , u )   K III  G (u, u ) 
I III II III III III III
   

can be solved for the stress intensity factors.

© ADINA R&D, Inc, 2016 25


Calculation of stress intensity factors
• Conceptually, the "amount" of each mode contained in the finite
element solution is measured by the terms G (u, u I ), G (u, u II ), G (u, u III ) .

• All of the G (u A , u B ) integrals are evaluated numerically, at the element


integration points. Outside of the tube, these integrals are zero.

• The accuracy of the procedure increases as the number of element


integration points that lie within the tube increases.

• The program also computes error measures such as


G (u I , u I )  1/ E 
1/ E 
so that the accuracy of the numerical integration can be assessed.

© ADINA R&D, Inc, 2016 26


Crack description using CRACK-SVS command
• The additional model input required to specify a crack is entered using
the CRACK-SVS command. This model input can be divided into the
following items:
– Specification of the crack front. The crack front is specified by
entering the element faces on the top surface of the crack, and the
element faces on the bottom surface of the crack.

The element faces


can be specified
either using element
face-sets, or by
specifying the
underlying geometry.

© ADINA R&D, Inc, 2016 27


Crack description using CRACK-SVS command
– Spacing of the crack advance stations: at corner nodes, by
distance, or by number of crack advance stations.

© ADINA R&D, Inc, 2016 28


Crack description using CRACK-SVS command
– Number, sizes and spacings of the radial domains

– Direction of increasing coordinate along the crack front

© ADINA R&D, Inc, 2016 29


Crack description using CRACK-SVS command
– Starting location of the crack front coordinate.
– Option to shift nodes adjacent to the crack front to the quarter-points.

By shifting the nodes to the quarter-points, the 1/ r stress singularity is


directly modeled in the finite elements. This is thought to improve the accuracy
of the solution.

However, in practice, shifting the nodes to the quarter-points in an unstructured


mesh causes the elements to become overdistorted. Our experience is that the
solution is not noticeably improved when the nodes are shifted to the quarter-
points.

© ADINA R&D, Inc, 2016 30


Modeling hints
• The crack faces must be flat and parallel to each other.

• The radial domains cannot be too small or too big.

© ADINA R&D, Inc, 2016 31


Modeling hints
• The number of crack advance stations should not be greater than the
number of corner nodes on the crack front.

• Using fewer crack advance stations than corner nodes can increase the
accuracy of the solution, but will smooth out the variation of stress
intensity factors along the crack front.

© ADINA R&D, Inc, 2016 32


Modeling hints
• The radius of curvature of a curved crack should be much greater than
the radius of the radial domains.

• Remember, all of the theory is derived for a planar crack.

© ADINA R&D, Inc, 2016 33


Example solution
Inclined through-crack in a thick plate

© ADINA R&D, Inc, 2016 34


Example solution
Geometry

© ADINA R&D, Inc, 2016 35


Example solution
Additional faces on crack plane

© ADINA R&D, Inc, 2016 36


Example solution
Free-form meshing 4-node tets used throughout, for
demonstration. No special
meshing near crack front.

Green / red mesh All nodes coincident


glued to blue mesh in green / red mesh
© ADINA R&D, Inc, 2016 37
Example solution
Split mesh to create duplicate nodes on top and bottom crack surfaces

MESH-SPLIT SURFACE-FACE
1 2
2 2
1 3
3 3
DATAEND

© ADINA R&D, Inc, 2016 38


Example solution
Define the cracks

CRACK-SVS 1 ADV-OPTION=ADV-NUMBER ADV-NUMBER=10,


RADIUS-NUMBER=3 RADIUS-MAXIMUM=5
2 2 TOP
1 3 BOTTOM
DATAEND
CRACK-SVS 2 ADV-OPTION=ADV-NUMBER ADV-NUMBER=10,
RADIUS-NUMBER=3 RADIUS-MAXIMUM=5
1 2 TOP
3 3 BOTTOM
DATAEND

© ADINA R&D, Inc, 2016 39


Example solution
Visualization of CRACK-SVS definitions

© ADINA R&D, Inc, 2016 40


Example solution

KI

K II

Very good results


are obtained, even
with 4-node tets.

© ADINA R&D, Inc, 2016 41

Potrebbero piacerti anche