You are on page 1of 7

MECH3300 Finite Element Methods

LECTURE 7: Finite element methods in fracture


mechanics
Estimation of stress intensity
For well-known crack geometries, K or G can be found from
formulae or charts. For more general cases, a finite element
model can be used, both to estimate K and to predict growth of
a crack.
Several methods can be used to make FE estimates of KI.

(1) Use the stress field ahead of a crack tip.


The 2D plane strain analytical solution for a crack in an infinite
plate becomes more true as the crack tip is approached. Both
direct stresses on the line ahead of the crack approach the
same values. While both xx and yy tend to infinity at the crack
tip, xx(2x) = yy(2x) and both expressions remain finite,
tending to KI, as x tends to zero.
y
xx
x
yy
Estimation of K from stresses continued
The x and y stresses can oscillate due to numerical error. Better
estimates can be obtained by finding the average (xx + yy)/2.

We can then estimate {(xx + yy)/2} (2x) This can be extrapolated to


the crack tip.

The singularity at the crack tip is not represented properly, even with a
very fine mesh. This has been shown to pollute the whole solution to
some degree.

The stress everywhere is in theory KI, so an error in KI affects the


whole stress field.

Special crack tip elements, including an appropriate singularity in their


displacement fields, have been devised for crack tips. They are present
in a few commercial packages (not STRAND7 or NASTRAN for
Windows).
Estimation of K from displacements
The opening of the crack, measured by normal displacement uy
can also be found from the analytical solution for a crack in a
semi-infinite plate. Near the crack tip, this solution is
increasingly valid, predicting that, in the limit, as L tends to zero
Eu y 2
KI
4(1 2 ) L(1 L )
2a
uy

a
This estimate shows less oscillation when extrapolated to
the crack tip, as errors in the solution affect the
displacements less than the stresses or strains.
Estimation of G from work or energy
Consider a 2D plane strain model of a crack, implied to be of unit
thickness.
Keeping the load P opening a crack constant, half the work done is
stored as elastic strain energy, and the other half grows the crack.
We find the displacement between the loading points, then
extend the crack by a by disconnecting a pair of elements at the
crack tip.
We again find this displacement giving 2
P(2 - 1) is the work done. P(2 - 1) /(2a) estimates GI which can
be converted to KI
Doing this with a half model, the crack is grown by removing a
restraint, and the work by crack area P(2 - 1)/a does not need to
be halved. P

a
Crack-tip elements
Crack-tip elements are designed to capture a singularity, like the
1/r function in the 2D plane strain solution for the stress field in
a semi-infinite plate.

One way to achieve this is to take a quadratic quadrilateral


element, collapse one side to a point so that 3 nodes are
superimposed, and shift the mid-side nodes on the adjacent
sides to 1/4 the side length.
3L/4

L/4
3 nodes
superimposed

This is not easy to do in most commercial packages.


Estimation of J
Some FE packages, designed especially for nonlinear analysis (eg
Abaqus and MSC/Marc) offer the possibility of estimating the J
integral - the energy release rate that accounts for the work of plastic
deformation.

This requires processing of strain energy per volume, stresses and


displacement derivative data from a ring of elements surrounding the
crack tip, following an elastoplastic analysis to predict the stresses
and strains.

Many packages can do the elastoplastic analysis, and predict the size
of the plastic zone at the crack tip, but the derivatives of displacement
needed to estimate J may not be available outputs (eg ux/y).

Details of how to estimate J with a finite element package are given in


Anderson, Fracture Mechanics.

You might also like