Vous êtes sur la page 1sur 41

Failure in Castings and Application to Chassis Parts

Paul A. Du Bois
consulting engineer
paul.dubois@gmx.net

FOIL, , june 2010

Oasys LS-DYNA UK Users Meeting 2012


25 January 2012
The Arup Campus, Solihull, UK
Oasys LS-DYNA UK Users Meeting 2012

Numerical simulation of failure


in castings and forgings

industrial references
modeling issues
specific aspects of castings and forgings
failure and damage models
validation techniques
conclusions
Oasys LS-DYNA UK Users Meeting 2012

Modelling issues

Massive components with complex geometry


Modeling with hexa is very costly
Tetra elements are a solution
Quadratic tetra (#16, #17) are equivalent to hexa but
lead to small timesteps
Linear tetra (#10) suffers from locking
Forging tetra (#13) is a possible compromise
Crack simulation requires fine meshes
4-5 elements over the thickness is a rock bottom
minimum
Selective mass scaling may have potential
Oasys LS-DYNA UK Users Meeting 2012

Some theoretical background on type 13 Tetra


manual: "1 point nodal pressure tetrahedron for bulk forming"

paper: J. Bonet & A.J. Burton. A simple average nodal pressure


tetrahedral element for incompressible dynamic explicit
applications. Comm. Num. Meth. Engrg. 14: 437-449, 1998

"... the element prevents volumetric locking by defining nodal


volumes and evaluating average nodal pressures in terms of
these volumes ...
... it can be used in explicit dynamic applications involving
(nearly) incompressible material behavior
(e.g. rubber, ductile elastoplastic metals) ..."

TET #13 = TET #10 + averaging nodal pressures


= TET #10 volumetric locking
Oasys LS-DYNA UK Users Meeting 2012

Example 1: Notched steel specimen


discretized quarter system:
5 mm

10 mm TET #13

30 mm u (t)

TET #10
MAT_PIECEWISE_LINEAR
_PLASTICITY
HEX #2
E = 206.9 kN/mm2
= 0.29
y = 0.45 kN/mm2
Et = 0.02 kN/mm2 (nearly HEX #1
ideal plastic)
Oasys LS-DYNA UK Users Meeting 2012

Example 1: Notched steel specimen

load-displacement curve: von Mises


should show a limit force stresses
TET #13
0 - 480
volumetric N/mm2
8 locking TET #10
TET #10
force in kN

6
HEX #1
4
TET #13HEX #2
2
HEX #2
0
0 0,1 0,2 0,3 0,4

displacement in mm
HEX #1
Oasys LS-DYNA UK Users Meeting 2012

Example 2: Rubber block compression

HEX #1 HEX #2 TET #10 TET #13


MAT_MOONEY-RIVLIN_RUBBER

A = 4.0 N/mm2
B = 2.4 N/mm2
= 0.499 nearly incompressible material
= 1.5E-06 kg/mm3
Oasys LS-DYNA UK Users Meeting 2012

Example 2: Rubber block compression

HEX #1 HEX #2 TET #10 TET #13


deformation

von Mises stresses (0 - 12 N/mm2)


Oasys LS-DYNA UK Users Meeting 2012

Example 3: Taylor bar impact

MAT_PIECEWISE_LINEAR_PLASTICITY:
= 8930 kg/m3, E = 117 kN/mm2, = 0.35, y = 0.4 kN/mm2, Et = 0.1 kN/mm2

deformation
HEX #1 HEX #2 TET #10 TET #13

12.6 14.1 10.4 13.9


mm mm mm mm
Oasys LS-DYNA UK Users Meeting 2012

Example 3: Taylor bar impact

HEX #1 HEX #2 TET #10 TET #13

pressure (-300 300 N/mm2) checkerboard mode


Oasys LS-DYNA UK Users Meeting 2012

Example 4: Structural component

load-displacement curve

500 TET #10


400

force in kN
300 TET #13
von Mises stresses
200
x - stresses
100

0
0 25 50 75 100 125 150
x time in ms
Oasys LS-DYNA UK Users Meeting 2012

Conclusion

TET #13 is superior to TET #10 when it comes to (near)


incompressible material behavior: less volumetric locking

CPU time penalty TET #13 vs. TET #10: ~ 25 %

Applications: Bulk forming, Rubber materials, Ductile metal parts


with plastic deformations

Currently, TET #13 works only for a subset of materials:


1,3,6,24,27,77,81,82,91,92,106,120,123,124,128,129,181,183,225,244
224 was updated for #13 tetra in June 2011

Keyword User's Manual: well described


Oasys LS-DYNA UK Users Meeting 2012

Specific aspects of cast components

How big is the difference between a prototype ( gravity cast )


and a series component ( die cast ) ?
How homogeneous is the cast component ?
Simulation of the casting process or testing on samples out of
different locations in the component can answer that question
Magnasoft determines local values of yield strength (+/-
20%), fracture strain and density (allows to identify
macrofaults but not microfaults)
We would like to see the initial porosity in the component,
this seems hard to obtain
Skin formation and through-the-thickness density gradients
Surface quality is an issue (sand blasted, milled)
Sulphur content may be an issue for forged steels
Forgings usually have higher ductility but are more
expensive
Oasys LS-DYNA UK Users Meeting 2012

Failure and damage models

8% bis 15% fracture strain


MAT_GURSON seems like the natural choice to model
castings but does not predict failure under shear
conditions
Need to perform tensile tests, cylindrical samples are
prefered due to the 3D nature of the structures
Gurson model was developed in the 1970s to assess
failure in castings
further development by Tvergaard and Needleman
implementation in LS-DYNA and regularisation of the
Gurson law goes back to Feucht and Fassnacht
recent generalisation by Nahshon and Hutchinson to
include shear failure (Lode parameter dependent )
Oasys LS-DYNA UK Users Meeting 2012

The Gurson-model in LS-DYNA (MAT_120)

The yield function is given as

e2 q tr

2
(, M , f ) 2 2q1f * cosh 2 1 q f *
0
M
1
2 M

e = equivalent von Mises stress

M = yield stress (matrix)

= stress tensor Undamaged Gurson yield surface and


f * = effective void volume fraction hardening law

For the matrix material associative von Mises M


plasticity is assumed for the undamaged state.
Energy equivalence is applied :
p vp vm p
M
1 f M
M M dt
Oasys LS-DYNA UK Users Meeting 2012

The Gurson-model in LS-DYNA : associated flow rule

Compatibility condition allows to compute the


volumetric plastic strain (zero if no damage) :

f 3 pq2
vp 3 q q f *
sinh
p 2 M
M 1 2

vp p
f 2 vm
p
vm
Damaged Gurson yield surface
(yields in hydrostatic loading)
And the plastic Poisson ratio : (0.5 if no damage)

vm 3 p
3 pq2
f 2f 1 M q1q2 f * sinh
3p 2 M
vm 3p p
p 3 pq2
f 2f 2 M q1q2 f * sinh
2 3p 2 M
vm 3p Typical Gurson stress-strain curve
Oasys LS-DYNA UK Users Meeting 2012

The Gurson-model in LS-DYNA : void evolution


2
1
M N
fN
f fN fG f f0 1 f vp M
2 sN
The growth of the void volume is e
sN 2

Nucleation of new voids intension: fN A Mpl

fN 1 pl 2
where A exp M N

sN 2 2 sN The effective void volume fraction is

A defined according to
N = mean nucleation strain f f fc

Mpl = eff. pl. strain (matrix) f * (f ) 1/ q1 fc
f
c (f fc ) f fc
N Mpl sN = std. deviation fF fc

fc is the critical void volume fraction


sN above which the voids start to combine
and grow.
Growth of existing voids: fG (1 f ) kkpl Failure is being initiated at
Vvoids V Vmatix VVmatix 1
where f fG f * (fF ) f fF
Vvoids Vmatix V V2 q1
Oasys LS-DYNA UK Users Meeting 2012

The Gurson-model in LS-DYNA : damage

Taylor series of cosh shows the yield function is of the Drucker-Prager type :

vm
2
q 2 3 p
f g 2 2q1f * cosh 1 q12 f *2
M 2
M
x2 M 3 pq2
cosh x 1 1 q1q2 f * sinh
2 3p 2 M 2 q22 q1f *
p
9 2
M 4 q22 q1f *
2
f g vm
2
q1f * q22 p 1 q1f * M2 3 pq
2 q1q2 f * sinh 2
4 3p 2 M
f g vm A2 p 2 1 d A0
2 2

d q1f * Plastic Poisson ratio is


Related to the damage !
A0 M2
9
A2 q1f * q22
4

This allows expressing classical damage in function of void volume fraction


Oasys LS-DYNA UK Users Meeting 2012

The Gurson-model in LS-DYNA : damage evolution


The damage evolution can now be computed in function of equivalent plastic strain for every value of the stress
triaxiality :


2
1
M N

d q1 f q1 f 0 q1q2 1 f
9 p * fN
f p M
* 2 2 sN
e
4 vm s N 2

Result is shown for uniaxial tension

For every value of the stress triaxiality a value of the plastic strain to
failure can then be computed by accumulating the damage :
p
t
vm pf t p
f f F

The graph shows no failure is predicted


under shear loads
Oasys LS-DYNA UK Users Meeting 2012

The Gurson-model in LS-DYNA : regularisation

Regularisation is done by making initial, nucleation, critical and failure void volume fractions dependent upon
element characteristic length :

, fN
Up to 4 regularisation curves are defined : f0 f0 lc fN lc , fc fc lc , fF fF lc ,
Oasys LS-DYNA UK Users Meeting 2012

Quasistatic tensile test / axisymmetric sample

Probenform:
DIN 50125 d=10mm
10 x 50
L0=50mm

E = 206 GPa
(197 211)
Rp02 = 604 MPa
(542 638)
RM = ca. 900MPa
(881-932)
A = 15,6% ??
(14,0 -15,5)

Forging material shows some variability but less then the castings
Oasys LS-DYNA UK Users Meeting 2012

Determine the hardening curve from sample ZR1


Oasys LS-DYNA UK Users Meeting 2012

Regularisation (Gurson parameters)


for 0.5mm 1.0mm and 2.0mm

Allows a good description of failure under uniaxial tension and triaxialities close to -1/3
Oasys LS-DYNA UK Users Meeting 2012

MAT_GURSON is a coupled damage model :

1 f * lc q1
Regularisation influences the damage
Damage influences the failure
Damage influences the strength

True stress is mesh dependent after localisation / plastic instability


Oasys LS-DYNA UK Users Meeting 2012

Validation techniques

MAT_GURSON will have a limited range of validity with


respect to state of stress and strain rate
More advanced failure and damage models such as
MAT_224 and MAT_GURSON_JC or
MAT_ADD_EROSION_GISSMO may be more suitable for
real life applications
a minimum material characterisation program should
include dynamic tension, compression, shear and
notched tension
validation testing is typically on more complicated
samples under bending loads
Oasys LS-DYNA UK Users Meeting 2012

Material testing at MPA Stuttgart


Quasistatic tension Sample shape : DIN 50125 10 x 50
Probe vernetzt in
0,5 mm
1,0 mm d=10mm
2,0 mm L0=50mm
Shear test 6 x 50
Quasistatic notched tension
Probe vernetzt in
0,5 mm
1,0 mm
2,0 mm
L0=50mm
Dynamic tension (strain rate 100 1/s)
Probe vernetzt in
0,5 mm
0,75 mm
1,0 mm
d=6mm
1,5mm
L0=30mm
Oasys LS-DYNA UK Users Meeting 2012

Compression test usually shows no failure :

Prfaufbau:
Oasys LS-DYNA UK Users Meeting 2012

Quasistatic and dynamic tensile tests

Dynamic tension, strain rate 1/100 s:


Oasys LS-DYNA UK Users Meeting 2012

Notched tensile test DIN 50125 10x50 R=2mm


L0=50mm

Notched sample : eleongation measured by


extensometer is not matched by *MAT_GURSON,
Triaxiality<-1 as opposed to -0.577 in plane strain
Oasys LS-DYNA UK Users Meeting 2012

Validation techniques

Try dynamic bending tests on samples with a E


moderately complex geometry :
Oasys LS-DYNA UK Users Meeting 2012

Determine critical elements


Oasys LS-DYNA UK Users Meeting 2012

Determine triaxiality
Oasys LS-DYNA UK Users Meeting 2012

Determine triaxiality

p
t = - 0.45
s vm
Oasys LS-DYNA UK Users Meeting 2012

Simulation with failure criterion


Oasys LS-DYNA UK Users Meeting 2012

Verify that element erosion is stable


Oasys LS-DYNA UK Users Meeting 2012

Application example

Any failure model with a dependency upon the state of


stress can explain many phenomena already
Example is for the forging 38MnVS6
Oasys LS-DYNA UK Users Meeting 2012

Application example
Element
No. 3/4

Element
No. 1/2
Oasys LS-DYNA UK Users Meeting 2012

Application example

Elements No. 3 /4
do not fail although
equivalent
plastic strain is
ca. 19 %

Element No. 1/2 fail


at an equivalent
plastic strain of
ca. 11 %
Oasys LS-DYNA UK Users Meeting 2012

Application example
Triaxialitt (History Variable 2)
Oasys LS-DYNA UK Users Meeting 2012

Application example
Oasys LS-DYNA UK Users Meeting 2012

Conclusions

The state of stress in cast components is usually 3D,


this precludes modeling with shell elements if failure
needs to be predicted
3D structures have a much wider range of triaxialities
then thin metal sheets
LS-DYNA has a wealth of useful failure and damage
models but experimental work and numerical validation
work are needed to populate these models with
realistic data
The relevance of test samples with respect to real
components needs to be assessed
Coupling between casting and crash simulations may be
desirable
Careful modeling and critical assessment of the
numerical results are the key to success

Vous aimerez peut-être aussi