1/61
Dr. Stéphane AVRIL
LMPF Research Group, ENSAM Châlons en Champagne, France
THE VIRTUAL FIELDS METHOD
PRINCIPLE, RECENT ADVANCES AND APPLICATIONS
Paris
Châlons en Champagne 12th April 2007, Université de Technologie de Compiègne
OUTLINE
• Introduction and basic concepts;
• Application 1: anisotropic elasticity in composites;
• Application 2: visco-elasticity in PMMA;
• Application 3: elasto-visco-plasticity in metals;
• Application 4: heterogeneous modulus distribution;
• Conclusion and main prospects.
3/61
Introduction and
basic concepts
Widespread use of full-field optical techniques in
experimental mechanics
5/61
Largely used for qualitative
analyses
Largely used for qualitative
analyses
7/61
Largely used for qualitative analyses
– Detection of shear bands or cracks ; – Model validation ;
– Verification of boundary conditions.
Interest in quantitative use for
model identification
9/61
Interest in quantitative use for
model identification
– Identification of sophisticated models in one single test ;
– Specimen geometry is not a constraint ; – Localization effects can be handled ;
– Identification of graded properties in the same specimen.
Interest in quantitative use for
model identification
11/61
- No standard characterization of uncertainty in full- field optical measurement techniques.
- Large amount of data to process require specific inverse approach.
Reason for poor use
for model identification
Resolution of an inverse problem Basic approach: updating
Parameters Ai
Model: f
Response Mj=f(Ai)
Measurements mj
Matching optimization
13/61
Particular case of full-field
measurements: alternative approach
Measurements m(x,y,z)
Direct
computation F(Ai,m)=0
Ai
Model f is derived from:
- the constitutive equations, - the conservation equations.
F(Ai,M)=0 Æ M=f(Ai)
THE VIRTUAL FIELDS METHOD
Application 1: anisotropic
elasticity in composites
15/61
* 0
* + =
−
∫ ∫
∂V
i i V
ij
ij ε dV T u dS
σ or
I Equilibrium equations
, j = 0
σ
ij + boundary conditions strong (local) weak (global) II Constitutive equationsIII Kinematical equations (small strains/displacements)
) u u
2 ( 1
i, j j
, i
ij
= +
ε
The virtual fields method Example in linear elasticity
kl ijkl
ij
C ε
σ =
0 dS
u T dV
V
* i i V
* ij
ij
ε + =
σ
− ∫ ∫
∂
Eq. I (weak form, static)
Substitute stress from Eq. II
*
0
*
+ =
− ∫ ∫
∂V
i i ij
kl
ijkl
dV T u dS
C ε ε
kl ijkl
ij
C ε
σ =
The virtual fields method
Example in linear elasticity
17/61
Equation valid for any KA virtual fields.
For each choice of virtual field: 1 equation.
Choice of as many VF as unknown parameters.
In elasticity:
Æ Linear system of equations to be inverted.
Æ Optimal choice of virtual fields for minimizing result uncertainty (Avril et al., Comp. Mech., 2004)
Æ More robust than updating approach (Avril et al., Int. J.
Sol. Struct., 2007)
*
0
*
+ =
− ∫ ∫
∂V
i i V
ij kl
ijkl
dV T u dS
C ε ε
The virtual fields method
Example in linear elasticity
Problem presentation
19/61
Glass epoxy ring cut from a tube:
Example
θ
r
x y
F
⎟⎟ ⎠
⎜⎜ ⎞
⎝
⎛
⎥ ⎥
⎦
⎤
⎢ ⎢
⎣
= ⎡
⎟⎟ ⎠
⎜⎜ ⎞
⎝
⎛
s r
ss r
r rr
s r
Q Q
Q
Q Q
ε ε ε σ σ σ
θ θθ
θ
θ
θ
0 0 0 0
Æ 4 parameters to identify Æ No standard test available
Need of two cameras
“back to back”
21/61
Deformation maps
using two cameras
Polynomial fit, degree 3, transform to cylindrical and analytical differentiation:
Strain maps
εr εθ εs
23/61
( )
dS M
u M T
dS e
Q
dS e
Q dS
e Q
dS e
Q
f
S s s ss
S r r
r r S
S r rr
) (
* )
( .
. .
.
*
*
*
*
*
∫
∫ ∫ ∫
∫ + =
+ +
+
S
r r
ε ε
ε ε ε
ε ε
ε ε
ε
θθ θ θ θ θ θ[ ] A × ( ) ( ) Q = B
4 virtual fields
dS M
u M T
dS
f
) (
* ) (
* ∫
∫
S: =
Se r r
ε σ
PVW Constitutive equations
⎟⎟ ⎠
⎜⎜ ⎞
⎝
⎛
⎥ ⎥
⎦
⎤
⎢ ⎢
⎣
= ⎡
⎟⎟ ⎠
⎜⎜ ⎞
⎝
⎛
s r
ss r
r rr
s r
Q Q
Q
Q Q
ε ε ε σ σ σ
θ θθ
θ
θ
θ
0 0 0 0
&
Identification
Qrr Qθθ Qrθ Qss
Reference* (GPa) 10 40 3 4
Identified (GPa) 11.4 45.4 2.62 6.78
Coeff. var (%) – 9 tests 29 10 29 4
Moulart R., Avril S., Pierron F., Identification of the through-thickness rigidities of a thick laminated composite tube, Composites Part A: Applied Science and Manufacturing, vol. 37, n° 2, pp. 326-336, 2006.
Results
25/61
Application 2: visco-elasticity in
polycarbonate
⎟⎟
⎟⎟
⎠
⎞
⎜⎜
⎜⎜
⎝
⎛
⎥⎥
⎥
⎦
⎤
⎢⎢
⎢
⎣
⎡
− +
⎟⎟
⎟⎟
⎠
⎞
⎜⎜
⎜⎜
⎝
⎛
⎥⎥
⎥
⎦
⎤
⎢⎢
⎢
⎣
⎡
−
=
⎟⎟
⎟⎟
⎠
⎞
⎜⎜
⎜⎜
⎝
⎛
xy yy xx
i xy i
xx i
xx i
xy
i xy i
xx
xy yy xx
r xy r
xx r
xx r
xy
r xy r
xx
xy yy xx
Q Q
Q Q
Q Q
i Q
Q Q
Q
Q Q
ε ε ε ω
ε ε ε σ
σ σ
0 0
0 0 0
0
0 0
)]
, ( cos[
) , ( )
, (
) cos(
) , ( )
, (
y x t
y x y
x
t y
x y
x
ϕ ω
ε ε
ω σ
σ
+
=
=
)]
, (
) [
, ( )
, (
) , ( )
, (
y x t
i t i
e y x y
x
e y x y
x
ϕ ω
ω
ε ε
σ σ
= +
=
Constitutive equations and assumptions
Inertial excitation with steady state and linear response.
27/61
Experimental arrangements
Polycarbonate panel 200mm x 150mm h = 3mm
CCD
O
M P
δ Q
λ
dα
δ = 2λ . dα
The deflectometry technique
29/61
Out of resonance 80 Hz Near resonance 100 Hz
Slope measurements
Out of resonance 80 Hz Near resonance 100 Hz
Deduced curvatures
31/61
⎪⎭
⎪⎬
⎫
⎪⎩
⎪⎨
⎧
⎟⎟
⎟⎟
⎟⎟
⎠
⎞
⎜⎜
⎜⎜
⎜⎜
⎝
⎛
⎥⎥
⎥⎥
⎥
⎦
⎤
⎢⎢
⎢⎢
⎢
⎣
⎡
− +
⎥⎥
⎥⎥
⎥
⎦
⎤
⎢⎢
⎢⎢
⎢
⎣
⎡
−
⎪ =
⎭
⎪⎬
⎫
⎪⎩
⎪⎨
⎧
s y x
xy xx
yy xy
xy xx
xy xx
yy xy
xy xx
s y x
k k k B
B B
B
B B
i D
D D
D
D D
M M M
0 2 0
0 0
0 2 0
0 0
ω
xy xy xy
xx xx xx
xx xy 2
xx
D ; B
D B
D ; D
) 1
( E D
= β
= β
= ν ν
= −
Four parameters to identify
Stiffness Damping
( j ) VF y ( j )
x
VF
1=
21 + ,
2=
21 +
Identification
E (GPa)
3 3,2 3,4 3,6 3,8 4 4,2 4,4 4,6 4,8 5
80 Hz 90 Hz 100 Hz Reference
Results
33/61
ν
0 0,05 0,1 0,15 0,2 0,25 0,3 0,35 0,4 0,45 0,5
80 Hz 90 Hz 100 Hz
Results
βxx (10-4 s)
0 0,2 0,4 0,6 0,8 1 1,2
80 Hz 90 Hz 100 Hz Reference
Results
35/61
βxy (10-4 s)
0 0,2 0,4 0,6 0,8 1 1,2 1,4
80 Hz 90 Hz 100 Hz
Results
Giraudeau A., Guo B., Pierron F., Stiffness and Damping Identification from Full Field Measurements on Vibrating Plates, Experimental Mechanics, Vol. 46, N°6, pp. 777-787, 2006.
- Improvement of the measurement temporal accuracy.
- Extension to anisotropic plates underway
Prospects
37/61
Application 3: elasto-visco-
plasticity of metals
Constitutive equations and assumptions
) ,
,
( X
g σ ε
σ & = &
Constitutive parameters to identify
39/61
Elasto-plasticity with plane stress and Von Mises criterion
⎪ ⎪
⎪
⎭
⎪ ⎪
⎪
⎬
⎫
⎪ ⎪
⎪
⎩
⎪ ⎪
⎪
⎨
⎧
−
−
−
⎥ ⎥
⎥ ⎥
⎦
⎤
⎢ ⎢
⎢ ⎢
⎣
⎡
− −
⎪ =
⎭
⎪ ⎬
⎫
⎪ ⎩
⎪ ⎨
⎧
s xy s
s xy y
s xx x
s y x
p s p s p s
E
ε σ ε σ ε σ
ν ν ν
ν σ ν
σ σ
&
&
&
&
&
&
&
&
&
3 2 3 2 3
2 0 1
0
0 0 1
1
2)
( p
Q ε ε
σ & = −
Perzyna’s model for viscoplasticity
⎪⎪
⎪⎪
⎭
⎪⎪
⎪⎪
⎬
⎫
⎪⎪
⎪⎪
⎩
⎪⎪
⎪⎪
⎨
⎧
− −
−
− −
−
− −
−
⎥⎥
⎥⎥
⎦
⎤
⎢⎢
⎢⎢
⎣
⎡
− −
⎪ =
⎭
⎪⎬
⎫
⎪⎩
⎪⎨
⎧
s xy n s s
s xy n y s
s xx n x s
s y x
p s H
p s H
s p
H
E
σ σ
γ σ ε
σ σ
γ σ ε
σ σ
γ σ ε
ν ν ν
ν σ ν
σ σ
) 1 3 (
) 1 ( 2
3
) 1 ( 2
3
2 0 1
0
0 0 1
1
0 0 0
2
&
&
&
&
&
&
Constitutive parameters to identify: E, ν, γ, σ , n, dH/dp.
41/61
Mechanical arrangement
L=28mm
x y
w=22mm
20 mm
Average strain rate: 1s-1 Images of the specimen recorded with high speed cameras
Deformation deduced by digital image correlation.
Fields of strain rate
43/61
Virtual field
y u
u * x = 0 ; * y = −
Across the measurement area:
0 ;
1 ;
0 * *
* = yy = xy =
xx ε ε
ε
Identification
( ) ∑ ∫∫
= ⎟⎟
⎠
⎞
⎜⎜
⎝
⎛ −
= N
l
l S
t
t y
t Sh
L t dtdS P
t y x E S n
E n
F
l
1
2
0
0 0
) ) (
, , , , , , 1 (
, ,
, γ σ σ γ
σ &
Æ minimization of a cost function defined from the principle of virtual work:
45/61
Identification
0 5 10 15 20 25
0 50 100 150 200 250 300 350 400
time (ms)
average stress (MPa)
from displacement fields from resultant load
- Numerical issues
- Larger strain rates Æ Hopkinson
- Heterogeneities: microscopic scale, welds...
Prospects
Pannier Y., Avril S., Rotinat R., Pierron F., Identification of elasto-plastic constitutive parameters from statically undetermined tests using the virtual fields method, Experimental Mechanics, Vol. 46, N°6, pp. 735-755, 2006.
Avril S., Pierron F., Yan J., Sutton M., Identification of viscoplastic parameters using DIC
47/61
Heterogeneous modulus
distribution
*
0
*
+ =
− ∫ ∫
∂V
i i V
ij kl
ijkl
dV T u dS
C ε ε
The virtual fields method for heterogeneous solids in linear
elasticity
*
0
*
+ =
− ∑ ∫ ∫
∂V
i i V
ij kl n
n
ijkl
dV T u dS
C
n
ε
ε
49/61
V1 V2 V3
Vn
Discretization of the solid
Cube with a stiff inclusion buried in it.
Silicone gel materials mimicking human tissue containing a tumour.
Experimental arrangements
1/8
51/61
Displacement fields measured by MRI
Scanning tomographic method:
Æ 3D bulk measurements!!
Principle of the identification
53/61
3D Results
Avril S., Huntley J., Pierron F. and Steele D., 3D-Heterogeneous stiffness
55/61
Conclusions
- Huge potential for some difficult and important engineering issues (heterogeneous materials, high strain rate, micro-scale…);
- Training required on optical FFMT: not a black box, primary influence on identification results
- Full-field processing: noise filtering, displacements to strains…
Prospects
57/61
Measurements m(x,y,z)
KEY POINT F(Ai,m)=0
Ai
Key point:
field reconstruction
Model f is derived from:
- the constitutive equations, - the conservation equations.
F(Ai,M)=0
Comparison of two approaches
59/61
Comparison of two approaches
Comparison of two approaches
61/61