CAPSUL - CIMNEcongress.cimne.com/icme2016/admin/files/filepaper/p104.pdfCAPSUL: a tool for...
Transcript of CAPSUL - CIMNEcongress.cimne.com/icme2016/admin/files/filepaper/p104.pdfCAPSUL: a tool for...
CAPSUL:a tool for computational homogenization of polycrystals
A. Cruzado1, B. Bellón1, J. Segurado1,2, J. LLorca1,2
1IMDEA Materials Institute & 2Polytechnic University of Madrid
2nd Int. Workshop on Software Solutions for Integrated Computational Materials EngineeringBarcelona, 12th-15th, April, 2016
SUMMARY
1. IMDEA Materials’ suite of simulation tools
2. CAPSUL: a suite of computational homogenization tools- RVE generation- Crystal plasticity models- Levenberg-Marquardt optimization algorithm- Fatigue indicator parameters
3. Examples of applications- Virtual testing of extruded AZ31 Mg alloys- Virtual testing of wrought IN718 superalloy in fatigue
4. Conclusions
IMDEA MATERIALS’ SUITE OF SIMULATION TOOLS
Object oriented, general purpose, parallel code for computational mechanics in solid, fluid, and structural applications including finite element and meshless capabilities.
http://www.materials.imdea.org/research/software
An open source library of material models (solids & fluids) for general numerical methods of continuum mechanics problems, which can be easily integrated into commercial codes.
Computational micromechanics tool to predict ply properties of fiber-reinforced composites from the properties and spatial distribution of the phases and interfaces to carry out in silico ply design and optimization.
An open source Kinetic Monte Carlo tool (coupled to a finite element code to include the effect of mechanical stresses and to an ion implant simulator) to simulate epitaxial growth and damage irradiation.
Suite of homogenization tools for polycrystals within the framework of crystal plasticity.
SINGLE CRYSTAL - POLYCRYSTAL HOMOGENIZATION
Virtual testing of polycrystall ine materials can now be achieved by means of computational homogenization.
Microstructural features: Grain size, shape and orientation distributions easily obtained by means of 2D and 3D characterization techniques (including serial sectioning, X-ray µtomography, 3D EBSD, X-ray diffraction, etc.)
Single crystal behavior: CRSS for each slip system and twinning (including latent and forest hardening) provided by
- Multiscale modelling- Mechanical tests of single crystals- Inverse problem: back up single crystal behavior from tests on polycrystals
• Homogenization of polycrystals• Nanoindentation
Key ingredients
transient in macroscopic flow strength behavior associated with a new flow localization pattern. But such transient flowstress behavior is eliminated upon sufficiently large deformation due to disruption of the irradiation induced defectstructure.
Regarding mesh convergence in the finite element simulations, simulations with two initial H are run using 403, 803,and 1603 elements in polycrystal calculations. Stress–strain curves are shown for the three mesh refinements forrl ¼ 3:61" 1021 m#3 in Fig. 7. While fine details of the strain localization may show changes to measurably finer meshes,gross behavior of the stress response appears to be reasonably well converged by 803 elements, which was taken as thereference mesh size in all of the above calculations.
Fig. 6. FE response to 10% strain. (a) Unirradiated, (b) rl ¼ 3:61" 1021 m#3.
Fig. 7. Mesh refinement results rl ¼ 3:61" 1021 m#3. h is the relative element size.
N.R. Barton et al. / J. Mech. Phys. Solids 61 (2013) 341–351348
CAPSUL
A tool to generate RVE of the microstructure (grain size, shape and orientation distributions) using as input statistical data obtained from 2D microscopy images.
A set of crystal plasticity constitutive models that take into account slip and twinning. Both monotonic and cyclic behavior can be considered by the combination of different laws for isotropic hardening, kinematic hardening and cyclic softening. The model is programmed as a UMAT subroutine for Abaqus.
An inverse optimization tool to obtain the crystal plasticity model parameters from the result of a set of mechanical tests (both microtests on single crystals or macroscopic tests on polycrystals).
A set of python scripts to generate cyclic loading conditions and to postprocess the results to obtain fatigue indicator parameters and other microfields.
a suite of tools for computational homogenization of polycrystals
http://www.materials.imdea.org/CAPSUL
RVE GENERATION microstructure generation
2D images of the microstructure 2D grain size distributionsegmentation
The 2D equivalent radius distribution is transformed into a 3D distribution assuming:
- Spherical grains
- The probability that the section of a sphere intercepted by a plane i is in the range given by
where is the maximum section of the sphere
pji
Aj
a 2 [ai�1, ai]
3D grain size distribution
StripStar, Basel University
RVE GENERATION microstructure generation
The set N of points for the tessellation, , is obtained by minimizing the error function
by means of the Monte Carlo method- : Experimental probability density function of the grain size ( )- : Trial probability density function of the grain size ( )
p = piO(p)
O(p) =NX
j=1
|PDFexp
(Vj
)� PDFtrial
(Vj
)| �V
[Vj , Vj +�V ]
�V = (Vmax
� Vmin
)/N
PDFexp
PDFtrial
[Vj , Vj +�V ]
The 3D RVE of the microstructure is an ensemble of N polyhedra (each one representing one grain) built from a Laguerre tessellation.
with distance to the nearest neighbor.
V (pi) = {dL(p, pi) < dL(p, pj), j 6= i}
dL(p, pi) = [dE(p, pi)]2 � r2i
ri pi
RVE GENERATION microstructure generation
Periodic RVEs (including orientation distribution information obtained from textures measurements by X-ray diffraction) are automatically generated and meshed with Gmesh or Dream3D (voxel models).
CRYSTAL PLASTICITY MODEL Crystal plasticity
S. R. Kalidindi, J. Mech. Phys. Solids 46, 267–271 and 273–290, 1998.
A crystal plasticity model was implemented as a UMAT in Abaqus/Standard
Multiplicative decomposition ; velocity gradient
Plastic deformation is accommodated by slip and twinning systems
F = FeFp L = Le + FeLpFe�1
Lp = Lpsl + Lp
tw + Lpre�sl
Nsl Ntw
f↵ volume fraction of twinned system ↵
Lptw =
NtwX
↵=1
f↵�tws↵tw ⌦m↵
tw
Lpre�sl =
NtwX
↵=1
f↵
0
@Nsl�twX
i⇤=1
�i⇤si⇤
sl ⌦mi⇤
sl
1
A
Lpsl =
✓1�
NtwX
↵=1
f↵
◆ NslX
i=1
�isisl ⌦mi
sl
The crystal was assumed to behave as an elasto-viscoplastic solid.
Plastic slip rate where
Twinning rate
Evolution of the shear strength
slip
twinning
re-slip
CONSTITUTIVE EQUATIONS
⌧ i = S : (si ⌦mi)
⌧↵ = S : (s↵ ⌦m↵)
minormal to the slip plane
sislip direction
Crystal plasticity
Phenomenological model: Monotonic deformation
�i = �0
✓|⌧ i|⌧ ic
◆ 1m
sign(⌧ i)
f↵ = f0
✓h⌧↵i⌧↵c
◆ 1m
with h⌧i =⇢
⌧ if ⌧ � 00 if ⌧ < 0
and f↵ = 0 ifNtwX
↵=1
f↵ � 0.80
⌧↵c = qtw�tw
NtwX
�=1
h0tw
✓1� ⌧�
⌧ twsat
◆atw
f↵�tw
⌧ i⇤
c = qsl�sl
Nre�slX
j=1
h0j
✓1� ⌧ j
⌧ jsat
◆asl
|�j |
⌧ ic = qsl�sl
NslX
j=1
h0j
✓1� ⌧ j
⌧ jsat
◆asl
|�j |+ qtw�sl
NtwX
�=1
h0tw
✓1� ⌧�
⌧ twsat
◆atw
|�� |
V. Herrera-Solaz, J. LLorca, E. Dogan, I. Karaman, J. Segurado. International Journal of Plasticity, 57, 1-15, 2014.
The crystal was assumed to behave as an elasto-viscoplastic solid.
The phenomenological model can account for the different cyclic deformation mechanisms, such as Bauschinger effect, cyclic softening/hardening and mean stress relaxation/ratcheting.
Plastic slip rate in the slip system i is given by
The slip resistance has three contributions:
- Isotropic hardening:
- Cyclic softening:
The backstress is introduced to include kinematic hardening
CONSTITUTIVE EQUATIONS Crystal plasticity
Phenomenological model for cyclic deformation
�i = �0
��⌧ i � �i��
⌧ ic
! 1m
sign(⌧ i � �i)
�↵
⌧ ic = ⌧0 + ⌧ iis + ⌧cs
⌧ iis =X
j
qij h0 sech2
����h0�
⌧s � ⌧0
����
⌧cs (�c) = � (⌧�s + h2�c)
⇣1� exp
�h1�c⌧�s
⌘
� =X
i
Z t
0|�i|dt
�c =X
i
tZ
0
���i��dt�
������
tZ
0
�idt
������
!
�i = c�i � d�i|�i|⇣
|�i|(c/d)
⌘mk
N. Ohno, J.-D. Wang. International Journal of Plasticity, 9, 375-390, 1983.
The model is based on the Orowan equation
where stands for the mobile dislocation density, bi the Burgers vector and
and is the average distance between barriers, the attempt frequency and the activation free energy.
where is the athermal contribution to the slip resistance
is the thermal contribution to the slip resistance
CONSTITUTIVE EQUATIONS Crystal plasticity
Physically-based model
⇢m
D. Rodríguez-Galan, I. Sabirov, J. Segurado. International Journal of Plasticity, 70, 191-205, 2015.
�i = ⇢imbivi
l ! �F
⌧ef = |⌧ i|� sia
vi =
8<
:
0 if ⌧ ief 0
¯l! exp
⇢��F
kT
✓1�
h⌧ ief
sit
ip◆q�sign(⌧ i) if 0 < ⌧ ief
sia
sta
LEVENBERG-MARQUARDT METHOD
Levenberg-Marquardt method:
are experimental data (i.e. stress-strain curves in different orientations) are the model predictions for a set of parameters (xi, yi)(xi, f(xi;�)) �
O(�) =nX
i=1
|yi � f(xi, �)| = ky � f(�)k
Optimization algorithm
Parametrization of the phenomenological constitutive models from experimental data is critical for the accuracy. This task is carried out using the Levenberg-Marquardt algorithm, well suited for non-linear problems.
Objective: find the set of parameters that minimizes Assuming a small perturbation of the model parameters and linearizing:
Levenberg and Marquardt minimized by adding a dumping factor to the expression of steepest descent, leading to the following linear set of equations:
whose solution provides the new set of parameters that minimizes .
� O(�)
f(� + �) ⇡ f(�) + J�
�
O(� + �) ⇡ ky � f(�)� J�k
O(�)�
(JT J + � diag(JT J))� = JT [y � f(�)]
� O
Jij =@f(xi,�)
@�j
FATIGUE LIFE PREDICTION Fatigue indicator parameters
Fatigue life is assumed to be controlled by crack initiation. Crack initiation is associated with the development of persistent slip bands in
grains favorably oriented for slip. The crack propagates within the slip band.
FIPp =2
3
Z t
0
����Lp : Lp
�1/2���dtAccumulated plastic strain
FIP iw =
Z t
0⌧ i�idtPlastic energy dissipated
Fatemi-Socie FIPFS
=�i
2
1 +K 0�
i
n,max
�y
!
The fatigue life is related to the range of the FIP, , in one fatigue cycle (averaged on a microstructural feature: grain, slip band, etc.)
This approach can take into the effect of microstructural features (grain size distribution, texture, etc.), defects (surface roughness, inlcusions, notches) as well as multiaxial stress states on the fatigue life.
N �FIP = FIPc
�FIP
VIRTUAL TESTING OF Mg ALLOYS Applications Mg alloys
Comparison withexperiments
0
50
100
150
200
250
300
0 0.02 0.04 0.06 0.08 0.1 0.12
C-ND T-RD T-ND T-ND/RD45
Stre
ss (M
Pa)
Strain
exp. sim.
simulation experiments
Error objective function
Numerical simulation (FEM; FFT) of mechanical tests
Error minimization
single crystalproperties
RVE microstructure: grain size, shape, texture
Single crystal plasticity model
V. Herrera-Solaz, J. LLorca, E. Dogan, I. Karaman, J. Segurado. International Journal of Plasticity, 57, 1-15, 2014.
Mg DEFORMATION MECHANISMS
Basal <a>, prismatic <a> and pyramidal <c+a> slip Tensile twinning
Applications Mg alloys
0
50
100
150
200
250
300
350
0 0.04 0.08 0.12 0.16
Tension-RDCompression-NDTension-NDTension-ND/RD 45
Stre
ss (M
Pa)
Strain
APPLICATION to AZ31 Mg ALLOY
AZ31 Mg alloy processed by hot rolling (25.4 mm thickness)
Average grain size (25 µm) Typical rolling texture Tensile and compression
tests along different directions.
Applications Mg alloys
DETERMINATION of AZ31 SINGLE CRYSTAL PROPERTIES
Parameters to be determined: initial strength, , saturation strength, , initial hardening modulus, , for basal, prismatic, pyramidal c+a slip and extension twining
Fixed parameters: - elastic constants of Mg single crystal- Latent hardening parameters- Hardening exponents- Strain rate sensitivity exponent
⌧0 ⌧s
h0
qsl�sl = 1.0 qsl�tw = 2.0asl = 0.6 atw = 1.0
m = 0.1
Input single crystal parameters(pure Mg, values in MPa)
Parameter Deformation initial RVEmode values 64 grains 512 grains 584 grains
64 voxels 512 voxels 4096 voxels
⌧0
Basal 1.75 11 20 23Prismatic 25 87 80 80
Pyramidal c+a 40 93 83 88Twinning 3.5 22 34 35
⌧s
Basal 40 13 23 25Prismatic 85 101 94 94
Pyramidal c+a 150 168 171 179Twinning 20 24 64 59
h0
Basal 20 1 20 20Prismatic 1500 2831 2831 2831
Pyramidal c+a 3000 3817 2990 2990Twinning 100 13 24 24
Table 3: Optimum values of the parameters that define the mechanical behavior of eachslip system and extension twinning in the AZ31 Mg alloy as a function of the RVE usedin the optimization process. Magnitudes are expressed in MPa.
midal and prismatic slip were found to take place at much higher stresses[52, 53].
The accuracy of the optimization process is clearly shown in Fig. 8(a), inwhich the experimental and computed stress-strain curves are very close forthe three orientations in the whole deformation range. It is worth noting thatthis excellent agreement is only possible because the physical mechanisms ofplastic deformation and the most important microstructural details are in-corporated in the computational model. In order to validate the optimizationprocedure, the tensile test in the RD-ND plane at 45� from both orientationswas simulated using the single crystal parameters obtained by optimizationand the RVE containing 584 grains. The numerical and experimental stress- strain curves are plotted in Fig. 8(b).The agreement between both is verygood and the average error per point similar to the one obtained for the fittedresults in the tensile ND tests (Fig. 8(a)).
The set of parameters obtained by the optimization procedure using threeindependent stress-strain curves as input (Table 3) can be compared withprevious data in the literature for rolled AZ31 Mg alloys. Our optimization
24
0
50
100
150
200
250
300
0 0.02 0.04 0.06 0.08 0.1 0.12
C-ND T-RD T-ND T-ND/RD45
Stre
ss (M
Pa)
Strain
exp. sim.
Initial simulations
PREDICTIONS WITH OPTIMIZED PARAMETERS
0
50
100
150
200
250
300
0 0.02 0.04 0.06 0.08 0.1 0.12
Compression ND Tension RD Tension ND
Stre
ss (M
Pa)
Strain
exp. sim.
0
50
100
150
200
250
300
0 0.02 0.04 0.06 0.08 0.1 0.12
Tension ND/RD 45
Stre
ss (M
Pa)
Strain
exp. sim.
V. Herrera-Solaz, J. LLorca, E. Dogan,I. Karaman, J. Segurado. International Journalof Plasticity, 57, 1-15, 2014.
The input has to provide information for all the deformation mechanisms (minimum two stress-strain curves, better three and, in the future, texture).
The results obtained have been shown to be independent of the initial values of the CRSS for each system.
ROBUSTNESS and UNIQUENESS
V. Herrera-Solaz, J. Segurado, J. LLorca. European Journal of Mechanics A/Solids, 53, 220-228, 2015.
Applications Mg alloys
FATIGUE OF WROUGHT IN718
Wrought IN718 subjected to fatigue at 400ºC. Average grain size ≈ 12 µm. The alloy subjected to cyclic deformation undergoes isotropic hardening,
Bauschinger effect, cycles softening, mean stress relaxation and ratcheting. Cyclic deformation is simulated by means of the simulation of an RVE of the
microstructure. The parameters of the phenomenological cyclic fatigue model are calibrated using the Levenberg-Marquardt method from the experimental cyclic stress-strain curves.
Applications Fatigue IN718
D. Gustafsson; J. J. Moverare; K. Simonsson; S. Sjöström. J. Eng. Gas Turbines Power., 133, 094501, 2011.
FATIGUE LIFE PREDICTION
Fatigue life predictions for R= 0 and R=-1 were carried out using
where the FIP was averaged in all the slip systems in each grain.
Applications Fatigue IN718
N �FIP iw = FIPc
FIPc =N
�FIP iw
CONCLUSIONS
CAPSUL is a suite of tools for crystal plasticity homogenization of polycrystals which can be used to take into account the effect of the microstructure on the mechanical properties of polycrystals.
It includes tools to built a realistic RVE of the microstructure from easily available experimental data (2D sections of the microstructure, X-ray diffraction textures) and a robust inverse homogenization strategy (based on the Levenberg-Marquardt method) to obtain the single crystal properties from experimental data (micropillar compression of single crystals, monotonic and cyclic polycrystal tests, etc.)
Current capabilities are focussed in the prediction of monotonic and cyclic deformation as well as fatigue life. Further developments will include creep and thermo-mechanical fatigue.
MICROMECH project (Microstructure-based Material Mechanical Models for Superalloys), EU 7th FP, Clean Sky JTI, grant agreement nº 62007
ACKNOWLEDGEMENTS
VIRMETAL project (Virtual Design, Virtual Processing and Virtual Testing of Metallic Materials), ERC Advanced Grant, EU H2020 programme, grant agreement nº 669141
MAGMAN project (Analysis of the microstructural evolution and mechanical behavior of Mg-Mn-rare earth alloys)