component scale process model for metal additive … of the future/3d... · modeling material...
TRANSCRIPT
VTT TECHNICAL RESEARCH CENTRE OF FINLAND LTD
Process Model for Metal
Additive Manufacturing
Tom Andersson, Anssi Laukkanen, Tatu Pinomaa
NAFEMS NORDIC "Exploring the Design Freedom of
Additive Manufacturing through Simulation"
22.11.2016 Kalastajatorppa
2 A – PROCESS SIMULATION I – METALS
12/12/2016 2
Contents
Multiscale modeling for metal additive manufacturing: concepts
and ICME toolsets
Thermomechanical finite elements based process model and
machine integration
Case analysis results & examples
Evaluation of material properties & performance
412/12/2016 4
PRODUCT PERFORMANCE
AND COST
Multiscale modeling for metal
additive manufacturing
Discrete modeling of
powder bed physics
Thermodynamics and
phase fields
Modeling material structure → properties and performance
Topology optimization
Thermomechanical modeling of
selective laser melting
→ Part specific optimized process
design
SLM process design and optimization
Powder and alloy design
Material property & performance design
Part geometry design
Microstructural FEM to predict materials (and
component) properties, behavior and performance
Powder bed model for
heat transfer and
solidification
Possible redo of shape
optimization to better design
the production phase
support structure
PF model for reactive wetting,
phase and microstructure prediction
512/12/2016
Model Generation &
I/O with SLM Machine
Currently with
SLM125, integration
via log and build files
(CLIs++).
Parsing scan strategy from CLIReproduce scan strategy 1-1 with the SLM125 machine
Extensible since the basis classes describing the AM build process have been developed and established
for the SLM125
Model creation via a “push of a button”:
1) interface to read and process machine build
files (and logs) and 2) create the
thermomechanical process model directly for
simulation with a solver of interest (presently
Abaqus, code_aster, some OS libraries)
Geometry parsed from bracket ,example
layers 1 & 100
12/12/2016 6
1. Scan strategy and
process parameters
parsed from cli & system
configuration and log
files
2. Re-construct scanning
strategy and process
parameters, transfer via
user subroutines and
databases to finite
element model.
3. Model definition + heat
source via separate
library (meshing, heat
transfer, properties etc.)
4. Material model definition
via separately library
(thermal and mechanical
models)
5. Solution using either
implicit or explicit
approaches
6. Post-processing for
engineering material
properties
712/12/2016 7
SLM transient thermal process model
The process model consists of the powder bed, laser heat source, and the
respective thermal initial and boundary conditions
Convection BC to
powder bed sides
Convection BC via
top surface
Base plate at uniform
temperature or convection to a
thermal sink
Laser heat source
Beam described as a
moving Gaussian
surface heat flux
Transient thermal solution utilizing a moving mesh motif over the respective process history and layers of
interest. Powder vs solid properties following common conventions, using packing density to homogenize
properties.
As output
thermomechanical
history, residual
stresses, strains
etc.
Derive a section of powder bed and
update its size as required. Adaptive
mesh refinement and coarsening.
812/12/2016 8
Kinetic model for diffusion affiliated phase
transformations
• Nucleation governed by ΔG for
precipitation acting as a driving
force.
• Nucleation rate via Berker-Döring,
including free energy reduction due
to phase transformation and free
energy increase due to interface
between phases. Growth via the
Zener equation.
• If cooling rate over threshold → time
independent transformation to
martensite (CCT like tracking)
• Otherwise, time dependent
nucleation and growth process.
“Tweaking” or calibration/validation
using CCT curves, but predictive.
• Enables tracking of local powder &
solid material state on the basis of
thermal history.
• In the future considered a template
for coarse graining smaller
resolution, e.g. phase field
simulation, results.
Thermodynamical model based on classical nucleation and
growth theories to address precipitation of metastable and
stable phases. 1st Nucleation and growth, followed 2nd by
growth and coarsening (model layout following Deschamps &
Brechet, 1999). Example of isothermal transformation in H13
steel:
Growth and size of precipitates
Solute concentrations
912/12/2016 9
SLM transient thermal process model, some
layers of a bracket geometry (~20)
Temperature isosurfaces, example from layer 1100,
20 by 20 cm powder bed in model.
Laser power P = 100 W, beam
velocity v = 1000 mm/s
Laser power P = 150 W, beam
velocity v = 1000 mm/s
Laser power P = 250 W,
beam velocity v = 1000 mm/s
Laser power P = 375 W,
beam velocity v = 1000 mm/s
Bracket geometry of this
case study, approx. 2k
layers in experimental build.Linking between local thermal
solution, process parameters,
scan strategy and part features
1012/12/2016 10
SLM transient thermal process model, test
samples (thermal field, layer 20)
Modeling porosity evolution during metal AM
process
Sample
No.
Scanning speed
mm/s
Power
W
VED
J/mm3
Measured
porosity (%)
Calculated porosity (%)
(simulation model)2 666.7 100 50 12.74 14.73 1200 180 50 2.67 2.74 1200 228.51 63.5 0.23 0.611 1200 200 83.3 0.15 0.424 993.7 250.51 84 0.09 0.08
Process parameters, measured porosities and
calculated porosities of five DOE samples
from second sample set
1212/12/2016 12
Microstructure founded computation of cycles to short crack initiation
Image based microstructural modeling of
fatigue crack initiation
Derivation of time to initiate short fatigue cracks on the basis of microstructural models of SLM microstructures (precipitate
hardened steel). The defects which completely deteriorate the fatigue performance of the microstructure are high aspect ratio
cracks.
strain
amplitudestrain
amplitude
microstructure
porosity
porositylarge
inclusion
One example
of defect
structure
Multiple small defect
types (inclusions,
porosity)
Crack-like
defects
1st principal stress
Equivalent plastic strain
Four different microstructures, fatigue
performance indicator
12/12/2016 13
Image based microstructural modeling of
fatigue crack initiation FS parameter has been demonstrated to correlate to multiaxial fatigue crack
initiation results both in high and low cycle fatigue (in shear dominated crack
initiation)
𝑃𝐹𝑆 =∆𝛾𝑚𝑎𝑥
𝑝∗
21 + 𝐾′ 𝜎𝑚𝑎𝑥
𝑛∗
𝜎𝑌
∆𝛾𝑚𝑎𝑥𝑝∗
is the maximum cyclic plastic shear strain over a finite volume of material,𝐾′
incorporates the influence of normal stress,𝜎𝑚𝑎𝑥𝑛∗ is the maximum stress normal to the plane of
∆𝛾𝑚𝑎𝑥𝑝∗
and 𝜎𝑌is the cyclic yield strength.
For estimation of cycles to initiation, the FS parameter is linked to the Coffin-
Manson (CM) strain life
𝑃𝐹𝑆 = 𝛾𝑓′ 2 𝑁𝑖
𝑐 ,
𝛾𝑓′ is strain affiliated coefficient application to crack initiation at the scale of interest, 𝑁𝑖 the
cycles to initiation and 𝑐 is the CM exponent.
Modeling porosity evolution during metal AM
process Modeling subsequent product performance (with
respect to fatigue)
Performance gains in product lifetime, interpretation of
defect significance
Samples
2, 3, 4,
11, 24
1412/12/2016 14
Raw data, 6 by 6 points for a
property surface
Power [W]
Velocity
[mm/s]
Melt pool
instabilities
Defect density = ratio of damaged material (pores, cracks,
delaminations, gas pores,…).
Defect density = 1 (“perfect solid”),
Defect density = 0 completely porous/damaged material
“Metallic foam”
“Perfect solid”
Non-load critical
components
Critical components
Cracks,
porosity
Experimental example
Defect density for different computed process
parameter sets
1512/12/2016 15
Summary and conclusions
Framework for modeling metal AM processes and enabling
streamlined model generation has been presented. The
toolset provides means to derive complex models with ease
as well as modify them e.g. to run DoE on the model.
The approach enables the definition of a machine realistic
model of the metal AM build process.
The model outcome can be exploited in evaluation of
engineering material properties as well as e.g. analysing in
more detail part performance with respect to fatigue.
H-adaptivity and mesh refining/coarsening is to be further
developed to decrease computing times and increase model
sizes.
Improve links with other tools to improve physical
descriptivines, e.g. solidification and powder bed models to
enhance and mitigate limitations of thermomechanical finite
elements.