Engineering Conferences InternationalECI Digital ArchivesHarnessing The Materials Genome: AcceleratedMaterials Development via Computational andExperimental Tools

Proceedings

Fall 10-3-2012

Predicting Microstructure-Property Relationshipsin Structural Materials via Multiscale ModelsValidated by In-Situ Synchrotron ObservationPeter D. LeeManchester X-ray Imaging Facility

Lang YuanManchester X-ray Imaging Facility

Chedtha PuncreobutrManchester X-ray Imaging Facility

S. KaragaddeManchester X-ray Imaging Facility

Follow this and additional works at: http://dc.engconfintl.org/materials_genome

Part of the Biomedical Engineering and Bioengineering Commons

This Conference Proceeding is brought to you for free and open access by the Proceedings at ECI Digital Archives. It has been accepted for inclusion inHarnessing The Materials Genome: Accelerated Materials Development via Computational and Experimental Tools by an authorized administrator ofECI Digital Archives. For more information, please contact franco@bepress.com.

Recommended CitationPeter D. Lee, Lang Yuan, Chedtha Puncreobutr, and S. Karagadde, "Predicting Microstructure-Property Relationships in StructuralMaterials via Multiscale Models Validated by In-Situ Synchrotron Observation" in "Harnessing The Materials Genome: AcceleratedMaterials Development via Computational and Experimental Tools", J.-C. Zhao, The Ohio State Univ.; M. Asta, Univ. of CaliforniaBerkeley; Peter Gumbsch Institutsleiter Fraunhofer-Institut fuer Werkstoffmechanik IWM; B. Huang, Central South University Eds,ECI Symposium Series, (2013). http://dc.engconfintl.org/materials_genome/12

Henry G. J. Moseley

b. 1887

d. 1915

Predicting MicrostructurePredicting Microstructure--Property Property

Henry G. J. Moseley

b. 1887

d. 1915

The University of Manchester

Predicting MicrostructurePredicting Microstructure--Property Property

Relationshops via multiscale models Relationshops via multiscale models

validated by synchrotron observationsvalidated by synchrotron observationsPeter D. Lee,Peter D. Lee, Lang Yuan, Chedtha Puncreobutr, S. Karagadde Lang Yuan, Chedtha Puncreobutr, S. Karagadde

Manchester X-ray Imaging Facilitywww.mxif.manchester.ac.ukwww.mxif.manchester.ac.uk

Nature versus Nurture in the MGI…Nature versus Nurture in the MGI…

Why simulate Microstructural Evolution (ICME)?To track genome evolution across length scales and processes

90

σmax (MPa)

3. Machining:3. Machining:

-30

90

30

σmax (MPa)

2. Heat Treatment:2. Heat Treatment:residual stresses50

290

170

L (µm)

1. Casting:1. Casting:Alloy/Microstructure dependent properties

AlloyAlloy

Genome?Genome?

Page 2

-30

90

303. Machining:3. Machining:residual stresses

εmax (××××10-3)

0

1.2

0.6

4. Service:4. Service:Cyclic stresses/strains

1

3

2

Nf (××××106)

Crack Crack Crack Crack InitiationInitiationInitiationInitiation

5. Component 5. Component Performance: Performance: Fatigue Life Prediction

A B

CLee&Hunt,MRS, MCWASP, Acta, 1994-7 Lee et al, Mat. Sci. Eng. A,, 2004Maijer, Lee et al, Met. Trans, 2004John Allison, ICME. JOM, 2006

IIII’’’’ve modelled alloy processing for 25 ve modelled alloy processing for 25 ve modelled alloy processing for 25 ve modelled alloy processing for 25

years years years years ---- What lessons have I learned?What lessons have I learned?What lessons have I learned?What lessons have I learned?

1. When using multi-scale, through process modelling (or ICME), there are sufficient unknown parameters one can tune, you

usually get the answer you want…

2. For structural materials, it is not only the innate alloy properties, but also how you manufacture the component that matters

I.e. Nurture can be more important than Nature if you want to get the most out of a

Material’s Genome. 3

Typical number of Nurturing steps…

Thermal Treatm

ent

Thermal Treatm

ent

Melt Homogenize TMP TMP Machining Weld ServiceSolidify

ASTM GSASTM GSASTM GSASTM GS

0 4 8

With Rolls-Royce; Special Metals & Wyman-GordonUniv. of Cambridge (Tin) and Birmingham (Ward)Kermanpur, Tin, Lee, JOM 56(3) 2004, 72-78. or Tin, Lee, et al Met. Trans. A., 2005.

Evolution of theMaterial GenomeDuring Nurturing!

Is optimising Nature, then providing good Nurturing enough?

• 1989 Kegworth air crash, caused by fan blade loss, manufacturing defect

• 1985, Manchester, failed combustor weld repair -porosity

Lesson 3 - Lifing is often limited by a deviant microstructural feature, rather than the average, even

though it may have the same genome…

My Conclusion…My Conclusion…My Conclusion…My Conclusion…

The Materials Genome Project needs to map out not only the average

behaviour, but also the distribution in behaviour, but also the distribution in behaviour, including the rebels

6

Example 1: Predicting deviant Example 1: Predicting deviant Example 1: Predicting deviant Example 1: Predicting deviant

microstructures in Nimicrostructures in Nimicrostructures in Nimicrostructures in Ni----based SX turbine based SX turbine based SX turbine based SX turbine

blades blades: or blades blades: or blades blades: or blades blades: or the Freckle Rebelthe Freckle Rebelthe Freckle Rebelthe Freckle Rebel

Beckermann, Flemings Symposium, 2001

Solidification of Ga-25wt%In alloy, G .5K/mm, R 8.1 8.1 8.1 8.1 µµµµmmmm////ssssX-ray In-situ observation,

Courtesy HZDR,DEµµµµMatIC Simulation

wt% Ga

N Shevchenko et al 2012

Mater. Sci. Eng. 33 012035www3.imperial.ac.uk/advancedalloys/

Solidification of Ga-25wt%In alloy, G .5K/mm, R 8.1 8.1 8.1 8.1 µµµµmmmm////ssssX-ray In-situ observation,

Courtesy HZDR,DEµµµµMatIC Simulation

Upwards liquid flow increases solute concentration in the channel with local

N Shevchenko et al 2012

Mater. Sci. Eng. 33 012035www3.imperial.ac.uk/advancedalloys/

wt% Ga

with local remelting, remelting secondary arms and stopping primary arms

1000

10000

Rayle

igh

nu

mb

erRayleigh number

10000

1000

To scale the microstructural model to a

macroscopic level, we can approximate it via the

Rayleigh number, to predict the rebels…

Rebels

Simulations

1

10

100

0 1 2 3 4 5 6 7

Rayle

igh

nu

mb

er

Data group from experiments

Ra_crit ≈ 35Experimental data

1. Tewari,19962. Sarazin, 19883. Bergman, 19974. Wang, 19885. Streat, 19746. Sarazin, 1990

Rayleigh number

100

10

1

Data group from experiments

0 1 2 3 4 5 6 7

Yuan L and Lee PD, Acta Mater., 2012

Average Guys

Example 2Example 2Example 2Example 2

Understanding why eating your Understanding why eating your Understanding why eating your Understanding why eating your

spinach is not always good for spinach is not always good for spinach is not always good for spinach is not always good for spinach is not always good for spinach is not always good for spinach is not always good for spinach is not always good for

your strength, or your strength, or your strength, or your strength, or predicting Fepredicting Fepredicting Fepredicting Fe----intermetallic Rebelsintermetallic Rebelsintermetallic Rebelsintermetallic Rebels…………

Why worry about deviant microstructures like pores and Fe-intermetallics?

They initiate failure during service

Yi, J.Z. et al. Met. Trans. 34A, 2003.

And they alter formabilityduring manufacturing

What needs to be simulated?

α α α α Al +Si

ββββΙΙΙΙ

ββββΙΙΙΙΙΙΙΙ

ββββΙΙΙΙΙΙΙΙΙΙΙΙ

α α α α Al + Al2Cu

α α α α Al

Grain nucleation and Growth

Solute partitioning(Si & H)at solid/liquid front

Intermetallic nucleationAnd growth

Page 13

α α α α Al

α α α α Al + Si

ββ ββΙΙΙΙ ΙΙΙΙ

ββ ββΙΙ ΙΙ

Eutectic growth

And growth

Growthrestricted by solid

For speed we model with 10µm elements, and approximated anisotropy

• Nucleation: nanometres

• Dendrite tip radius: 1 micron

• Coarsening: 10 microns

• Solute diffusion: 10 microns

Empirical fn

Fs>0.5, approximate...

Borderline...

Fs>0.5, ~Scheil • Solute diffusion: 10 microns

• H diffusion: 100 microns

• Pores: 10-100’’’’s microns

• Intermetallics: 10-100’’’’s microns

• Grain size: 100-1000 microns

Fs>0.5, ~Scheil btwn dendrites

�

�

�

� Page 14

Pββββ

Synchrotron CT Characterization of Fe Intermetallic Morphology compared to model prediction

100µµµµm

100µµµµm

400µµµµm

Typical Typical Typical Typical predictionpredictionpredictionprediction

How do we capture the genome and span scales?Via model-based constitutive equations

Maximum

Intermetallic Length

Microns

1000’s micromodelpredictions

Regression Fit

1000

So

lid

ific

ati

on

Tim

e)

ln Lmax

= b0

+ b1ln tsCFe

init

and statistical variation

0

So

lid

ific

ati

on

Tim

e(L

n t

s )

Initial Fe Content CFeinit

Challenges: 1. Improved statistical tracking of multi-variant distributions2. Fitting highly coupled phenomena (e.g. Pressure) Page 16

Coupling deviant microstructure to Lifing

CastingHeat

Treatment

Cyclic

Loads

Manufacturing: Service:

Residual StressMicrostructure Cyclic

Machining

Fatigue Life

290

Lmax (µm)

170

Page 17

Residual StressMicrostructure

and Defect

Cyclic

Stress/Strain

290

Lmax (µm)

50

170

Fatigue Life

90

σmax (MPa)

-30

30

90

σmax (MPa)

-30

30+ +

1.2

εmax (××××10-3)

0

0.6+ = ???

50

Li et al, MCWASP 2006, MMTA 2007

Crack Initiators

Crack outline

1 mm

Crack under opening load of 100 MPa

Fe-intermetallic

σa

Surface Si

10 µm

σa

PoreSi

σa

20 µm

σa

Page 18

( )2 2

2

1 10

0 1

1 amax i

a y

s t

t t

f i p f

CN N N k a

kC a

β

σ

α σε

λ σ σλ

− ×

− + − +

= + = + + −

Gao YX et al., Acta Mat, 2005Li P et al., AEM, 2006

εpl (××××10-3)

0.24.010

εpl (××××10-3)

0.22.04.8

σmax (MPa)

80140200

Maximum size of deviant microstructures: Pores or Fe-rich intermetallics - Lmax

Challenges: models for flow stress of each phase,interface strength and adding debonding model, etc…

Does it work?

Accurate prediction of failure location was achieved (deviant feature - pores).

AB

3

Nf (××××106)

1

2

Crack InitiationCL

AB

C

C

In situ observation shows Fe reduces hot

forming/increases hot-tearing of A319, why?

0.2wt.%Fe 0.6wt.%Fe

Comparison of analysis techniques

Imaging DVC Tracked Quantification

We can directly compare to predicted influence

of triaxiality on localisation of damage

0.330.260.20

0.010.00

0.13

III

Triaxiality(σσσσh/σσσσeq)

1st hypothesis, the intermetallics reduce

interdendritic flow – we can use image based

modelling to directly simulate the flow

275 µm 275 µm

Dendrite Dendrite + Intermetallics

Flow Simulation results – <10% reduction in flow

Geometry file k [um^2] k [d] Input pressure [Pa] Output pressure [Pa] Flow rate [um^3.s^-1] Viscosity [Pa.s]

Without Plate 178.5231 180.88853 130000 100000 4.87E+11 0.001

With Plate 167.14487 169.35954 130000 100000 4.86E+11 0.001

Synchrotron imaging with direct simulations

demonstrated flow is only a minor effect, so now

we need another hypothesis!

275 µm

However, the synchrotron observations helped

answer other questions:

1. when/where do the intermetallics nucleate, and

2. do pores nucleate on intermetallics

Cooled at

3°°°°C/min

Do pores nucleate on intermetallics?

Cooled at 3°°°°C/min

ConclusionsConclusionsConclusionsConclusions

In situ observation shows us the kinetics of microstructural formation, clarifying dominate mechanisms and causality

Nurture can matter as much as Nature

We need to look out for the rebels when during the Materials Genome Initiative

28

Acknowledgements

• Ford, Tata Steel, GE

• Diamond Light Source & I12 Team

• EPSRC (grant EP/I02249X/1)

• Research Complex at Harwell

• The MXIF Team:• The MXIF Team:• UoM

• RCaH

• Diamond

• Imperial College

Questions?

29