Multiscale Modeling of Materials using

Machine Learning

M. Arul Kumar
Associate Professor
Mechanical Engineering Department
IIITDM Kancheepuram

1
1
 Course Objectives and expected outcomes

To provide an opportunity for the students to appreciate how

the integration of machine learning with multiscale modeling
of materials advances the field to meet rapidly growing
industrial demands.

• Understand the concept of multiscale modeling of materials and the basics of

machine learning algorithms.
• Ability to develop numerically efficient computational models for engineering
applications.

2
 Course content
Review of Materials Science and Machine Learning concepts
• Crystal systems/structures, symmetry, anisotropy, tensorial representation
of crystal properties
• Multiscale hierarchical microstructure (polycrystals, grains, sub-
structures, defects, and atoms) and its effect on properties
• ML algorithms: supervised, unsupervised and reinforcement learning methods

Conventional Multiscale Modeling of Materials

• Atomistic (density functional theory and molecular dynamics) and micro-scale (defect
dynamics) modeling
• Meso-scale (phase field and meso-plasticity) and macro-scale (finite element methods)
modeling
• Integration of length and time scales for materials modeling
3
 Course content
Machine Learning for Materials Science
• Materials Data: Source, size, composition, version, and uncertainty
• Materials modeling: Model selection, data scaling and normalization, and model evaluation
• Model development: Calibration, validation, benchmarking, and uncertainty quantification

Machine Learning and Multiscale Modeling of Materials

• Concurrent and sequential transfer of data, theory, and correlations
• Data-driven and physics-driven machine learning approaches
• Bayesian approach to multiscale modeling of materials: Surrogate models

4
 Reading Materials
• R. J. Asaro and V.A. Lubarda: Mechanics of Solids and Materials,
Cambridge University Press, 1st edition, 2006.
• D. Barber: Bayesian Reasoning and Machine Learning, Cambridge
University Press, 2nd edition, 2012.

• U.F. Kocks, C. N. Tome, and H. R. Wenk: Texture and Anisotropy, Cambridge University Press,
1st edition, 2000.
• J. Fan: Multiscale Analysis of Deformation and Failure of Materials, John Wiley & Sons, 1st
edition, 2011.
• A. C. Mü ller and S. Guido: Introduction to Machine Learning with Python, O’Reilly, 1st edition,
2016.
• Aurelien Geron: Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow:
Concepts, Tools, and Techniques to Build Intelligent Systems, Shroff/O’Reilly, 3rd edition, 2022.
5
 Logistics

Schedule: (3-1-0-4)
• Theory classes:
Email:
• Monday: 9-9.50 am
marulkr@iiitdm.ac.in
• Wednesday: 10-10.50 am Office: L-219-J
• Friday: 11-11.50 am Google classroom
• Practice session: Thursday: 5-5.50 am Whatsapp-group
• Back-up hour: -

6
 Logistics

Schedule: (3-1-0-4)
• Theory classes:
Email:
• Monday: 9-9.50 am
marulkr@iiitdm.ac.in
• Thursday: 5-6.50 am Office: L-219-J
• Practice session: Friday: 11-11.50 am Google classroom
• Back-up hour: - Whatsapp-group

7
 Logistics

Course evaluation rubrics

Email:
• Term-project: 25 marks marulkr@iiitdm.ac.in
Office: L-219-J
• Mid-sem: 25 (15 + 10)marks
Google classroom
• End-sem: 50 (20 + 30) marks Whatsapp-group

8
 Expectations

Punctuality

Academic honesty

Discipline
9
The blind men and the elephant

John Godfrey Saxe, (1865)

10
Let’s start the course

“Nothing in life is to be feared,

it is only to be understood. Now
is the time to understand more,
so that we may fear less.”

- Marie Curie

11
Materials are everywhere!

12
 Materials Science: Interdisciplinary field
Helps to realize concepts! Smart materials make smart phones!

www.metallurgyfordummies.com

American Ceramic Society Bulletin, 97 (2018) 9

13
 Materials Science and Engineering

How it is looking?

Ro2man et al., Mater Today (2021)

www.theengineerspost.com

What do we get?
How to make?

www.thomasnet.com
TRUMPF

How it behaves/respond?
www.mac-tech.com 14
 Material science aids to
select or design application-specific-materials

15
 Material science aids to
assess the quality of the manufactured components

cerEficaEon

16
 Material science aids to
quantify the properties/performance and limits

Bhopal gas tragedy (1984) Liberty ships accidents (World war – II)

17
Materials in aerospace industry

18
 Structural materials

Metals and Ceramics

alloys and glass

Polymers
and
plastics
19
 Structural materials: Metals
Metals o Crystalline in solid form De la Piro technia (Technique
o Opaque, ductile, malleable, of fire/warm process)
and dense and good conductors - V. Biringuccio (1540)
alloys of heat and electricity.

Metal is a mineral body, by nature either liquid or somewhat hard.

The latter may be melted by the heat of the fire, but when it has
cooled down again and lost all heat, it becomes hard again and
resumes its proper form. In this respect it differs from the stone
which melts in the fire, for although the latter regain its hardness, yet
it loses its pristine form and properties.

First printed book (in Italian) on metallurgy

English translation in 1945: Cyril Smith
(Manhattan project) and Martha Gnudi
20
 Structural materials: Metals and alloys

Source: Makin Metals

Stone age Copper age Bronze age Iron age ?? 21

Lattice structure of crystalline metals

Source: www.ggspdt.com 22
 Multi-scale nature of materials

370 B.M. Morrow et al. / Materials Science & Engineering A 613 (2014) 365–371 component

Q. Sun et al. / Scripta Materialia 90–91 (2014) 41–44 43

Microstructure

Matter
Defects
Fig. 7. Key frames from in situ straining video showing interaction of twins under tension. (a) t ¼0 s. (b) t¼ 15.8 s. (c) t ¼34.2 s. (d) t¼ 55.7 s.

electron diffraction pattern of a twinning boundary along the ½1 21

Atoms
Figure 1. (a) Morphology of a {1 0 !1 2} twinning tip. The theoretical {1 0 1! 2} twinning plane is denoted by a yellow solid line. (b) Selected-area
! 0" zone axis. (c) HRTEM image of the twinning tip highlighted by a red circle in
(a). Three kinds of facets are shown, including a straight {1 0 !1 2} interface, a BP and a PB, which are marked by yellow and red lines, respectively.
(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

23
he deformation is very small and the specimen is elastic. Necking in a uniaxial cylindrical ten
Multiscale
Correspondingly, nature
a linear of materials
relationship exists betweenand
the itssymmetric.
effect on properties
For flat specimens, diffu
minate in fracture. But usually aft
diffuse neck, the deformation continu
load until the development of a loc
Elastic
leads ultimately Plastic
to ductile fracture. T
is a narrow band with about equal to
Initial
and inclined at an angle to the speci
width of the specimen. A sudden d
tic

Plastic stress can be observed when the spe

Elas

In this stage, the data obtained from

no longer valid, since the uniaxial d
tion does not stand.

2.3 | True stress–strain cur

FIGURE 2 Engineering stress‐engineering strain curves from
niaxial tensile test [Colour figure can be viewed at The engineering stress and the eng
24
 Computational modeling of materials
C. Muyan and D. Coker: Finite element simulations for investigating the strength characteristics 1351

(Muyan and Coker, 2020)

Figure 18. Damage evolution – IFF (A) or IFF (B) – in the pressure side, internal flange and suction side of the blade (from top to bottom in
a panel) at 50 % of extreme flapwise loading.

Continuum scale FEM simulations

Empirical constitutive laws 25
Multiscale modeling of materials

Scale bridging

Capturing physics

Robustness

Computational efficiency

Predictability

Ease of implementaEon
Raabe et al., 2010 26