This course aims to provide a succinct overview of the emerging discipline of Materials Informatics at the intersection of materials science, computational science, and information science. Attention is drawn to specific opportunities afforded by this new field in accelerating materials development and deployment efforts. A particular emphasis is placed on materials exhibiting hierarchical internal structures spanning multiple length/structure scales and the impediments involved in establishing invertible process-structure-property (PSP) linkages for these materials. More specifically, it is argued that modern data sciences (including advanced statistics, dimensionality reduction, and formulation of metamodels) and innovative cyberinfrastructure tools (including integration platforms, databases, and customized tools for enhancement of collaborations among cross-disciplinary team members) are likely to play a critical and pivotal role in addressing the above challenges.
Role of Structures in PSP Linkages
Loading...
From the course by Georgia Institute of Technology
Materials Data Sciences and Informatics
63 ratings
From the lesson
Materials Knowledge and Materials Data Science
• Understand property, structure and process spaces
• Learn about Process-Structure-Property Linkages
• Learn what does Materials Knowledge mean
• Learn about a role of Data Science in Materials Knowledge System
• Overview approaches and main components of Data Science
• Learn about a new discipline - Materials Data Sciences
Meet the Instructors
Dr. Surya KalidindiProfessor
The George W. Woodruff School of Mechanical Engineering
[MUSIC]
Hello everyone.
Welcome to Materials Data Science and Informatics Course.
I'm Surya Kalidindi, the Instructor for this class.
The title of today's lecture is Role of Structure in PSP Linkages.
The learning outcomes for this class are to understand the key role
played by material structure and material development efforts.
And to also understand the challenges involved in
quantifying the material structure.
I would like to start with this slide from a previous module from the guest lecture
by approximate doll.
Where he showed you the different stakeholders
in the materials development activities.
The stakeholders the materials suppliers,
the original equipment manufacturers and the design engineers.
And they all exchange they all conduct various studies and exchange information,
so that they can make a product design.
As he pointed out, the transactions within the materials, specialists, and
the design engineers happens along this interface.
At that interface it's essentially a material selection activity,
because the materials develop first of producing properties,
tables of properties, and the design engineer
is simply selecting materials based on their available properties.
Our goal is to push this design front to
include material structure within the design space.
We would like the designers to use all of their available materials,
potentially available materials, in coming up with the products.
That greatly enhances the design space.
Because you're not only designing the product but
you're actually designing the material with the product.
However, it is also significantly more challenging,
because now the design space has exploded.
And now we have many variables to deal with.
The reason we want to do this is there is a Amajor payoff, expected payoff.
And essentially, all advanced technologies are currently impeded
by lack of advanced materials that can provide the performers the need,
so once we can actually make these materials Into products
that have superior performance, the potential payoffs are unlimited.
Now, let's start by looking at what is currently done in labelling materials.
The current practice in labelling materials is very rudimentary.
It had designations such as these numbers that tell us something about the material.
Oftentimes they tell us the chemistry involved in the materials.
For as this particular example let's look at 7075-T6 Aluminum
which is essentially an aerospace alloy, aluminum alloy.
In this particular case, that designation tells us that the chemical
composition of this alloy is in that particular range, and has these particular
elements in addition to aluminum, which is basically the bulk of the alloy.
Now, it also tells us, this particular designation, T6, also tells us
that the material is expected to have these particular property ranges.
One thing you'll notice is that the property ranges are pretty broad.
Usually there's an expectation that this particular
temper is achieved by a particular set of heat treatments.
But it doesn't have to be.
There can other ways to get to the particular set of combinations of
properties.
So, as you can see the main deficiency in this kind of labelling materials is that
it provides only information on chemical composition and final processing step.
It does not account at toward any information on the structures
that are these materials.
And as we said before, materials structure is an extremely important in controlling
the properties of the materials.
And the third deficiency in this kind of labelling is that the properties
that are guaranteed properties, are a limited set of properties.
In other words, just because I know these particular properties,
they may or may not be the best properties to be looking at for
a particular applications I have in mind.
So, these are the deficiencies of the current system.
And to do something better, as we said before we have to embrace the concept
of structure and include the definition of structure in the material definition.
In order to do that, one has to embrace statistical metrics or
statistical measures of material structure.
So, some of the statistical measures are,
some of the currently used statistical measure are shown here.
For example, you can think of grain size distribution,
not just an average grain size, but a grain size distribution.
Or you can think of a grain orientation distribution.
These are parts of grain orientation distribution.
Yet another statistic of interest might be misorientation distribution.
So, there are many distributions one can define given a structure like this.
However, these distributions are statistical in nature
are exactly what we need in defining the material.
The problem however is that the set of
statistical measures one can define is an extremely large set.
What we really would like is a relatively small set of salient statistical measures.
And this again is where data science can be extremely useful and valuable to us.
S, the reason for data science approach is as the following: it
is schematically shown here, this is just a schematic.
And in this version what we start with is a certain number of statistical measures.
So, if we believe that there are N statistical measures that adequately
capture the material structure, then the microstructure space is an space.
So, in this space every space is a microstructure.
So, every point here is a microstructure.
So this is a set of all possible vital structures.
When you apply a process on the microstructure
essentially you’re moving from one point in space to another point and
that is represented by these start lines.
That's essentially when you apply.
When you take your microstructure like this.
And you apply a process for
example, then you are moving from one location in this base to another location.
This back lines tell you how they access particular microstructures.
So far, imposed on this plot is also composed of performance.
Because every microstructure is associated with their property.
So for example, a microstructure here might have a property let's say.
And makes modular of say10GPA.
And a micro-structure here might have a property of 100GPA.
So, for a particular application a property of this side may be
better than a property on this side.
And so, one can impose that information and
superimposed that information on to the microstructure space.
So basically, what I'm trying to say in all these plots is that
such a space is a most convenient space where we can impose about the process
information in terms of this flat lines as well as the performance information in
terms of the contours that are going from blue to red in this direction.
So once we superimpose all this information on the structure space
we can do the inverse problems and see what microstructures
correspond to the properties of information and what can I take?
What parts can I take to get to the microstructures of interest?
So in summary, the keys for accelerated material innovation
are standardized presentation of material internal structure.
This representation of quantification of structure has to be objective,
has to be broadly applicable and has to be low-dimensional for
it to be of practical utility in material.
We also need high throughput protocols.
The reason we need high throughput protocols is because all these spaces of
interest are very large.
And we really cannot afford to explore the spaces through very careful experiments.
We have to learn how to skim through this space fast,
find locations of interest and then pay detailed attention to particular locations
where there is valuable information.
Yet another key to what we want to do is uncertainty quantification.
Given that we are going to use high throughput protocols and
this are not going to be exact in what we're looking for.
We need to quantify uncertainty and
use that information in decision in making this.
And in doing that, we want to employ data driven processes.
Because we believe data driven processes lead to best decisions or
objective decisions, and are likely to save us time and money.
Another key to accomplishing what we envision are the collaborations.
As we discussed before the meticulous development activity
leading to improved product design.
A very multidisciplinary,
it needs a lot of different expertise from different stakeholders.
And we need to learn how to collaborate faster and better.
And this is going to be produced by cyber infrastructure and
we will talk about that later in this class.
A final key to accelerated materials innovation comes from
digital recording of workflows.
One of the important aspects of work we are discussing so
far is that the workflows are fairly complicated.
They involve a lot of different physics, a lot of different length scales,
lot of different pools, lot of different expertise.
We need to actually understand and record which workflows work better.
We not only need to understand, learn from our successes.
But we also need to learn from our failures.
We will be able to establish the best workflows.
In summary, Materials Data Sciences and
Informatics is going to help us With all these items that are listed here.
And we're going to learn exactly how we're going to do that in the coming lessons.
So, in summary for this lesson.
We learned that structure plays a central role in establishing PSP linkages.
We need to quantify structure in a statistical framework.
We absolutely need a low-dimensional representation for
it to be practically useful in materials development efforts.
And the quantification of
structure is foundational to Materials Data Sciences and Informatics Approaches.
Thank you.
[MUSIC]
Explore our Catalog
Join for free and get personalized recommendations, updates and offers.
Coursera provides universal access to the world’s best education,
partnering with top universities and organizations to offer courses online.
© 2018 Coursera Inc. All rights reserved.
