Integrating Machine Learning with a Genetic Algorithm for Materials Exploration

2021.11.17 Joseph D. Kern, Georgia Institute of Technology
Table of contents below.

The genetic algorithm is a computer algorithm inspired by nature: selecting parents of a generation via some fitness function, crossing-over the parent genes (reproduction), and randomly mutating the genes of the children. In this talk, we will explore how this algorithm can be used for materials discovery by:

1. Generating swift material property prediction using machine learning (ML)
2. Creating an algorithm to design new materials from combinations of prior ones
3. Integrating the ML property predictors with the design algorithm to discover new materials

PolyGA, an implementation of the genetic algorithm for the polymer domain, will be used as the basis for this exploration.

Table of Contents:
00:00 Integrated Computational Materals Engineering in the Classroom
02:33 Outline
03:36 ICME and Industry 4.0
05:30 Making the jumps
06:48 The influence of chemistry and temperature
09:27 Composition Effects
10:57 The materials data challenge
11:55 CALPHAD: A phase-based approach
13:27 CALPHAD: A phase-based approach
16:16 A teaching tool for fundamentals
17:06 Thermo-Calc Academic Lesson Ideas
19:04 Thermo-Calc Academic on nanoHUB
19:46 Case Study: Solidification Cracking
28:51 Introduction to Solidification - Equlibrium
29:41 Introduction to Solidification – Non-Equlibrium
30:27 Mass percent Cu
31:52 Introduction to Solidification – Scheil Derivation
32:40 Scheil-Gulliver Derivation
32:55 What does this describe?
33:42 Solidification Cracking Theory
34:49 Case Study: Additive Manufacturing
39:25 Summary – AM ICME Case Study
40:01 Summary
40:54 Thank you!
45:17 Summary – AM ICME Case Study