Open Conference Systems, UCUR 2017

Font Size: 
Effects of Strain on Twin Formation and Transmission Models Created Using Machine Learning Techniques
Andrew D Orme, David T Fullwood, Christophe Giraud-Carrier

Building: Classroom Building
Room: CB 510 & 511 - Poster Sessions
Date: 2017-02-17 09:30 AM – 10:50 AM
Last modified: 2017-02-06

Abstract


Statistical and probabilistic models have proven to be invaluable in understanding the vital
deformation mechanisms, such as twinning, required for ductile formation in brittle materials such as
magnesium. These models are limited due to the effort required to define and verify correlations
between attributes, especially when considering the impacts of including further attributes in a
previously defined model. It has been proposed that using machine learning algorithms on data
collected via electron backscatter diffraction (EBSD) is an effective method to generate models relating
twin formation and transmission in MG A31 to attributes of the microstructure. This study assumed a
simplistic constant relationship between twin formation events and strain. However, various factors
which influence twin formation and propagation are known to evolve with strain, rendering the models
created previously inaccurate with changes in strain. The work presented here investigates strain and
its role in twin activity evolution using machine learning methods and presents models which account
for changes in strain.

Keywords


Magnesium Twinning; Machine Learning; Modeling

Conference registration is required in order to view papers.


Add comment