Publication: Machine learning for glass modeling
Program
KU-Authors
KU Authors
Co-Authors
Tandia, Adama
Mauro, John C.
Advisor
Publication Date
2019
Language
English
Type
Book Chapter
Journal Title
Journal ISSN
Volume Title
Abstract
With abundant composition-dependent glass properties data of good quality, machine learning-based models can enable the development of glass compositions with desired properties such as liquidus temperature, viscosity, and Young's modulus using much fewer experiments than would otherwise be needed in a purely experimental exploratory research. In particular, research companies with long track records of exploratory research are in the unique position to capitalize on data-driven models by compiling their earlier internal experiments for research and product development. In this chapter, we demonstrate how Corning has used this unique advantage to develop models based on neural networks and genetic algorithms to predict compositions that will yield a desired liquidus temperature as well as viscosity, Young's modulus, compressive stress, and depth of layer.
Description
Source:
Springer Handbooks
Publisher:
Springer
Keywords:
Subject
Materials Science