Advanced search
Start date

Finding empirical equations to predict glass properties using genetic programming-based Symbolic Regression

Grant number: 19/26460-4
Support type:Scholarships abroad - Research Internship - Post-doctor
Effective date (Start): June 01, 2020
Effective date (End): March 31, 2021
Field of knowledge:Engineering - Materials and Metallurgical Engineering - Nonmetallic Materials
Principal researcher:Edgar Dutra Zanotto
Grantee:Daniel Roberto Cassar
Supervisor abroad: John Christopher Mauro
Home Institution: Centro de Ciências Exatas e de Tecnologia (CCET). Universidade Federal de São Carlos (UFSCAR). São Carlos , SP, Brazil
Research place: Pennsylvania State University, United States  
Associated to the scholarship:17/12491-0 - Kinetic processes in glass and formulation of new glasses using machine learning, BP.PD


The interface between Data and Glass Sciences is a fertile and promising ground. The interest in this interface is in part because glasses are very relevant in many modern applications (both technological and non-technological), and in part because their chemical composition can be changed much more freely than that of crystals, making it possible to fine-tune their properties to meet specific materials design constraints. Machine learning algorithms can make use of the collection of available data to produce accurate predictive models. However, many of these algorithms produce "black-box" models, which are not interpretable and may have lower accuracy when predicting data outside of the training domain. The objective of this project is to design, implement, run, and test Genetic Programming-based Symbolic Regression (GPSR) algorithms to find empirical mathematical expressions with good predictive power for glass properties. The expectation is that these mathematical expressions will have a better extrapolation accuracy when compared with other models. GPSR was selected because the final result is interpretable and can be used both to predict glass properties as well as to increase our understanding of the natural laws that govern such properties. Few works have advanced this particular interface of knowledge.

News published in Agência FAPESP Newsletter about the scholarship:
Articles published in other media outlets (0 total):
More itemsLess items