Machine Learning and Scraping to create a curated dataset of oxide glass viscosity.

• Grant number: 25/01793-1
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: April 01, 2025
• End date: March 31, 2026
• Field of knowledge: Engineering - Materials and Metallurgical Engineering - Nonmetallic Materials
• Principal Investigator: Daiane Aparecida Zuanetti
• Grantee: Eric Trevelato Costa
• Host Institution: Centro de Ciências Exatas e de Tecnologia (CCET). Universidade Federal de São Carlos (UFSCAR). São Carlos , SP, Brazil
• Associated research grant: 13/07793-6 - CEPIV - Center for Teaching, Research and Innovation in Glass, AP.CEPID

Abstract

In this project we intend to: 1) develop a computer code that harvests datafrom patents revolving around the viscosity and composition of oxide glasses to createan easy-to-access data frame with this information for, through machine learningmethods, to identify better or optimal glass compositions, for example, in futurestudies. Other students of the team also plan to: 2) expand the data frame mentionedwith other glass properties such as refractive index and liquidus temperature. Theseare important for multiple reasons, since viscosity data combined with liquidustemperature can be used to estimate the glass forming ability (GALEAZZO; SHIRAIWA,2022), for example. Viscosity is also the key parameter controlling the fining time toeliminate bubbles, the homogenization time to eliminate striae, and the conformationtechnique of glass products. As patents generally do not provide information on all theproperties of each composition, the databases will have to be built independently,using different filters, and this task will be carried out by other students who are alsolinked to the project, in parallel (these students are Gustavo Laranja Thomaello -TLiquidus and Thomaz Yeiden Busnardo Aguena - Refractive Index). The objective 3) isto combine the data frame generated with other databases publicly available to inferstatistically how the data frame generated by our code is consistent and can increasethe performance of predictive machine learning models acting on these databases.The project also aims to enhance the quality of research of scientific initiationstudents by tutoring computer programming, management of databases, and applyingdescriptive, analytical, and predictive statistical and machine learning methodologiesapplied for materials development.