Chemical-Based Prediction of the Glass Transition Temperature ofPolymers using Machine Learning Approach

Understanding and the ability to predict polymer properties such as Glass transition temperature hasalways been a crucial steps in polymer science. Tg is one the most important termal transitions ofpolymers, which defines their behavior and properties. Classical modeling techniques apply very hardto complex problems such as prediction of Tg; in order to overcome this challenge a new methodusing Aritificial neural networks has been implemented and various settings and configurations formultilayer perceptrons has been tested. The best architectures gave coefficient of determination ofaround ۰.۹ and show good predictive power considering the diversity of data. Moreover, theperformance exceeded that of previous parameterizations developed for this purpose and alsoperformed better than existing machine learning models.

کلیدواژه ها:

Polymer Characterization ، Property Prediction ، Machine learning و Artificial neuralnetwork ، glass transition