Paper Image

Machine learning prediction of thermal conductivity

Published on:

6 November 2023

Primary Category:

Materials Science

Paper Authors:

Yagyank Srivastava,

Ankit Jain

Bullets

Key Details

Researchers more than doubled available high-quality thermal conductivity data from 166 to 398 materials

Multiple machine learning models were tested for thermal conductivity prediction

A new graph neural network delivered the most consistent performance

Best model achieved ~55-60% mean absolute percentage error on test data

Machine learning shows promise to accelerate screening, but accuracy is still limited

AI generated summary

Machine learning prediction of thermal conductivity

Researchers investigated using machine learning to predict materials' thermal conductivity, which is important for applications like thermoelectrics and heat dissipation. They performed first-principles calculations to expand the dataset of materials with known thermal conductivity from 166 to 398 entries. Multiple machine learning models were tested, with a novel graph neural network delivering the most consistent performance. The best model achieved mean absolute percentage error around 55-60% on test data. While not perfect, machine learning shows promise to accelerate thermal conductivity screening versus purely first-principles approaches.

Answers from this paper

Comments

No comments yet, be the first to start the conversation...

Sign up to comment on this paper

Sign Up