Home https://server7.kproxy.com/servlet/redirect.srv/sruj/smyrwpoii/p2/ Science https://server7.kproxy.com/servlet/redirect.srv/sruj/smyrwpoii/p2/ Order! AI finds the right material

Order! AI finds the right material



Order!  AI finds the right material

Material representations compatible with machine learning models play a key role in the development of models that exhibit high accuracy in property prediction. Credit: College of Engineering

Engineers are always looking for materials with very specific properties for their projects. Unfortunately, there are far too many opportunities for researchers to guess and control until they find what they are looking for. Even if the simulated materials, instead of testing them in the laboratory, it would take far too long to find a suitable material.


Fortunately, researchers have created algorithms using artificial intelligence that find the right material for any project. In a recently published paper, a team from Carnegie Mellon University and University of Calgary researchers has improved one of these algorithms so that researchers can quickly and accurately find materials with the desired properties.

“Because the material space is so large, it is very difficult experimentally and computationally to characterize the material properties,”

; said Amir Barati Farimani, an assistant professor of mechanical engineering at CMU. “So we create algorithms or models that can quickly predict the material properties.”

To use artificial intelligence or AI, researchers must first train the algorithm using known data. Then the algorithm learns to extrapolate new ideas from this information. Barati Farimani and his team trained the algorithm with data on chemical composition of materials. In particular, they included information about the role that electrons play in the determination of material properties. These chemical data have created a new material description for the algorithm, according to Barati Farimani.

Since this algorithm can predict the properties of a wide variety of materials, it has many applications. For example, the algorithm could find a material with thermal properties suitable for solar panels. In addition, it could identify materials for the manufacture of substances and batteries. To use this algorithm, a researcher can simply get the pre-trained models for deep learning to find the property they are looking at.

The way in which these algorithms are improving is becoming faster and more accurate. If the algorithm is not accurate enough, the results will be useless. If the algorithm is too slow, researchers will never be able to access the results. Currently, the team has found that their algorithm is better than other leading algorithms.

“You can use this algorithm and train a deep learning model and predict them in a fraction of a second,” said Barati Farimani. “The essence is to prove that it predicts different kinds of materials with high accuracy – then any industry can use it.”

Their paper was published in Materials for physical review. CMU post-doctoral scholar Mohammadreza Karamad, Ph.D. student Rishikesh Magar and researcher Yuting Shi were also listed as co-authors. Other authors include Samira Siahrostami and Ian D. Gates of the University of Calgary.


Algorithm predicts the composition of new materials


More information:
Mohammadreza Karamad et al. Orbital graph coincidence neural networks for the prediction of material properties, Materials for physical review (2020). DOI: 10.1103 / PhysRevMaterials.4.093801

Provided by Carnegie Mellon University Mechanical Engineering

Citation: Order! AI finds the right material (2020, October 16) retrieved October 16, 2020 from https://phys.org/news/2020-10-ai-material.html

This document is subject to copyright. Except for any fair trade for the purpose of private examination or research, no parts may be reproduced without written permission. The content is provided for informational purposes only.




Source link