Prediction of rare-earth-free permanent magnet via ML

1.

Prediction
of rare-earth-free
permanent magnet via
ML

2.

What is magnet?
Rare-earth magnet dominate the
magnet market
Efficiency of magnet is assessed using
hysteresis loop data

3.

Application of magnet

4.

Rare-earth elements
Rising prices and demand pose the challenge
of finding other materials for magnets

5.

Problem
The magnetic properties of nanoparticles strongly depend on
the size, composition, and shape
Small number of qualified workers
Experimental selection
of nanoparticles for specific purposes:
- take a lot of time
- resource-intensive

6.

Solution
Use of alternative cheaper raw materials
Low threshold for entry into data driven material
science:
- Increase in the number of employees
Prediction of a hysteresis loop using minimal
input data:
- more effective than the experimental way
Parallel world of ML

7.

What are we planning to do?
Data collection
Fe1.7Co0.3P
Machine learning
Size 28 nm
Hc, MR, Ms

8.

Competitors

9.

Consumers
Arnold Magnetic Technologies
Adams Magnetic Products Co.
Hitachi Metals Limited
BGRIMM Magnetic Materials and
Technology Co. Limited
Our Web service

10.

Roadmap
R2 > 0.6
Data collection
Proof of concept
October – November
January
Model validation and
paper draft
May – June
Web service
250-300 lines
Descriptors engineering and
model training
November – January
Model selection and optimization
MVP
February – May
August
↑R2
↑lines

11.

Thank you for
attention!
Naboychenko Olga
Chemistry and Artificial Intelligence
[email protected]