Nb3Sn image analysis using AI/ML

by Dr Al Baskys (LBNL)

Thursday 8 Sept 2022, 09:00 → 10:00 US/Pacific

Zoom (Wisconsin)

This is a "return" seminar by LBNL to UWEC on AI/ML.  Please note the University of Wisconsin Zoom link below:

https://wisconsin-edu.zoom.us/s/5930039950

 

Title:
Nb3Sn image analysis using AI/ML

Algirdas Baskys1, Kevin Gillespie1, Jean-Francois Croteau1, Ian Pong1
1Berkeley Center for Magnet Technology, Lawrence Berkeley National Laboratory, 1 Cyclotron Rd, Berkeley, CA 94720

 

Abstract:

Materials development is a critical aspect of the accelerator magnet design as the energy envelope and the magnetic field strength of the accelerator magnets increases. The present-day workhorse conductor for accelerators, Nb-Ti, is widely accepted to be at its performance limits and cannot be used for future upgrades or the next generation colliders with higher energy and intensity. Due to cost and production maturity considerations, the most likely material candidate for the next generation of accelerator magnets is Nb3Sn. However, the performance of the commercially produced state-of-the-art Nb3Sn may not be sufficient for future accelerator magnets. An indication of this is that, at present, no commercial conductor meets the performance targets set out in 2015 for the 16 T dipole magnets considered for the Future Circular Collider [1].

Recent advances in Nb3Sn with Artificial Pinning Centers (APC) led by X. Xu [2] at Fermilab and Hf doped wires by S. Balachandran et al. at Florida State University [3] have exceeded or shown potential to meet and exceed the critical current density (Jc) requirements for a 16 T dipole. However, a better understanding of these advancements is still needed to get the most out of their potential and to widen the processing window. The interlink between the APC and Hf additions and optimal heat-treatment, manufacturability, and mechanical properties needs much more investigation to bring the design to maturity. As the relevant feature sizes move towards the nanoscale, these materials are becoming more complex and time-consuming to characterize. This is where the use of machine learning (ML) can significantly accelerate the characterization of Nb3Sn conductors to provide feedback for development. The talk focuses on three specific tasks that could significantly accelerate materials characterization by automating the analysis that is currently done manually and is very time-consuming: (1) grain size analysis, (2) nanoprecipitate distribution in APC wires, (3) cracked Nb3Sn subelement identification. These problems present unique challenges and consequently need different approaches, which will be discussed in this talk.

 

References:

[1] A. Ballarino and L. Bottura, “Targets for R&D on Nb3Sn Conductor for High Energy Physics,” IEEE Trans. Appl. Supercond., vol. 25, no. 3, pp. 1–6, Jun. 2015, doi: 10.1109/TASC.2015.2390149.

[2] X. Xu, M. D. Sumption, and X. Peng, “Internally Oxidized Nb3Sn Strands with Fine Grain Size and High Critical Current Density,” Adv. Mater., vol. 27, no. 8, pp. 1346–1350, Feb. 2015, doi: 10.1002/adma.201404335.

[3] S. Balachandran et al., “Beneficial influence of Hf and Zr additions to Nb-4at.%Ta on the vortex pinning of Nb3Sn with and without an O source,” Supercond. Sci. Technol., vol. 32, no. 4, 2019, doi: 10.1088/1361-6668/aaff02.

 

Acknowledgments:

Research supported by National Energy Research Scientific Computing Center (NERSC).