Сloud platform for scalable finite element modeling of magnetostatic fields to generate datasets for magnetic characteristics reconstruction

The problem of accelerated generation of representative numerical simulation datasets, used in the reconstruction of magnetic characteristics of materials in information-measuring systems for short samples, is considered. In previously developed measuring solutions, increased sensitivity is achieved, among other things, by introducing a parallel magnetic shunt. However, the interpretation of measurement data requires solving an inverse problem and constructing models tailored to the specific geometry of the setup. When the geometry or parameters of the magnetic system change, a large number of scenarios must be recalculated using the finite element method. Direct parallelization of solving a single finite element problem is challenging due to the global connectivity of the sparse system of linear algebraic equations and the high cost of interprocessor communications. Therefore, parallelization at the level of independent computational tasks during a parametric sweep is proposed. A cloud-based microservice architecture is proposed that implements a fully automated cycle: mesh generation, formulation of the magnetostatic problem, numerical solution, centralized storage of results, and generation of a training dataset. The implementation was carried out within the Yandex Cloud infrastructure. It is experimentally demonstrated that the average calculation time for a single point is 22.78 seconds, of which mesh generation takes 3.64 seconds and the problem solving takes 19.13 seconds. The time required to generate a dataset of 900 characteristics is reduced from 105 hours to 9 minutes when the number of parallel containers is increased to 900, confirming the near-linear scalability of the proposed approach.

1. Sandomirskiy S.G. Calculation and analysis of the demagnetizing factor of ferromagnetic bodies. Minsk: Belaruskaya navuka; 2015. 244 p. (In Russ.).

2. Surnyaev V.A., Blazhko I.O., Blazhkova E.N., et al. Application of a magnetic shunt to increase the sensitivity of a device for testing samples of magnetostrictive materials. Engineering Journal of Don. 2021;(9). (In Russ.). URL: https://ivdon.ru/en/magazine/archive/n9y2021/7200

3. Surnyaev V.A., Surnyaev D.A. Measuring device of the magnetic characteristics of the magnetostrictive materials. Engineering Journal of Don. 2016;(4). (In Russ.). URL: https://ivdon.ru/en/magazine/archive/n4y2016/3956

4. Bakhvalov Y.A., Gorbatenko N.I., Grechikhin V.V. A Method of Solving Inverse Problems of Magnetic Measurements. Measurement Techniques. 2015;58(3):336–340. https://doi.org/10.1007/s11018-015-0711-5

5. Aharoni A. Demagnetizing factors for rectangular ferromagnetic prisms. Journal of Applied Physics. 1998;83(6):3432–3434. https://doi.org/10.1063/1.367113

6. Beilina L., Smirnov Yu.G. Nonlinear and Inverse Problems in Electromagnetics: PIERS 2017, 22–25 May 2017, Saint Petersburg, Russia. Cham: Springer; 2018. 145 p. https://doi.org/10.1007/978-3-319-94060-1

7. Dyakin V.V., Kudryashova O.V., Rayevskii V.Y. On the Well-Posedness of the Direct and Inverse Problem of Magnetostatics. Part 2. Russian Journal of Nondestructive Testing. 2018;54(10):687–697. https://doi.org/10.1134/S1061830918100030

8. Obornev E.A., Obornev I.E., Rodionov E.A., Shimelevich M.I. Application of Neural Networks in Nonlinear Inverse Problems of Geophysics. Computational Mathematics and Mathematical Physics. 2020;60(6):1025–1036. https://doi.org/10.1134/S096554252006007X

9. Elgy J., Ledger P.D. Reduced order model approaches for predicting the magnetic polarizability tensor for multiple parameters of interest. Engineering with Computers. 2023;39(6):4061–4076. https://doi.org/10.1007/s00366-023-01868-x

10. Voskoboinikov M.L., Feoktistov A.G. Comparative analysis of scientific workflow management systems. Informational and mathematical technologies in science and management. 2024;(3):102–111. (In Russ.). https://doi.org/10.25729/ESI.2024.35.3.009

11. Geuzaine Ch., Remacle J.-F. Gmsh: A three-dimensional finite element mesh generator with built-in pre- and post-processing facilities. International Journal for Numerical Methods in Engineering. 2009;79(11):1309–1331. https://doi.org/10.1002/nme.2579

12. Geuzaine Ch. GetDP: A general finite-element solver for the de Rham complex. PAMM. 2007;7(1):1010603–1010604. https://doi.org/10.1002/pamm.200700750

13. Kiselev E.A., Yarovoy A.V. Application of container virtualization for implementing parallel computing on supercomputer clusters. Software & Systems. 2026;(1):058–069. (In Russ.). https://doi.org/10.15827/0236-235X.153.058-069

14. Boettiger C.D. An introduction to Docker for reproducible research. ACM SIGOPS Operating Systems Review. 2014;49(1):71–79. https://doi.org/10.1145/2723872.2723882

15. Oprescu A.-M., Kielmann Th. Bag-of-Tasks Scheduling under Budget Constraints. In: 2010 IEEE Second International Conference on Cloud Computing Technology and Science, 30 November – 03 December 2010, Indianapolis, IN, USA. IEEE; 2010. P. 351–359. https://doi.org/10.1109/CloudCom.2010.32

16. Ari I., Muhtaroglu N. Design and implementation of a cloud computing service for finite element analysis. Advances in Engineering Software. 2013;60-61(3-4):122–135. https://doi.org/10.1016/j.advengsoft.2012.10.003

17. Butrylo B., Musy F., Nicolas L., et al. A survey of parallel solvers for the finite element method in computational electromagnetics. COMPEL: The International Journal for Computation and Mathematics in Electrical and Electronic Engineering. 2004;23(2):531–546. https://doi.org/10.1108/03321640410510721

18. Shafiei H., Khonsari A., Mousavi P. Serverless Computing: A Survey of Opportunities, Challenges, and Applications. ACM Computing Surveys. 2022;54(11s). https://doi.org/10.1145/3510611