From 1586ba66c49b83abdae8b5be082e8e2477c12e9a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?St=C3=A9phane=20Lanteri?= Date: Tue, 16 Jun 2026 14:59:42 +0200 Subject: [PATCH] Adding BSC-Inria project --- .../hpc_and_ai_for_electromagnetics.md | 81 +++++++++++++++++++ 1 file changed, 81 insertions(+) create mode 100644 collections/_projects/hpc_and_ai_for_electromagnetics.md diff --git a/collections/_projects/hpc_and_ai_for_electromagnetics.md b/collections/_projects/hpc_and_ai_for_electromagnetics.md new file mode 100644 index 00000000..26fb62e3 --- /dev/null +++ b/collections/_projects/hpc_and_ai_for_electromagnetics.md @@ -0,0 +1,81 @@ +--- +layout: post +title: Extreme-scale AI and computing for high-fidelity electromagnetics +date: 2026-05-26 +updated: 2026-05-26 +navbar: Research +subnavbar: Projects +project_url: +status: running +topics: + - numerics + - ai + - apps +keywords: computational electromagnetics (CEM), reduced order models (ROM), deep neural network (DNN) +head: + - delapuente_j + - lanteri_s +members: + - agullo_e + - castillo_o + - circuns_m + - farnos_j + - guillet_c + - giraud_l + - lopez_h + - marait_g + - modesto_d + - sylvand_g +--- + +## Research topic and goals + +Electromagnetic (EM) waves are ubiquitous in our daily environment. +They find applications of social and environmental relevance, as well +as in technology for industry and defense. The scales resolved involve +the micrometer scale in optoelectronics to the hundreds of km involved +in radar detection. Computational electromagnetics refers to the use +of numerical modelling for the study of EM wave interaction with +objects and matter. The participant groups from BSC and Inria have a +long-standing expertise on the development and implementation of +computational electromagnetic methods including conforming and +high-order finite element methods (FEM), method of moments – boundary +element methods (MoM/BEM) and discontinuous Galerkin (DG). These are +considered as high-fidelity methods and the corresponding so- called +fullwave solvers are materialized in in-house simulation software +based on numerical algorithms that have been extensively adapted to +high performance computing hardware possibly combing multicore CPU +chips and GPU accelerators. From the point of view of applications, an +even more important objective of computational electromagnetics is to +shape the EM wave interactions, or to solve the inverse problem for +unveiling the values of physical parameters of the underlying problem +whose mathematical modeling relies on the system of Maxwell +equations. These contexts are treated with an outer loop driving a +numerical optimization workflow, meaning that many fullwave +simulations need to be performed therefore drastically increasing the +computational load and ultimately preventing the study of large-scale +problems. + +Our overarching objective is to design, develop and demonstrate +disruptive numerical approaches leveraging high performance computing +and AI-driven methodologies for addressing the current performance +bottlenecks that are faced when running inverse design or parameter +inversion studies for large-scale EM wave interactions. From the +methodological point of view, we have identified three main topics +that will underly specific objectives of our joint research: (1) +Numerical Linear Algebra (NLA); (2) HPC and AI; (3) Data-driven +reduced-order modeling (ROM). + +## Visits and meetings + +## Impact and publications + +### Papers + +### Funding + +### Impact on other projects + +## Future plans + +## References