Contexte et atouts du poste
High-resolution modeling of wave propagation plays a crucial role in several applications, including geophysics, non-destructive testing, or helioseismology.
While efficient high-order discretization techniques have been developed to improve modeling accuracy, they often come at the expense of increased computational cost, e.g., with Discontinuous Galerkin methods.
Furthermore, in the context of inversion with Full Waveform Inversion (FWI) method, physical properties of a medium are iteratively reconstructed using repeated forward simulations.
This iterative process increases the computational cost, especially in large-scale applications.
The objective of this postdoctoral position is to investigate strategies for reducing the computational cost of both forward modeling and inversion, compromising between accuracy and cost.
Mission confiée
The successful candidate will contribute to the development strategies for frequency-domain wave modeling and inversion by efficiently combining standard numerical methods with model reduction approaches.
Research directions include the following,
1.
Investigate and develop suitable reduced-order models for time-harmonic wave propagation (e.g, Proper Orthogonal Decomposition (POD), Singular Value Decomposition (SVD), and other projection-based methods) depending on the context, [1,2,3].
The reduction strategies should account for both the frequency content of the wavefield and the heterogeneities of the medium for efficiency.
For instance, efficient methodologies for frequency sweeping and parametric (e.g., for viscosity model) wave problems will be studied, as well as the ability to evaluate wavefields for arbitrary sources, which represents a key-factor in reducing costs.
2.
Exploit reduced-order models in the context of inversion.
Here, as the model parameters change with iterations, and it is expected that the reduction basis must adapt to reflect these changes.
The key is that this adaptation should be performed as infrequently and minimally as possible, [4].
3.
In the context of inversion, explore the reduced-order parametrization of the data and of unknowns to mitigate the ill-posedness [5].
In particular combining with other techniques (e.g., FrgWI) to alleviate some restrictions.
4.
Longer-term perspectives can include the study of methods in the context of nonlinear wave propagation, e.g., [6].
These developments will be validated on representative test cases in seismic imaging and helioseismic inverse problems, in collaboration with domain specialists.
The computational developments and new methodologies will be implemented in the open-source software hawen ( with the help of the development team.
[1] R.
D.
Slone, J.-F.
Lee, and R.
Lee, A comparison of some model order reduction techniques, Electromagnetics, 22 , pp.
275–289.
[2] R.
Hawkins, M.
H.
Khalid, M.
Schlottbom, and K.
Smetana, Model order reduction for seismic applications, SIAM Journal on Scientific Computing, 47 , pp.
B1045–B1076.
[3] Y.
Hüpel, U.
Römer, M.
Bollhöfer, and S.
Langer, Efficient low rank model order reduction of vibroacoustic problems under stochastic loads, arXiv preprint arXiv:2408.08402, .
[4] J.
Besset, A Model Order Reduction Strategy for Parametrized PDEs: A New Paradigm for Efficient Subsurface Imaging., PhD thesis, Universit´e de Pau et des Pays de l’Adour, 2025.
[5] L.
Borcea, J.
Garnier, A.
V.
Mamonov, and J.
Zimmerling, Reduced order modeling for first order hyperbolic systems with application to multiparameter acoustic waveform inversion, SIAM Journal on Imaging Sciences, 18 , pp.
851–880.
[6] P.
J.
Schmid, Dynamic mode decomposition of numerical and experimental data, Journal of fluid mechanics, 656 , pp.
5–28.
Principales activités
The postdoc will join a dynamic research team at Inria, hosted at the University of Pau and Pays de l’Adour (southwest of France), with strong expertise in wave propagation, scientific computing, and inverse problems.
The team collaborates closely with geophysicists and astrophysicists, applied mathematicians, and HPC experts across academic and industrial partners.
Compétences
Keywords: Scientific computing, Reduced-order modeling, wave equations, Full-Waveform Inversion (FWI), frequency domain, large-scale linear systems, numerical modeling, PDEs.
Avantages
Rémunération
2788€ per month (before taxs)