David Ryckelynck

**Fields of interest** Scientific Machine Learning; Manifold learning in engineering; Model order reduction; Image-based digital twins; Patient specific modeling for predictive surgery; Shape learning; Anomaly detection; Physically guided machine learning; CNN and Graph neural networks **Education** Doctor of Science (DSc), University of Paris VI, Prof. E. Arlacon, Prof. F. Chinesta, Prof. P. Ladevèze, Prof. D. Dureissex (2006) PhD in Mechanics, Ecole Normale Supérieure de Cachan, France, Prof. J.-P. Pelle, Prof. G. de Saxcé (1998) Master of science in computational mechanics, Ecole Normale Supérieure de Cachan, France (1992) Ecole Normale Supérieure de Cachan (rank 5th/300) (1989) **Professional Experience** 12/2013-present Full Professor, Mines Paris PSL University, Institut Mines Télécom 12/2007-11/2013 Maître de Recherche, Mines Paristech 10/2004-12/2004 Assistant-professor at the Department of Structural Mechanics of ETSII Universidad Politécnica de Madrid 09/1998-11/2007 Assistant professor, Arts et Métier Paristech **Professional Service** Head of a minor on Data Science for Computational Engineering at Mines ParisTech, option Ingénierie Digitale des Systèmes Complexes (2017-present) Contribution to DATA program of PSL: Machine Learning and AI for Sciences (2022-present) Head of PhD programme MINDS, "Contrats doctoraux en Intelligence Artificielle - Etablissement" in Artificial Intelligence, funded by ANR-French Gouvernment Agnecy (2020-2024). Editorial board member of Mathematical and Computational Applications journal, <https://www.mdpi.com/journal/mca/editors> . Head of the research team CoCaS (Computational Mechanics for Structures) in Centre des Matériaux (2011-2015). Head of the PhD programme in mechanics at Mines ParisTech from (2009-2013). Member of the Scientific Committee of Computational Methods for Coupled Problems, Prof. M. Papadrakakis , Prof. E Oñate and Prof. B. Schrefler, (2004-present). Member of the Scientific Committee of the international conferences ESAFORM (2007-2009). Board member of CSMA (French Computational Mechanic Association) (2011-2015). Member of MECAMAT (French group of Mechanics of Materials) (2017-present). Scientific expert on computational mechanics at IRT SystemX (2014). **Lectures at Mines ParisTech, Institut Mines Télécom** PSL Intesive week, Computer Vision and Time Series for Physics and Engineering (2022), DATA program of PSL, in col. with A. Allauzen. Minor, Option Ingénierie Digitale des systèmes complexes, 200h, in col. with Prof. Elie Hachem, in French (new lecture proposed in 2018) Major, StatNum, Numerical statistics, (since 2018), Ecole des Ponts et Chaussées, Paris, in col. with V. Lefieux. Minor, Finite element method, 30h, enseignement specialisé, in French (2008-present). Major, Mechanics of solid materials, 30h, in French (2011-2016), Métiers de l’Ingénieur Généraliste (MIG), on electric vehicles, in French (2012), International lecture on Model order reduction, 30h, in English (2013-2017). **Invited Lectures** (Last Five Years): Plenary lecture, Manifold Learning in Mechanics of Materials, EuroMech Colloquium 642, Rome 2024 Plenary lecture, Manifold Learning for Fast Image-Based Modeling in Mechanics of Materials, CMCS ECCOMAS, Eindhoven 2023 Plenary lecture, Manifold Learning in Mechanics of Materials, CMDS14, Paris, 2023 Geen AI, EELISA FAU Erlangen Nürnberg, Dec. 2022 Plenary lecture at Congrès Français de Mécanique, Nantes, 2022 Plenary lecture at Data Analytics &amp; AI IMT 2020 <https://youtu.be/CHuDHu3Yq-M> Plenary lecture at CMCS 2019, ECCOMAS, Glasgow, 2019 Invited lecture, Artificial Intelligence, 2019. Plenary lecture at 7th GACM Colloquium on Computational Mechanics in Stuttgart 2017. Invited lecture at Centrale Supélec 2016. **Background** : David Ryckelynck is working on model-based/physics-based engineering assisted by machine learning. He did seminal works on hyper-reduction methods, also termed hyperreduction, in the field of applied mathematics and computational mechanics. This method couples machine learning and projection-based model order reduction. Without counting for self citations, hyper-reduction has been cited 544 times in peer reviewed papers, according to Web of Science in 2019. Hyper-reduction methods can be seen as an extension of non supervised machine learning for dimension reduction to the field of model order reduction for the approximate solution of partial differential equations. He also did seminal works on “a priori model reduction” with 227 citations in peer reviewed paper without counting self citations in 2019. He has been the supervisor of more than 25 PhD theses in computational mechanics. He has supervised Nissrine Hakkari and Fabien Casenave for their degree of Doctor of Science (Habilitation à diriger des recherches). He was the head of a research team involving 12 permanent researchers at Mines ParisTech, from 2011 to 2015. He is the head of a new lecture on Ingenierie Digitale Des Systemes Complexes (Data Science for Computational Engineering) at Mines Paris PSL University. This lecture has been created by Pr. David Ryckelynck and Pr. Elie Hachem in 2017. In this lecture, a team of 10 lecturers are teaching to students of Mines Paris the way that simulation data and experimental data can be used in machine learning. We focus our attention to hybrid approaches that combines machine learning and usual modeling methods in engineering. His main topic of research is now image-based digital twins via machine learning, including dimensionality reduction for both images and simulation data. He is developping new data augmentation methods for engineering applications. **Awards** Prix de thèse Maths Entreprises 2022 for Thomas Daniel (PhD 24th september 2019 Mines Paris PSL University) from Agence Maths Entreprises (AMIES), France. **Best papers** A priori hyperreduction method: an adaptive approach, Ryckelynck, D, JOURNAL OF COMPUTATIONAL PHYSICS Volume: 202 Issue: 1 Pages: 346-366 Published: JAN 1 (2005). [⟨10.1016/j.jcp.2004.07.015⟩](https://dx.doi.org/10.1016/j.jcp.2004.07.015). [⟨hal-00021102⟩](/hal-00021102) Hyper-reduction of mechanical models in involving internal variables. David Ryckelynck. *International Journal for Numerical Methods in Engineering*, Wiley, 2009, 77, pp.75-89. [⟨10.1002/nme.2406⟩](https://dx.doi.org/10.1002/nme.2406). [⟨hal-00359157⟩](/hal-00359157) On the "A Priori" Model Reduction: Overview and Recent Developments. David Ryckelynck, Francisco Chinesta, Elías Cueto, Amine Ammar. *Archives of Computational Methods in Engineering*, Springer Verlag, 2006, 13 (1), pp.91-128. [⟨10.1007/BF02905932⟩](https://dx.doi.org/10.1007/BF02905932). [⟨hal-01007164⟩](/hal-01007164) On the Reduction of Kinetic Theory Models Related to Finitely Extensible Dumbbells. Amine Ammar, David Ryckelynck, Francisco Chinesta, Roland Keunings. *Journal of Non-Newtonian Fluid Mechanics*, Elsevier, 2006, 134 (1-3), pp.136-147. [⟨10.1016/j.jnnfm.2006.01.007⟩](https://dx.doi.org/10.1016/j.jnnfm.2006.01.007). [⟨hal-01007150⟩](/hal-01007150) A new extension of the natural element method for non-convex and discontinuous problems: the constrained natural element method (C-NEM). Julien Yvonnet, David Ryckelynck, Philippe Lorong, Francisco Chinesta. *International Journal for Numerical Methods in Engineering*, Wiley, 2004, 60 (8), pp.1451-1474. [⟨10.1002/nme.1016⟩](https://dx.doi.org/10.1002/nme.1016). [⟨hal-01508695⟩](/hal-01508695) Model order reduction assisted by deep neural networks (ROM-net), <a name="scopus-author-name-link__author-name-lin"></a>Daniel, T., Casenave, F., Akkari, N., Ryckelynck, D., Advanced Modeling and Simulation in Engineering Sciences, 2020, 7(1), 16, [⟨10.1186/s40323-020-00153-6⟩](https://dx.doi.org/10.1186/s40323-020-00153-6). [⟨hal-02539647⟩](/hal-02539647) **Patents**: Développement d’une méthodologie de morphing de maillage entre pièce CAO et pièce réelle (Safran-Mines ParisTech PSL) B-025180FR01 -FR 2112804 , Date 08/2023. Méthode de détermination des efforts exercées sur une zone d’intérêt d’une pièce mécanique d’aéronef à partir d'un modèle éléments-finis (Safran-Mines ParisTech PSL) B-025813FR01, Date 12/2022. Data Augmentation method for stress predictions using deep classifiers (Safran-Mines ParisTech PSL) \[BVT 2013318, Date 12/2020\] Data Selection for stress predictions using deep classifiers (Safran-Mines ParisTech PSL) \[BVT 2013320, Date 12/2020\] Model selection for turbine blades using deep learning (Safran-Mines ParisTech PSL) \[BVT 2102262, Date 12/2020\] Patent relating to the CPTh substrate has been registered by Griset S.A, LNE and Armines at the Institut National de la Propriété Industrielle (inpi) \[Application No.: F0956865, Publication No.: 2951020, Date of publication: April 8, 2011\]. **Supervisor of 30 PhD for:** Pierre Belamri, Germain Vu, Amélia Ferhat, Matteo Bastico, Daria Mesbah, Simon Le Berre, Axel Aublet, Hamza Boukraichi, Pablo Alvarez Pereira, Hugo Launay, Thomas Daniel, Youssef Hammadi, Harris Farooq, Laurent Lacourt, William Hilth, Jules Fauque, Tuan Dinh Trong, Clément Olivier, Yang Zhang, Samuel Jules, Noémie Rakotomalala, Mélanie Leroy, Antoine Andrieu, Bahram Sarbandi, Sophie Cartel, Laëtitia Petroni, Abderrahmen Kaabi, Florence Vincent, Laurent Vanoverberghe, Djamel Missoum Benziane. **Supervisor of 3 Doctors of Science (Habilitation à diriger des recherche) for:** Fabien Casenave, Nissrine Akkari, Cédric Gourdin