Thesis : Analysis and optimization of the performance of a catalysis system using numerical modeling: Study of key parameters and prediction of yields using machine learning

5 November 2026
14h
Amphi A207, pôle API

Thesis defense: Ly NGUYEN

Title: Analysis and optimization of the performance of a catalysis system using numerical modeling: Study of key parameters and prediction of yields using machine learning

Date and venue: Thursday, 05 November 2026, 2:00 PM in the A207 amphitheatre, Télécom Physique Strasbourg (300 Boulevard Sébastien Brant, 67400 Illkirch-Graffenstaden). 

The defense will be held in English.

Jury 

Thesis director: Prof. Cédric WEMMERT, ICube, University of Strasbourg, France

Reviewers:

  • Prof. Hanane AZZAG, University Sorbonne Paris Nord, France
  • Prof. Raphaël COUTURIER, University Marie et Louis Pasteur, France

Examiners:

  • Prof. Mustapha LEBBAH, Université Paris-Saclay, France
  • Prof. Germain FORESTIER, Université de Haute-Alsace, France

Supervisors:

  • Dr. Xavier JURADO, AIR&D, France
  • Dr. Nicolas REIMINGER, AIR&D, France
  • Dr. Jean-Mario NHUT, ICPEES / CNRS, Strasbourg, France
     

Abstract:

This thesis is a multidisciplinary project combining machine learning and chemistry. Catalytic processes involve complex, nonlinear interactions among several factors, which makes their optimization experimentally costly and time-consuming. The aim is therefore to apply machine learning methods that give useful insights into catalytic processes and accelerate their development. First, pretrained language models were used to convert textual catalyst descriptions into continuous numerical embeddings for prediction. Evaluated across four reaction datasets, they outperformed one-hot encoding. Next, Bayesian-guided experimental design was applied to a catalytic plastic process under induction heating. Bayesian optimization identified promising conditions while reducing the number of experiments needed. Finally, the study focused on dynamic forecasting of reaction performance for catalytic methane decomposition under induction heating.

In close collaboration with

Logo INRIA
Logo INSERM
Logo IHU
Logo Hôpitaux Universitaires de Strasbourg

With the support of our partners

Logo France 2030
Logo Investissements d'avenir
Logo ANR
Logo Union européenne
Logo Région Grand Est
Logo Institut Carnot