Hybrid AI-kinetic models for catalyst regeneration

Essential for numerous industrial processes and energy systems, catalysts gradually lose their efficiency due to phenomena such as coke formation or sintering. To extend their lifespan, which is crucial for a sustainable economy, tens of thousands of tons of catalysts are regenerated each year to restore their performance. This regeneration process thus has a significant impact on the energy efficiency and sustainability of the industries involved, such as biofuel production, for instance.
In the case of bifunctional catalysts with an acidic phase and a hydrogenating phase, the choice of regeneration conditions becomes critical, as during this process, one of the active phases may be degraded, thus significantly affecting the performance of the regenerated catalyst. Current knowledge models on coke types and their oxidation do not allow accurate predictions of the consequences of a specific choice of regeneration conditions.
This PhD project focuses on the development of hybrid models combining machine learning techniques with physical models to predict the performance of regenerated catalysts. By leveraging complex data, such as X-ray diffraction patterns, the candidate will contribute to the implementation of the first hybrid kinetic models covering multiple catalysts. The thesis work will be structured around three key objectives:
•    Developing a hybrid learning strategy that integrates kinetic models with machine learning techniques (such as Neural ODE approaches).
•    Investigating the limits of model performance when experimental data is limited (few-shot learning)
•    Applying these approaches to real-world industrial regeneration processes to optimize operational conditions

Keywords: Kinetic modelling, machine learning, hybrid modelling, catalysis

• Academic supervisor    Prof. Jan VERSTRAETE, IFPEN, ORCID : 0000-0003-4536-5639
• Doctoral School    ED206 CHIMIE, Université Lyon 1,
• IFPEN supervisor    PhD Thomas PIGEON, ORCID : 0000-0002-7828-5128
• PhD location    IFP Energies Nouvelles, Lyon, France  
• Duration and start date    3 years, starting in the fourth quarter 2026 (Novembre 3)
• Employer    IFPEN
• Academic requirements    University master’s degree in chemical engineering or in machine learning   
• Language requirements    English level C1 (CECR)   
• Technical requirements    Programming skills, python required, Background in mathematical, physical or numerical modeling, Knowledge in machine learning and artificial intelligence, Interest in heterogeneous catalysis

To apply, please send your cover letter and CV to the IFPEN supervisor indicated here below.