Zaman YAZBECK defended her PhD on March. 19th, 2026 at 10:30 AM.
Place : amphitheatre Marc Seguin at INSA Lyon, 27 Av. Jean Capelle O, 69100 Villeurbanne
Jury :
Rapporteurs :
M. Thierry Poinot, Professeur des Universités, Université de Poitiers, LIAS
M. Dani Juričić, Professeur des Universités, University of Nova Gorica Slovénie, Jožef Stefan Institute
Examinateurs :
Mme. Marie-Cécile Péra, Professeur des Universités, Université Marie et Louis Pasteur, FCLAB / FEMTO-ST
M. Bertrand Morel, Dr. Ingénieur de Recherche LITEN, CEA Grenoble, Examinateur
Encadrement :
M. Minh Tu Pham, Professeur des Universités, Ampère INSA Lyon, directeur de thèse
M. Federico Bribiesca Argomedo, Maître de Conférences, Ampère INSA Lyon, co-encadrant
Invités:
Mme. Pauline Kergus, Chargée de Recherche, Université Marie et Louis Pasteur, LAPLACE, CNRS
M. Vincent Dimitriou, Digital Technologies Manager, GENVIA
Abstract :
Solid oxide electrolysis cells (SOECs) are promising devices for high-efficiency, low-carbon hydrogen production, but their industrial deployment is limited by complex multi-physics behaviour, durability issues and the difficulty of monitoring internal states with only a few sensors. This PhD thesis develops a model-based framework for Prognostics and Health Management (PHM) of SOEC stacks, combining dynamic modeling, parameter identification, optimal experiment design and fault diagnosis.
A lumped, zero-dimensional stack model is first derived, coupling electrochemistry, gas-phase mass balances and a reduced thermal description consistent with available measurements. Practical identifiability are analyzed using sensitivity differential equations. On this basis, truncated Levenberg-Marquardt and Gauss-Newton algorithms are proposed, using singular value decomposition of a normalized sensitivity matrix to focus updates on informative parameter directions and to quantify confidence intervals, and are validated on synthetic and experimental stack data.
Building on this calibrated model, a square-root Unscented Kalman Filter is designed to reconstruct internal gas partial pressures, stack temperature and deviations in inlet fuel flow from limited measurements (stack voltage, outlet air temperature and outlet hydrogen flow). Adaptive thresholds are used to detect and distinguish steam and hydrogen starvation at stack level, with validation on real SOEC experiments. Finally, the fast dynamics associated with starvation faults are combined with slower, degradation-driven slow varying dynamics related to aging, in order to analyze their interaction and to prepare a unified framework for diagnosis under aging conditions.
Altogether, the thesis provides a coherent PHM-oriented methodology that improves the reliability, safety and diagnosability of SOEC stacks for industrial hydrogen production.
Keywords: Solid oxide electrolyzer (SOEC), diagnostics, modeling, parameter estimation, sensitivity analysis, nonlinear Kalman filter.