진학프로석·박사 채용 접수 플랫폼

Université Gustave Eiffel is looking for a candidate to apply for a Postdoctoral Fellowship in the framework of the Marie-Sklodowska Curie Programme 2026.

The Candidate and Université Gustave Eiffel's supervisor will apply together to develop the following research project : AI-based Online Prognostics Model for Remaining Useful Lifetime Estimation of Power Modules

This postdoctoral position is part of a proposal to be submitted for funding under the Marie Skłodowska-Curie Actions (MSCA) European Postdoctoral Fellowship program. The fellowship is contingent upon the selection and approval of the proposal by the European Commission. The successful candidate will work closely with the supervisor to co-develop (from March 30th, 2026 to September 9th, 2026) and submit the funding application, which will include a detailed research plan and a personalized Career Development Plan.

If the application is successful, the project will start at the earliest in May 2027.

1- Motivation and Background
The reliability of power electronic modules remains a key concern for ensuring safe and predictable operation in demanding environments such as transportation and energy conversion. While significant progress has been made in condition monitoring and accelerated lifetime testing, the integration of online prognostic models enabling real-time estimation of the Remaining Useful Lifetime (RUL) is still an open challenge.
This project addresses the need for data-driven and hybrid modeling approaches that combine physics-based knowledge with artificial intelligence (AI) algorithms for accurate, interpretable, and robust health state estimation during field operation.

2- State of the art
The approach generally adopted for estimating the lifetime of a power device is based on accelerated aging tests carried out in the laboratory under standard mission profiles [1], which feed into physics-based models (such as the Coffin-Manson model) [2]. These tests are accelerated in both time and amplitude through a stressing parameter which, in the case of thermomechanical aging tests, is the temperature variation (ΔT). The models developed so far can only handle a single damage mechanism at a time, known a priori, and are unable to accurately extrapolate accelerated results to normal operating conditions. In addition, the results based on lifetime models refer to the time duration until a certain percentage of accumulated failure, which do not provide the time-to-failure of each individual item of interest. These limitations represent major scientific challenges. The modeling approach based on physics-of-failure analysis has greatly contributed to understanding the mechanisms leading to degradation and ultimately to the failure of components [3], [4]. These models quantify the strains induced at the module level during operation, providing a way to describe the damage present in critical areas. They rely on multi-scale finite element simulations with multiphysics couplings between electrical, thermal, and mechanical domains. In addition, it is necessary to consider the continuous evolution of technologies, the regular introduction of new materials, increasingly complex geometries, and the many challenges these pose, particularly for numerical simulation (material properties, geometrical singularities, etc.). All these difficulties, inherent to the study of lifetime and reliability of power modules, make it extremely challenging to obtain accurate lifetime estimations for these devices. Finally, the development of reliable models for robust prognostic tools is essential to reduce the cost of reliability testing, optimize design margin, and to facilitate data-driven preventive maintenance.

3- Objectives
The work initiated in the PhD thesis of M. Ghrabli [1] proposes a new approach to estimating the remaining lifetime of a power module by monitoring the degradation of bonding wires at each load cycle. The methodology is illustrated on Fig. 1. Unlike conventional methods, which are limited to periodic profiles, this approach enables predictions under variable loadings. It combines experimental data, finite element simulations, and probabilistic models. Results depicted on Fig. 1 and Fig. 2 show high extrapolation and interpolation capabilities of the obtained model. Based on these results, the main objective is to develop and validate an online prognostics model capable of estimating the remaining useful lifetime of power modules under varying thermal and electrical stresses.

Specific goals include:
- Development of a hybrid model combining degradation indicators and AI-based algorithms.
- Integration of the model into an online monitoring framework.
- Experimental validation on a dedicated test platform with realistic mission profiles.

To achieve these goals, the project will be carried out through steps detailed bellow:
Step 1 : Validation on real industrial datasets
Test the developed algorithms on additional datasets provided by ECPE industrial partners to assess their robustness, quantify generalization performance, and identify potential limitations.
Step 2: Integration into online monitoring platforms to evaluate real-time feasibility, computational constraints, and monitoring accuracy.
Integrate the models into online monitoring frameworks: at the module level through the SATIE experimental platform, and at the system (converter) level through the facilities at Aalborg University.
Step 3: Improving extrapolation capabilities
Enhance the models’ extrapolation ability, particularly under low thermal stress conditions, by incorporating physics-informed constraints, synthetic data generation (surrogate model), or hybrid modelling.
[1]M. Ghrabli, M. Bouarroudj, L. Chamoin, and E. Aldea, “Physics-informed Markov chains for remaining useful life prediction of wire bonds in power electronic modules,” Microelectronics Reliability, vol. 167, no. February, p. 115644, 2025, doi: 10.1016/j.microrel.2025.115644

4-Planned secondments

Secondments will be decided with the candidate

5-Planned duration of the project

24 months