At the crossroads between disciplines
A third-year PhD student at CEA-Leti, Youssof Fassi embodies a new generation of engineers who combine expertise in physical systems and artificial intelligence. After graduating with a Master's degree in mechatronics from INSA Lyon, he rounded out his studies with an M.Sc. in Aerospace Engineering from the Georgia Institute of Technology in the USA. Building on this dual background, he chose to join CEA-Leti for a PhD on predictive maintenance. His thesis focuses on optimizing systems through machine learning by bridging data and domain knowledge.
His research has already led to multiple publications, two of which drew particular recognition from both the scientific and industrial communities. Presented at the major international conferences APEC and PCIM, they each earned him the “Best Presentation Award" and the “Outstanding Presentation Award" respectively.
“These are the field's premier industry conferences, and a fantastic opportunity to share your work with the community, meet people and create synergies," he notes.
Artificial intelligence for predictive maintenance
The approach developed by Youssof centers on a key concept: combining the data collected on a system with the physical knowledge that describes it. This hybrid method, which he is exploring through two related lines of research, opens up new perspectives for predictive maintenance.
His first paper posits a breakthrough in the field of power electronics: using ultrasound to characterize the behavior of power converters.
“This is an entirely non-invasive, contactless technique that can be used to determine a system's performance without disrupting its operation", he explains. Experiments and algorithm development work carried out at CEA-Leti resulted in the filing of a patent, underscoring the innovative nature of this approach.
His second line of research is a collaborative endeavor with Aalborg University, in Denmark, as part of the European project TEAMING (e-powerTrain prEdictive mAintenance using physics inforMed learnING)*. During a six-month residence, Youssof worked on developing predictive maintenance algorithms for three-phase inverters, which are critical systems linking the battery and motor on electric vehicles. His goal was to minimize the need for sensors, computing power and data acquisition resources while maintaining reliable damage detection functionality. The result is a lightweight, robust and non-intrusive solution able to monitor changes to inverters over time using minimal hardware and software resources.
Unrestricted real-time remote monitoring
In both cases, the industrial applications of Youssof's research center on real-time monitoring of a fleet of power converters.
“This kind of approach is ideal, particularly for difficult working conditions such as off-shore facilities, where maintenance is complex and expensive," he explains. It reduces the need for dedicated instrumentation by using a non-intrusive, remote monitoring solution that does not interfere with system operation.
These advances address major challenges for smart grids and electric mobility, enabling power converters to be monitored for changes over the long term. Ultimately, they should make it possible to detect early signs of damage while limiting the need for additional hardware and computational resources.