KEYNOTE I
Title: Prognostics and Health Management: from Achievements to Future Challenges
Speaker:
Professor Kamal Medjaher, University of Technology Tarbes Occitanie Pyrénées, France
Abstract:
Since the early 2000s, Prognostics and Health Management (PHM) has emerged as a key discipline for improving the reliability, safety, and availability of industrial systems while reducing maintenance costs. It has attracted increasing interest from both academic researchers and industrial stakeholders, leading to major advances in concepts, methodologies, and applications across sectors such as manufacturing, energy, transportation, and aerospace. Early efforts focused on identifying fundamental challenges, clarifying core concepts such as health assessment and remaining useful life estimation, and structuring a unified workflow from data acquisition to maintenance decision-support.
Building on this foundation, research has progressively shifted toward the development of advanced methodologies, tools, and algorithms to enhance the performance and applicability of PHM solutions. Contributions can be broadly categorized into three main groups: physics-based approaches relying on first-principles models, data-driven approaches exploiting historical and real-time data, and hybrid approaches combining both paradigms. This diversity reflects the need to leverage heterogeneous sources of knowledge and data when dealing with complex degradation phenomena, while recent advances in Artificial Intelligence have further strengthened PHM capabilities in handling multimodal information and limited failure data.
In this keynote, we will revisit the main challenges currently faced by the PHM community, including data limitations, model robustness, uncertainty management, and deployment in real industrial environments. We will highlight selected achievements from our research projects that address these issues through AI-based and hybrid modeling solutions. Finally, we will discuss future directions, with a focus on emerging approaches such as physics-informed machine learning and digital twins to enable more reliable, interpretable, and scalable PHM systems.
Speaker’s Biography:
Professor Kamal Medjaher received his PhD in Control and Industrial Informatics from the University of Lille in 2005. He is currently a Full Professor at the University of Technology Tarbes Occitanie Pyrénées (UTTOP), where he teaches at the National School of Engineering in Tarbes (ENIT) and is a member of the Production Engineering Laboratory (LGP). His research focuses on Prognostics and Health Management (PHM), with an emphasis on developing advanced methodologies, tools, and algorithms for condition monitoring, anomaly detection, fault diagnostics, and prognostics of industrial systems. His contributions support the advancement of smart, condition-based, and predictive maintenance strategies. Professor Medjaher supervised 17 PhD students and 5 postdoctoral researchers and led numerous national and international research and innovation projects, as well as collaborative initiatives in PHM and predictive maintenance. He actively contributes to the scientific community as an Associate Editor for the Journal of Intelligent Manufacturing (JIM) and the International Journal of Prognostics and Health Management (IJPHM), serves on committees for leading international conferences, and regularly evaluates research projects for national and international organizations. He is also a member of several PhD defense and Habilitation committees and a frequent reviewer for top-tier journals and conferences.
KEYNOTE II
Title: Prognostics and Health Management Technology for Power Conversion System Based on Transient Electromagnetic Fingerprint
Speaker:
Dr. Y. Q. Chen, Deputy Director of National Key Laboratory, China Electronic Product Reliability and Environment Testing Research Institute, China
Abstract:
The Power Conversion System (PCS) serves as the “heart” of modern smart electronics within aviation, rail, electric vehicles (EVs), and drones. Subjected to extreme electrical and thermal stress, the PCS exhibits elevated failure rates due to arcing, Joule heating, and electro-migration, thereby challenging conventional reliability methodologies that depend on costly and often inaccurate pre-life predictions. Prognostics and Health Management (PHM) addresses these deficiencies by fusing failure physics with data analytics. PHM facilitates proactive health assessment and remaining useful life (RUL) prediction by integrating physics-based models with sensor data, enabling the detection of incipient faults and the prescription of corrective measures. This report explores the frontiers of PCS reliability and PHM technology. Firstly, the current development status of PCS across key application sectors is reviewed, including an in-depth analysis of emerging quality and reliability challenges. Secondly, recent advancements in the physics-of-failure (PoF) of key PCS components are elaborated upon, focusing on degradation mechanisms and corresponding PoF models for SiC and GaN power devices. Thirdly, cutting-edge research in sensing and prediction technologies is expounded, with particular emphasis on transient electromagnetic fingerprint analysis as an early indicator of degradation. Finally, practical application cases and software tools utilizing PHM technology in smart electronics equipment are presented.
Speaker’s Biography:
Yiqiang Chen is a Full Research Fellow at the Science and Technology on Reliability Physics and Application of Electronic Component Laboratory, China Electronic Product Reliability and Environment Testing Research Institute. A recipient of the State Council Special Allowance, he has been recognized as a National Talent in Science and Technology. He serves as an expert for the National Science and Technology Achievements Evaluation Centre, the Ministry of Science and Technology, and the National Natural Science Foundation of China, and so on. His research focuses on failure mechanisms and modelling of key devices—including microwave and power semiconductors—as well as prognostics and health management (PHM) for power conversion systems (PCS) and systems-on-chip (SoC). He has successfully developed eight types of monitoring chips alongside quantitative evaluation and power cycle test systems, providing critical technical support for electronic product reliability and security. He has received numerous awards, including the China Electronic Information Science and Technology Innovation Team Award and the Ministry of Industry and Information Technology Science and Technology Innovation Team Award, along with several First Prizes for provincial and ministerial Science and Technology Progress. He has led over 30 major projects, such as the National Key R&D Program and the National Natural Science Fundation. To date, he has published more than 130 SCI-indexed papers in leading IEEE journals (including TPE, TIE, EDL, TED, and TMTT) and holds approximately 100 international and domestic patents, including 12 US patents. Additionally, he has co-authored three books, 20 standards, and five software suites.
KEYNOTE III
Title: From AI driven Prognostics, towards Prognostics-Aware Safe Control Design and Learning
Speaker:
Dr. Mayank Shekhar Jha, Associate Professor at École Polytechnique de l’Université de Lorraine (Polytech Nancy) and researcher at CRAN, France
Abstract:
This keynote aims to provide both a synthesis of the current state of prognostics and a vision for the development of prognostics-aware safe operation in future intelligent systems. This talk will present a perspective on the evolution of prognostics, from early data-driven methods to current deep learning approaches, and toward their integration into safe control design and learning. This talk will visit the fundamental concepts of prognostics and health management, highlighting how data-driven methods have been used to model degradation, estimate remaining useful life, and support decision-making under uncertainty. It will then review the development of deep learning in prognostics, emphasizing the transition from handcrafted features and shallow models to more advanced architectures capable of learning rich temporal and multimodal representations.
Building on this overview, the keynote will argue that prognostics should not remain confined to prediction and maintenance support alone. A major next step for the field is to embed prognostic knowledge into the control loop itself, so that control reconfiguration and control design can explicitly account for system health and anticipated degradation. In this perspective, the talk will introduce the emerging concept of safe control design and learning, and discuss how prognostics can contribute to making control strategies more adaptive, resilient, and safety-aware.The presentation will finally outline the main opportunities, advantages, and challenges of this integration, including uncertainty management, real-time implementation, robustness, explainability, and certification.
Speaker’s Biography:
Dr. Mayank Shekhar Jha is Associate Professor at École Polytechnique de l’Université de Lorraine (Polytech Nancy) and researcher at CRAN, a joint research unit between the University of Lorraine and the French National Scientific Research Center (CNRS) since 2018. He obtained PhD in 2015 at Ecole Centrale de Lille in France and has previously held post-doctoral research position at the Institut National des Sciences Appliquées de Toulouse (INSA Toulouse) France and Research Associate position at Rolls Royce Technology Centre at University of Sheffield, United Kingdom in 2017. Dr. Jha has authored around 35-40 publications in prestigious international conferences and journals, leads a Work package (WP) in recently accepted project funded by National Agency for Research (ANR) in France titled “Self-Organizing, Smart and Safe heterogeneous Robots Fleet by collective emergence for a mission (SOS)”, has been Co-PI of three and PI of one industrially funded scientific projects with French National Space Agency (CNES) as well as Co-PI of one project with Dassault Aviation in last 5 years. Dr. Jha is an external collaborator and visiting researcher at NASA Ames Research Centre. Dr. Jha serves on the editorial board of Scientific Reports, Nature and Associate Editor of Aerospace Science and Technology, Elsevier. Dr. Jha’s current research interests include Safe Reinforcement Learning, Deep Learning based prognostics and Adiabatic Quantum Computing. (http://w3.cran.univ-lorraine.fr/mayank-shekhar.jha/?q=content/mayank-shekhar-jha)
KEYNOTE IV
Title: PHM Technologies Integrate for a Long Time Artificial Intelligence
Speaker:
Dr. Jean-Baptiste LEGER, Chief Technology and Scientific Officer, iQanto, a subsidiary of Groupe Snef, France
Abstract:
AI applied to time series data allows for learning correlation patterns between variables, parameters, conditions, and other data to explain real-world physical phenomena. AI enables learning drift and mathematical models of evolution to propagate over time and calculate remaining lifespan. AI applied to reliability allows for improving failure models and integrating conditions and events influencing failure laws to adjust the probability of failure based on the actual operational context. AI applied to fleet-wide data allows for comparing the behavior and capabilities of each asset within a fleet of similar assets. Leveraging recent advances and improvements in generative-AI, PHM solutions now incorporate new features to compare the condition of parts of an asset using photos, to assess the condition of assets, to calculate remaining useful life with richer contextual data and to benefit from many other assistance for better health management and prescriptions.
Speaker’s Biography:
Dr. Jean-Baptiste LEGER was CEO and co-founder, in 1999, of the PREDICT start-up innovative company. He sold PREDICT and joined Groupe Snef in 2018. He is now Chief Technology and Scientific Officer in charge of Industrial AI solutions at iQanto, a subsidiary of Groupe Snef. He graduated from Lorraine University, France and his PhD thesis, defenced in 1999, was on Formal Modelling Framework for Proactive Maintenance Systems mainly based on Monitoring, Predictive Diagnosis and Prognosis. He has more than 30 years of experience on usage of Artificial Intelligence in Industry and began on machine learning for CBM (Condition Based Monitoring) and advanced predictive algorithms for PHM (industrial Prognostics and Health Management). He has worked for nearly 200 companies and institutes. He was member of several International Scientific Society such as PHM Society, SAE, IEEE, IFAC, BNAE, AFNOR. He has contributed to 12 European Research & Innovation collaborative projects dedicated to AI for industry since 1995. After his PhD, he continues to contribute to scientific research and he published 12 papers in reviews and books and more than 35 papers in conferences. He was invited for evaluation of 7 PhD thesis mainly based on predictive AI algorithms for industry. His experience includes fault detection and isolation, condition monitoring, fault tolerant control, prognostic, health management, intelligent maintenance and e-maintenance. He started work in 1994 and contributed to develop remote monitoring and diagnostic systems for hydropower plants in Spain, Portugal and Norway. He has worked for more than 130 companies and institutes. He is currently working on formal approach of CBM, PHM, HUMS (Health and Usage Monitoring System) and IVHM (Integrated Vehicle Health Management). He has contributed to a lot of European Research & Innovation projects (E! RobCrane, 1999-2002: Overhead crane bridge intelligent control system, NeCST- 2004-2007: Networked Control Systems Tolerant to faults, PAPYRUS, 2010-2013: Plug and Play monitoring and control architecture for optimization of large scale production processes, PRIME, 2011-2014: Platform for enhancing Reliability of Industrial Measurements, POWER-OM, 2012-2015, Power consumption driven Reliability, Operation and Maintenance optimisation, STOICISM, 2013-2016: Sustainable Technologies for Calcined Industrial Minerals in Europe, T-REX, 2013-2016: Lifecycle Extension through Product REdesign and REpair, REnovation, REuse, REcycle strategies for Usage & REusage oriented business models) and French Research & Innovation projects (BMCI, 2009-2012: Health Management for Naval Intelligent Maintenance & Control, EXPERRE, 2012-2014: Real-time Simulation and Monitoring for Grids and Electric Embedded All Electric Vessels, CONNEXION, 2012-2016: French Cluster for New Control Solutions for Nuclear Power Plant). He participated as member of the steering committee to European projects (Twin-Control, 2015-2018: Twin-model based virtual manufacturing for machine tool-process simulation and control) and French ones (Innov’Hydro, 2015-2018: New predictive technologies and fleet-wide health assessment for hydropower plants ; udd@orano: Usine du Futur; Cassiopée: ship energy performances; TNTM: decarbonization of maritime). He was member of the Board of Directors of the Joint Laboratory ANR between CNRS, Lorraine University and PREDICT. This Joint Laboratory named PHM Factory develops robust and industrial technologies of PHM. He was member of the French Research Council CNRS MACOD working group and AFNOR (French Standard) Maintenance working group. He is co-Director of the European Committee of the PHMSociety (International Society of academics, experts and engineers on PHM), member of the SAE (Society of Automotive Engineers) HM-1 IVHM standardisation working group, member of the IEEE Std. P1856 PHM working group, member of the IFAC TC 5.1 on Manufacturing Plant Control, member of the IFAC A-MEST working group (Advanced Maintenance Engineering, Service and Technology), member of the BNAE (French Aeronautic and Space Standardisation) Testability working group and Vice-President of Diag21, a French joint Academic and Industrial Association working on Diagnostic, Prognostic and Testability. Since 2003, he has acted in 15 conference program committees. Since 2000, he was invited for 7 PhD thesis evaluations from Grenoble National Polytechnic Institute, Nancy University, Metz University, Marseille University and Besancon University. He gave invited talks during summer schools. He published 12 papers in reviews and books since 1999 and more than 35 papers in conferences (5 plenary sessions) and workshops since 2007. He was Local-Chair of the PHM Europe Conference 2014, the Vice-Chair of the PHM Europe Conference 2016 and he is the Chairman of the PHM Europe Conference 2018.
Tutorial Speaker
Title: From Machine Health to Agent Health: A PHM Framework for Monitoring Drift and Failure in Agentic AI
Speaker:
Dr. Sagar Jose, Principal Engineer and AI Technical Lead, Digital Excellence Center, Assystem, Paris, France
Abstract:
Agentic AI systems can now plan, use tools, and execute long-horizon tasks with increasing autonomy. In doing so, they exhibit behaviors that are well known in the PHM literature: drift, degradation, intermittent faults, cascading errors, and failed recovery. Yet these behaviors are still often treated as prompt-level issues or isolated AI failures rather than as system health problems.
This tutorial presents a PHM framework for agent health monitoring. Building on earlier work in machine diagnostics and prognostics, it introduces a methodological transfer from machine health monitoring to agent health monitoring: using trajectory-level observations of agent behavior across long-horizon tasks in changing environments to construct a latent health map from historical data on successful and failed runs. This map distinguishes healthy execution, degraded behavior, and distinct failure modes. The session covers monitoring, diagnosing, and improving the reliability of agentic systems in industrial and safety-critical settings.
Speaker’s Biography:
Sagar Jose, PhD, is Principal Engineer and AI Technical Lead (LLM & Agentic AI) at Assystem in Paris, where he works on agentic AI systems for safety-critical and regulated engineering environments. His work focuses on traceable agentic workflows, retrieval and reasoning systems, evaluation and observability, and the design of robust AI architectures for real-world industrial use. Before joining Assystem, he developed data-scarce diagnostics and prognostics methods for data-scarce industrial assets, including multimodal fault diagnostics and graph-based prognostics for hydrogenerator fleets in collaboration with Hydro-Québec. He received the Best Paper Award at PHM 2024 for this work. His current research bridges Prognostics and Health Management with agentic AI, with a focus on monitoring, fault detection, and recovery in long-horizon autonomous systems.