
Special Paper Session: Graph Neural Networks for Prognosis and Health Management
Predictive maintenance (PdM) has consistently attracted the interest of the industrial community due to its significant potential for reducing maintenance costs while increasing equipment reliability. Current PdM research has focused much attention on fault perception (anomaly detection), whereas the processes of identifying the fault source and estimating its future evolution are more complex functions, influenced by many other factors. Most existing approaches have not been able to effectively manage existing knowledge to determine the cause-and-effect relationship between failures. Furthermore, the complete analysis of the correlation between identified failures and the corresponding root causes represents a major challenge for industry. Graph Neural Networks (GNN), associated with the concept of cognitive intelligence, represent an interesting avenue to explore. These methods may prove superior in semantic causal inference, heterogeneous association, and visual explanation (interpretation). In addition, these methods can achieve promising performance in reasoning tasks in a PdM process by revealing the dependency relationship between systems and/or equipment components. Recently, GNNs as the emerging neural networks have been widely used to model and analyze the data. However, there still lacks a guideline on leveraging GNNs for realizing intelligent fault diagnosis and prognostics.
We expect this special session to result in coherent and focused contributions on the topic of Graph Neural Networks for PHM. By carefully selecting complementary contributions from academia and industry, we expect this special session will be a forum to exchange knowledge between these two worlds. This special session dedicated to GNNs applied in the PHM will provide a platform for researchers and practitioners to exchange their ideas and findings, discuss the latest developments, and explore future directions in the field of the Graph Neural Networks approaches for Prognostics and Health Management.
For a better organization of this special session, please send a notification e-mail to the organizers as soon as you submit your contribution via the submission platform.
Organized by:
Ryad Zemouri (zemouri.ryad@hydroquebec.com) Research Center of Hydro-Quebec, Canada
Hung Pham (Pham.QuangHung2@hydroquebec.com) Research Center of Hydro-Quebec, Canada
Special Paper Session: AI-Driven Fault Detection and Prognostics in Rotary Machinery
Rotary machinery plays a critical role in various industries, from energy and aerospace to manufacturing and transportation. Ensuring their reliability and efficiency requires advanced fault detection, anomaly diagnosis, and prognostics techniques. While traditional condition monitoring approaches rely on vibration analysis, thermal imaging, and acoustic emissions, the integration of artificial intelligence (AI), machine learning (ML), and Digital Twin technologies has revolutionized predictive maintenance strategies.
This special session aims to bring together researchers and industry experts to discuss the latest advancements in AI-driven fault detection and prognostics for rotary machinery. Topics of interest include, but are not limited to:
• AI and ML applications in fault detection for bearings, turbines, compressors, and rotating shafts
• Digital Twin-based anomaly detection and predictive maintenance strategies
• Sensor fusion and data analytics for health monitoring of rotary systems
• Uncertainty quantification and explainable AI in machinery prognostics
• Industrial case studies and real-world applications of AI-driven maintenance
By fostering discussions between academia and industry, this session will provide a platform to exchange insights, address challenges, and explore future directions in intelligent fault diagnosis and predictive maintenance for rotary machinery.
Organized by:
Junyu Qi (junyu.qi@reutlingen-university.de) Electronics & Drives, Reutlingen University, Reutlingen 72762, Germany
Shazaib Ahsan (ahsans1@myumanitoba.ca) Department of Mechanical Engineering, University of Manitoba, Canada
Dr. Dandan Peng (dandanpeng2@gmail.com) Department of Electrical and Electronic Engineering, The Hong Kong Polytechnic University, China
Dr. Xihui (Larry) Liang (xihui.liang@umanitoba.ca) Department of Mechanical Engineering, University of Manitoba, Canada
Special Paper Session: Enhancing Predictive Maintenance with Digital Twins
This session focuses on the development of digital twins for prognostics and health management applications. Machine learning (ML)-based data-centric models are frequently employed for health assessment, including anomaly detection, fault diagnostics, and prognostics for predicting remaining useful life. However, the development of these models is often hindered by the scarcity of training data, particularly for unreported damages. Physics-based models, which leverage an understanding of damage dynamics and progression, offer an alternative by reducing the data requirements of data-driven models. Despite their potential, these models can exhibit high modeling errors due to assumptions and simplicity.
The session explores the concept of the digital twin, a dynamic virtual representation of a physical asset, as a solution to these limitations. Digital twins provide an opportunity to enhance prediction accuracy by integrating both data-centric and physics-based approaches. The discussion highlights how digital twins can address challenges such as the unavailability of data for all failure modes, the black-box nature of ML models, high modeling uncertainty at the fleet level, and the assumptions inherent in physics-based models.
Organized by:
Pradeep Kundu and Himanshu Gupta, KU Leuven, Belgium
![]() |
|||||
![]() |
![]() |
![]() |
![]() |
||
|
![]() |
![]() |
![]() |