ENBIS Spring Meeting 2026

28 May 2026, 08:00 → 29 May 2026, 21:30 Europe/Madrid

Granada, Spain

Maria Luz Gamiz

Welcome to the ENBIS Spring Meeting 

Granada, Spain, May 28-29, 2026

Reliability, maintenance and safety in the data-driven era

In a context of increasing system complexity, digital transformation, and growing societal demands for safety, resilience, vulnerability reduction, and sustainability, the field of reliability and maintenance engineering is undergoing a profound evolution. This year’s Spring Meeting aims to serve as a cross-disciplinary platform for researchers, practitioners, and industry experts to exchange cutting-edge developments and strategic insights. At its core, the meeting focuses on how data, models, and algorithms can transform the design, monitoring, and maintenance of engineering systems, both in theory and in practice. We welcome both contributions that advance traditional maintenance and reliability models and those that push boundaries using AI-based methods to support smarter and more adaptive systems.

Beyond technical depth, the conference will encourage discussions on strategic challenges and opportunities in the field, including the deployment of digital twins, integration of AI into safety-critical systems, and the incorporation of sustainability into reliability-centered maintenance planning. The goal of this Spring Meeting is to facilitate a dynamic exchange of ideas that promotes scientific advancement and its effective translation into practical solutions, fostering meaningful collaboration with industry and society in the reliability and maintenance field.

Topics of interest include, but are not limited to:

• Data-driven reliability and maintenance.
• Stochastic processes for reliability and failure modelling.
• Condition monitoring, diagnostics, and prognostics.
• Predictive maintenance and digital twins.
• Machine learning and neural networks in reliability engineering.
• Structural health monitoring and sensor system design.
• Uncertainty and sensitivity analysis for safety evaluation of critical structures.
• Value of information analysis for improved decision-making in structural safety.
• Resilience engineering and infrastructure systems management.
• Sustainable maintenance.
• Bayesian methods in reliability and maintenance modelling.
• Partially Observable Markov Processes for Condition-Based Maintenance.
• Deep Learning for Remaining Useful Life (RUL) Prediction.
• Hybrid AI-Statistical Models for Fault Diagnosis.
• Reliability and Maintenance Strategies for Industry 5.0.

Organizing committee

María Luz Gámiz, University of Granada, Spain (Chair)
Juan Eloy Ruiz Castro, University of Granada, Spain
Rocío Raya Miranda, University of Granada, Spain
Fernando Navas Gómez, University of Granada, Spain
Delia Montoro Cazorla, University of Jaen, Spain
Nuria Caballé Cervigón, Complutense University of Madrid, Spain
Lucía Bautista Bárcena, University of Extremadura, Spain

Scientific program committee

Inma T. Castro, University of Extremadura, Spain (Chair)
Mari Carmen Segovia García,University of Granada, Spain
Rosa Elvira Lillo Rodríguez, Carlos III University of Madrid, Spain
Massimiliano Giorgio, University of Naples Federico II, Italy
Jacqueline Asscher, Kinneret College, Technion, Israel
Nikolaus Haselgruber, CIS consulting in industrial statistics GmbH, Austria

 

ENBIS Spring Meeting 2026 Highlights

Special Session:

Maintenance in the Fusion Industry

 

Mission and relevance of the event:
This special session highlights how cutting-edge reliability engineering and maintenance technology enables safe, reliable operation of fusion facilities and other Big Science installations. These assets include first-in-the-world prototypes that operate under vacuum, cryogenics, high magnetic fields and radiation, while needing high availability and strong safety margins. Unexpected downtime severely impacts costs and availability, components have long procurement times, and most operations require remote handling, making reliability, maintainability, inspection planning and logistics inseparable from design and operation.

By bringing together top voices from fusion industry, major Big science installations and academia, the session aims to build a shared roadmap—priority areas, benchmark problems, transferable methods, data standards, and collaboration opportunities—that accelerates learning and knowledge cross-pollination. Participants will leave with concrete lessons learnt from industry, new research questions to tackle jointly, and collaboration opportunities to help scaling fusion technology and Big Science operations for tomorrow.

 

Reasons to attend:

• Hear first-hand case studies from leading fusion and Big Science organisations on how they manage reliability, maintainability and safety at real scale.
• Learn what data is actually available (and missing), which methods work in practice, and how to make GenAI and digital-twin approaches auditable for safety-critical use.
• Network with statisticians, engineers and decision-makers, identify collaboration opportunities, and help shape a shared research and benchmarking agenda.

 

Session Organizers:

Prof. Juan Chiachio | Email: jchiachio@ugr.es
Department of Structural Mechanics & Hydraulic Engineering
Andalusian Research Institute in Data Science and Computational Intelligence (DaSCI)
University of Granada, Spain

Prof. Manuel Chiachio | Email: mchiachio@ugr.es
Department of Structural Mechanics & Hydraulic Engineering
Andalusian Research Institute in Data Science and Computational Intelligence (DaSCI)
University of Granada, Spain

Plenary speakers

Professor Nikolaos Limnios (Université de Technologie de Compiègne, SorbonneUniversity Alliance (France))

Nikolaos Limnios, is an Emeritus Professor (Exceptional class Full Professor) at University of Technology of Compiègne (UTC) Sorbonne University, and former Director of the Laboratory of Applied Mathematics. He has obtained his diploma in 1979 at AUTh Greece, PhD in 1983, and Dr of Sciences (Doctoratd'Etat) in 1991 at UTC France. In 1988 he was appointed assistant professor (Maitre de conferences), and in 1993 a Professor at UTC in the Laboratory of Applied Mathematics. His research interest include stochastic processes and statistics with applications in biostatistics, in reliability/maintenance, statistical seismology, stochastic mechanics, etc. He is the Editor-in-Chief of book series “Mathematics and Statistics” in iSTE with J. Wiley. He has published more than 150 journal papers and about 15 books on theory and applications of stochastic processes and statistics; including the books: Semi-Markov chains and hidden semi-Markov Chains toward applications, Springer, 2008 (with V. S. Barbu); Stochastic Systems in Merging Phase Space, World Scientific, 2005 (with V. S. Koroliuk); Semi-Markov Processes and Reliability, Birkhauser, 2001 (with G. Oprisan).

 

Anne Barros, PHD, is professor in reliability and maintenance modelling at Ecole CentraleSupélec, University of Paris-Saclay, France. Her research focus is on resilience analytics, degradation modelling, prognostics, condition based and predictive maintenance. She got a PHD then a professorship position at University of Technology of Troyes, France (2003-2014) and spent five years as a full-time professor at NTNU, Norway (2014-2019). She is currently heading the research department of Industrial Engineering and the Chair of Risk and Resilience of Complex Systems (in the Risk Resilience and Reliability – R3 Group "http://r3.centralesupelec.fr/") with the ambition to improve resilience assessment and maintenance modelling methods for complex systems.

 

ENBIS Next-Gen Session

As part of the ENBIS Spring Meeting, a special session for PhD students (ENBIS Next-Gen Session) will be organized. This session will be an excellent opportunity for doctoral students to

1. Present the current progress of their PhD research
2. Share ideas and challenges with peers
3. Receive feedback from senior researchers in this area

Application submission deadline: March 6.

PhD students interested in participating in this session should:

1. Submit their abstract via the conference website 
2. Send an email to the session organizer, Antonio Piscopo, including a short CV.

Special Issue:

delegates will be invited to submit papers to a Special Issue in

Proceedings of the Institution of Mechanical Engineers, Part O: Journal of Risk and Reliability

guest edited by the conference chairs. Papers will be subject to the standard peer review processes of the journal.

Venue

IMAG – Instituto de Matemáticas de la Universidad de Granada
Rector López Argüeta Street
18071 Granada, Spain
https://wpd.ugr.es/~imag/

Contact information

For any question about the meeting venue and scientific programme, registration and paper submission, feel free to contact: Inma Torres and Maria Luz Gamiz.

Flyer_doctorate.pdf

Thursday, 28 May

08:30 → 08:40 Registration & Information desk (8:30 - 12:30)

09:00 → 10:00 Keynote: Decision Markov Processes under Constraints for Maintenance with Extension to Decision semi-Markov Evolutions, Prof. Nikolaos Limnios

09:00 Decision Markov Processes under Constraints for Maintenance with Extention to Decision semi-Markov Evolutions.

Decision or controlled processes are of particular interest for maintenance problems, especially for determining optimal control actions to maximize the reward or minimize the cost of an operating system experiencing failures and when its temporal behavior is described by a Markov process or a semi-Markov process. This presentation includes two parts. In the first part, we introduce the fundamentals of Markov decision processes, followed by a constrained Markov decision process with absorbing states. The objective is to find a policy that maximizes the expected average reward over an infinite horizon before system failure, subject to a constraint on the asymptotic average failure rate. We propose a linear programming algorithm to obtain a stationary optimal policy, belonging to the class of history-dependent policies. Finally, we apply this method to the optimization of the maintenance policy of an aeronautical system, where the system failure rate must remain below a critical value provided by the Perron-Frobenius eigenvalue of the transition matrix.
Given the complexity of this controlled process problem, and to simplify the presentation and treatment of this model, in the second part, we present controlled semi-Markovian random evolutions in discrete time and prove some limit theorems, in particular averaging and diffusion approximation.

References
1.Boussemart M., Limnios N. (2023). Constrained Markov Decision Process for the Industry, In Ch. 16, Vol. 1 - Computational Methods and Mathematical Modeling in Cyber physics and Engineering Applications, Koroliouk D., Lyashko S. I., Limnios N. (Eds), ISTE, J. Wiley, London & N.J.
2.Limnios N., Swishchuk A. (2023). Discrete-Time Semi-Markov Random Evolutions and Their Applications, Birkhäuser/Springer.

Speaker: Nikolaos Limnios (Université de Technologie de Compiègne, Sorbonne University Alliance, France)

10:00 → 11:20 Condition Monitoring, Diagnostic & PHM

10:00 TTS R Package: Non-parametric Time-Temperature Superposition for Materials Lifetime Prediction

We present a novel non-parametric method (MNAT) within materials reliability studies to apply the Time-Temperature Superposition (TTS) principle, enabling prediction of long-term viscoelastic material behavior from short-term accelerated tests—critical for polymers, composites, adhesives, and advanced materials where full-scale durability testing spans decades.
This work presents the TTS R package (CRAN) implementing the MNAT methodology, a non-parametric approach based on first-derivative curve shifts that outperforms traditional Arrhenius/WLF models across glass transition regions. Unlike parametric methods requiring restrictive temperature assumptions, MNAT delivers smooth master curves via GAM regression with bootstrap confidence bands, following Explainable AI principles for full model transparency.
The open-source TTS package democratizes accelerated lifetime testing for Industry 4.0 materials engineering, from adhesives to composites. Real Ocean Engineering validation confirms adhesive viability vs. welding (weight/emissions reduction).

Speaker: Prof. Salvador Naya (Universidade da Coruña)

10:20 Data-Driven Condition-Based Maintenance Optimization

Studies on condition-based maintenance optimization typically consider systems with known deterioration processes. In this talk, we make the often more realistic assumption of an unknown deterioration process and we discuss various approaches for determining when to carry out preventive maintenance based on limited condition data. Contrary to many existing studies that consider a certain parametric form for the deterioration process with uncertainty its parameters, we also assume that the parametric form of the deterioration process is unknown, making the approaches fully data-driven.

For the approaches that we consider, the maintenance policies are determined based on observed condition data of K runs-to-failure. We use logistic regression to estimate the failure probability of the system as a function of its condition, and based on this we determine in which states to carry out preventive maintenance. We compare the resulting policies to the oracle policies under the assumption that the exact deterioration process is known, and analyze how the performance of our approaches depends on the amount of data that we have.

In addition to only using available condition data, we analyze to what extent the data-driven policies can be improved by creating additional, synthetic runs-to-failure. We generate these additional runs-to-failure by resampling from the deterioration increments that have been observed in the past. It turns out that this leads to improved accuracy and reduced variance compared to existing methods.

Another extension that we consider is that of a system with two condition indicators, i.e., the system deteriorates according to a bivariate deterioration process. For each run-to-failure, the two conditions are measured periodically, until the system fails. In this case, the failure probability is estimated as a function of the two condition indicators. We also address how a suitable maintenance policy can be determined in the setting with two condition indicators.

Speaker: Bram de Jonge (University of Groningen)

10:40 Process Analytical technology by ultrasounds for industrial product control

Abstract

Process Analytical Technology (PAT) is a regulatory and scientific framework introduced by the Food and Drug Administration (FDA) to promote innovation and efficiency in pharmaceutical manufacturing by enabling in-line, real-time process monitoring and quality control, consistent with quality by design (QbD) principles, rather than solely through conventional laboratory testing (quality by testing approach). In this work, a new PAT tool was developed for implementation in an injectables manufacturing plant. Considering the sterility and cleanliness constraints associated with injectable products, a non‑invasive ultrasound technique, using a pulse‑echo configuration, was selected as the most suitable approach. Additional advantages include rapid analysis, robustness and straightforward implementation. Herein we present a method for quality control of saline physiological solution for infusion (NaCl 0.9%) developed with the new PAT.

Keywords: Injectables, Process Analytical Technology (PAT), Quality by Design (QbD), Ultrasound

Introduction

In modern pharmaceutical production, PAT is a key enabler of continuous manufacturing, supporting automated feedback and feed-forward control loops that ensure product consistency 1. PAT is closely aligned with to QbD principles, which emphasizes proactive quality assurance based on process knowledge and identification of Critical Quality Attributes (CQAs) and Critical Process Parameters. By establishing relationships between CQAs, CPPs and product quality, manufacturers can operate within a design space approved by regulatory authorities, ensuring consistent quality while allowing flexibility in process optimization 2.
Regulatory agencies, including FDA and EMA encourage PAT adoption, especially when associated with Real-Time Release Testing (RTRT) [3]. Aligned with these principles, our team developed an acoustic PAT tool and applied it to determining NaCl concentration in physiological saline solution for infusion.

Setup

The experimental setup (Figure 1) includes an ultrasonic probe operating in pulse‑echo mode, a pulser-receiver to send an electrical pulse to the probe and receive the echoes after propagation in the solution, a temperature probe, a signal‑processing board, and dedicated software. This configuration enables non‑invasive, real‑time measurement of acoustic velocity in the solution.

Results

To evaluate the CQA NaCl concentration, calibration was performed using solutions ranging 0.45-1.35% (w/v) NaCl, at temperatures from 15 to 40 ºC (1ºC increments). Sound speed was measured under each condition. The data was fitted under a full quadratic 3D polynomial model. The resulting response surface (Figure 2) shows excellent predictive performance, with an RMSE of 0.032 and an adjusted R² of 0.99.

Conclusion

The ultrasound‑based PAT tool proved robust and reliable for determining NaCl concentration in physiological saline solution for infusion. The strong predictive performance confirms acoustic velocity as a suitable surrogate for this CQA. The non‑invasive nature, compatibility with sterile and closed systems, and rapid data acquisition make this PAT highly advantageous for the pharmaceutical industry, particularly for injectables. Moreover, the method aligns with regulatory priorities for PAT adoption and supports future integration into automated control strategies and RTRT.
Overall, our ultrasound PAT represents a practical, scalable technology capable of enhancing process understanding, improving quality control, and supporting the transition toward next‑generation pharmaceutical manufacturing.

Acknowledgements:

This project was developed with the support of the Amélia de Mello Foundation, through the Jorge de Mello – Indústria e Inovação Fellowship, awarded to Basinnov Life Sciences in collaboration with the University of Coimbra.

References

1. FDA, CDER, CVM, ORA. Guidance for Industry: PAT – A Framework for Innovative Pharmaceutical Development, Manufacturing and Quality Assurance (2004). https://www.fda.gov/downloads/drugs/guidances/ucm070305.pdf, last accessed 2026/01/05.
2. Henriques J, Sousa J, Veiga F, Cardoso C, Vitorino C. Process analytical technologies and injectable drug products: Is there a future? Int J Pharm 10 (554), 21-35. 2019.
3. EMA. Guideline on Real Time Release Testing (formerly Guideline on Parametric Release). EMA/CHMP/QWP/811210/2009-Rev1 (2012) https://www.ema.europa.eu/en/documents/scientific-guideline/guideline-real-time-release-testing-formerly-guideline-parametric-release-revision-1_en.pdf, last accessed 2026/01/05.

Speaker: Dr João Henriques (Basinnov Life Sciences, 1070-325, Lisboa, Portugal)

11:00 A Geometrical Method for High Dimensional On-line Process Monitoring

I will review recent work where we take an extrinsic manifold fitting point of view, contrary to a manifold learning (i.e., embedding) approach, to develop an on-line monitoring scheme with a simple geometric interpretation which requires neither decorrelation of process dynamics nor dimensionality reduction. The new monitoring framework for online or ``phase II” SPC monitors deviations from the reconstructed manifold using a novel univariate (scalar) distribution-free control chart. The average run length performance of the new method is compared to manifold learning methods that first embed the data into a lower-dimensional space before monitoring the embedded observations via explicit out of sample extensions. The manifold fitting SPC method has a provable controllable Type I error probability and, contrary to manifold learning SPC methods, it can operate in a high-dimensional regime where the ambient dimension exceeds the number of "phase I" in-control observations. An application to image data will be presented as well directions for further research.

Speaker: Prof. Enrique del Castillo (Penn State University)

11:20 → 11:50 Coffee Break

11:50 → 12:50 Data-Driven & AI for Reliability (I)

11:50 A Data-Driven Optimization Framework for Joint Reliability and Energy Efficiency in 5G Smart Factory Monitoring

Ensuring both the safety and operational continuity of smart factories requires reliable real-time hazard detection and energy-efficient predictive maintenance. This work addresses the joint challenge of guaranteeing ultra-reliable, low-latency communication for safety-critical sensors (Factory Safety Detectors - FSDs) while minimizing energy consumption for a large-scale network of equipment health monitors (Equipment Monitoring Units - EMUs). We propose the Factory Resource Optimization (FRO) framework, a data-driven iterative algorithm that dynamically allocates 5G network and edge computing resources. By decomposing the problem into optimal channel pairing and power allocation, FRO ensures that stringent latency bounds for safety-related data are never violated, thereby enhancing system reliability and safety. Simultaneously, it significantly extends the operational lifespan of EMUs by minimizing their transmission energy, directly supporting sustainable predictive maintenance cycles. Simulation results demonstrate that our approach achieves over 40% energy savings for monitoring units while fully meeting sub-millisecond latency targets for safety detectors. This work provides a scalable model for integrating communication optimization into data-driven reliability and maintenance strategies for Industry 5.0.

Speaker: Abdellah AMZIL (Computer, Networks, Mobility and Modeling Laboratory (IR2M), Faculty of Sciences and Techniques, Hassan First University of Settat)

12:10 A Data-Driven Health Indicator for Condition Monitoring of Electric Vehicles

The increasing availability of high-frequency telemetry data in electric vehicles (EVs) creates new opportunities for monitoring system health. Within the broader context of Prognostics and Health Management (PHM), these data streams have the potential to facilitate the transition from reactive maintenance to evidence-based condition assessment. However, in real-world industrial settings, reliable labels for faults, maintenance interventions or remaining useful life (RUL) are often incomplete, inconsistent or unavailable. This restricts the effectiveness of supervised predictive modelling and necessitates the development of statistical approaches that can operate without explicit degradation benchmarks.
This work proposes a data-driven framework for developing a Vehicle Condition Indicator (VCI): an interpretable, synthetic measure of vehicle health derived from heterogeneous, multisensory telemetry. The framework's primary methodological contribution is the definition of a statistically robust indicator that can capture deviations from normal behaviour while remaining stable under different operating conditions.
The framework integrates multivariate statistical monitoring and unsupervised learning techniques into a consistent modelling approach. In particular, baseline operating regimes are identified through dimensionality reduction and clustering to characterise nominal system behaviour. This modelling step enables normal variability due to usage conditions to be separated from structurally atypical behaviour. Any deviations from these patterns are quantified using metrics based on distance and density, and their temporal evolution is analysed to distinguish persistent shifts, which may be associated with degradation processes, from transient anomalies.
The resulting VCI is not intended to be a black-box score. It is designed to be interpretable and can be broken down into subsystems, such as the battery, power electronics or braking system. This breakdown provides diagnostic proxies that support engineering analysis, even when labelled failures are absent. Rather than directly estimating RUL, the indicator provides a continuous, comparable measure of relative condition which can be monitored over time and across vehicles operating under different duty cycles, thus supporting PHM decision processes without relying on fully supervised degradation models.
A case study based on real-world EV telemetry data illustrates the statistical properties and practical relevance of the proposed indicator. The analysis shows how the VCI captures emerging behavioural changes before explicit faults are recorded, enabling consistent health ranking across vehicles with different usage profiles. Although it was developed for EVs, the framework can easily be transferred to other complex industrial systems characterised by high-dimensional telemetry and scarce labels. By basing condition monitoring on interpretable multivariate statistics rather than purely predictive modelling, this work aims to stimulate discussion within the business and industrial statistics community on robust and scalable approaches to data-driven health assessment.

Speaker: Stefano Puttero (Politecnico di Torino)

12:30 Anomaly Detection in Knitting Machines for Medical Compression Socks Using Performance Indicators based on Quadrant Analysis, and PCA-Based K-means Clustering

Chronic Venous Insufficiency (CVI) affecting the lower extremities is common among adults. Compression textile products, such as compression socks, play a crucial role in the treatment and prevention of CVI and other lower-limb disorders [1], [2]. By applying controlled external pressure, these products support venous return and alleviate symptoms such as pain, edema, and venous hypertension [2], [3], [4]. Since the clinical effectiveness of compression therapy depends on an appropriate pressure profile and patient adherence, manufacturing defects that affect fit, comfort, or pressure distribution may compromise therapeutic performance and user comfort [5], [6].
This study originates from a practical industrial issue: the occurrence of appearance-related and dimensional defects in medical socks detected during quality inspections. Identifying the causes of these defects is essential to improve product quality and optimize quality control procedures.
The analysis is based on annual production data from circular knitting machines used to manufacture compression socks, including total production, number of defects, and number of machine events. From these data, three performance indicators were derived: defect rate, event rate, and defect rate per event. Together, these indicators describe quality outcomes, operational stability, and the severity of machine-related events.
As a first level of analysis, an engineering-oriented quadrant analysis was developed using event rate and defect rate per event. Quadrant-based approaches are widely used in engineering and management as intuitive tools for performance assessment, prioritization, and anomaly detection, particularly when combining frequency- and impact-related metrics [7], [8], [9]. Reference thresholds based on mean values enabled the classification of machines into four categories: critical, high risk, inefficient, and optimal. This representation provides a transparent tool for anomaly detection and constitutes the first set of results of the study.
To improve and complete the quadrant-based analysis, Principal Component Analysis (PCA) was applied to the performance indicators. PCA is widely used in engineering and industrial contexts to capture latent multivariate structures and support visual classification of complex systems [10]. Previous studies have shown its effectiveness in identifying machine clusters and validating grouping results across datasets [11], [12]. In this study, the first two principal components explained more than 96% of the total variance, representing overall operational burden and event severity. The PCA projection revealed coherent machine groupings as well as borderline and extreme behaviors not evident in univariate analyses.
Subsequently, K-means clustering was applied to the PCA scores to obtain a data-driven classification of machine performance. The resulting clusters consistently isolated machines with anomalous behavior, such as high event severity or frequency, while also identifying machines with stable and near-optimal performance.
The proposed methodology represents an essential step toward improving quality control in compression sock manufacturing. By providing a reliable classification of machine behavior, this work lays the groundwork for future studies on defect-specific correlations, reprocessing monitoring, and predictive quality control strategies. More broadly, it demonstrates how performance indicators and multivariate analysis can support continuous improvement in the production of medically relevant textile products. Finally, feedback from industrial experts supports the interpretation of the results and their practical significance.

[1] Y. Shi, R. Liu, J. Lv and C. Ye, "Biomedical therapeutic compression textiles: Physical-mechanical property analysis to precise pressure management," Journal of the Mechanical Behavior of Biomedical Materials, vol. 151, 2024.
[2] E. Akçagün, F. Siddique and A. Yilmaz, "Development of seamed compression socks and comparison with class i socks using existing mathematical models," Journal of The Faculty of Engineering, vol. 28, no. 3, 2023.
[3] C. S. Lim and A. H. Davies, "Graduated compression stockings," Canadian Medical Association Journal, vol. 186, no. 10, pp. 391-398, 2014.
[4] B. Sarı and N. Oğlakcıoğlu, "Analysis of the parameters affecting pressure characteristics of medical stockings," Journal of Industrial Textiles, vol. 47, no. 6, p. 1083–1096, 2018.
[5] B. Yang, Y. Xiong, S. Liu, F. Wang and X. Tao, "Assessment of spatial and temporal pressure distribution of compression stockings during 8-hour wear by using a smart bionic morphing leg mannequin," Measurement, vol. 262, p. 120057, 2026.
[6] J.-M. Gong, J.-S. Du, D.-M. Han, X.-Y. Wang and S.-L. Qi, "Reasons for patient non-compliance with compression stockings as a treatment for varicose veins in the lower limbs: A qualitative study," PLoS One, Public Library of Science, vol. 15, no. 4, pp. 1-11, April 2020.
[7] H. Pires Ferreira and P. Odete Fernandes, "Importance-performance Analysis Applied to a Laboratory Supplies and Equipment Company," Procedia Computer Science, vol. 64, pp. 824-831, 2015.
[8] J. Sarker, J. Hong and C. Im, "Innovation Strategies for Textile Companies in Bangladesh: Development Using Quadrant Analysis Based on a Productivity Index," Sustainability, vol. 14, no. 24, pp. 1-15, December 2022.
[9] A. Shahin and T. Hamidreza, "Investigating the supplier-producer relationship quality: a gap-quadrant analysis approach," Research in Production and Operations Management, vol. 14, no. 4, pp. 31-48, 2024.
[10] F. Alves de Almeida, G. Ferreira, P. Balestrassi, G. Belinato, R. P.A.R and M. Achilles, "Principal Component Analysis: An Overview and Applications in Multivariate Engineering Problems," in Uncertainty Modeling: Fundamental Concepts and Models, Brasília, DF, Brazil, University of Brasília, 2022, pp. 172-194.
[11] M. Chattopadhyay, S. Mazumdar, P. K. Dan and P. S. Chakraborty, "Application of principal component analysis in machine-part cell formation," Management Science Letters, vol. 2, no. 4, pp. 1175-1188, 2012.
[12] W. Hachicha, F. Masmoudi and M. A. Haddar, "Principal component analysis model for machine-part cell formation problem in group technology," Applications, 2006.

Speaker: Liseth Victoria Pasaguayo Baez (École des Mines de Saint-Étienne)

12:50 → 13:50 Data-Driven & AI for Reliability (II)

12:50 Variable-Domain Multivariate Functional PCA for PHM and RUL Prediction

Functional data analysis methods are increasingly used in Prognostic Health Management (PHM) to model degradation from multi-sensor systems. Multivariate Functional Principal Component Analysis (MFPCA) is effective in this context, but its standard formulation assumes that all units are observed over a common time domain. In practice, operational data often exhibit variable domain lengths due to heterogeneous usage conditions and lifespans. We propose a Variable-Domain Multivariate Functional Principal Component Analysis (vd-MFPCA) framework that extends MFPCA to handle multivariate functional observations defined over unit specific domains. By estimating a domain dependent covariance structure, the method extracts principal components that adapt to the observed operational domain of each unit. This enables more flexible modeling of degradation trajectories across heterogeneous units. The proposed method is applied to the NASA C-MAPSS aircraft engine dataset. Results show that vd-MFPCA improves Remaining Useful Life prediction accuracy compared to conventional MFPCA while offering clearer functional interpretations of degradation patterns. The approach provides a practical and interpretable framework for PHM applications and RUL predictions involving variable-length, multi-sensor functional data.

Speaker: Mr CEVAHIR YILDIRIM (UC3M)

13:10 Hybrid modelling through Latent Differential-Regression Analysis (LDRA) for predicting long-term equipment degradation in the Chemical Process Industry

Long-term equipment degradation decisively affects production cycles of chemical process industries (CPI), and has a major impact on plant safety, operation and economy. Equipment degradation is caused by underlying phenomena that evolve over time with a rate of change that depends on the operating conditions. To tackle this problem, a Latent Differential Regression Analysis (LDRA) methodology is introduced to predict and analyze long-term equipment and process degradation in CPI. As the degradation state is typically not observable, LDRA follows a hybrid modeling approach that combines knowledge-based feature engineering to infer an Equipment Health Indicator (EHI) and data-driven models to find the process variables related to EHI degradation.
The LDRA methodology was tested using real data from an industrial plant, where fouling takes place in several heat exchangers located in the reaction section. The case study illustrates how the proposed methodology unfolds in a real and challenging application and the results that it provides.
For this case study, LDRA successfully identified the concentration of a key component as being critically related to fouling. Furthermore, the models showed good prediction capabilities during both steady and unsteady operation periods, strengthening the hypothesis that fouling is caused by the accumulation of the identified component. Therefore, the results provided useful insights into the fouling phenomenon and allowed the plant personal to narrow down troubleshooting on a specific component of the process.
The methodology is general and can be applied to other long-term degradation modes commonly found in the CPI, such as catalyst deactivation, corrosion, mechanical degradation of packing beds and catalysts, coking, among others. We thus foresee that the proposed modeling approach based on LRDA and a case-dependent EHI can find wide application in CPI.

Authors acknowledge support from CERES – Chemical Engineering and Renewable Resources for Sustainability Research Center, funded by FCT – Fundação para a Ciência e Tecnologia (UID/00102/2025) and PRR – Recovery and Resilience Program, of the Portuguese Republic (UID/PRR/00102/2025).

Speaker: Tiago Rato (University of Coimbra)

13:30 Unsupervised Anomaly Detection in AIS Data from Fishing Vessels

The increasing availability of AIS (Automatic Identification System) data in the fishing sector offers new opportunities for monitoring and statistical analysis of maritime activity, but it also poses significant methodological challenges. We present an open problem motivated by the study of fishing vessels operating along the coast of Norway, aiming to promote the development and discussion of rigorous statistical approaches in a real industrial context.
The dataset integrates heterogeneous information at the vessel level: structural characteristics (length, tonnage, engine power, etc), target species, and declared catch volume. From a statistical perspective, this defines a multivariate problem with mixed-type covariates and potentially complex dependence structures. As an initial approach, we apply Isolation Forest to detect anomalous catches conditioned on vessel characteristics. This unsupervised method is scalable and flexible but raises relevant issues in industrial contexts, such as result stability, interpretability of anomaly scores, and uncertainty quantification.
Additionally, complete vessel trajectories obtained from AIS signals are available. These trajectories are treated as functional data, representing space–time curves over fishing campaigns. Functional Isolation Forest is used to identify atypical navigation patterns.
The main difficulty arises when attempting to analyze jointly both types of detected anomalies: on one hand, those related to vessel characteristics and catches and, on the other, those associated with navigation trajectories. To date, both analyses have been conducted separately. However, in an industrial context, it makes more sense to study both aspects in an integrated manner, as an anomalous catch may be linked to an unusual navigation pattern.
Therefore, it is necessary to develop a common framework that combines information from both sources. This involves integrating different anomaly measures into a single analysis framework, taking operational context into account (such as fishing area, season, or target species), and defining clear criteria for evaluating results, particularly in settings where reliable labels indicating true anomalies are unavailable.
We propose this case as an open challenge intended to foster discussion and collaboration on anomaly detection in complex and heterogeneous data. The goal is to advance toward more robust and well-founded methods that can be reliably applied in real industrial contexts, combining methodological contributions with proposals to improve result interpretation and validation.

References:
1. Mangé et al., Anomaly detection in ship trajectories using machine learning and dynamic time warping. Eng. Appl. Artif. Intell. 2025, 157.
2. Staerman et al., Functional Isolation Forest. Asian Conf. on Mach. Learning. 2019, 101.
3. Liu et al., Isolation – based anomaly detection. ACM Trans. Knowl. Discov. Data. 2012, 6, 1.

Speaker: Fernando Navas-Gomez (University of Granada)

13:50 → 15:30 Lunch

15:30 → 18:05 Special Session: Maintenance in the Fusion Industry

15:30 Always On, Always Safe: The engineering challenge of Continuous Operation in the IFMIF-DONES fusion neutron source

Large scientific infrastructures increasingly depend on availability-driven Operations \& Maintenance (O\&M) models, where every additional hour of uptime translates directly into scientific output. In IFMIF-DONES, this link is especially explicit: higher infrastructure availability enables longer irradiation time and therefore greater throughput and quality of materials data for fusion-relevant qualification campaigns.

Achieving sustained high availability is, however, far from a conventional O\&M problem. Key systems will operate in radiological environments where direct human intervention is constrained, access windows must be minimized, and maintenance actions must be carefully planned, executed, verified, and documented. This creates a demanding operational paradigm: maximize continuity of operation while preserving strict safety, traceability, and configuration control.

This presentation outlines an O\&M approach tailored to IFMIF-DONES, focused on “availability without compromising safety.” A central pillar is the design and integration of remote handling capabilities for inspection, maintenance, rescue and recovery tasks, enabling effective interventions in contaminated or high-dose areas and supporting maintainability as a design requirement rather than an afterthought.

Building on this foundation, we discuss how data-driven and AI-enabled methods can further reduce unplanned downtime and optimize maintenance scheduling. Examples include predictive maintenance to anticipate degradations before failure, and reinforcement learning–inspired strategies to support complex operational decision-making and robotic task execution, always under rigorous supervision and interlock logic.

The talk concludes with a practical view of how these elements—remote maintainability, predictive analytics, and intelligent assistance—can be combined into a coherent O\&M concept for IFMIF-DONES that is measurable, auditable, and aligned with the facility’s core mission: delivering maximum irradiation availability in a controlled and safe manner.

Speaker: Moisés Weber (Consorcio IFMIF-DONES España)

15:55 Ensuring Reliability Across Generations: Maintenance and Asset Management at CERN

CERN's mission is to design, build, and operate particle accelerators and experimental facilities that serve a diverse international scientific community. The accelerator complex, experimental areas, and associated infrastructure encompass a wide variety of technologies and components, ranging from cutting-edge instruments to legacy systems with decades of operational history.

Given the long operational lifetimes of its installations, CERN faces unique challenges in ensuring the retention and transfer of critical knowledge across multiple generations of engineers and scientists. Effective asset and maintenance management is therefore essential to guarantee reliability, performance, and safety throughout the lifecycle of the facilities.

This presentation will provide an overview of CERN's current approach to asset and maintenance management, including organizational structure, strategic priorities, and the integration of advanced tools and methodologies. Emphasis will be placed on how data-driven decision-making, maintenance strategies, and knowledge management practices are applied to maintain operational excellence.

Attendees will gain insight into the challenges and solutions of managing one of the world's most complex and long-lived scientific infrastructures, with potential applications to other large-scale, technology-intensive organizations.

Speaker: Goran Perinic (CERN)

16:20 Reliability Gets Hot at 100 Million Degrees: RAMI Challenges and Opportunities in Fusion Energy

Fusion energy is rapidly transitioning from experimental research to a highly capitalized industrial ambition, with private and public actors targeting near-term deployment of fusion power plants. The cumulative funding in private fusion surged to 13 billion EUR since 2020. The fusion private ecosystem is now larger than ever, with 77 companies and growing. However, beyond the well-known physics and engineering challenges, a critical gap remains largely unaddressed: the absence of robust Reliability, Availability, Maintainability, and Inspectability (RAMI) frameworks to support credible lifetime and performance assessments.

Today, there is limited comprehensive methodology to justify key parameters such as plant availability or maintenance strategies in fusion environments. This lack of validated systematic and well-founded approach directly impacts the ability to estimate Levelized Cost of Energy (LCOE), undermining confidence in economic viability and scalability. In many cases, RAMI considerations are still secondary to performance driven design, creating a significant risk for future deployment and operation of an expected series of thousands of plants to be operated.

At the same time, this gap represents a major opportunity. As fusion projects move toward industrialization, the demand for structured RAMI methodologies, predictive tools, and data-driven models will grow rapidly (including, of course, Artificial Intelligence tools). Organizations capable of delivering credible and standardized lifetime assessments, availability projections, and maintenance strategies will play a key role in enabling investment decisions and de-risking fusion projects. RAMI is poised to become a cornerstone discipline in the transition from experimental devices to reliable fusion power plants.

Speaker: Javier Serrano (Fusion for Energy (F4E))

16:45 A Novel Approach to Reliability Data Selection and Integration for IFMIF-DONES Accelerators

The motivation of this work is the development of new methods for selection and combination of multi-source reliability data to obtain useful results within the WPENS project for the IFMIF-DONES facility.

Typically, the component failure rates used in RAMI studies are estimated by the heterogeneous datasets extracted from several sources of different content, structure, and accuracy. This poses an important problem: databases might not be complete in the sense of that the information is limited (e.g. only mean values are available), or simply, is missing. With the objective of integrating multiple inputs into one model of the failure rate, a novel methodology consisting of two phases is introduced. Firstly, a Bayesian result is adopted to prove that the failure rate can be modelled through a Gamma distribution, and secondly, two common approaches of the reliability analysis are proposed (method of moments and minimizing the Kullback–Leibler divergence) to fuse Gamma distributions coming from different components used in the accelerator.

The performance of this methodology is validated by means of a depth simulation study by introducing different sample sizes and data affected by homogeneity and heterogeneity. Finally, it is applied to an empirical dataset.

Speaker: Dr Christian Acal (University of Granada)

17:10 Ontology-Driven Natural Language Query System for Fusion Maintenance Planning

The increasing complexity of fusion facilities demands robust data governance strategies capable of integrating heterogeneous knowledge sources while ensuring semantic consistency, traceability, and regulatory compliance. This talk presents an ontology-based data governance model as the foundational architecture for a Natural Language Query (NLQ) system designed to support maintenance planning in fusion facilities. We argue that the successful implementation of such an ontology-driven approach is contingent upon a high level of maturity in "classical" corporate data governance. Without a solid foundation of data quality, stewardship, and standardized master data management, the higher-level semantic framework lacks the structural pillars necessary for reliable knowledge representation.

The proposed governance framework establishes a formal ontological structure that harmonizes maintenance taxonomies, component hierarchies, failure mode classifications, and procedural constraints within a unified semantic layer. By grounding advanced data governance in ontological principles, the model enforces controlled vocabularies, domain axioms, and inference rules that preserve data quality and interoperability across distributed maintenance information systems. Built upon this governance backbone, the NLQ system translates unstructured maintenance queries posed by engineers and planners into structured SPARQL expressions, leveraging the ontology as both a semantic mediator and a validation mechanism.

The system architecture integrates large language models with ontology-driven entity recognition and relation extraction modules, enabling context-aware disambiguation of maintenance-specific terminology such as remote handling sequences, radiation exposure thresholds, and component lifecycle states. Evaluation on a corpus of representative maintenance planning queries for IFMIF-DONES subsystems demonstrates that this governance-anchored approach yields significant improvements in query precision and recall compared to schema-agnostic baselines, while simultaneously enforcing compliance with safety and quality assurance standards. The results confirm that the transition from mature corporate governance to ontology-based semantic layers constitutes a critical enabler for intelligent, linguistically accessible decision-support tools in next-generation fusion engineering environments.

Speaker: Francisco Carmona Martínez (Quantia Ingeniería y Consultoría SL)

18:05 → 18:30 Coffee Break

18:30 → 19:30 Resilience, Sustainability & Industry 5.0

18:30 Integrating Real-Time Fatigue Monitoring into Adaptive Human–Robot Collaboration Systems

The increasing use of collaborative robots in warehouse logistics is changing intralogistics systems, improving productivity, but is also creating significant challenges for operator safety, ergonomics, and operational reliability. The introduction of new technologies implies continuous cognitive and physical pressures on operators in modern warehouses, increasing the impact of human fatigue on system performance and the risk of accidents. In this context, traditional models of human-robot collaboration are mostly static and unresponsive to changes in workers' psychophysiological states. To overcome this gap, an integrated approach will be proposed by combining real-time biometric fatigue monitoring with adaptive robotic behavior and dynamic risk assessment.

The development and validation of a socio-technical framework will allow estimating operator fatigue using specific indicators derived from non-invasive eye-tracking data; hence, a quantitative cognitive fatigue index will be defined and integrated into a decision-making model that dynamically adjusts task assignment and robot behaviour in collaborative warehouse operations. Unlike traditional Human Robot Collaboration (HRC) systems that rely on fixed safety thresholds, a model enabling the robot to proactively adapt its contribution to the task in response to changing human conditions and the relative Risk Assessment will be proposed.

Speakers: Mario Di Nardo (Università Telematica Pegaso), Prof. Teresa Murino (Materials and Production Engineering of the University of Naples)

18:50 Data-Driven Model Selection and Performance Assessment of Solar Chimney Systems Using CFD and Multivariate Analysis

The increasing complexity of engineering systems and the growing demand for sustainable energy solutions require robust data-driven approaches for performance evaluation and decision-making. Solar Chimney Power Plants (SCPPs) constitute a promising renewable energy technology; however, their optimal design remains challenging due to the interaction of multiple physical and environmental variables, as well as external operating conditions.

In this work, we propose a methodology that combines Computational Fluid Dynamics (CFD) simulations with multivariate statistical analysis to support model selection and performance assessment of different SCPP configurations. A dataset is generated from numerical simulations considering key variables such as air velocity, temperature, pressure, and mass flow rate, along with geometric design parameters of the system.

The proposed approach focuses on the joint analysis of multiple performance indicators in order to identify the most suitable configuration under different operating conditions. Statistical techniques, including correlation analysis and multivariate exploratory methods, are used to study the relationships among variables and to evaluate the global behaviour of each design alternative. Additionally, the methodology allows the detection of dominant factors influencing system performance.

In particular, Chernoff faces are employed as an integrative visualisation tool to represent multiple variables simultaneously, enabling an intuitive and comprehensive comparison of competing models. This approach facilitates the identification of patterns that may not be evident through traditional univariate or bivariate analysis.

The results provide a practical and interpretable framework for model selection in complex engineering systems, supporting decision-making based on multiple criteria. This methodology is aligned with current trends in data-driven engineering and offers potential extensions to reliability assessment, uncertainty analysis, and optimisation in other industrial applications.

Speaker: Alberto Hananel Baigorria (Universidad Católica Santo Toribio de Mogrovejo)

19:10 Modeling a Human-Centric Framework to Link Cognitive Factors and Maintenance Reliability in HRC Systems

In the Industry 5.0 transition, Humans and Collaborative Robots (Cobots) work together in different operations.. In maintenance ,the role of Human-Robot Collaboration (HRC) is becoming increasingly crucial for achieving sustainable efficiency and human-centered automation. While Collaborative Robot (Cobot) design takes into account worker safety, ergonomics, and process precision, it is necessary to consider aspects such as psychological stress in human operators that may lead to uncorrect operations and uncertainty. In maintenance activities, this stress can compromise the operations and the related quality, leading to increased defect rates, accelerated component degradation, and economic losses. Despite many studies on Human Reliability Analysis (HRA) and Human–Robot Interaction (HRI), there is no integrated quantitative framework that connects trust, stress, and repair reliability within maintenance operations.
The ongoing study proposes a dynamic model linking operator trust and stress to maintenance performance and post-repair degradation in HRC environments. The model will consider and quantify the effect of trust level on operator stress, determine the probabilistic relationship between stress and repair accuracy, and evaluate the overall economic implications, including rework and downtime costs. The analytical core of the framework is based on a Weibull reliability formulation extended with a cognitive stress factor. The resulting equations describe the probability of repair failure as a nonlinear function of trust-induced stress.

Speaker: Mario Di Nardo (Università Telematica Pegaso)

Friday, 29 May

08:30 → 08:40 Registration & Information desk (8:30 - 12:30)

09:00 → 10:00 Keynote: Risk and Resilience of Complex Systems (RRSC), Prof. Anne Barros

09:00 Risks and Resilience of Complex Systems (RRSC)

At CentraleSupélec, research on resilience is conducted by the Laboratoire de Génie Industriel (LGI) and its research team R3 (Risk Resilience Reliability). Today, resilience has become a key issue for all so-called critical infrastructures. Although each sector has its own specific characteristics, the methodological frameworks used to study resilience appear to be very similar. This opens the possibility for synergies and cross-sector reflections among stakeholders.
Thus, although relatively young, the R3 team (created in 2009) has grown significantly in recent years and now maintains major national, international, academic, and industrial partnerships in the following sectors: energy, telecommunications, transportation, space infrastructures, healthcare systems, manufacturing industry, and smart cities.
A sign of this methodological cross-disciplinarity is the Chair on Risks and Resilience of Complex Systems (RRSC), founded in 2009 with EDF. The chair was renewed in 2019 with the addition of two new partners, Orange and SNCF, and in 2024 with NaTran and RTE. The chair's objective today is to study the resilience of critical energy, transport, and telecommunications infrastructures, focusing on their interdependencies and on the challenges associated with climate change and geopolitical risks.
In order to be as comprehensive as possible, the ambition is to address all phases of the resilience process: prevention , degradation and crisis, restoration and recovery, and learning.

In the scientific project of the chair, 3 main areas have been defined:
1. Systems Modelling for Risk Resilience and Reliability
2. Decision Optimization for Risk Resilience and Reliability
3. Environmental and Geopolitical Challenges of interdependent infrastructures

In this presentation, after a brief overview of the chair scientific project, we will focus on some selected works achieved in areas 1 and 2 and 3 that may be of interest for the audience.

Speaker: Anne Barros (CentraleSupélec, Paris Saclay University, France)

10:00 → 11:20 Stochastic Modelling & Statistical Estimation

10:00 Advanced FIT-Rate Estimation for Serial Systems

FIT(failures in time )-rates are a typical reliability measure for the constant part of the bathtub curve for non-repairable systems. The FIT-rate is the parameter of the exponential distribution in units of 109 hours. It is the inverse of the mean times between failures (MTBF). It is assessed at so-called MTBF-tests. For a serial system, FIT-rates of the individual components are added up. Frequently, individual FIT-rates are given as interval estimations at a certain confidence level (CL), e.g., 60% or 90% CL. Although it is common practice, adding up such FIT-rates can lead to an overly conservative estimation of the overall FIT-rate for a serial system.

We introduce a novel method to calculate FIT-rates at a given CL for serial systems. This method is based on data pooling. Raw data from individual MTBF-tests are combined. Situations of MTBF-tests with different durations are covered as well. Corner cases of various lengths of MTBF-tests, like balanced and unbalanced cases, are investigated. This data pooling forms the basis for the further FIT-rate calculation of a serial system. This new estimator is substantiated by a probability theoretical model. Statistical properties of this new estimator are examined. Special focus is put on the coverage probability of this new estimator, such that it ensures the stated CL.

The novel method is applied to multi-chip semiconductor products and supported by simulations. We see wide range of potential of this method for further applications where the raw data of MTBF-tests are available. This enables manufacturers of complex systems to achieve a competitive FIT-rate based on existing data.

Speaker: Dr Horst Lewitschnig (Infineon Technologies Austria AG)

10:20 Maximum Likelihood Inference for Non-Linear Wiener Processes under Multiple Arithmetic Reduction of Degradation (ARD) Maintenances

As time passes, complex industrial systems suffer degradation phenomena that will inevitably lead them to failures. Several degradation models have been proposed to model these phenomena [5]. The most usual ones are based on stochastic processes such as the Wiener process, the Gamma process or the Inverse Gaussian process. In the case of complex systems that must operate without interruptions for long periods of time, waiting for a failure to initiate a repair process is not a viable option. Hence, industrial agents need to carry out preventive inspections, in which they assess the degradation level of the system, along with preventive maintenances, aiming to reduce the degradation level.

Overall, the goal of degradation data analysis is to estimate the lifetime of industrial assets. Mathematically, this is done by estimating the residual time before the degradation level hits a certain threshold after which the product will be unable to perform its task. In order to do so, it is crucial to have a good estimation of the model parameters. In the case of systems for which inspections and/or maintenance actions are expensive, it is harder to estimate the effect of the maintenance actions as less data means less precision. Furthermore, in some cases, the degradation may present non-linear patterns that need to be considered. For instance, the degradation might be accelerating [4], deccelerating or hit a plateau [2]. Thus, it is important to develop general modelling frameworks capable of taking into consideration as much information as possible.

The present work presents a maximum likelihood based inference method for a degradation model where the underlying degradation process is a Wiener process and the maintenance effect is modelled by an Arithmetic Reduction of Degradation (ARD) model [1] [3]. The originality of this work is that the drift and volatility of the Wiener process are both assumed to be non linear. The ARD framework stipulates that each maintenance reduces the degradation level by a quantity proportional to a certain amount accumulated in the history of the studied asset. We consider multiple maintenance types and associate each of them to a coefficient representing their efficiency. Indeed, the same piece of system may be subject to different types of maintenance actions. For instance chemical treatments may have different efficiencies depending on the used product. It is worth to note that the present method covers all possible schemes of observation.

The presentation is divided in four parts. Firstly, we introduce our model in a framework where we assume both the drift and the volatility to be as general as they can be and we deduce the expression of the likelihood for an arbitrary dataset. Secondly, we assume a parametric shape for the drift and volatility that lead to a set of estimators for the model parameters.
Thirdly, the estimators are implemented with the Julia language and their quality is assessed through simulations. Finally, the estimators are applied to real data from energy production systems.

References
[1] L. Bautista, I. T. Castro, C. Bérenguer, O. Gaudoin, and L. Doyen. Parameter estimation in an ARD∞ wiener degradation and maintenance model under different observation schemes. Proceedings of the Institution of Mechanical Engineers, Part O: Journal of Risk and Reliability, 240(1):261–277, 2026.
[2] M. Fouladirad, M. Giorgio, and G. Pulcini. A transformed gamma process for bounded degradation phenomena. Quality and Reliability Engineering International, 39(2):546–564, 2023.
[3] S. Mercier and I. T. Castro. Stochastic comparisons of imperfect maintenance models for a gamma deteriorating system. European Journal of Operational Research, 273(1):237–248, 2019.
[4] G. A. Whitmore and F. Schenkelberg. Modelling accelerated degradation data using wiener diffusion with a time scale transformation. Lifetime Data Analysis, 3(1):27–45, 1997.
[5] Z.-S. Ye and M. Xie. Stochastic modelling and analysis of degradation for highly reliable products. Applied Stochastic Models in Business and Industry, 31(1):16–32, 2015.

Speaker: Tom Berger (Laboratoire Jean Kuntzmann, Université Grenoble Alpes)

10:40 Discretization of the Gamma Process: Structural Properties and Reliability Applications

The discretization of the gamma process plays an important role in both theoretical investigations and practical implementations of stochastic modeling. The gamma process is a continuous-time, non-decreasing Lévy process with independent increments and is widely used in applications such as reliability engineering, survival analysis, and degradation modeling.

In practice, however, observed degradation data are often collected in a categorized or discretized form. In many engineering applications, the data must also be categorized in order to apply decision-making frameworks such as Markov decision processes (MDPs). For this reason, studying the discretized gamma process becomes essential for linking theoretical models with real-world data.

Discretization enables the approximation of continuous-time dynamics while attempting to preserve the key probabilistic properties of the original process. In particular, it is important to examine whether fundamental characteristics—such as the Markov property, independent increments, stationarity of increments, and infinite divisibility—are maintained after discretization. Understanding the preservation or alteration of these properties is crucial for ensuring model validity and analytical consistency.

Furthermore, analyzing these structural properties provides insight into transition behavior, dependence structure, and long-term dynamics, all of which directly influence statistical inference and predictive performance. Therefore, a rigorous study of the discretized gamma process and its characteristic properties serves as an important bridge between continuous-time stochastic theory and practical modeling applications. Additionally, the impact of this discretization will be investigated in the context of maintenance and reliability analysis, where degradation modeling plays a central role.

Speaker: Elham Mosayebi Omshi (University of Technology of Troyes)

11:00 Efficient Post-Shrinkage Estimation Strategies in High-Dimensional Cox’s Proportional Hazards Models

Regularization methods such as LASSO, adaptive LASSO, Elastic-Net, and SCAD are widely used for variable selection in statistical modeling. However, these approaches primarily focus on variables with strong effects and often overlook weaker signals, which may lead to biased parameter estimates and reduced predictive performance. To address this limitation, corrected shrinkage strategies have been proposed to incorporate both strong and weak signals within the estimation process. Nevertheless, existing developments have largely been restricted to linear models, and their applicability to survival data has received limited attention, despite the common presence of both strong and weak covariate effects in biomedical survival studies.

To bridge this gap, we propose a novel class of post-selection shrinkage estimators within the Cox proportional hazards model framework. We establish the asymptotic properties of the proposed estimators and evaluate their performance through simulation studies that explicitly account for weak signals. The results demonstrate improvements in estimation accuracy and predictive performance. Finally, we illustrate the practical utility of the proposed approach using two real-world datasets, highlighting its advantages over existing methods.

Speaker: Reza Arabi Belaghi (SLU)

11:20 → 11:50 Coffee Break

11:50 → 12:50 Advanced Methods in Reliability, Maintenance and Degradation Analysis (I)

11:50 Imperfect diminishing maintenance modelling and reliability analysis for a multi-process degradation system

This work proposes a framework for analysing reliability and optimizing maintenance in a system subject to multiple degradation processes. Unlike models that assume perfect maintenance, this study incorporates an imperfect maintenance mechanism with diminishing efficiency, where the ability to reduce accumulated degradation decreases with each successive maintenance intervention.
Analytical survival functions and failure-time distributions for two distinct failure scenarios are derived. The first one is an aggregate-threshold scheme modelled via Wiener processes, where system failure is driven by a linear combination of all the degradation processes. In the second scenario, an aggregate-or-critical scheme modelled using Gamma processes is considered, where failure occurs if either the total degradation or a specific critical component exceeds its respective threshold. These mathematical derivations explicitly account for the discontinuities in degradation paths caused by imperfect repairs.
Building on these reliability models, our work develops two condition-based maintenance (CBM) strategies: one using the conditional Residual Useful Life (RUL), and another based on direct degradation thresholds. Numerical experiments show that optimizing the inspection intervals and decision thresholds of these policies provides significantly lower long-run average cost rates compared to traditional periodic maintenance strategies.

Speaker: Lucia Bautista Bárcena (Universidad de Extremadura)

12:10 A maintenance policy for a partially observable parallel system with two components subjected to stochastically dependent degradation processes]{A maintenance policy for a partially observable parallel system with two components subjected to stochastically dependent degradation processes

Degradation modelling based on stochastic processes has become a key tool in reliability analysis and condition-based maintenance of engineering systems. While a large body of literature addresses single-component systems, many practical assets consist of multiple components whose degradation processes are influenced by common environmental or operational factors. In such systems, component deterioration is often correlated, and maintenance decisions become particularly challenging when only partial information on the system state is available. This situation arises when some components can be directly inspected while others remain hidden or inaccessible, requiring maintenance policies that account for both dependence among components and incomplete state observability.
This paper investigates maintenance decision-making for a parallel two-component system in which one component is fully observable and the other is latent. The degradation of both components is modelled through stochastic degradation processes driven by a shared random effect, capturing the influence of unobserved environmental or operational conditions and inducing correlation between the two degradation trajectories. At inspection times, the degradation level of the observable component is perfectly known, whereas the state of the latent component must be inferred indirectly from the information provided by the observable one.
The system is assumed to operate over a finite time horizon with a single inspection and maintenance opportunity. At the inspection time, the decision-maker observes the degradation state of the accessible component and selects an optimal maintenance action. System failure is defined through a threshold-based degradation model, linking component deterioration to system lifetime. The objective is to determine the optimal inspection timing and the corresponding maintenance action that minimise the expected cost per unit of time over the system lifetime.
The analysis focuses on a one-inspection policy in order to clearly isolate and characterise the informational contribution of the observable component to the maintenance decision process. This modelling choice allows the role of correlation and partial observability in supporting optimal decisions to be examined in a transparent way. Nevertheless, restricting the number of inspection can also provide operational and/or economic benefits in some industrial applications (such as, for example, offshore energy or large-scale plants) where the mobilisation of specialised inspection teams and the downtime associated with maintenance interventions may justify policies with limited inspection opportunities.
Obtained results offer insights into the value of indirect information in condition-based maintenance policies.

Speaker: Dr Nicola Esposito (Scuola Superiore Meridionale)

12:30 Optimal repair policies for monotone systems with simultaneous failures

Industrial systems composed of multiple interacting components are often exposed to external events that can simultaneously affect several parts of the system. Such events, commonly referred to as common shocks, introduce strong dependence between component failures and significantly complicate maintenance planning. When multiple components fail at once, the system may experience large operational disruptions and costly repair interventions, making the design of efficient maintenance policies a challenging problem.
In this talk we show optimal repair strategies for multi-component systems subject to dependent failures. The dependence structure is modeled using the Lévy–Frailty Marshall–Olkin (LFMO) distribution, which captures both individual component failures and simultaneous failures caused by common shocks through an underlying stochastic degradation process..
The maintenance decision problem is formulated as a sequential decision-making problem. Starting from the continuous-time reliability model, we construct a Semi-Markov Decision Process (SMDP) under an average-cost performance criterion. The SMDP is then transformed into an equivalent discrete-time average-cost Markov Decision Process (MDP), allowing the use of dynamic programming methods to compute optimal or near-optimal maintenance policies.
Because the transition dynamics of the system can be complex or analytically intractable, we also develop a simulation-based approach to approximate the underlying MDP. In particular, we employ reinforcement learning techniques based on the Stochastic Approximation Value Iteration Algorithm (SAVIA) to estimate optimal policies using simulated trajectories of the LFMO process.
In addition to the policies obtained through the MDP formulation, we also study and compare several alternative maintenance strategies. These include simple rule-based policies that depend on aggregate system indicators, such as the number of failed components. Although such policies are generally less optimal than those derived from the MDP framework, they are computationally inexpensive and scale well to larger systems. Simulation experiments show that, despite their simplicity, these heuristic policies can sometimes achieve competitive performance, making them attractive practical alternatives when solving the full MDP becomes computationally challenging.
Computational experiments illustrate how the dependence structure among failures affects maintenance decisions and demonstrate the effectiveness of the proposed framework for designing and evaluating maintenance policies in complex reliability systems.

Speaker: Cristóbal Reyes Pavez (Universidad Adolfo Ibáñez)

12:50 → 13:50 Advanced Methods in Reliability, Maintenance and Degradation Analysis (II)

12:50 Reliability Verification with Economic Constraints

The verification of a reliability target is one of the last and most expensive steps in the development process of a technical product. It ensures a certain level of product reliability prior to market release and provides a limit for expected warranty costs. Depending on product complexity and diversity as well as production volumes, a sufficient verification of a challenging target may become economically impossible. In such cases, current industrial practice is to select critical product variants and determine for this choice an unconstrained reliability verification plan without consideration of economic consequences.
This talk discusses the role and importance of reliability verification in product development and shows how economic constraints can be considered in stochastic reliability models for planning of optimal reliability verification. An example from the automotive industry illustrates the practical applicability of this innovative approach.

Speaker: Nikolaus Haselgruber (CIS Consulting in Industrial Statistics GmbH)

13:10 Monotonicity-Based Parent-Child Inference for the Efficient Computation of Signatures and Minimal Cut/Path Sets in Coherent Systems

In binary reliability systems, the minimal cut-sets, path-sets and signature vector [1, 2], provide the basis for computing key reliability measures, such as the survival function and the mean time to failure [1, 3]. However, these quantities are typically obtained by exhaustive evaluation of the structure function over the Boolean state space [1], whose size grows exponentially with the number of components: if $n$ is the number of components in the system, there are $2^n$ possible states of the system [2]. Thus, as system size increases, this becomes computationally burdensome, and exact calculation becomes practically impossible even for moderate-sized systems.
In this work, we propose a new method to compute minimal cut-sets, path-sets and the signature vector for monotonous binary systems. We make no assumptions about the structure of the system and consider its structural function a black box. The proposed approach explores the Boolean state space by failure level, defined as the number of failed components. At each iteration, a level is selected from an adaptive biased distribution, then an “unchecked” state within that level is sampled uniformly at random and evaluated using the structure function, and then its status (working or failed) is propagated through the state lattice using monotonicity-based parent-child inference [2]. Indeed, if the sampled state is working, all of its ancestors toward lower failure levels are inferred working; and if it is failed, all of its descendants toward higher failure levels are inferred failed. In this way, a single exact evaluation can determine the status of many additional states without further structural-function calculations. With this, we propose confidence intervals for the signature vector. And in the cases where all states can be checked, we can compute the minimal cut- and path-sets, and also the signature of all intermediate states.
Preliminary results across multiple tested systems indicate a substantial reduction in computation time when using our algorithm.
[1] F. Samaniego, System Signatures and their Applications in Engineering Reliability. Springer, 2007.
[2] J. Navarro, Introduction to System Reliability Theory. Springer, 2022
[3] J.-L. Marichal, P. Mathonet, and T. Waldhauser, “On signature-based expressions of system reliability,” Journal of Multivariate Analysis, vol. 102, no. 10, pp. 1410–1416, 2011.

Speaker: Simon Avila (Universidad Adolfo Ibañez)

13:30 Estimation of return levels with the generalised F distribution #5

The generalised F distribution (GenF) has been called an “umbrella” that covers several conventional distributions. In this work, we compare GenF with the Generalised Extreme Value (GEV) distribution with focus on estimation of return levels. Upper-tail asymptotics of the GenF are examined in relation to GEV. Potential numerical challenges in fitting the GenF are discussed. Data sets of river discharge and significant wave height data are analysed, as well as synthetic data from simulation studies.

Speaker: Jesper Rydén (Swedish University of Agricultural Sciences)

13:50 → 15:30 Lunch

15:30 → 17:15 Next-Gen Session

15:35 A Finite-Time Prescriptive Maintenance Policy for Systems Exhibiting Bathtub-Shaped Degradation Rates

Real-world degradation processes often exhibit three distinct phases, namely a running-in (accommodation) phase, a steady-state (normal) phase, and a catastrophic wear phase. As a consequence, the degradation rate function, here intended as the first derivative of the mean degradation function, is typically bathtub-shaped: it decreases during the first phase, remains approximately constant during the normal phase, and increases again during the last phase. In such settings, maintenance policies should explicitly account for this behavior, since model misspecification can lead to systematically suboptimal decisions and, in turn, increased expected costs. Motivated by these considerations, we introduce a non-homogeneous gamma degradation process inspired by Giorgio et al. (2023) that accounts for a bathtub-shaped degradation-rate function, and we use it to develop a prescriptive maintenance policy. In this work, we consider a prescriptive maintenance policy for a degrading system in which degradation is influenced by a usage rate, that represents the operating intensity of the system, and where prescriptive actions include its adjustment.
We consider a nominal usage level corresponding to standard operating conditions. At each inspection, the usage rate is selected from a finite set for the next interval, where reductions slow degradation but incur a per-unit-time cost that increases with the deviation from nominal. The system is planned over a finite operating mission with periodic, perfect inspections and non-self-announcing failures. At each inspection, the policy selects the next usage rate and may perform preventive or corrective replacement.
The problem is formulated as a finite-time Markov decision process and solved via backward recursion. Numerical studies vary the number of available usage rate levels and the length of the operating mission. Results, compared to those obtained by adopting a simpler policy that allows for corrective and preventive replacements only, show that within the considered experimental scenarios, allowing for usage rate reductions significantly reduces the expected cost. Moreover, the results suggest that the optimal action at each inspection strongly depends on the degradation phase experienced by the system at that time.

Keywords: prescriptive maintenance, non-homogeneous Gamma process, bathtub-shaped degradation rate, finite-horizon maintenance policy, average total cost.

Esposito, N., A. Piscopo, B. Castanier, and M. Giorgio (2024). Adaptive Maintenance Policy For Gamma Degrading Units With Bathtub-Shaped Degradation Rate Function In The Presence Of Random Effect. In Advances in Reliability Safety and Security, Cracow (PL), 29–37.
Gertsbakh, I. B., and Kh. B. Kordonskiy (1969). Models of Failure. Springer.
Giorgio, M., A. Piscopo, and G. Pulcini (2023). A new Wiener process with bathtub‐shaped degradation rate in the presence of random effects. Applied Stochastic Models in Business and Industry 40, 574–597.
Meissner, R., A. Rahn, and K. Wicke (2021). Developing prescriptive maintenance strategies in the aviation industry based on a discrete-event simulation framework for post-prognostics decision making. Reliability Engineering & System Safety 214, 107812.
Sun, Q., P. Chen, X. Wang, and Z. Ye (2023). Robust condition‐based production and maintenance planning for degradation management. Production and Operations Management 32, 3951–3967.

Speaker: Antonio Piscopo (Scuola Superiore Meridionale; Université d'Angers)

15:55 Connections between information measures and activation functions

In this work, we introduce a generalized measure of uncertainty, namely the cumulative information $\psi$-measure, in order to provide a unified perspective to the uncertainty framework. Indeed, it is a variability measure which reduces to several well-known information measures for appropriate choices of the function $\psi$. In particular, cumulative versions of Shannon and Tsallis entropies, Gini’s dispersion indices, cumulative information generating functions, and additional new classes of uncertainty measures arise as special cases. We also investigate a suitable relative version, referred to as the relative cumulative information $\psi$-measure. We establish fundamental properties of the proposed measures, including monotonicity properties, bounds, and a set of related inequalities. We derive covariance-based and quantile-based representations, too. Illustrative examples are presented to clarify the behaviour of the measures under different model assumptions and choices of $\psi$. A notable feature of the proposed approach is the natural connection between the function $\psi$ and certain increasing activation functions, commonly employed in neural networks. For instance, the function $\psi$ associated with the cumulative paired Shannon entropy can be derived from the sigmoid activation function, and conversely. Motivated by this link, we introduce the generalized logistic-linear family of activation functions, which include classical models such as sigmoid and linear activations as particular cases within a parametric structure. Finally, we develop an application motivated by a neural network scenario to demonstrate the flexibility and novelty of the proposed framework. The results highlight the potential of the cumulative information $\psi$-measure as a bridge between information theory and modern neural networks.

References:
M. Capaldo, A. Di Crescenzo, and A. Meoli,
``Cumulative information generating function and generalized Gini functions,'' Metrika, vol. 87, pp. 775--803, 2024.
DOI: 10.1007/s00184-023-00931-3.

M. Capaldo, A. Di Crescenzo, and G. Pisano,
``Information measures and activation functions,'' submitted, 2026.

A. Di Crescenzo and M. Longobardi,
``On cumulative entropies,'' Journal of Statistical Planning and Inference, vol. 139, no. 12, pp. 4072--4087, 2009.
DOI: 10.1016/j.jspi.2009.05.038.

M. Rao, Y. Chen, B. C. Vemuri, and F. Wang,
``Cumulative residual entropy: A new measure of information,''
IEEE Transactions on Information Theory, vol. 50, no. 6, pp. 1220--1228, 2004.
DOI: 10.1109/TIT.2004.828057.

Speaker: Giulia Pisano (Università degli Studi di Salerno)

16:15 Bayesian Hierarchical Modeling for Reliable Large-Scale Sensors Deployments and Applications

Reliability of large populations of sensors is a major challenge in modern industrial production and applied functional monitoring systems. The massive deployment of low-cost Micro-Electro-Mechanical Systems (MEMS) sensors across several technological domains requires calibration strategies that ensure metrological reliability while remaining feasible at industrial scale. However, traditional laboratory calibration procedures become impractical when sensors are produced in very large quantities, motivating the development of statistical approaches for large-scale calibration.

This work develops a Bayesian statistical framework for the virtual calibration and actual applications of large sensor batches. A earlier published model exploits information from a reference, laboratory-calibrated batch to infer the calibration properties of new lots of the same production process. By calibrating only a small subset of sensors from each batch, the model estimates
key parameters that characterize the entire lot, providing robust reliability assessments while drastically reducing direct calibration effort. Building upon this formulation, a hierarchical extension of the model is developed by introducing a Beta hyperprior distribution
on the probability of detecting out-of-tolerance sensors, enabling the integration of prior industrial knowledge with explicit control of parameter variability. This approach softens prior deterministic assumptions about batch quality and enhances the flexibility and robustness of reliability estimates. The choice of hyperprior parameters therefore plays a
crucial role in ensuring that the model coherently reflects prior industrial knowledge.
The study evaluates sensor batch reliability metrics under varying model parameters and introduces alternative metrics to address identified limitations. Model validation was conducted on a case study
of 100 digital MEMS accelerometers calibrated at INRiM. Models based on weakly informative hyperpriors rely more strongly on the information conveyed by the likelihood. Consequently, when the observed data indicate a smaller percentage of defective sensors than that expected according to the prior, higher batch reliability and lower uncertainty associated with the statistically calibrated sensors are obtained. Conversely, when the data suggest a higher percentage of defective sensors, the same models produce a substantial increase in the associated uncertainty. This result highlights how the agreement between the prior assumption and the data-driven evidence plays a fundamental role in the assessment of the batch reliability and uncertainty.

Future research will extend the model by including utility and cost functions to formally represent the producer’s risk in rejecting a batch and the consumer’s risk in accepting it, integrating
probabilistic calibration results with decision-making criteria. These studies will be crucial for balancing the benefits of using less informative hyperpriors, which may lead to higher estimated reliability of lots, against the drawback of a larger posterior variance, that is, a reduced confidence in the estimation of the true number of defective sensors.

Speaker: Anna Ballario (Politecnico di Torino)

16:35 Implementation of Automatic differentation on Maintenance optimization problems

Abstract

Automatic Differentiation (autodiff or AD) is a technique for computing derivatives that differs from both symbolic and numerical differentiation. It relies on the implementation of the chain rule to functions composed of simpler smooth functions whose Jacobian matrices are known.

Given the implementation of a function composed of smooth operations, Automatic Differentiation enables the computation of its partial derivatives by propagating derivatives through the computational graph of the program. Using this approach, we can differentiate a function that is part of a code and does not have an explicit mathematical expression.

In this work, we demonstrate the application of this technique to basic maintenance optimization problems, where the objective is to identify maximum or minimum points of a cost function approximated using Monte Carlo methods. This approach enables efficient gradient-based optimization even when the cost function is estimated through stochastic simulation. The implementation is carried out using modern automatic differentiation libraries such as JAX and Autograd.

References

1. Atilim Gunes Baydin, Barak A. Pearlmutter, Alexey Andreyevich Radul, and Jeffrey Mark Siskind.
   Automatic differentiation in machine learning: a survey, 2018.
2. Waltraud Kahle, Sophie Mercier, and Christian Paroissin. Degradation Processes in Reliability. Mathe-
   matics and Statistics Series, Mathematical models and methods in reliability set, Vol. 3. June 2016.
3. Charles C. Margossian. A review of automatic differentiation and its efficient implementation. WIREs
   Data Mining and Knowledge Discovery, 9(4), March 2019.
4. Toshio Nakagawa. Maintenance Theory of Reliability. Springer Series in Reliability Engineering. Springer
   Science & Business Media, London, 2005.

Speaker: Alberto Arias (University of Extremadura)

16:55 A Piecewise Hidden Markov Model for Degradation and Maintenance

Hidden Markov Models (HMMs) are increasingly recognized in reliability engineering as valuable tools for monitoring systems where the true operational state is not directly observable and must be inferred from certain indicators provided by a control system. Accurate estimation of these hidden health states and prediction of failures are crucial for minimizing unexpected downtime and optimizing operational costs. While recent literature has demonstrated the utility of HMMs in estimating dependability measures, such as identifying failed components and assessing system reliability, the explicit integration of maintenance actions into these models remains an evolving area of study.

Traditional HMM applications often assume time-homogeneous transition dynamics, which may not capture the discontinuous physical effects of a maintenance intervention. Specifically, a maintenance action introduces an instantaneous restorative shock and often alters the subsequent aging rate of the system, which does not agree with the homogeneous degradation assumption.

To address this gap within a structured probabilistic framework, this paper presents a piecewise, time-inhomogeneous HMM approach designed to incorporate trend changes in system degradation. We model the degradation level over a finite state space, partitioned into functional and failure states. The system's operational evolution is decomposed into distinct chronological phases.

For robust parameter estimation, we propose an adapted Expectation-Maximization (EM) algorithm. A new algorithm is discussed to decode the hidden state trajectory across these varying temporal regimes. The proposed piecewise framework allows for the modeling of systems where the transition matrix structurally changes due to external occurrences. This framework provides a practical method to isolate natural degradation from repair actions, for instance, facilitating a rigorous evaluation of maintenance effectiveness in HMM-based reliability studies to determine the optimal timing for future scheduled interventions. Furthermore, this approach can be considered for other situations where the transition matrix changes due to external events, such as the arrival of a disruptive shock.

Speaker: Demian Benjamin Stuursma (Universidad de Granada)

17:15 → 17:45 Coffee Break

17:45 → 18:15 JMP-Session

17:45 Predicting Product Reliability: A Case Study

This live demonstration uses real-world hard drive data to explore key concepts in product reliability analysis. After a brief introduction to reliability data, data preparation, and accelerated life testing, we will fit and compare models to predict failure rates across different products.

Using JMP’s Life Distribution platform and built-in Reliability Profilers, we will examine how model choice and sample size influence reliability predictions. We will then combine individual product components in a Reliability Block Diagram to evaluate the reliability of an overall server system.

The demonstration is designed for an audience with little or no prior JMP experience. All capabilities shown are available in JMP Student Edition, the free academic license, making JMP a powerful and accessible environment for teaching and applying reliability concepts without the need for coding.

Speaker: Dr Volker Kraft (JMP Statistical Discovery)

18:15 → 18:25 Closing

20:30 → 21:30 Gala Dinner