Tutorial 1
Predicting the Future with Confidence: A Tutorial on Uncertainty-Aware Prognostic
Prof. Dr. Marcos E. Orchard (Universidad de Chile, PHMS Fellow), Prof. Dr. Nick Eleftheroglou (iSP, TU Delft)
Outline:
Which fundamental property most clearly defines the essence of prognostics?
At its core, prognostics is the science of predicting the future. Yet the future, by definition, carries inherent uncertainty. Consequently, any prognostic approach must not only quantify and represent this uncertainty but also propagate it through time with efficiency and rigor.
Despite this fundamental requirement, many state‑of‑the‑art prognostic methods continue to neglect uncertainty. Their performance is often judged solely on accuracy, overlooking the critical importance of well‑calibrated uncertainty estimates—an element indispensable for meaningful and trustworthy decision‑making in real‑world operations.
In this tutorial, we lay out the foundations of uncertainty quantification in prognostics and present a practical framework for managing uncertainty. This framework enables predictions that are not only technically sound but also directly informative for operational planning and maintenance decision‑making.
We will showcase the value of uncertainty quantification and management through a live demonstration centered on electric vehicles, illustrating how uncertainty‑aware prognostics can optimize operational strategies. The demo will highlight how properly quantified and managed uncertainty provides distinct operational advantages and how the core principles of prognostics translate directly into enhanced real‑world performance.
The tutorial concludes with a discussion of open challenges and emerging research directions in the vital field of uncertainty‑aware prognostics, highlighting opportunities to advance decision‑making methodologies and real‑world applications.
Profiles:
Dr. Marcos Orchard (Google Scholar: h-index 42 and +7,200 citations) is Professor with the Department of Electrical Engineering of University of Chile, Director of the Center for Sustainable Acceleration of Electromobility (CASE), and Principal Investigator at the Advanced Center for Electrical and Electronic Engineering (Universidad Técnica Federico Santa María). He has authored and co-authored more than 200 papers on diverse topics, including the design and implementation of failure prognostic algorithms, statistical process monitoring, and system identification. His research work at the Georgia Institute of Technology was the foundation of novel real-time failure prognosis approaches based on particle filtering algorithms. His current research interests include the study of theoretical aspects related to the implementation of real-time failure prognosis algorithms and prognostic-decision-making, with applications to battery management systems, electromobility, mining industry, and finance. Prof. Orchard is Fellow of the Prognostic and Health Management Society, Editor-in-Chief of the International Journal of Prognostic and Health Management, and Past-President of the Chilean Association of Automatic Control (NMO of IFAC).
Dr. Nick Eleftheroglou is an Assistant Professor in the Faculty of Aerospace Engineering at Delft University of Technology and head of the Intelligent System Prognostics for Operations and Maintenance Group (https://www.groupisp.com/home), which currently consists of five PhD candidates, two postdoctoral researchers, as well as master’s students and visiting researchers. He also serves as an Associate Editor for the International Journal of Prognostics and Health Management. He received his Diploma in Mechanical and Aeronautics Engineering cum laude from the University of Patras, Greece, in 2015, and earned his PhD cum laude from TU Delft in October 2020. His research focuses on Prognostics and Health Management (PHM), with emphasis on the fundamentals of prognostics, such as uncertainty management, and on developing PHM methodologies that provide enhanced reliability, robustness, and feasibility for operations and maintenance.
Tutorial 2
Why Uncertainty Matters: Trustworthy Machine Learning for Prognostics and Health Monitoring
Will Jacobs (Sheffield University)
Outline:
Machine-learning models are increasingly used in equipment health monitoring (EHM) to support maintenance and operational decisions. However, these models are typically trained on limited, noisy, and operationally biased data, making uncertainty quantification (UQ) essential for trustworthy deployment.
This tutorial introduces practical approaches to uncertainty quantification for machine learning applied to prognostics and health monitoring. Key sources of uncertainty in data-driven EHM – including measurement noise, operational variability, data sparsity, model uncertainty, and distribution shift- are discussed, alongside commonly used uncertainty quantification techniques such as probabilistic models, ensembles, and Monte Carlo methods. The focus is on how uncertainty information can be used to support risk-aware health assessment, degradation monitoring, and maintenance planning. Examples are drawn from aerospace engine health monitoring applications.
Profile:
Will Jacobs is a Research Fellow at the Rolls-Royce University Technology Centre at the University of Sheffield. He received his PhD in Bayesian system identification from the University of Sheffield and has over a decade of experience in prognostics and health monitoring applied to aerospace systems. His work focuses on data-driven and physics-aware approaches to equipment health monitoring, with an emphasis on uncertainty-aware decision support. He has led and contributed to numerous industry- and research-council-funded projects in collaboration with aerospace partners.
Tutorial 3
Data-Driven Optimization for Maintenance and Spare-Parts Logistics: From Data to Actionable Decisions
Heraldo Rozas (University Chile)
Outline:
Recent advances in prognostics have improved our ability to estimate the remaining useful life of industrial components from sensor data, yet these insights are still rarely translated into actionable maintenance and logistics decisions. This tutorial addresses that gap by demonstrating how prognostic outputs can be seamlessly integrated into optimization models for predictive maintenance and spare-parts planning.
We begin with a brief introduction to prognostic models, outlining how remaining life estimates are generated and how their uncertainty can influence downstream decisions. Building on this foundation, we explore a progression of optimization frameworks. We first discuss a classic single-component renewal model that uses prognostic information to determine cost-effective maintenance intervals. We then scale this approach to fleets of components through a deterministic mixed-integer linear programming formulation that coordinates maintenance activities across assets under operational constraints. Next, we extend this deterministic model to a chance-constrained stochastic program that incorporates uncertainty in both the objective function and the constraints, thereby minimizing expected costs while limiting the risk of undesired events. Finally, we discuss how to make reliable maintenance decisions when prognostic estimates are inaccurate by using distributionally robust optimization to account for potential shifts in the estimated remaining lifetime distributions.
The tutorial includes Python examples for all models discussed, giving participants practical notebooks that convert prognostic outputs into actionable maintenance decisions.
Profile:
Heraldo Rozas is an Assistant Professor in the Department of Electrical Engineering at the University of Chile, an Associate Researcher at the Advanced Center for Electrical and Electronic Engineering, and an Invited Researcher at the Complex Engineering Systems Institute. He received the B.S. and M.S. degrees in Electrical Engineering from the University of Chile and his Ph.D. in Industrial Engineering from the Georgia Institute of Technology, where he was a Graduate Research Assistant in the Predictive Analytics & Intelligent Systems (PAIS) Research Group. He also worked as an Assistant Researcher at NASA’s HOME Space Technology Research Institute.
His research focuses on transforming data into actionable decisions for industrial applications, including predictive maintenance, spare parts logistics, electromobility, lithium-ion batteries, and power systems.
LinkedIn: https://www.linkedin.com/in/heraldo-rozas/
Website: https://sites.google.com/view/heraldo-rozas/