Prof. Frank Coolen
Durham University
Title: Decision Support for System Reliability using the Survival Signature
Abstract:
The survival signature of a system consisting of multiple types of components, is simply the probability that the system functions given the exact number of components of each type that function. For large systems or networks with relatively few component types, the survival signature is a powerful tool for quantification of the system reliability. This talk will start with an introductory overview to the survival signature, including a discussion of related research challenges.
In practice, quantification of system reliability is typically performed in order to support decisions. We will discuss two scenarios involving decisions with regard to spare components. The survival signature concept is particularly suited for such decision support as it typically involves consideration of the numbers of spares per component type.
BIO:
Prof Frank Coolen is Professor of Statistics at the Department of Mathematical Sciences, Durham University (UK). He studied at Eindhoven University of Technology, achieving MSc (`Mathematical Engineer') and PhD (Statistics) in 1989 and 1994, respectively. He then joined Durham University as Lecturer, and was promoted to Professor in 2005.
Frank has contributed to a wide variety of research topics, with main focus on the development of Nonparametric Predictive Inference (NPI) and Reliability Theory. He proposed NPI in the 90s, in an attempt to base statistical inference on minimal structural assumptions, enabled by using imprecise probability for uncertainty quantification. NPI has been developed for many applications in Statistics, Operations Research, Risk and Reliability, and related fields. In Reliability Theory, a main contribution was the concept of Survival Signature, which enables reliability quantification of (very) large systems and networks with relatively few different component types. Most of his research is joint work with his wife and colleague Prof Tahani Coolen-Maturi.
Frank has (jointly) supervised 37 PhD students to completion of their studies, and is currently supervising 15 PhD students. He is on the editorial or advisory boards of 11 journals, including Journal of Statistical Theory and Practice, Journal of Risk and Reliability, and the Communications in Statistics journals.
Dr. Chris Lindley
University of Sheffield
Title: On the use of statistical learning theory for structural health monitoring
Abstract:
The use of statistical models in engineering can be limited by the availability of data. This issue is particularly prevalent in Structural Health Monitoring (SHM), where labelled data can be expensive to obtain since it requires data from most, if not all, operational conditions that a structure might experience. As a result, there has been a surge of interest in transfer learning within the research community, aiming to use existing data to improve the learning on new structures that have yet to experience previously unseen conditions. The issue of generalisation can be addressed by employing the bounds found in Statistical Learning Theory (SLT); in particular, by estimating an upper bound on the expected risk of a transfer learner.
This short talk will present an overview of the problem of statistical generalisation and model selection in SHM. How STL can be used in practice will be demonstrated in simple simulated case studies, whereby the estimated risk bounds provide guarantees for generalisation in scenarios where small amounts of data are available.
BIO:
Chris was awarded his Ph.D. in Mechanical Engineering from the University of Sheffield in 2024. He conducted his research on nonparametric Bayesian techniques for Structural Health Monitoring (SHM) as part of the Dynamics Research Group, under the supervision of Prof. Keith Worden and Prof. Nikolaos Dervilis. He is currently a research associate primarily investigating the uses of statistical learning theory in transfer learning and the potential benefits it could bring in the development of Digital Twins.
The workshop will take place in the institutes library, room 116, on Wednesday, May 21, 2025. Start is at 10:00 a.m. If you want to participate online please contact Torsten Ilsemann at least one day before the presentations.
