* Videos of the ICASP13 keynote lectures are now available online (Click here for URLs and details) *
Engineering Risk Analysis Group
Technical University of Munich, Germany
Daniel Straub is currently Associate Professor for engineering risk and reliability analysis at the Technical University of Munich (TUM), Germany. His interest is in developing physics-based stochastic models and methods for decision support in infrastructure, environmental and general engineering systems, with a particular focus on Bayesian techniques. At TUM, he has developed a successful MSc study line in engineering risk and reliability, with around 70-80 students annually. Daniel is active in multiple industries, including structural design and assessment, offshore and marine engineering, geotechnical engineering, natural hazards, automotive as well as aero- and astronautical engineering. He is past president of IFIP WG 7.5 and Geosnet, and is a member of the advisory board of the TUM Institute of Advanced Studies and multiple editorial boards. His awards include the ETH Silbermedaille and the Early Achievement Research Award of IASSAR. He is also an Honorary Professor at the University of Aberdeen, UK.
Reliability. Uncertainty. Sensitivity. Decision!
Reliability analysis and other uncertainty quantification (UQ) techniques, such as sensitivity analysis and advanced data analysis, have become mature and are increasingly utilized in engineering practice. Early pioneers emphasized the role of decision analysis as a main driver for employing these techniques, but the community has been reluctant to employ formal decision analysis in engineering practice and even research. As a consequence, many reliability and UQ methods have been proposed that make implicit assumptions on preferences of decision makers and often lead to sub-optimal and possibly inconsistent decision support. Additionally, increasingly automated and autonomous technical processes and systems demand for a more explicit consideration of decisions in the context of reliability and uncertainty. This talk will show how formal decision analysis can provide fundamental insights and practical answers to many challenges in engineering research and practice. I will start out by briefly reviewing the status quo and discuss reasons for the negligence of decision analysis. The main part of the talk will demonstrate the application of decision analysis and its potential through a number of example cases that are based on recent research and consulting projects. In particular, I will present decision-theoretic approaches to (a) sensitivity analysis of engineering models, (b) reliability demonstration of autonomous vehicles, (c) predictive maintenance of structural systems, and (d) optimal infrastructure design.
National University of Singapore, Singapore
Kok-Kwang Phoon is Distinguished Professor and Vice Provost (Academic Personnel), National University of Singapore. His main research interests include statistical characterization of geotechnical parameters and reliability-based design in geotechnical engineering. He is the lead editor of 3 books: Reliability of Geotechnical Structures in ISO2394 (CRC Press/Balkema, 2016), Risk and Reliability in Geotechnical Engineering (CRC Press, 2015), and Reliability-based Design in Geotechnical Engineering (Spon Press, 2008). He was bestowed with numerous research awards, including the ASCE Norman Medal in 2005, the John Booker Medal in 2014, and the Humboldt Research Award in 2017. He is the Founding Editor of Georisk, Board Member of the International Society for Soil Mechanics and Geotechnical Engineering (ISSMGE), and Vice-President of the International Association for Structural Safety and Reliability (IASSAR).
The “Site Challenge” in Geotechnical Engineering
Site investigation is a key cornerstone of geotechnical practice. Several past studies have demonstrated that there are sufficient data in global databases to construct generic probability density functions (PDFs) for soil parameters. These global databases contain information from sites worldwide. At each site, geotechnical data are typically multivariate, unique (to some degree), uncertain, sparse, and incomplete. These characteristics - Multivariate, Uncertain and unique, Sparse, and InComplete - can be abbreviated by MUSIC. The purpose of this talk is to discuss how site-specific PDFs can be constructed for MUSIC data (or MUSIC-X data containing spatial correlations). Although “site-specificity” is widely accepted by practitioners as a fundamental feature that distinguishes geotechnical from structural engineering, its rigorous characterization remains an open question at this point in time. Overcoming this "site challenge" can potentially transform geotechnical engineering practice and reduce its reliance on engineering judgment to handle site-specific features that cannot be readily quantified using data from the site of interest and possibly “comparable” sites elsewhere.
Rice University, USA
Leonardo Dueñas-Osorio is an Associate Professor of Civil and Environmental Engineering at Rice University in Houston, Texas. He obtained his Master’s degree from the Massachusetts Institute of Technology in 2001, and Ph.D. from the Georgia Institute of Technology in 2005, both in Civil and Environmental Engineering. He develops analytical and computational methods, including quantum algorithms, for the performance assessment of structural and Infrastructure systems and their upkeep decision support. Leonardo is actively involved with the Structural Engineering Institute and the Infrastructure Resilience Division of the American Society of Civil Engineers (ASCE) through their risk assessment, lifeline interdependencies, and decision analysis committees. He is also an Associate Editor for Natural Hazards Review and a member of the editorial board of Structural Safety. His research has been recognized with the 2008 U.S. National Science Foundation (NSF) “CAREER award”, the 2015 Earthquake Engineering Research Institute (EERI) “Outstanding Earthquake Spectra Paper Award”, and the 2017 International Association for Structural Safety and Reliability (IASSAR) “Early Achievement Research Award”. Leonardo’s research is mainly funded by NSF, the U.S. Department of Defense, and the Office of Public Safety and Homeland Security of the City of Houston.
Computer-aided Reliability and Resilience Verification for Infrastructure Networks
As automation, interdependence and operational uncertainty across critical infrastructure networks continue to increase, engineers need system-level state estimation methods that are correct and computationally feasible to guide safety decisions. This talk introduces new algorithms for the efficient reliability and resilience assessment of networked systems, which offer exact bounds or a priori guarantees on the quality of their estimates. In particular, I present how to leverage powerful solvers for generic constrained satisfaction problems in a formulation that is scalable, naturally applicable to network reliability assessment, and that issues probably approximately correct (PAC) certificates. This logic-based formulation appeals to the intuitive counting of system states, revealing a unique connection to quantum tensor network estimators, which I showcase along with implementable adiabatic algorithms on quantum simulators and computers. And while reliability is a fundamental property of networked engineered systems, so too is their resilience. Hence, I present emerging strategies that exploit the decentralized structure of the constraints governing interdependent infrastructure networks, particularly when they need to be optimally restored in minimum time at a minimum cost after contingencies materialize. I end the talk with a prospective account of methods for reliability verification and resilience reasoning consistent with uncertainty quantification requirements for evolving infrastructure systems.