Recent advancements in performance-based design and seismic risk assessment of steel structures
*NEW. A recording of the talk is available on our YouTube Channel. Here is the link.
Introduction (Professor D.G. Lignos)
This seminar presents some of the recent work from the Resilient Steel Structures Laboratory (RESSLab) at EPFL, Switzerland. The central research mission of the laboratory is to develop new approaches that advance our physical understanding on key deteriorating mechanisms causing earthquake-induced structural collapse. This knowledge is then leveraged to propose new concepts that minimize damage in steel and composite-steel structures during earthquake shaking. The expertise of the lab in the domains of performance-based design of steel structures, seismic risk assessment along with applications of data-driven methods and surrogate modelling is exemplified by the two talks in this seminar. Specifically, the first presentation focuses on methodological developments that enable “instability-free” seismic performance of steel moment resisting frames afforded by innovations in the seismic design of welded moment connections with inelastic column web panel zones, a concept that defies the current design paradigm. The second presentation discusses the methodological advancements pertaining to assessment of seismic collapse capacity and risk using data-driven surrogates as a complement to physics-based response history analyses, with a particular emphasis of keeping the data requirements low so as to reduce the computational burden of uncertainty quantification involved in seismic risk assessment of steel structures.
Presentation 1: Welded moment connections with inelastic column web panel zones for enhanced seismic performance of steel moment frames (Dr A. Skiadopoulos)
In capacity-designed steel moment resisting frames (MRFs), the participation of the beam-to-column web panel zones in the energy dissipation is generally limited during an earthquake event. In such a design context, beam local buckling is likely even at modest lateral drift demands, thereby engendering structural repair costs in the aftermath of earthquakes. This presentation discusses recent advancements in the state-of-knowledge regarding the seismic design and behaviour of steel MRFs with highly dissipative panel zones. A new panel zone design model that addresses the limitations of all available models in the literature will be presented. The panel zone model can effectively enable a balanced seismic design of welded moment connections with inelastic panel zones. One step further, the current detailing of welded connections is revisited. Simplifications in their fabrication process are proposed by intentionally keeping a customized bevelled backing bar in place, without impairing the connection’s ductility under cyclic loading. The proposed connection weld detail is substantiated by continuum finite element analyses and full-scale experiments. Quantitative seismic response characteristics of steel MRFs with highly dissipative panel zones through large-scale system-level parametric studies are finally presented and implications on structural collapse and repairs are discussed.
Presentation 2: Data-Centric Machine Learning in Earthquake Engineering and Surrogate Modelling of Steel Moment Resisting Frames (Dr N. Bijelić)
While the performance-based earthquake engineering (PBEE) approach is increasingly being utilized by practicing engineers, one of the major impediments to its widespread adoption is the heavy burden of the underlying nonlinear simulations. In that sense, data-driven machine learning (ML) approaches have been gaining research interest in the earthquake engineering domain including efforts to sidestep the nonlinear response history analyses while maintaining their predictive power. However, most of the past research efforts focused on algorithmic developments or contrasting of the performance of different ML tools – i.e., the so-called “model-centric” approach – while putting less emphasis on the most effective use of data. In contrast, this talk discusses a “data-centric” approach to seismic collapse risk assessment. The specific objective herein is to demonstrate the opportunity for reducing the required number of physics-based simulations to trace structural collapse by leveraging domain-specific data augmentation. To this end, two recent methodological advancements are presented: 1) a surrogate-agnostic data engineering methodology for seismic collapse risk assessment, termed the automated collapse data constructor (ACDC) technique, and 2) a novel approach for seismic collapse fragility and risk estimation termed the data-driven collapse classification (D2C2) method. The utility of the proposed methodologies is demonstrated through examples from extensive case studies of seismic collapse risk estimation of steel MRF structures ranging from four to twenty stories in height. In particular, the results suggest that the ACDC and D2C2 methodologies allow dramatic improvements in the predictive capability of data-driven surrogates while at the same time significantly reducing data requirements. Moreover, a major implication and finding demonstrated in this talk is that “big data” is not paramount to effectively leverage ML tools for assessment of earthquake-induced collapse risk.
About the speakers
Professor Dimitros G. Lignos
Dimitrios Lignos is a Full Professor and Chair of the Civil Engineering Institute at the École Polytechnique Fédérale de Lausanne (EPFL). He joined EPFL in 2016 from McGill University, Canada, where he was a tenured Associated Professor. Prior to that he was a post-doctoral researcher at Kyoto University, Japan (2010) and Stanford University, USA (2009). He holds degrees in Structural Engineering (Stanford University, USA, M.S. 2004, Ph.D. 2008) and Civil Engineering (NTU, Athens, 5-year Diploma 2003). Prof. Lignos’s research involves integrated computational modeling and large-scale experimentation for the fundamental understanding and simulating of earthquake-induced collapse of steel and composite-steel structures. His awards include the 2022 Raymond Reese Research Prize and the 2019 Walter L. Huber Prize from American Society of Civil Engineers (ASCE) among others. As a member of the Project Team 2, he was responsible for the revision of the steel and composite steel concrete structures in Eurocode 8 Parts 1-1 and 1-2 and the new Chapter 9 of Eurocode 8 Part 3 for the development of seismic assessment models for existing steel and composite steel structures. Prof. Lignos is a member of the Canadian Standards Association (CSA) S16 technical committee for Steel Structures. He is regularly involved as a NEHRP consultant in research-to-practice projects related to the nonlinear modelling and analysis of structures applicable to the engineering practice through the Applied Technology Council (ATC) and the National Institute of Standards (NIST) and Technology in the US.
Dr Andronikos Skiadopoulos
Andronikos Skiadopoulos is currently a postdoctoral researcher in the Resilient Steel Structures Laboratory (RESSLab), at the Swiss Federal Institute of Technology Lausanne (EPFL). He holds a degree in Structural Engineering (EPFL, Ph.D. 2022) and Civil and Environmental Engineering (National Technical University of Athens (NTUA), 5-year Diploma 2016). His research focuses on modeling and experimentation of steel structures subjected to extreme hazards. He is the recipient of the 2022 Raymond C. Reese Research Prize from the American Society of Civil Engineers (ASCE) and the 2022 Outstanding Ph.D. Thesis Distinction in Civil and Environmental Engineering from EPFL.
Dr Nenad Bijelić
Nenad Bijelić is a postdoctoral scholar in the Resilient Steel Structures Laboratory (RESSLab) at EPFL, Switzerland. He obtained his Ph.D. (2018) and M.S. (2014) from Stanford University, USA, and B.S. (2010) from University of Zagreb, Croatia. His research is in the area of probabilistic risk modeling and natural hazards engineering focusing on dynamics of nonlinear systems and application of statistical and machine learning tools. Pertinent to this seminar, Dr. Bijelić co-convened a session on theory-guided machine learning approaches in earthquake engineering at the 2021 Earthquake Engineering Research Institute (EERI) Annual Meeting, USA, and is organizing a technical session “Sense and sensibility of machine learning in the natural hazards engineering” at the upcoming World Conference on Earthquake Engineering (WCEE2024, Italy). In addition, Dr. Bijelić co-lead the EERI-StEER joint reconnaissance team following the December 29, 2020, magnitude 6.4 earthquake in Petrinja, Croatia. His PhD research was supported by the 2012 International Fulbright Science and Technology Award, and the 2015 Shah Family Fellowship on Catastrophic Risk.
The event is open to all and is free to attend. Please join the meeting via Teams* by clicking on this link.
The talk is organised by SECED Young Members. For further information, please contact Marco Baiguera (
* Microsoft Teams is freely available on all major platforms, including Windows, Linux, MacOS, Android and iOS. To download and install the software, follow this link. SECED is not sponsored by or otherwise affiliated with Microsoft, and we do not endorse any Microsoft products.
|Event Date||14/06/2023 12:30 pm|
|Event End Date||14/06/2023 2:00 pm|