Accelerating Risk Analysis in Natural Hazards with SciML (Session 41) and PBWE from Buildings to Communities (Session 44)
Dear Colleague,
I am excited to announce that I will be organizing two sessions at ICOSSAR’25, scheduled to take place at USC in Los Angeles from June 1-6, 2025. The sessions I will be organizing are as follows:
Session 41: Accelerating Computational Models by Scientific Machine Learning for Uncertainty Quantification and Risk Assessment in Natural Hazards Applications. Co-organizers: Somdatta Goswami; Dimitrios Giovanis; Bowei Li; Michael Shields; Alexandros Taflanidis
Session 44: Performance-Based Wind Engineering from Building to Community Scale. Co-organizers: Teng Wu; Ahsan Kareem; Tracy Kijewski-Correa; Adam Zsarnoczay
Detailed descriptions of the sessions can be found below.
We warmly invite you and your team to submit your work to these sessions.
The deadline for abstract submission is December 31, 2024.
You can access the abstract submission system here.
We very much hope you will accept our invitation.
Best regards, and I hope to see you at USC next year!
On behalf of the session organizers
Seymour Spence, University of Michigan
Session 41: Accelerating Computational Models by Scientific Machine Learning for Uncertainty Quantification and Risk Assessment in Natural Hazards Applications
Abstract: The convergence of Scientific Machine Learning and Computational Modeling offers transformative potential for simulating and analyzing engineering systems exposed to natural hazards. Traditional numerical modeling, although robust, faces challenges such as high computational costs, especially with models characterized by nonlinear behaviors, high dimensionality, and heterogeneous parameters. Deep learning methodologies present innovative opportunities to enhance and expedite numerical model evaluation, allowing for more efficient propagation of uncertainty, risk assessment, and performance-based design in natural hazard applications.
This mini-symposium focuses on developing and integrating advanced scientific machine learning aimed at accelerating traditional numerical solvers for uncertainty quantification, risk assessment, and performance-based design of complex systems exposed to earthquakes, floods, and extreme winds. Emphasis is placed on improving computational efficiency, predictive accuracy, and scalability to facilitate system robustness and resilience.
We welcome contributions on topics including:
– Deep learning-powered surrogate and reduced-order modeling for uncertainty propagation.
Physics-informed Neural Operators and Neural Networks for accelerating the evaluation of complex computational models.
– Multi-fidelity simulation, surrogate modeling, and transfer learning for uncertainty quantification, risk assessment, and performance-based design of structures/systems.
– Bayesian Neural Networks and Bayesian Inference in natural hazards engineering.
– High-performance computing with deep learning for uncertainty quantification, risk assessment, and large-scale optimization of systems subject to natural hazards.
– Practical applications demonstrating improved computational efficiency, particularly in seismic, flood, and wind engineering applications including computational fluid dynamics.
This symposium aims to foster collaboration among researchers and practitioners, providing insights into theoretical advancements and practical applications of Scientific Machine Learning to enhance the resiliency and sustainability of engineered systems in the face of natural hazards.
Session 44: Performance-Based Wind Engineering from Building to Community Scale
Abstract: Performance-based Wind Engineering (PBWE), and, more generally, risk/resiliency-based wind assessment, is becoming an area of ever-growing importance within the realm of wind engineering. The fundamental shift from prescriptive approaches to probabilistic system-level performance/risk/resiliency-based approaches has the potential to significantly reduce the substantial annual losses that communities across the globe suffer due to events such as hurricanes, tornadoes, and thunderstorm downbursts. While the benefits of PBWE are well understood, there are still many theoretical, computational, and experimental issues that need to be addressed before this approach can become a reality and significantly change wind engineering practices. Moreover, the growing interest in high-fidelity simulation of the impacts of extreme winds on communities opens new doors to the application of PBWE workflows at a community scale. This session will bring together researchers and practitioners who work in this domain. Areas of interest include, but are not limited to:
– Computational Fluid Dynamics (CFD) and its role in estimating all aspects of building performance (e.g., wind pressures, wind-driven rain, community-scale wind simulation).
– Computational modeling and experimental testing of the structural and envelope systems of buildings in extreme winds.
– High-fidelity vulnerability and catastrophe modeling.
– Experimental and numerical modeling of non-stationary wind events and their effects on natural and built infrastructure.
– Probabilistic and computational wind-borne debris modeling.
– Machine learning for enabling PBWE workflows, uncertainty quantification, and community-level simulations.
– Digital twins to enable real-time forecasting of the impacts of extreme winds on communities.
Contributions addressing theoretical and computational developments, numerical algorithms, experimental testing, and practical applications are welcome. The session will provide an opportunity for researchers, academics, and practicing engineers active in these topical areas to share their experiences and latest research results.