Dr. Chao Dang
TU Dortmund University, Germany
Title: Forward Uncertainty Quantification by Bayesian Active Learning
Abstract:
Uncertainty quantification (UQ), in its broadest sense, addresses the identification, characterization, and management of uncertainties in both computational models and real-world systems. Forward UQ investigates how input uncertainties propagate to model outputs and plays a central role in stochastic response analysis, reliability analysis, and sensitivity analysis. In practice, however, these analyses are often computationally prohibitive due to the high cost of evaluating complex numerical models.
In this talk, forward UQ is presented within a unified Bayesian inference and active learning framework – referred to as Bayesian active learning. The focus lies on both time-independent and time-dependent reliability analysis, as well as response probability distribution estimation. By interpreting quantities of interest as Bayesian inference problems and leveraging posterior statistics to guide stopping criteria and learning functions, the proposed approach significantly improves computational efficiency while maintaining high accuracy. A brief outlook on ongoing work, including dynamical reliability analysis and Bayesian model updating, will also be provided.
BIO:
Dr. Chao Dang is currently a postdoctoral researcher at the Chair of Reliability Engineering, Department of Mechanical Engineering, TU Dortmund University, Germany. He received his Bachelor’s (2016) and Master’s (2019) degrees in Civil Engineering from Hunan University, China, and earned his doctoral degree with distinction from Leibniz University Hannover in 2023, focusing on forward uncertainty quantification. His research interests include structural reliability analysis, stochastic structural dynamics, Bayesian model updating, and Bayesian numerical methods. Dr. Dang has published more than 45 peer-reviewed journal papers, with over 1,400 citations and an h-index of 24 (Google Scholar). He serves as an early-career editorial board member for several journals, including Computers & Structures, ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems, Part A: Civil Engineering and Part B: Mechanical Engineering, Journal of Reliability Science and Engineering, Urban Resilience and Earthquake Engineering, and Smart Construction.
Dr. Zhouzhou Song
TU Dortmund University, Germany
Title: Reliability Analysis of Stochastic Dynamical Systems Using Probabilistic Function-on-Function Nonlinear Autoregressive Surrogate Model (F2NARX)
Abstract:
Evaluating the reliability of structural dynamical systems under uncertainties arising from materials, manufacturing processes, and external excitations is a critical task in engineering practice. Surrogate models have gained increasing attention for efficient reliability analysis in complex and computationally intensive problems. However, constructing highly accurate surrogate models for stochastic dynamical systems under limited computational budgets remains challenging, particularly when estimating small first-passage failure probabilities.
This presentation introduces a Function-on-Function Nonlinear AutoRegressive model with eXogenous inputs (F2NARX) surrogate modeling approach tailored for nonlinear stochastic dynamical systems. Compared to state-of-the-art NARX models, the proposed F2NARX framework achieves significantly improved efficiency—up to orders of magnitude—while maintaining higher overall accuracy. Building on its probabilistic prediction capabilities, an active learning strategy is developed to efficiently estimate first-passage failure probabilities and is further embedded into subset simulation for rare-event analysis. The results demonstrate that the proposed methodology provides accurate and robust estimates of small first-passage failure probabilities for stochastic structural dynamical systems in a computationally efficient manner.
BIO:
Dr. Zhouzhou Song is a Humboldt Postdoctoral Fellow at the Chair of Reliability Engineering, TU Dortmund University, Germany. He received his B.Eng. degree with honors in Mechanical Engineering from Huazhong University of Science and Technology in 2019 and completed his Ph.D. with honors in Mechanical Engineering at Shanghai Jiao Tong University in 2024. His research focuses on developing machine learning–empowered methods for reliability assessment and design optimization of stochastic dynamical systems. He serves as a member of the scientific committee of ESREL 2026 and as an early-career editorial board member of the Journal of Reliability Science and Engineering (since 2025).
The workshop will take place in the institute’s library, Room 116, on Tuesday, March 3, 2026. The event starts at 10:00 a.m. (CET). If you would like to participate online via Webex, please contact Mengze Lyu at least one day before the presentations.
