Poster Presentations:
Graduate Student Researcher
University of Florida
Error quantification and data requirement for simulation of non-gaussian multivariate wind processes through data-informed stochastic models
Co-Authors: Srinivasan Arunachalam (Verisk Analytics), Arthriya Subgranon (University of Florida), and Seymour M.J. Spence (University of Michigan)
Abstract: The accurate simulation of multivariate wind processes is crucial for the probabilistic assessment of building systems under wind excitation. In particular, stochastic wind models based on the translation process are particularly valuable, as they can generate non-Gaussian stationary wind processes that match target marginal distributions. Integrating the translation process with the proper orthogonal decomposition (POD) method further enhances computational efficiency through mode truncation, while preserving adequate accuracy. Recent research has demonstrated the potential of using experimental wind tunnel pressure data to calibrate these stochastic wind models, capturing complex building-specific aerodynamic phenomena across various wind directions and geometries. However, quantifying potential errors that can arise from calibrating these models with typical short-duration wind tunnel records remains challenging. To address this gap, an extensive experimental campaign was conducted at the NHERI Boundary Layer Wind Tunnel at the University of Florida. This study evaluated wind tunnel records of varying lengths, wind directions, and surrounding configurations to quantify errors related to wind tunnel record variability, the calibration process, numerical modeling, and mode truncation. The results provide critical insights into the requirements and guidelines for confidently using wind tunnel data to calibrate data-informed stochastic wind models, that will soon be included in the SimCenter toolset.
Graduate Student Researcher
Florida State University
Self-Adaptive Evolutionary Optimization Approach to Enhance CFD Prediction of Wind Pressures on Buildings
Co-Authors: Pedro Fernandez-Caban (FAMU-FSU College of Engineering)
Abstract: Accurately predicting wind pressures in flow-separated regions, leveraging Steady Reynolds Averaged Navier–Stokes (RANS) turbulence models, remains a significant challenge in computational fluid dynamics (CFD), particularly for bluff-body structures. Steady RANS turbulence models, widely used for predictions of mean pressure coefficients, require the calibration of closure coefficients that directly influence model accuracy in the flow separation zones where negative suction pressures develop. This study introduces a self-adaptive evolutionary algorithm (SAEA) to enhance the predictive performance of the k-ω SST turbulence model by automating the selection of optimal closure coefficients. The SAEA is designed to minimize discrepancies between CFD-derived results and experimental wind tunnel data through an adaptive search process. Before integrating the SAEA, a parametric study was conducted to quantify the sensitivity of the closure coefficients in influencing mean wind pressures on a 1:20 Texas Tech University (TTU) WERFL building. The proposed SAEA, driven by experimental data collected at the Boundary Layer Wind Tunnel (BLWT) of The University of Florida (UF) NHERI Experimental Facility (EF), will iteratively refine these parameters to optimize the model's ability to capture complex flow separation phenomena. Steady RANS simulations were performed, revealing significant effects of the closure parameters on the predictive accuracy. This approach aims to reduce manual recalibration efforts, improve model robustness, and provide a computationally efficient and reliable tool for simulating wind pressures in flow-separated regions.
Assistant Professor
Texas Tech University
Hurricane-induced Damage Risk Assessment for Building Envelope Systems: A Real-time Forecast Framework
Co-Authors: Seymour Spence (University of Michigan)
Abstract: In past decades, hurricanes have shown a tendency to increase in frequency and intensity. This poses an increasingly serious concern for coastal community resilience under hurricanes. As such, advancement in emergency response and management under extreme wind events of this kind is urgently needed. In this research, a real-time damage risk framework is developed for building envelope systems, to provide comprehensive information support for emergency response and management. In particular, the damage to the building envelope system is quantified through a recently developed multi-demand coupled model with a full range of uncertainty. The intensity, i.e., site-specific wind speed and direction, for damage assessment is forecasted in real-time, based on the National Hurricane Center (NHC) advisory with the additional consideration of track stochasticity, and a parameterized wind field model with uncertainty. This scheme produces wind intensity samples when an imminent hurricane is identified, which is fed into the damage model to assess damage risks. To address the computational demand, the Kriging modeling scheme is used to establish an efficient mapping from site-specific wind speed and direction to damage statistics, which is further used to calculate risks. This framework is illustrated through a building with 8,100 envelope elements subjected to hurricane IRMA, showing its potential in informing emergency response and management under extreme wind events.
Postdoctoral Researcher
Kansas State University
Disruption Propagation in Interdependent Infrastructures: A Heterofunctional Graph Theoretic Approach
Co-Authors: Balasubramaniam Natarajan (Kansas State University)
Abstract: Modern cities depend on critical interdependent infrastructure systems (CIS), such as power, water, stormwater, wastewater, transportation, and communication systems, to provide essential services. CIS, by nature, are highly interconnected and interdependent. Recent worldwide events such as hurricanes, cyber-threats, earthquakes, and other disruptions have shown the vulnerability of failures in single infrastructure or partial CIS to cascade into multi-infrastructure failures, leading to service degradation, loss of life, property, and monetary losses. Therefore, studying the impact of disruptions in CIS has received significant research interest. Current techniques rely on models that explicitly incorporate dependencies in various forms, such as interdependency matrices and dependency graphs. However, several aspects critical to the study of CIS, such as the nature of resources, the extent of disruption, socio-economic factors, network physics, and system recovery, which are critical in studying the resilience of these systems, have yet to be simultaneously incorporated into these models. Therefore, in this work, we develop a novel heterofunctional graph framework to generate rich representations of CIS while considering social factors and community interactions. Community assets and emergency services such as residential buildings, schools, industries, EV charging stations, gas stations, hospitals, and fire services are geographically linked as consumers or producers of resources for infrastructure models. Further, unlike traditional multi-layer networks, the developed graphs provide a mechanism to study cascading failures and can be extended to more infrastructure systems, community assets, and emergency services. The developed framework can also be used for various applications, such as identifying critical infrastructure assets and asset hardening.