PhD Student Auburn University
LES Evaluation of Terrain Adjusted Inflow Conditions on Gable Roof Buildings
Co-Authors: David Roueche (Auburn University) and Jay Khodadadi (Auburn University)
Abstract: It is established that wind loads drive roof uplift, façade pressures, and cladding design in low rise structures, however the influence of terrain adjusted flow conditions remains to be investigated. This research focuses on the impact of terrain-induced turbulence over gable-roof buildings and assesses the boundary layer flow acceleration/deceleration over sloped terrain, additional flow separation and twist effects.
The computational framework was developed using DesignSafe WE-UQ platform for a benchmark TPU gable-roof building over a 3D symmetric hill to ensure proper meshing, realistic inlet boundary conditions and LES numerical setup conditions. To ensure fully developed boundary layer flows, a turbulent Divergence-Free Method (DFM) inlet was adopted from WE-UQ. Computational verification and preliminary validation studies were conducted in two stages: First, RANS simulations were conducted for normal and terrain adjusted inflow conditions on the model-scale building. Second, LES simulations on the isolated 3D hill were validated using literature wind-tunnel studies. Currently, the TPU building is placed above the hill for LES simulations to establish realistic terrain adjusted inflow conditions and currently being executed using TACC HPC nodes.
The study shows the development and changes observed in mean and peak roof pressures from windward to leeward sides of the hill at five specified locations. The LES results also revealed additional flow deceleration on the leeward hill side that RANS failed to capture due to blockage effects. Ultimately, this work seeks to develop terrain-adjusted inflow conditions for low-rise structure, accounting for the complex aerodynamic coupling of building locations over sloped hill terrain.
Masters Student Oklahoma State University
Quantifying the Contribution of Wind-borne Debris Impact to Structural Failure through Multi-scale Modeling of Transmission Infrastructure
Co-Author: Brian Giffin (Oklahoma State University)
Abstract: Contemporary research efforts have sought to assess the structural performance of transmission infrastructure assets such as transmission towers under extreme wind loading scenarios, but observational evidence suggests that impacts from wind-borne debris, such as fallen trees and vegetative debris, plays a potentially significant role in contributing to structural damage. The present study seeks to quantify the extent to which debris impact and localized damage to structural members in transmission towers influence overall structural performance. This is achieved through a multi-scale computational modeling strategy, encompassing: determination of localized damage to structural components caused by debris impact using LS-DYNA; and assessment of overall structural performance under wind and debris loading using OpenSees, considering the influence of component-scale damage. At the component-scale, quoFEM is used to simulate a range of impact scenarios considering variable debris size, velocity, and impact location. These results are then used to construct component-level fragility curves relating to the intensity of a single impact event to member-scale damage (quantified as a percent reduction in buckling capacity). At the structure-scale: the Structural Wind-borne debris Impact Risk assessment Library (SWIRL) is used in combination with OpenSees to assess the overall structural performance of transmission towers under combined wind and debris impact loads, accounting for reductions in the capacity of individual members caused by discrete impact events. Finally, quoFEM is used to execute the aforementioned workflow to develop fragility curves characterizing the performance of transmission towers under the combined effects of wind and debris hazards.
Professor Louisiana State University
CFD-based Integrated Aero-Hydro-Structural-Mooring Dynamics Modeling of Floating Wind Turbines
Co-Author: Pengxu Zou
Abstract: Developing critical technologies for floating offshore wind turbines (FOWTs) is imperative to meet the rapidly increasing demand for clean energy. Compared with fixed-bottom turbines, FOWTs enable access to deep-water regions, offering larger power capacities and reduced visual impact. However, understanding their complex dynamic responses remains challenging due to strong nonlinear aero–hydro–structural-mooring interactions, which require advanced multi-physics modeling, laboratory testing, and field measurements. In this study, a high-fidelity, fully coupled aero–hydro–mooring numerical model for FOWTs is developed using OpenFOAM. An integrated and numerically robust coupling strategy is implemented to consistently resolve the interactions among aerodynamic loads, hydrodynamic wave effects, and mooring system dynamics. To accurately handle the large-amplitude motions of the floating platform under extreme environmental conditions, an overset mesh technique is employed. In addition, simulations using the widely adopted open-source program OpenFAST are conducted for comparison. A series of numerical case studies are performed under both operational and extreme wind–wave conditions. Under operational conditions, good agreement is observed between the OpenFAST predictions and the high-fidelity CFD results. However, under extreme conditions, noticeable discrepancies arise. The results indicate that strong wind–wave coupling effects significantly influence the dynamic responses of the FOWT system, revealing nonlinear interaction mechanisms that cannot be fully captured by mid-fidelity or uncoupled modeling approaches. These findings demonstrate the capability of high-fidelity CFD-based methods to resolve complex coupled dynamics, which provide valuable insights for the analysis, planning, and design of next-generation floating offshore wind turbines.
PhD Student Rice University
High Fidelity Simulation-based Fragility Assessment Approaches to Account for the Effects of Neighboring Structures
Co-Author: Jamie E. Padgett (Rice University)
Abstract: Although the literature on risk assessment of coastal structures subjected to hurricane-induced storm surge and waves has grown significantly, the fragility models used as the basis for such analyses (as well as the hazard simulations that inform estimates of the hazard intensity at each structure) overwhelmingly neglect the effects of neighboring structures that alter the loads experienced across a portfolio. Our research addresses this gap by developing a new set of topology-aware fragility models that utilize the spatial layout of neighboring structures as an additional input parameter, along with hazard intensity measures and structural parameters. Numerical full-scale computational fluid dynamics (CFD) analyses are leveraged to capture the load modification effects from neighboring structures by including multiple structures in the analysis domain. The CFD models are validated using experimental data available from DesignSafe Data Depot to achieve confidence in the numerical simulations. Additionally, the repetitive analyses requirement for fragility derivation, considering different combinations of intensity measures, structural parameters, and layout parameters, is fulfilled by leveraging DesignSafe HPC resources. The analysis results inform the development of closed-form topology-aware fragility models derived via statistical learning algorithms. Additionally, lower-fidelity simplified tools to estimate pseudo-modified intensity measures due to the presence of neighboring structures, or fragility modification factors, are also posed as alternative approaches for incorporating these traditionally neglected effects in existing risk assessment pipelines. The resulting HPC and machine learning enabled fragility and risk assessment framework paves a path to rigorously, yet affordably, account for realistic spatial layouts of coastal structural portfolios.
PhD Student Oregon State University
Numerical Investigation of Wave-Induced Uplift Forces on Elevated Coastal Structures Using DualSPHysics
Co-Authors: Daniel Cox (Oregon State University), André Barbosa (Oregon State University), Jorge Romero-Loyola (Oregon State University), and Pedro Lomonaco (Oregon State University)
Abstract: The accurate estimation of wave-induced uplift forces is essential for developing physics-based fragility functions and improving design guidance for elevated coastal buildings subjected to extreme weather events. Existing flood fragility models used in community-level risk assessment tools, such as HAZUS-MH, primarily rely on inundation depth and neglect hydrodynamic forces from waves and surge, despite evidence that these forces strongly influence structural damage and failure. Recent uplift load formulations relate vertical pressures beneath elevated decks to horizontal wave pressures on seaward faces through empirically derived coefficients; however, these relationships remain insufficiently tested under irregular, directional wave conditions representative of hurricanes.
This study uses smoothed particle hydrodynamics (SPH) modeling with DualSPHysics to investigate wave-induced uplift loading on elevated coastal structures. The numerical modeling simulation results are compared to data from the NHERI Directional Wave Basin at Oregon State University. The SPH framework allows for analysis of wave impact and flow separation beneath elevated decks. These processes are known to drive variability in uplift forces and are difficult to capture using depth-based models.
Simulated horizontal and vertical pressure time series are integrated to quantify uplift forces and examine the sensitivity of uplift response to wave breaking regime and air gap. Specifically examining force fluctuations and pressure redistribution at the leading edge of the structure, to ultimately develop consistent design parameters. The resulting insights support the development of improved fragility functions that explicitly account for wave height and impact processes, informing future updates to coastal building design standards.
PhD Student University of California, San Diego
Comparative Assessment of Gaussian Process and LSTM-Based NARX Surrogates for Large-Scale FE Models of Dam-Water-Foundation Rock Systems
Co-Author: Joel P. Conte (University of California, San Diego)
Abstract: High-fidelity finite element (FE) models of large-scale civil infrastructure, such as concrete gravity dams, representing coupled dam–water–foundation rock (DWFR) systems are widely used to predict seismic response and assess seismic safety. To enhance their reliability, these models are commonly calibrated within a Bayesian framework to enable uncertainty quantification. However, repeated direct evaluations of such large-scale FE models are computationally prohibitive, particularly in Bayesian model updating, which requires a large number of forward-model evaluations. Surrogate models (SM) are therefore widely adopted to emulate FE model responses and reduce computational cost, provided that the resulting loss of accuracy is rigorously quantified relative to direct FE simulations.
This study presents a comparative assessment of Gaussian Process Regression (GPR)–based and Long Short-Term Memory (LSTM)–based nonlinear autoregressive models with exogenous input (NARX) as surrogates for predicting the seismic response of a DWFR system. As a first step toward surrogate-assisted Bayesian inference, the investigation focuses on a linear 2D FE model of a concrete gravity dam interacting with a compressible water reservoir and an elastic foundation-rock domain. The FE model serves as the physics-based reference from which input–output datasets are generated. Ground acceleration time histories recorded at the foundation-rock surface are used as inputs, while total acceleration and added hydrodynamic pressure responses at multiple sensor locations along the dam faces are considered as outputs.
The SMs are evaluated in terms of time-history prediction accuracy under one-step-ahead and free-run modes, robustness to variations in excitation characteristics and training data size, and computational cost. The resulting accuracy–cost trade-offs provide a quantitative basis for assessing the suitability of GPR-NARX and LSTM-NARX surrogates for Bayesian model updating of DWFR system.
Assistant Professor Rice University
Stochastic emulation for seismic risk assessment using recorded ground motions: evaluation of applicability and significance of intensity measure selection
Co-Author: Alexandros Taflanidis (University of Notre Dame)
Abstract: Surrogate modeling has emerged as a powerful tool for accelerating seismic risk assessment, accomplishing emulation of structural response using a limited number of nonlinear time-history analyses. One approach to develop a surrogate model for this type of application is by establishing a stochastic mapping that links a low-dimensional input vector, composed of ground-motion intensity measures (IMs) and structural parameters, to peak engineering demand parameters (EDPs), inheriting the spirit of traditional seismic fragility analysis. A major consideration when establishing this mapping is the accurate prediction of record-to-record variability, which represents an aleatoric source of uncertainty in surrogate model development. When this uncertainty is heteroskedastic across the parameter space, a special surrogate modeling approach called stochastic emulation needs to be utilized. The characteristics of the aleatoric uncertainty, particularly its scale and heteroskedasticity, depend on the selection of IMs. This work investigates the three-way coupling between IM selection, characteristics of aleatoric variability, and surrogate performance, thereby providing recommendations for IM selection and the required properties of stochastic emulators to achieve the desired performance. This is facilitated by a Gaussian process-based stochastic emulator that flexibly describes the heteroskedasticity of second-order moments associated with the aleatory uncertainty. The results indicate that the choice of IM can have a substantial impact on the aleatoric uncertainty characteristics of the EDPs, leading to different preferences for emulator characteristics and varying accuracy expectations.
PhD Student Lehigh University
Real-Time Hybrid Simulation of Fluid-Structure Interaction Using AI-Based Surrogate Models
Co-Authors: Dimitrios Kalliontzis (University of Houston), James Ricles (Lehigh University), and Liang Cao (Lehigh University)
Abstract: Real-time hybrid simulation (RTHS), also referred to as real-time cyber-physical simulation, is an experimental methodology for evaluating the dynamic response of structures subjected to hazards such as wind, earthquake, and wave loading. In RTHS, the structural system is divided into an analytical substructure, modeled numerically, and an experimental substructure, tested physically in the laboratory, with the two substructures kinematically linked, satisfying equilibrium across their interface, and integrated in real-time.
Extending RTHS to fluid-structure interaction (FSI) problems presents significant challenges. Experimental representation of fluid effects requires specialized facilities and complex multi-degree-of-freedom actuation systems, leading to substantial cost and experimental complexity. In addition, high-fidelity computational fluid dynamics (CFD) solvers are incompatible with the strict real-time computational constraints required for stable RTHS execution.
To address these limitations, this work introduces an artificial intelligence (AI)-based surrogate modeling framework to represent the fluid domain within RTHS. A neural network (NN) surrogate is trained using high-fidelity simulation data generated using the Arbitrary Lagrangian-Eulerian with skeleton-based structural models (ALE-SSM) approach. Large training datasets are generated using high-performance computing resources at the Texas Advanced Computing Center (TACC). The trained NN model predicts fluid forces in real-time, enabling RTHS of FSI systems without the need for physical fluid modeling or real-time CFD solvers.
The proposed approach enhances the feasibility of RTHS for complex FSI problems and provides an efficient pathway for extending RTHS performed at the NHERI Lehigh Experimental Facility (EF) to more complex multi-physics systems that are otherwise experimentally or computationally prohibitive.