Poster Presentations:
Postdoctoral Fellow
UC Berkeley
Computational Tools for Structural Health Monitoring at Scale
Co-Authors: Khalid Mosalam (UC Berkeley)
Abstract: What computational aspects must be considered for an "at-scale" structural health monitoring platform, and how might a software library achieve them? The BRACE2 platform (Bridge Rapid Assessment Center for Extreme Events) is a dynamic web application that monitors instrumented structures in California. Currently, it evaluates an inventory of 22 bridges. Part of its computational infrastructure is the `mdof` library, which exposes an automation-friendly interface for vibration-based, data-driven models of structures. Through its use on the BRACE2 platform, the library undergoes iterative development to systematize the computation of modal parameter identification and response reconstruction in a manner that is agnostic to structural properties, and thus amenable to large-scale monitoring. This presentation walks through the methods available through `mdof` and how it interfaces with the BRACE2 platform to compute the structural health metrics for each event evaluation. A preview is provided of the dataset that results from the inventory-level assessment, which is proposed as a set of baseline structural performance metrics that can be used to train ML models.
Graduate Student Researcher
Texas A&M University
An Integrated Model for Predicting Power Outages during Various Types of Extreme Weather Events in Texas
Co-Authors: Sangkeun M. Lee (Oak Ridge National Laboratory) and Stephanie G. Paal (Texas A&M University)
Abstract: Climate change is causing an increase in severe weather patterns, leading to a growing vulnerability of the U.S. power grid to disruptions. For instance, Hurricane Beryl in July 2024 caused power outages for approximately 2 million households, primarily in Houston, Texas. Similarly, the winter storms in February 2021 caused significant power outages for 4.5 million households across Texas. Moreover, Hurricane Ida, which landed in August 2021, resulted in more than 1 million customers in Louisiana being left without power. All of these unprecedented storms motivate the development of a grid outage prediction model to prepare for the impacts of different extreme weather events leading to power disruptions.
This research focuses on the state of Texas due to its large population and potential susceptibility to power outages. Data was collected regarding the number of customers experiencing power outages in each county using the Environment for Analysis of Geo-Located Energy Information (EAGLE-I) developed by Oak Ridge National Laboratory (ORNL) from 2014 to 2023. The National Weather Service (NWS) identified at least 20 types of severe weather classes and the top 10 weather types were selected for this study based on the cumulative number of power outages. In addition, this study used weather data from the National Aeronautics and Space Administration (NASA) to gather information on 30+ weather parameters. This approach led to developing a predictive model to estimate power outages resulting from Texas's top 10 severe weather events. Although the study focuses on Texas, it has the potential to be applied nationwide on a comprehensive scale.
Graduate Student Researcher
Kansas State University
SWMM-TK: an open-source toolkit for enhancing stormwater infrastructure resilience under increased flooding risk
Co-Authors: Rahul Madbhavi (Kansas State University), Vaishali Sharda (Kansas State University), and Bala Natarajan (Kansas State University)
Abstract: The frequency of urban flooding has increased significantly due to global climate change, posing a critical threat to the resilience of stormwater and wastewater infrastructure. The Storm Water Management Model (SWMM) has been widely employed to simulate and analyze the relationships between environmental variables and the performance of urban drainage systems. However, SWMM's reliance on detailed infrastructure data, which might be unavailable, presents a challenge for its effective use in many areas.
To address this limitation, we introduce the Storm Water Management Model Toolkit (SWMM-TK), an open-source toolkit designed to automate the setup of SWMM models using publicly available data such as Digital Elevation Models (DEM) and road network. The toolkit leverages road network data as a proxy for the underground pipeline network, assuming stormwater and wastewater pipelines follow the road infrastructure.
SWMM-TK generates a stormwater collection network by pruning the road-based network to form an acyclic graph. A graph traversal algorithm then creates a directed network, placing pumps strategically based on elevation changes during the traversal process toward the outfall node. In addition to network generation, the toolkit incorporates essential hydraulic and hydrologic parameters such as watershed characteristics, pipeline properties, infiltration methods, and rainfall data. These inputs are used to run simulations in SWMM, ultimately identifying areas vulnerable to flooding.
This toolkit is being tested in rural and urban communities in Kansas to simulate a range of infrastructure designs, loading, and vulnerability. The resulting network is then compared to real data from these communities to evaluate how accurately the synthetic network can predict locations vulnerable to flooding.
Future iterations of SWMM-TK will include an automated pipe-sizing feature to optimize pipe dimensions based on flow data. We demonstrate the toolkit’s utility across several case studies that generate SWMM models using publicly available data, highlighting its effectiveness in urban areas with limited stormwater or wastewater network information.
Undergraduate Student Researcher
Texas A&M University-San Antonio
X2Sim: Rapid digital twin creation from text and videos for natural hazard modeling
Co-Authors: Jonathan Gaucin (The University of Houston), Krishna Kumar (The University of Texas at Austin), Cheng-Hsi Hsiao (The University of Texas at Austin), and Justin Bonus (University of California, Berkeley)
Abstract: Traditional methods for modeling natural hazards such as landslides, floods, and storm surges are computationally intensive and time-consuming, thus limiting their applicability in effective disaster preparedness and response in real-world scenarios. To address this challenge, we present a novel framework, through TACC HPC resources, for rapidly creating digital twins that significantly reduces the time, manual input, and computational resources needed for simulating the interaction between natural hazards and real-world 3D objects. Our X2Sim framework utilizes an agentic text-to-simulation Large Language Model (LLM) to generate digital twins, integrating two distinct 3D object generation methods: (i) A text-to-3D point cloud diffusion model that swiftly creates 3D point clouds from natural language descriptions, enabling rapid digital twin prototyping, and (ii) An efficient method for constructing high-fidelity point clouds from video input, allowing for more detailed digital twin representations of existing structures. These digital twins are integrated into a Graph Network-based Simulator (GNS) that models the dynamics of particle and fluid interactions, enabling the simulation of complex natural hazard scenarios. Our X2Sim system allows for adjusting simulation parameters, offering a robust tool for exploring various disaster scenarios and their impacts on the digital twins. While our digital twin framework may not match the accuracy of high-fidelity numerical methods, it significantly reduces computation time and complexity, making it viable for near-real-time applications. The X2Sim approach offers a valuable balance between speed and precision in digital twin creation and simulation, providing a streamlined, low-intervention workflow for researchers and practitioners in natural hazard modeling and disaster preparedness.
Adjunct Lecturer
Stanford University
Structural Fire Analysis using OpenSeesPy
Abstract: OpenSeesPy is a Python 3 interpreter of OpenSees (Open System for Earthquake Engineering Simulation) and is widely used for the nonlinear simulation of structures under earthquake hazards. Recently, OpenSeesPy has extended its capabilities to support performance-based analyses of structures exposed to fire hazards, albeit with limited functionality. This course module introduces students to the essential commands in OpenSeesPy for conducting structural fire analyses, categorized into thermo-mechanical uniaxial materials, thermo-mechanical fiber sections, thermo-mechanical elements, and fire-induced structural temperature loading. Targeted at graduate students in structural engineering with foundational knowledge in programming, nonlinear structural analysis, and thermal effects on materials, the course aims to equip students with a solid understanding of structural behavior under fire conditions. Key topics include time-gas temperature fire curves and fire loads, which are crucial for understanding the dynamic behavior of fires in enclosed spaces. The course will also delve into the properties of materials at elevated temperatures, emphasizing the degradation of strength and stiffness in common construction materials like steel and concrete when exposed to fire. By integrating the principles of thermo-mechanical analysis, students will learn to simulate the complex interactions between thermal loads and structural responses, enabling them to accurately assess the fire resistance of structures. Developed as part of the NHERI SimCenter tools, this course module offers a hands-on, interactive learning environment, fostering the development of critical skills essential for the next generation of engineers and researchers in structural fire safety.