NHERI Computational Academy
July 23-26, 2024 | Texas Advanced Computing Center | Austin, TX
The 2022 NHERI Hackathon winning team, the Disaster Doctorates (from left: Te Pei, Shitao Shi, Louis Lin, Saanchi Singh Kaushal, and Shayan Razi)
Updates
The 2024 NHERI Computational Academy application window is now open! Download the 2023 NHERI Computational Academy agenda to learn more about what to expect this year, or check out the Program Information below.
Application Deadline: April 29, 2024
Program Information
NHERI Computational Academy Overview
DesignSafe and the NHERI SimCenter are pleased to announce the NHERI Computational Academy on July 23-26, 2024, at the Texas Advanced Computing Center.
The NHERI Computational Academy (NCA) includes hands-on training on DesignSafe and SimCenter tools and explores real-world case studies through project-based learning. The projects are designed to provide new graduate students and postdocs working on natural hazards research opportunities o learn about DesignSafe and SimCenter tools and to use these tools in new and innovative ways. The NCA covers a range of tools and workflows for data analysis, regional-scale modeling, uncertainty quantification, artificial intelligence, high-performance computing, and data visualization.
The NCA offers lectures that include hands-on training as well as project-based learning employing real-world case studies. Participants will be able to explore projects focussed on hurricanes, flooding, and earthquakes hazards with mentoring provided by domain experts. The academy covers these main topics:
- Jupyter notebooks for research
- Data analysis, artificial intelligence, and machine learning for natural hazards
- Deploying high-performance computing applications through DesignSafe/SimCenter
- Uncertainty quantification
- Regional-scale modeling of multi-hazard scenarios
Participants will develop new workflows that advance their research and educational goals with support from experts in Natural Hazards and TACC staff. Workflows have the potential for publication and deployment on DesignSafe and SimCenter for broader use in the natural hazards community. Participants are encouraged to choose from a suite of NCA projects based on their interests. The projects are an exciting opportunity to improve knowledge of NHERI tools, discover new research workflows, and explore interesting research topics.
NCA Projects
Project 1: Optimizing ADCIRC simulations using AI/ML
The project will involve modeling storm surges using full-scale ADCIRC. The team will then train ML models such as XGBoost and random forest to develop a fast and accurate surrogate model. The coastal impact of storm surges across Texas will be evaluated.
Project 2: AI/ML surrogate modeling
This project will explore using GNS as a differentiable AI/ML simulator to solve inverse problems for structural and geotechnical issues, such as earthquake-induced landslides or structural behavior. Using gradient-based differentiable simulators we aim to accelerate optimizing the design of infrastructure and safety features.
Project 3: Equitable emergency response and recovery during flooding
Situational awareness data inequity could translate to inequitable emergency response and recovery assistance. This project will collect historical community flood reports and contextualize them with sociodemographic and flood inundation data to quantify and visualize situational awareness data equity during a past flood event in Houston. Insights gained during this project will enable inequity-aware emergency response and recovery strategies and promote long-term community resilience.
Project 4: LLMs for liquefaction analysis
Leveraging ChatGPT to identify new knowledge in liquefaction Accessing ground motion databases and Next Generation Liquefaction dataset requires sophisticated knowledge of API requests or SQL query construction. The project aims to develop a plugin interface to facilitate complex queries and discover new knowledge through Large Language Models, such as ChatGPT.
Project 5: Employing AI/ML to predict wind loads on low-rise buildings
Low-rise residential buildings are vulnerable to wind damage during extreme weather events like hurricanes. Accurately predicting the wind loads on these structures for various exposure conditions and building shapes is crucial for mitigating potential wind-related damage. This project leverages aerodynamic datasets obtained from boundary layer wind tunnel measurements to develop ML-based models for predicting wind loads on archetype low-rise buildings. You will kick off by collecting the necessary dataset via Web API and then investigate the variation of wind pressure on the building surface across different parameters. Then, using PyTorch libraries, you will create deep learning models and train them using the prepared datasets. In the final phase, you will assess the performance of each model in predicting both mean and peak wind pressure patterns for unseen representative configurations.
Project 6: Sensitivity analysis for regional-scale risk assessment.
Running regional-scale hazard risk analysis requires substantial effort in collecting different kinds of input information ranging from asset characteristics (e.g., number of stories, built year of buildings and bridges), hazard characteristics (e.g., rupture, soil property, water elevation) to local design standards. With a case study example, this project will investigate the consequential impact of each input to the final decision variables, and identify what information is more important than others. Such exercise can better guide the data collection effort and facilitate reasonable assumptions for the regional analysis. To this aim, global sensitivity analysis methods will be investigated. After this session, the participants will be able to
- Run regional-scale risk simulation through R2D
- Understand the benefits and limitations of global sensitivity analysis
- Perform basic sensitivity analysis using python and quoFEM
Project 7: From the Table-Top to the Computer and Back: Using HydroUQ for Experiment-Simulation Validation
Design against hazards such as tsunamis and landslides are reliant on simulations, but are they correct? Learn how to perform an easy table-top experiment and replicate it in HydroUQ to check your modeling abilities. Then, design a blind-study by modifying the numerical simulation set-up and running a case in HydroUQ. Afterwards, try to match it on the table-top.
Together, perform the initial non-blind experiment and simulation. Individually, perform a shared blind numerical-experiment case. Finally, compare results as a team to observe the biases introduced unintentionally by individuals in simulations and experiments.
This project is suited for those in the fields of landslides, avalanches, tsunamis, and storm-surges where the Material Point Method excels. Students who want to bridge experimental, simulated, and real hazards will benefit, as well as those who’d like to learn more about HydroUQ, the Material Point Method, and GPU-acceleration. Free access will be provided on TACC Frontera and Lonestar6 to high-end nodes equipped with 4 Quadro RTX 5000 and 3 A100 GPUs.
Project 8: Transportation Infrastructure Performance under Seismic Hazards with SimCenter Tools
This project applies the workflow of seismic hazard analysis, transportation inventory generation, infrastructure performance assessment, and traffic flow simulation in SimCenter’s R2D Tool. Participants will apply data mining and processing for regional transportation inventory generation and traffic demand data collection. SimCenter tools will be used to assess likely physical damage on the transportation infrastructure (e.g. roads and bridges) and determine the traffic service disruptions caused by seismic hazards. Extending SimCenter’s Pelicun framework by incorporating user-defined fragility functions for infrastructure performance assessment will be explored.
Project 9: Open project
Teams will be formed based on interest. Choose at least two areas. Choose area of interest:
- Scientific Machine Learning
- AI/ML algorithms
- Computer Vision
- Large Language Models
- Cyber Infrastructure development
- Natural Hazards
- Civil Engineering applications
Dates and Venue
The NHERI Computational Academy (NCA) will occur at the Texas Advanced Computing Center (TACC) at the University of Texas at Austin July 23-26, 2024.
Registration Fees and Travel Support
The registration fee for the NCA is $150. We will offer domestic travel support up to $1,500 can be requested on the application form.
Accommodation
You may stay near TACC at any of these hotels(opens a new window).
Eligibility
The course will use Python as the main programming language. It is expected that participants will have some basic programming experience. If you are familiar with other languages such as MATLAB or Java, refresh your Python knowledge by following Google's Python course(opens a new window).
Explore
Learn about the NHERI Computation Academy by exploring past interviews and projects.
2021 DSA Hackathon Project Videos (view full YouTube playlist(opens a new window))
Automated Model Calibration for Cyclic Tests Using OpenSees & Jupyter Notebooks
Using Deep Learning for One-dimensional Full-Waveform Inversion
Improving ADCIRC UX w/ DS - Parameter Customization & Outputs Visualization
Contact Us
Professor Krishna Kumar
The University of Texas at Austin
krishnak@utexas.edu