Development of digital twins to improve hurricane prediction

More than half of the US population lives in counties or parishes in coastal watersheds. The coastal area along the Gulf of Mexico is one of the most populous and energy-intensive areas, making it a national hub for many large-scale carbon-to-capture storage facilities. I am.

The proximity of both the community and the energy infrastructure to the ocean makes it extremely vulnerable to the devastation that can be caused by floods and wind damage from severe weather events in the bay. Hurricane season..

Clint Dawson, a professor of aerospace engineering and engineering mechanics (ASE / EM) and director of the Computational Hydraulics Group at the Oden Institute for Computational Engineering and Science in UT Austin, to make hurricane storm surge predictions more accurate. It is working. More than ever. Thanks to a new grant from the Ministry of Energy (DOE), Dawson leads an interdisciplinary research project, a computational “digital twin” that bridges the gap between artificial intelligence (AI) and mechanical multiphysics simulation and knowledge discovery. Develop the framework. A learning (ML) technology called MuSiKAL.

Simply put, a digital twin is a virtual representation of an object or system throughout its lifecycle through periodic real-time data updates provided by sensors that span the object or system. Using simulation, machine learning, and other decision-making technologies, digital twins help predict future performance and behavior.

Dawson’s team has modeled storm surge forecasts for 20 years, from Hurricane Katrina, Rita, Ike and Harvey to Hurricane Aida, the biggest storm of the season. And storm surge experts will first tell you that each brings a unique set of features. However, you can learn from each lesson that may help you in the future.

Currently, when a hurricane model is running, measurements are collected in very unobtrusive places such as the coastline and the sea, but these points are all points in all areas that may be affected. It does not represent.

“We need a model that provides additional information. If that data is available, we can better inform the model we are currently running,” Dawson said. “Then you can go back and compare the model with the data to get a more accurate image.”

Digital twins have already been developed in a variety of situations, from the latest aircraft design to systems that help manage the entire city. In the context of extreme weather modeling, this technology can, with the help of AI and ML, combine knowledge of previous storms to enable faster prediction of storm behavior in real time.

“These models are very complex and can take hours to simulate on a supercomputer. Machine learning We may be able to make faster predictions in real time based on data collected from very similar previous hurricanes, “Dawson said.

Through its Advanced Scientific Computing Research (ASCR) program, DOE supports a collaborative team of experimental and computational scientists at the University of Texas at Austin, Louisiana State University, Notre Dame University, and the Northwest Pacific National Laboratory. They are headed by Dawson along with fellow ASE / EM professors and Tan Bui-Thanh, a core faculty member of the Oden Institute.

Other participating UT experts include Bridget Scanlon and Alexander Sun from UT’s Department of Economics and Geology, and Dev Niyogi and Zong-Liang Yang from the Jackson School of Geosciences.

DOE has recently invested in the development of an Earth system model for climate research. Dawson said he is looking forward to working on research that is directly related to climate forecasting.

“I think this will be a groundbreaking project, which is in good agreement with the expertise we have built up over the last 20 years,” Dawson said. “It’s very exciting to work with the Department of Energy to develop longer-term projections of what will happen to the energy sector and society as a whole for the future climate.”

The Ministry of Energy’s fund for integrated computational and data infrastructure for scientific research will provide $ 5.2 million for the entire project, and UT Austin will receive $ 3 million.

Use machine learning and radar to better understand the risk of storm surges
For more information:
Coastal emergency risk assessment page cera.coastalrisk.live/

Provided by Oden Institute for Computational Engineering and Sciences