Four Oden Institute faculty, Tan Bui-Thanh, Leszek Demkowicz, Moriba Jah, and Ali Yilmaz received 2019 W. A. "Tex" Moncrief Grand Challenge Awards, based on their highly compelling research proposals related to the Grand Challenges in computational engineering and sciences that affect the competitiveness and international standing of the nation.
The Oden Institute will provide these faculty with necessary resources to cover release time from teaching for one or more semesters to work on their research. Stipends of up to $75,000 per award per semester are provided to cover salary and other expenses.
Bui-Thanh, an Oden Institute core faculty member and associate professor of aerospace engineering and engineering mechanics, will use his award to advance the subset of machine learning known as “deep learning.” For computer vision, speech recognition, and natural language processing, deep learning has proven effective, but its success in the scientific computing community has fallen short due to limitations in training neural networks. Through his Grand Challenge grant entitled “Advanced Deep Learning Strategies for Forward and Inverse Problems,” the Bui team will further their long-term goal to establish a research program that addresses advanced optimization methods, automatic hyperparameter determination, and uncertainty quantification techniques for deep learnings.
Demkowicz, an Oden Institute core faculty member and professor of aerospace engineering and engineering mechanics, will apply his award toward achieving the holy grail of hp -adaptive Finite Element (FE) methods: to build a self-adaptive FE method (and code) that automatically decides which elements should be refined and how to refine them. His project “A Scalable MPI hp-Adaptive Finite Element Software Library for Complex Multiphysics Applications” builds on an existing open MP implementation and a unique data structure based on a single (object) array storing element nodes (vertices, edges, faces and element interiors). This project contributes toward his long-term goal: to develop an MPI/open MP version of an existing open MP hp-adaptive finite element software, document it, and put the updated versions into public domain under BSD License. The Grand Challenge grant allows for a start on the documentation of the hp technology and the hp library.
Jah, an Oden Institute core faculty member and assistant professor of aerospace engineering and engineering mechanics, will focus his Grand Challenge research on three basic tasks: (a) high/multi-fidelity physics modelling of the space environment interactions with artificial satellites; (b) computational representation of behavioral scientists, social scientists, anthropologists, and (geo)political scientists; and (c) hard/soft information fusion and data science/analytics. The goal is to significantly improve the ability to quantify and assess the behavior of operational satellites, and to be able to predict these for any given scenario in space with reliable uncertainty, regardless of the nature or nationality of the satellite owner/operator. This underscores the ability to make decisions regarding space activities and the challenge is linked to meaningful Space Security and Sustainability, a major concern of the U.S. Government and of high interest to the United Nations Institute for Disarmament Research (UNIDIR).
Yilmaz, an Oden Institute core faculty member and professor of electrical and computer engineering, will use the Grand Challenge grant to develop and validate computational methods and software that enable simulation-based prediction of electromagnetic side-channel vulnerabilities of embedded cryptographic modules. Cryptographic techniques prevent unauthorized access to information, protect privacy, and authenticate other nodes. Embedded computing systems like those now used by devices directing transportation, health care, defense, manufacturing, and infrastructure have added challenges. Since they perform out in the world and remain physically accessible to adversaries, they become vulnerable to attacks differently than traditional information-processing systems. Yilmaz will develop state-of-the-art computational methods to increase the predictive power of simulations when modeling embedded cryptographic systems. The results should lead to advances in the design of such systems and ultimately improve cyber-physical security.