Dramatic advances in data sciences, machine learning, and scientific computing, as well as the growing ability to collect scientific data, has led to a need for improved predictive modeling and design of complex systems. In order to better characterize the predictability of computational models and product performance, a new research center at the University of Notre Dame, the Center for Informatics and Computational Science (CICS), will develop mathematical, statistical, and scientific computing techniques to address the challenges associated with uncertainty quantification.
In explaining the new center, Nicholas Zabaras, Viola D. Hank Professor of Aerospace and Mechanical Engineering and founding director of the CICS, said, “At the CICS, we want to be able to deliver computational models that quantify uncertainties across multiscale and multiphysics models, but also identify the most informative computational or experimental data needed to minimize uncertainties in predictions or design performance. From aerospace to pharmaceuticals, our work will be applicable to many industries for improving product reliability as well as for accelerating product development by minimizing unnecessary experimental testing or maintenance.”
Data-driven computational modeling is an imperative component of modern simulations that can significantly increase their predictive accuracy and such modeling is essential when designing products that need to perform with relative consistency in often uncertain environments. The center will emphasize unifying themes in the mathematical and statistical sciences, as well as scientific computing, which are necessary for the predictive modelling and design of complex systems within disciplines such as physics, chemistry, biology, and engineering.
“Predictive modeling of physical systems has significant benefits in many fields,” said Robert J. Bernhard, Vice President for Research and Professor of Aerospace and Mechanical Engineering. “Through the CICS, faculty will have the opportunity to develop techniques that not only better predict performance but also predict the variations that occur due to environmental conditions and manufacturing and quantify the uncertainty that will naturally occur in models.”
Members of the CICS will develop data-driven computational methods to predict system performance using limited and noisy data while accounting for model limitations and variability in environmental conditions and material properties. The center aims to build collaborations with both government and industry partners. To learn more about the CICS, its director, and how to get involved, please visit cics.nd.edu.