High resolution simulations for system analysis

Team: Anthony Patera with Phuong Huynh, David Knezevic, and Thomas Leurent

New generation of computer simulations that are component-­based, parametrized and fast. The underlying technology accelerates high resolution finite element simulations by orders of magnitude without compromising accuracy.

Current computer simulations used in engineering design are very complex to create and very time consuming to execute. The procedures developed in this project offer two key advantages: first, the ability to build --- once ---parametric computational components and to then re-use/assemble these components for the rapid and accurate analysis of a wide variety of systems; second, the ability to rigorously and rapidly certify the accuracy of these predictions relative to a detailed--physics high--resolution finite element truth. 

The technology from this project was spun out into a startup company, Akselos.

Related articles: 

To demonstrate speed of the simulation methods, Prof. Patera's lab has implemented an android app that performs full 3D simulations: 

Wired -- http://www.wired.com/gadgetlab/2010/08/supercomputing-app-android/

MIT News -- http://web.mit.edu/newsoffice/2010/supercomputer-smart-phones-0901.html

Gizmodo -- http://gizmodo.com/5616794/supercomputing-on-an-android-smartphone

Related video:

Video of Reduced Basis Approximation for Large Component Based Structures: http://augustine.mit.edu/methodology/video_clips/prscrbe_movie.mp4