MNTL Master Calendar

Back to Listing

Forked-Accelerated Simulation Technique (FAST) for cardiovascular blood flow modeling

Event Type
Department of Mechanical Science and Engineering
Join Zoom Meeting Meeting ID: 876 5153 3618 Password: 490410
Oct 2, 2020   12:00 pm  
Professor Mahdi Esmaily, Mechanical & Aerospace Engineering, Cornell University
Lindsey henson
Originating Calendar
MechSE Seminars

Despite their prominent role in scientific discovery and engineering analysis, computer simulations remain infeasible or unaffordable for solving the following two groups of problems: 1) time-critical problems that set a ceiling on the acceptable time-to-solution (TTS), and 2) problems that require running many simulations (e.g., optimization, uncertainty quantification, and parameter identification). Although parallel processing has offset these limitations, certain application areas remain where the current state-of-the-art tools are still inadequate. Cardiovascular blood flow simulation is one of the application areas where the two limitations above are fully materialized. Simulating a surgical design is practical only when the TTS is shorter than the available time between diagnosis and operation. Further customization through optimization entails running many simulations that put an even more stringent constraint on the TTS. This talk aims to address the above challenges through the introduction of Forked-Accelerated Simulation Technique (FAST). The underlying idea of FAST is to perform many inexpensive surrogate simulations in parallel to construct a physics-based reduced-order model that can predict an output of interest in virtually real-time. FAST offers a new dimension for parallelization since the bulk of computations is embarrassingly parallelizable. This property enables the fast solution of elliptic partial differential equations in general and incompressible Navier-Stokes in particular. Comparing FAST to standard CFD in simulating unsteady blood flow in complex patient anatomy shows that it can produce a prediction with a few percent errors in seconds rather than hours.

About the Speaker
Dr. Esmaily is currently an Assistant Professor in the Department of Mechanical and Aerospace Engineering at Cornell University. He received his Ph.D. from the University of California, San Diego, where he developed computational methods to study and improve a surgery performed on pediatric patients. After completing his Ph.D., he moved to the Center for Turbulence Research at Stanford University as a PostDoc to study the effect of radiation on particle-laden turbulent flow.  Dr. Esmaily's research interests center around the study of emerging applications in cardiovascular mechanics and biological flows and the development of computational techniques for such problems. The overarching goal of Dr. Esmaily's lab is to build high-fidelity predictive tools that can be applied to the study of cardiovascular diseases and contribute toward improving current treatment methods and surgical techniques.

Host: Professor Laura Villafane

link for robots only