You are here

Seminar: Rohit Deshmukh

Model Order Reduction of Multi-Scale Flows
Friday, April 17, 2020, 3:00 pm

Seminar Speaker
Dr. Rohit Deshmukh
The Ohio State University, Post Doctoral Researcher

Seminar Title
Model Order Reduction of Multi-Scale Flows

Seminar Location
Online only via Zoom: go.osu.edu/ARCseminar


Abstract
Reduced Order Models for complex, dynamical systems are long sought to enable design, uncertainty quantification, control, and analysis. Conventional modal decomposition techniques, which often serve as a foundation for model reduction, are limited for multi-scale problems. We present our evolving efforts at developing efficient reduced order models while satisfying physical constraints, maintaining numerical stability, and scaling to large systems with O(108-109) full order model degrees of freedom. We will discuss two approaches, borrowing ideas from both machine learning and computational mechanics, which seek to exploit spatio-temporal sparsity in dynamical systems. The first is based on the process of sparse coding, which seeks to identify a small number of multi-scale modes to span a large spectrum of features in a system. These modes promote the stability of projection based reduced order models. The second approach aims to scale the model order reduction to large multi-scale systems. Spatially local and redundant features are learned from data and then used to enrich the low-order finite element spaces. Finally, we will discuss our current and planned future efforts in data-driven modeling and analysis.


About the Speaker
Dr. Rohit Deshmukh is a Post Doctoral Researcher in the Multi-Physics Interactions Research Group at The Ohio State University. He earned a doctoral degree from OSU in the Autumn of 2016. Prior to joining Ohio State in 2011, he had a short stint working as a Noise Vibration Harshness Engineer at Tata Motors Limited, India. He received bachelor’s and master’s degrees in Mechanical Engineering from the Indian Institute of Technology Madras. His research interests include nonlinear dynamics, machine learning, reduced order modeling and fluid-structure interactions.


Hosted by Dr. Jack McNamara, Professor of Mechanical and Aerospace Engineering. 


 

Tags: Seminar