Enhancing Computational Mechanics with Deep Learning

Seminar | May 13 | 4-5 p.m. | 3110 Etcheverry Hall

 Sam Raymond, Doctoral Candidate, Center for Computational Engineering, MIT

 Department of Mechanical Engineering (ME)

Abstract: Machine Learning and numerical simulation represent two distinct approaches to researchers and engineers wanting to predict and understand the behavior of complex systems. Machine Learning trusts entirely in the data to hold the key and numerical simulation relies solely on the fundamental laws of nature. While each approach has been wildly successful in solving problems in science and engineering, they both possess flaws that make extending their uses potential limited. The work presented here offers an alternative that combines these two fields into a new, hybrid approach, where synthetic data is used to train networks and create new solutions to old problems that would have been impossible solve with a single approach. This new workflow is applied to two problems in the biomechanical space. The first in building devices capable of growing human organs in the lab, the second to understand the chemo-mechanical properties of kidney stones to improve treatment approaches for physicians.

Biography: Sam Raymond is currently a Doctoral Candidate in the Center for Computational Engineering at MIT. His work spans the fields of computational mechanics, machine learning, medical engineering, material science and engineering, and fracture mechanics. Sam’s focus is on the combination of these fields, leveraging their different approaches to solve new problems. Currently Sam is working on projects such as the understanding of how surface-surface shearing across heterogenous material behaves using advanced computational, in-house developed simulation engines, using deep learning to learn the equations governing microfluidic chips for better design tools and understanding the complex mechano-chemical behavior of kidney stones in efforts to build a deep learning system to assist in the treatment process. Sam was recently awarded the Dr. Mikio Shoji Award for innovation in information technology for his recent work combining computational mechanics with deep learning.

 ggu@berkeley.edu, 510-643-4996