FURI | Summer 2020
A Novel Use of GAN’s to Efficiently Simulate Fracture in Polymer Composites
Can a neural network be used to speed up finite element analysis in certain material simulations? The researchers have built a pipeline to generate thousands of finite element analysis maps and preprocess the data to be used by machine learning algorithms. Multiple architectures have been trained on the data and their effectiveness was measured. Preliminary data demonstrate that simple machine learning models can quickly estimate the output of finite element analysis. Research is ongoing, however, small models can produce the first estimates of finite element output which can be refined with traditional analysis. This results in a marked speedup.
Hometown: Walnut Creek, California, United States
Graduation date: Fall 2020