Pharmaceutical Stress vs. Strain Simulations
Motivation
The goal of the project was to predict the behavior of pharmaceutical powders under extreme pressure. This is normally done through computationally-intensive simulations.
Approach
I created a training dataset of simulation inputs and outputs with the help of German PhD student Kostas Giannis. Then, I built a neural network to predict simulation outputs in Python using the Keras Sequential API.
Results
I tried many different neural network architectures before finding one that worked well. Once I was finished, I presented the results to all of my professors at Fresno State.
See slides