pandas

NumPy

SciPy

OpenCV

scikit-image

Matplotlib

seaborn

PyTorch

Predictive and Generative Modeling of Mixed-Dimensional Aerogels with Programmable Properties

Scanning Electron Microscopy (SEM) is an imaging technique that allows scientists/ researchers to take high magnification images of sample surfaces. This is done by shooting electrons at the surface of a sample and reading the intensity of reflections. In this project we hypothesized that the porous microstructure of aerogels can determine its mechanical, thermal, and acoustic properties.

I built this tool to extract the distribution of Aspect Ratios, Widths, and Heights exhibited within the porous microstructures of samples. Paired with the recipe and mass loading of our samples, these extracted parameters were used to train an ANN model to predict the structural parameters exhibited by any fabrication recipe. This was used to "synthesize" synthetic datapoints based on any composition within our design space. Additionally, these parameters were used to characterize our samples, for use in simulation, and predicting target properties. Within the scope of this project, we constructed a diffusion model to generate the microstructures of virtual samples. The extracted structural parameters of generated microstructures were used to validate our generation model and evaluate its accuracy.