Problem Statement: Micro-slicing was performed on structures to study them and generate networks that create images based on a given set of input parameters.
Tools: Python, Fusion 360, OpenCV
Concepts: Micro slicing techniques, Neural Networks, Machine Learning, Image Processing.
To
July 2024
From
January 2024
Tasks:
In this project, multiple tasks were assigned, ranging from CAD modeling to the construction of neural networks, which were necessary to carry out the entire process.
Firstly, various types of microstructures were imported, and the appropriate planes were created based on analysis and observations using software known as PARAVIEW.
The images then had to be extracted using Python based on the inferences obtained on PARAVIEW.
Simulations were run to create a database of these images which were then classified based on the input parameters defined from Literature Review.
Tasks:
After obtaining a dataset of images, a GAN model (InceptionNet) was chosen to train the model and analyze the microstructures according to their respective input parameters.
This trained model was then used give an output of a sliced microstructure which was determined by the inputs given by a user.
This whole project and process helped in identifying a microstructure of any damaged component for which the image could be printed and accordingly extrapolated in PARAVIEW to obtain the complete structure in a given 2 planes.
The entire consolidated report along with the codes is attached below for reference.
