Machine Learning Directed Study: Report 2

The report details Anson's projects in a directed study on machine learning, focusing on gathering and preparing data from 3D models of satellite assemblies sourced from GrabCAD. The models were processed using Blender to create STL files, resulting in 108 unique parts. An algorithm was developed in Julia to calculate essential properties like moments of inertia, volume, and characteristic length, which are crucial for analyzing debris. The dataset was normalized based on volume, as it produced the least variation among properties. Principal component analysis (PCA) was applied to identify key properties that capture data variation, revealing that the moments of inertia, particularly Iz, are significant. K-means clustering was utilized to categorize the inertias into six clusters, highlighting distinct shapes among the debris. The report emphasizes the need to expand the dataset with more varied data and to derive additional properties for comprehensive analysis. Future steps include acquiring actual debris scans and exploring more complex properties like aerodynamic drag. Overall, the project aims to enhance understanding of satellite debris characteristics.

This content was originally posted on my projects website here. The above summary was made by the Kagi Summarizer