Machine Learning Directed Study: Report 1

The report details the initial steps taken in a machine learning directed study focused on analyzing 3D models of orbital debris, specifically using a 6U CubeSat model sourced from GrabCad. The author processed the model's individual parts to derive essential properties using CAD software, automating the data extraction with an AutoHotkey script. The resulting dataset, although initially cluttered, was refined to include key characteristics such as mass, volume, density, area, and bounding box dimensions, leading to a manageable format for analysis in Matlab. While the initial analysis aimed to perform Principal Component Analysis (PCA), it faced challenges due to the small dataset size and high data variation, prompting the author to postpone PCA until more data can be gathered. Instead, k-means clustering was applied to the dataset, focusing on mass and volume, revealing three distinct groups but highlighting the need for more data in one cluster. The report identifies the current dataset's limitations, including a lack of material diversity and additional properties that could enhance analysis, such as aerodynamic drag. Future steps involve expanding the dataset through 3D scanning methods and revisiting PCA for more advanced analysis. The report concludes with a link to the code and raw data repository for further exploration.

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