Code: Using machine learning to predict dimensions and qualify diverse part designs across multiple additive machines and materials

Published: 29 April 2022| Version 1 | DOI: 10.17632/h8tzpxkvdc.1
Contributors:
,
,
,
,

Description

This code was used for dimensional metrology of additively manufactured parts scanned using a high resolution document scanner, and for training and testing machine learning models to predict feature dimensions and part quality. Results are published in "Using machine learning to predict dimensions and qualify diverse part designs across multiple additive machines and materials," by Davis J. McGregor, Miles V. Bimrose, Chenhui Shao, Sameh Tawfick, and William P. King.

Files

Steps to reproduce

This code is for use in Python. The README.md contains detailed instructions. Running the Code - Setup a virtual environment - Open `workbench_measure.ipynb` and follow the directions there to measure the four provided example images - This notebook explains how the automated measurements are collected and creates four CSV files with raw measurement data - Data is cleaned with code that is not provided, but can be done manually - Open `workbench_analyze.ipynb` and follow the directions there to analyze the measurement data and train ML models to predict geometry, classify features, and classify parts - The example data should help researchers understand the code and functionality, but you may need to obtain a larger dataset to analyze

Institutions

University of Illinois at Urbana-Champaign

Categories

Machine Learning, Dimensional Metrology, Advanced Manufacturing

Licence