What is it about?
Test plans for wind tunnels are often designed prior to testing which can lead to oversampling some test configurations and undersampling others. We demonstrate a method that analyzes data during a test campaign and suggests what test configuration to complete next.
Featured Image
Photo by Hermeus on Unsplash
Why is it important?
Wind tunnel testing is often expensive and time limited. Strategically choosing test configurations can decrease the amount of testing needed or increase the amount of knowledge gained from a test campaign entry. Our work showed that by combing a series of machine learning methods: autoencoders, clustering, and support vector machines, we could identify regions of flow behavior and choose sample configurations that better define them.
Perspectives
This work's primary contribution is its adaptive sampling architecture, but I've also found it useful for data mining wind tunnel data. Often, us wind tunnel experimentalists have more data than we can analyze, but the approach outlined in the paper summarizes many results in a greatly simplified way.
Zachary McDaniel
Purdue University
Read the Original
This page is a summary of: Machine Learning for Practical Real-Time Test Campaign Design, July 2025, American Institute of Aeronautics and Astronautics (AIAA),
DOI: 10.2514/6.2025-3729.
You can read the full text:
Contributors
The following have contributed to this page
