[WIP] Machine Learning to Active Flow Control – Supersonic Nozzle Aeroacoustics!

Just wanted to put this out as this work in progress evolves! This is something really cool that I hope will make some interesting impacts in the active flow control community.

In 2020, I’ve built and tested a solenoid array driver to individually toggle individually 108 solenoids. This work was published here in the AIAA Scitech 2021 conference. You can access my personal version of the paper on ResearchGate, if you don’t have an AIAA subscription. In summary, we utilized this individually addressable solenoid array to tackle the active flow control actuator placement problem, which turns out is rather difficult because flows display very non-linear behavior. We found an unintuitive pattern of actuators that improved the cost function (circulation of the vortex in the wake of a cylinder, which is proportional to drag); something we would never have guessed.

So now we are going to attempt to apply this technique to an even more difficult flow control problem: Supersonic jet noise suppression. The principle is the same, we have several channels that we can address individually, and we will use an algorithm that measures the overall sound pressure level (OASPL) of a supersonic jet and tries to minimize it by changing the actuator locations and parameters. This is how the actuator array looks like:

This is a work-in-progress post. As things progress, I’ll update this. I’m excited to see the results!

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s