Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

P: Fachverband Plasmaphysik

P 6: Atmospheric Plasmas and their Applications II

P 6.2: Vortrag

Dienstag, 1. April 2025, 11:30–11:45, ZHG006

Applied machine learning for electron density measurements of an atmospheric plasma torch — •Christos Vagkidis, Alf Köhn-Seemann, Stefan Merli, Mirko Ramisch, Andreas Schulz, and Günter Tovar — IGVP, University of Stuttgart, Germany

Atmospheric plasma torches are considered a promising approach for the decomposition of waste gases. In order to enhance their performance, it is crucial to accurately measure the plasma properties. One of the most important properties of the plasma is the electron density.

In this work, a deep neural network is used to predict the electron density distribution of an atmospheric plasma torch. The neural network is trained on data obtained from 3D simulations, carried out with the COMSOL Multiphysics software. In the simulation domain, a microwave beam is propagating through the plasma and the beam power is monitored after the interaction with the plasma. A 1D cut of this power, calculated perpendicularly to the direction of propagation, is used as training data for the neural network.

Experimental data are obtained through a similar set-up. A network analyzer is used to measure the microwave beam power. By moving the detecting antenna of the network analyzer perpendicularly to the plasma torch the beam power is measured. The beam power profile is then fed into the neural network, which in turn estimates the electron density of the torch with very good accuracy.

Keywords: machine learning; comsol multiphysics; microwave diagnostics; plasma torch