What question did this study set out to answer?

The study aims to characterize argon-based atmospheric pressure plasma jets and develop machine learning models to predict plasma parameters.

April 18, 2026

Machine Learning (ML) Modelling and Characterization of Atmospheric Pressure Plasma Jet (APPJ) of Argon

Key Points

The study aims to characterize argon-based atmospheric pressure plasma jets and develop machine learning models to predict plasma parameters.
Generated plasma jet using a resonant transformer circuit in a narrow quartz capillary.
Conducted high-resolution optical emission spectroscopy measurements under controlled gas flow.
Collected extensive spectral data for statistical analysis of plasma characteristics.
Employed Boltzmann plot methodology to determine electron temperatures.
Applied a probabilistic machine learning approach with Bayesian neural networks for parameter estimation.
Observed intense neutral and ionized argon emissions, with nitrogen molecular bands in spectra.
Found a direct correlation between applied voltage and argon line intensities.
Determined electron temperatures ranging from 5400 to 5800 K.
Measured electron densities of (1.30‒1.33) × 10^17 cm–3.
Demonstrated successful predictive performance of the machine learning model with low uncertainties.

Abstract

Plasma, characterized by partially ionized gas containing reactive species, has become indispensable for advanced surface modification and biomedical applications. Argon-based non-thermal atmospheric pressure plasma jets (APPJs) enable precise material processing through controlled generation and efficient production of reactive species. This experimental and computational study characterizes an argon-based atmospheric pressure plasma jet generated in a narrow 2 mm quartz capillary. The discharge was initiated using a resonant transformer circuit. High-resolution optical emission spectroscopy measurements were performed under controlled gas flow conditions at 850–900 mbar. Extensive spectral data collection enabled detailed statistical analysis. The spectra exhibit intense neutral argon lines, ionized argon signatures, and nitrogen molecular bands from ambient air entrainment. Quantitative analysis revealed a direct correlation between applied voltage and Ar I line intensities, while gas flow rate variations produced complex non-monotonic intensity patterns. Electron temperatures of 5400–5800 K were determined using Boltzmann plot methodology. Stark broadening measurements of the 763.51 nm line yielded electron densities of (1.30‒1.33) × 1017 cm–3. A probabilistic machine learning approach employing Bayesian neural networks demonstrated outstanding predictive performance for estimating plasma parameters from operational variables, with remarkably low associated uncertainties. This integrated diagnostic framework provides new insights into plasma parameter relationships while offering practical tools for optimizing plasma-assisted surface treatments and chemical processes and the developed algorithms will be used to determine the parameters.

اسأل الذكاء الاصطناعي

Bookmark