A support vector regression method for efficiently determining neutral profiles from laser induced fluorescence data

Dustin M Fisher; Ralph F Kelly; Deep R Patel; Mark Gilmore

doi:10.1063/1.5036649

A support vector regression method for efficiently determining neutral profiles from laser induced fluorescence data

Rev Sci Instrum. 2018 Oct;89(10):10C104. doi: 10.1063/1.5036649.

Authors

Dustin M Fisher¹, Ralph F Kelly¹, Deep R Patel¹, Mark Gilmore¹

Affiliation

¹ Electrical and Computer Engineering Department, University of New Mexico, Albuquerque, New Mexico 87131, USA.

PMID: 30399686
DOI: 10.1063/1.5036649

Abstract

A support vector regression (SVR) method is integrated with a collisional radiative (CR) model of helicon plasmas in the Helicon-Cathode (HelCat) linear plasma device to determine Ar i profiles based on metastable-pumped Laser Induced Fluorescence (LIF) measurements. A machine learning approach to the CR model allows for an efficient exploration of the input parameter space and can inherently incorporate probe and LIF measurement errors in profile inputs to which a CR model would normally be sensitive. A training set is created for mapping CR model outputs to Ar i input profiles using radial points as SVR input features and parameters of a sigmoidal-type function as output features. This SVR method may easily be adapted to other LIF pumping schemes and may even be used in conjunction with a CR model to validate electron temperature and density plasma profiles if neutral or ion profiles are already known.