Dimension reduction of non-equilibrium plasma kinetic models using principal component analysis

Abstract

The chemical complexity of non-equilibrium plasmas poses a challenge for plasma modeling because of the computational load. This paper presents a dimension reduction method for such chemically complex plasmas based on principal component analysis (PCA). PCA is used to identify a low-dimensional manifold in chemical state space that is described by a small number of parameters: the principal components. Reduction is obtained since continuity equations only need to be solved for these principal components and not for all the species. Application of the presented method to a CO 2 plasma model including state-to-state vibrational kinetics of CO 2 and CO demonstrates the potential of the PCA method for dimension reduction. A manifold described by only two principal components is able to predict the CO 2 to CO conversion at varying ionization degrees very accurately.

A new methodology for the reduction of vibrational kinetics in non-equilibrium microwave plasma: application to CO2 dissociation

“…For instance, in 28 closer to the one presented here; specifically, the kinetics are described by the vibrational and translational temperatures for the dissociation and recombination regimes of N 2 + N in nonequilibrium hypersonic flows. Approaches where internal modes of molecules are divided into several groups have been reported as well [24][25][26] .…”

“…The reduction potential of the model is based on the lumping of all the asymmetric vibrational excited levels into a fictitious species, referred to as CO 2 *. This leads to a substantial simplification of the STS kinetic model, as most of the reactions come from the state-to-state interactions of vibrational levels 28 . Tables 2, 3, 4 and 5, respectively.…”

A Comprehensive Chemical Model for the Splitting of CO2 in Non‐Equilibrium Plasmas

“…The chemistry which is presented in ref . is also at the base of other works . Implementing such a large model can be challenging.…”

“…[6] is also at the base of other works. [11][12][13] Implementing such a large model can be challenging. First attempts to implement the chemistry in our own model resulted in significantly different results.…”

MG-local-PCA Method for the Reduction of a Collisional-Radiative Argon Plasma Mechanism

“…In case of global MG-PCA, the full dataset is kept as one principal cluster. Principal component analysis can also be combined with non-linear regression methods to obtain a further reduction and a better accuracy as suggested by Yang et al, 45 Biglari and Sutherland 4 and Mirgolbabaei and Echekki 27, 28 for combustion simulations, as well as by Peerenboom et al 39 for plasma flows. The potential of performing PCA in a kernel featured space has been also explored by Coussement et al 10 Although coupling PCA with non-linear regression has the potential of ensuring very strong compression, the approach strongly relies on the ability of the chosen non-linear model to reproduce the system physics.…”