Regression-based interpretation of Langmuir probe measurements

Abstract

A new approach is presented for interpreting low level Langmuir probe measurements in terms of physical plasma parameters such as density or temperature. Instead of relying on analytic expressions as in most analyses, the method uses regressions combined with a suitably prepared solution library consisting of precomputed probe characteristics for selected plasma parameters. In machine learning language, this amounts to generating a training data set, constructing and training a model, and validating it over a domain of physical parameters of interest. This study aims at establishing the feasibility and limits of the method by using synthetic data sets that can be generated quickly from analytic approximations. The ultimate goal is to use this approach with model training on data sets constructed with detailed kinetic simulations capable of accounting for more physical processes, and more realistic geometry, than are possible with analytic models.