Possibility of using machine learning methods to reconstruct solid body parameters during laser-induced desorption analysis

Abstract

The possibility of using machine learning methods for solving the inverse problem of the laser-induced desorption quadrupole mass-spectrometry (LID-QMS) diagnostic is studied. The formulation of the problem is given, and a general scheme of its solution is proposed. A test model of gas transport in a solid body is considered, which is used to construct a database of gas transport parameters in the sample. The application of the synthetic data and machine learning methods, viz. the interpolation technique, the method of K nearest neighbors, and the neural networks, for solving the LID-QMS inverse problem is investigated. The advantages and disadvantages of each approach are discussed.