APPLYING ARTIFICIAL INTELLIGENCE TECHNIQUES IN QUADRUPOLE MASS SPECTROMETER DATA ANALYSIS

The aim of the study was to demonstrate the capabilities of classification and division into clusters of measurement data of the compact quadrupole mass spectrometer MS7-200 with direct sample injection at atmospheric pressure, developed at the IAI RAS. To cluster the results of mass spectrometric measurements, the principal component analysis (to reduce the dimensionality of the obtained data) and the k-means machine training method were used. Nineteen samples of fermented milk products were used as measurement objects, divided into two groups. The first group included samples of products from farms, the second included samples of industrial products. The task was to automatically recognize the belonging of the measured sample to the group. Based on the measured mass spectrometric data of these samples, two separable clusters were constructed in the two-dimensional space of principal components, corresponding to two groups of samples