The paper considers the problem of classifying discharge and thermal defects in power transformers according to chromatographic analysis of dissolved gases, for which an expanded feature space has been formed based on concentrations of key gases and diagnostic ratios according to the International Electrotechnical Commission IEC 60599 standard. A comparison of various machine learning methods was carried out, among which the random forest algorithm showed the best results, which ensured maximum accuracy and stability of classification. The developed classifier complements the existing decision support system, providing automatic identification of the nature of defects based on chromatographic analysis of dissolved gases. The results of the study demonstrate the effectiveness of artificial intelligence methods in improving the reliability of transformer equipment diagnostics.

Keywords: power transformer, chromatographic analysis of dissolved gases, defect diagnostics, partial discharge, automated machine learning, ensemble methods, random forest, extra-trees

The results of the study of fluctuations of the center of mass of the car UAZ "Hunter" during its movement on the support surface with different height of road irregularities: asphalt-concrete surface and off-road are given. The study uses Simulink simulation.

Keywords: Center of mass fluctuations, vehicle, simulation, Simulink