The problem of selection of diagnostic signs is solved by methods of statistical, spectral and entroic processing of signals and the logical analysis of their results. Results of data processing allowed to choose the most efficient characteristics (diagnostic signs) of signals providing separability of signals in the chosen priznakovy space i.e. classification. The problem of algorithmization of diagnostics of electrodrive fittings (further EPA) is solved as a problem of a pattern recognition of five classes of a condition of EPA on the basis of comparison of the current Boolean combinations to seven standards.

Keywords: electrodrive fittings, spectral analysis, entropy of Shannon, permutation entropy, basic vector, pattern recognition