The solution of problems of diagnostics of windings of electric machines is associated with the necessity of selection of quasi-periodic test signals against the background noise. In order to highlight useful signals, as a rule, the differences in spectral compositions of signals and noises are used. Ideally, the shape of the optimal filter frequency response should coincide with the shape of the spectrum of the useful signal, which determines the complexity of such a filter. The aim of the

more »

... arch is to increase the accuracy of measurements and simplify the algorithmic support of measuring systems by developing a mathematical tool that makes it possible to uniquely identify and take into account errors caused by the finiteness of the measurement intervals in the processing. Determining a one-to-one relationship between local variations of signal time parameters and alterations in its spectrum parameters is believed to be the reserve of increase of sensitivity of methods of processing of quasi-periodic signals in the conditions of constant growth of computing capabilities of measuring instruments. Variations in the values of the parameters of the signals lead to a violation of the original distribution of the harmonic components, some of the latter being subjected to the greatest alterations changes, and the some other – to the smallest ones. It is proposed to increase the accuracy of measurements due to the replacement the low-sensitivity registration of alterations in the time parameters of signals with the registration of alterations in the parameters of the characteristic harmonic components of the spectrum, which have a maximum sensitivity to deviations of the controlled parameter and a minimum sensitivity to deviations caused by the instability of the measuring equipment. The mathematical tool corresponding to the practice has been developed, that makes it possible to determine unambiguously the errors caused by finiteness of measurement intervals of quasi-periodic signals. Automatic accounting of these errors makes it possible to do without complex correlation processing of quasi-periodic signals that require large computing resources (time and speed of data processing, the amount of RAM) and to ensure the accuracy of measurements.