EN
In this paper four processing methods of cyclic data runs, namely : Fourier interpolation and transformation, least-squaresapproximation by means of trigonometric polynomials , synchronous averaging and movable approximation , were presented by using, as an example, discretely measured runs of torque and angular speed. It was demonstrated that for extrac-ting a useful signal out of disturbance background and decomposing the disturbances the appropriate methods are the synchronous averaging and multiple movable approximation. To emerge significant spectral lines from a spectrum the least-squares approximation byusing trigonometric polynomials can be applied. The method contains a criterion for signal filtration andis insensitive to disturbances, run truncation and sampling irregularity. Fast Fourier Transform (FFT) may have an auxiliary significance for preliminary spectral analysis. Its main disadvantage is that in order tobuild a smoothing-out filter its parameters should be determined in advance, e.g. by applying one of theremaining methods