The paper is devoted to the analysis of increasing the accuracy of the spectral characteristics obtained after processing experimental data of the vibrations of a physical model of the offshore drilling platform in the experimental tank. The increase in accuracy is provided by the recognition and correction of breakdowns and errors in the course of median filtering of vibration recordings of different positions of the platform relative to the propagation of waves in the tank and with different deepening of the platform support columns. The median filters are known to belong to nonlinear transformations; therefore, their characteristics are determined by the structure of their processing procedures. In this case, these are narrow-band, random oscillations with maximum spectra at wave frequencies, at fundamental frequency of the platform oscillation and harmonics, at frequencies that are multiple to the fundamental frequency. To use median filters in processing such signals, the amplitude-frequency characteristics of the median filter were determined as a function of the ratio of the characteristic period of oscillations to the equivalent filter base, i.e. to multiplying a number of points in the filter core by the discreteness of measurements. As a result, the condition for the effective application of median filtering in the processing of narrow-band random signals has been determined, which is used in analyzing the spectra of the processes under study. There are given characteristic examples of the spectra of angular oscillations (by roll and depth) of a semisubmersible platform with different drafts. It has been shown that at maximum draft, when the model is most stable, median filtering suppresses high-frequency noise components of the vibration spectra by 5-10 times and, at the same time, the main features of the spectra are not distorted, i.e. the main energy-carrying components are reproduced without visible distortion. In the other limiting case, with minimum platform draft, changes in the high-frequency part of the spectra are negligible due to increased intensity of angular oscillations. In this case, the oscillation spectra also reproduce the main features of dynamic processes. These data indicate the selective sensitivity of the median filter to the structure of the operating procedure and possibility of improving the accuracy of reproduction of spectral characteristics by suppressing the high-frequency noise components of a signal.
USE OF MEDIAN FILTERS IN PROCESSING EXPERIMENTAL DATA ON FLUCTUATION DYNAMICS OF FLOATING STRUCTURES OF VARIOUS TYPES
