PSI - Issue 32

I.O. Glot et al. / Procedia Structural Integrity 32 (2021) 216–223

219

4

Shestakov A.P./ Structural Integrity Procedia 00 (2021) 000 – 000

Fig. 4. Spectra of signals from sensors HFS1, HFS5 during lifting and dropping of ore, respectively. The frames on the right show the spectra for the frequency intervals indicated by the ovals. 4. Analysis of spectral characteristics of vibrograms One of the parameters characterizing the mechanical state of a structure is a set of natural frequencies. Determination of natural frequencies can be performed by the spectra of vibrograms. In this paper, the spectrum is calculated using the fast Fourier transform. Figure 4 shows the spectra of signals from sensors HFS1, HFS5 when lifting of ore (top) and dropping of ore into the storage (bottom). From these graphs it follows: both spectra are not smooth functions; The spectrum obtained from the HFS1 sensor shows electrical noise in multiples of 50 Hz in the form of narrow peaks. The small width of such peaks is due to the fact that electrical noise, as a rule, are constant harmonic functions that exist over the entire measurement interval. This feature allows you to eliminate electrical noise with minimal impact on the useful signal. Not smooth spectrum and the presence of electrical noise greatly complicate the process of determining the natural frequencies of the structure. Figure 4 (right), on an enlarged scale, shows the spectra in the regions corresponding to natural frequencies. Obviously, without additional processing, the determination of the frequency from the spectrum maxima is not applicable in this case. One of the options for solving the problem associated with non-smoothness of the spectrum is the use of filtering methods, such as a moving average. However, the application of these methods also leads to smoothing of signals caused by electrical noise. These noises begin to distort the spectrum more and their elimination becomes more difficult. In addition, the smoothing methods require setting the interval over which the function is averaged. The final shape of the spectrum and, as a consequence, the number and value of natural frequencies depends on the choice of this value.

Fig. 5. Constant harmonic signal (left) and its spectrum (right).