Wavelet transformations in monitoring and diagnostics of self-oscillations of a cold rolling mill

Self-excited oscillations remain a frequent problem of cold rolling of thin strips. When self-oscillations occur, the amplitude increases by a jump of more than a hundred times, that is from 2-3 m/s2 to 200-250 m/s2. Oscillation frequencies (from 100 to 500 Hz, i.e. the number of loading cycles) can reach several thousands per a few seconds. Intense self-oscillations normally occur on thin strips with a thickness of less than 1 mm and high rolling speeds. The control of the occurrence of self-oscillations includes special systems, according to the level of vibrations or technological parameters, for example, by fluctuations in the tension of the strip. In control systems, classical methods include the decomposition of the controlled signal into a Fourier series. For stationary processes, Fourier series decomposition provides complete information about the signal structure and is sufficient to diagnose the technical condition of machines and mechanisms. As for a non-stationary non-harmonic signal, the information about the spectral composition of the diagnostic signal is insufficient, since it does not allow for determining the moment of occurrence of an undesirable process. In metallurgy and rolling production, processes are frequently not stationary. The cold rolling process is continuous, but discrete. It takes 5-6 minutes to roll one coil. The process of self-oscillation, even when using control systems, can last no more than 2-3 seconds. A high frequency of vibrations and a high level of stress can lead to the formation of cracks in the elements of the mill stands, for example, CVC plates. A signal conversion wavelet can become an alternative way of monitoring and diagnostics. If the FFT allows you to determine the presence of self-oscillations only in the developed state, the wavelet transform allows you to determine the origin of the self-oscillation process at low oscillation amplitudes. It also allows you to determine it in the absence of fluctuations by changing the signal of the technological parameters.