عنوان مقاله [English]
Rolling element bearing are one of the most commonly used components in the rotating machines and early detection of bearing faults can prevent potential catastrophic failures. Various methods in time, frequency and frequency-frequency domains are expressed for the fault diagnoses of rolling element bearings. In this paper, a frequency-frequency domain method by means
of cyclostationary concept is used to calculate spectral correlation density (SCD) function. SCD is used to detect bearing fault signature in vibration signal of the machine. In this method, signal is assumed stationary in the cyclic periods. Using this method, three-dimensional diagram of spectral correlation density function is generated on a dual frequency axis for spectral and cyclic frequency. Using this method, the hidden cycles in the presence of noise, which cannot be seen in conventional Fourier transform, become clear. There is a difference between meaning of spectral frequency and cyclic frequency. The spectral frequency shows resonance frequency excited by periodic impacts, but cyclic frequency shows the frequency of impacts itself. SCD can be seen as a tool for generalization of the detection of amplitude-modulated signals. SCD can show both carrier and modulating frequency of the signal. The possibility of application of this method in rolling element bearing fault detection is shown in a practical example, and the results have been presented. The practical example is a bearing with inner ring fault used in a centrifugal pump. The vibration is measured by an accelerometer in high frequency band and then transferred to the computer. Necessary codes are written in the Matlab software. Despite its computational complexity, this method is preferable compared to classical methods such as envelop, since it does not need to determine the frequency band for the filter and it can show more details. The spectral correlation density function proves to be a more accurate method and provides comprehensive information about the signal.