New revolution-order transform for analysing non-stationary vibrations

Vibration analysis in all types of industry is an integral part of modern predictive maintenance programmes and fault diagnosis systems for rotating machinery. The basic technique on which they are based is the classical Fourier Transform approach, FFT However, for machinery operating in non-stationary conditions, with variable load and/or speed, the FFT approach is not directly applicable for analysis. Presently, two techniques are used: Order Tracking Spectral Analysis (OTSA) and Time-Frequency Transform (TFT). OTSA is used to avoid a spread of the spectral components of a spectrum obtained from variable speed machinery, however, it presents the disadvantage of averaging the amplitude of the spectral components during acquisition time. The TFT are three-dimensional functions that allow to visualise the frequency and amplitude variations of the spectral components. However, when the analysed vibration is composed of many spectral components and with large changes of the machine speed during measurement, they become very difficult to analyse. The present paper presents a new transform called Revolution-Order Transform, which uses the advantages of both OTSA and TFT techniques. Basically, the technique involves changing the signal sampled at constant time increments to a signal sampled at constant angular increments and then processes it using any TFT Both the classical and the new proposed method are evaluated and compared using simulated numerical signal and experimental data.