An alternative method for order tracking using autopower spectrum

Order tracking is a method of analysis used by engineers in the diagnosis of rotating machinery. In many applications, order analysis of non-stationary signals is required. The direct extraction of the amplitude information from the shorttime Fourier transform may lead to inaccurate vibration-level estimation in the case of fast changes in the signal frequency content. This article discusses spectral smearing, which is the main reason of the problem, and its sensitivity to the characteristics of the signal (frequency and amplitude variations) and to the input parameters of discrete Fourier transform analysis (window size and type). Through the years, many different approaches to perform order analysis have been developed; this article introduces a novel method for order tracking based on the short-time Fourier transform, which applies a compensation of the smearing effect based on an invariant information contained in autopower spectrum. The limitations and capabilities of the proposed method with respect to other existing techniques are discussed: considering the accuracy of the results, low requirements of computational resources, and ease of implementation, this method proves a valid alternative to currently used techniques.