LAUSR

Rotating components are commonly used in industry and their failures may cause unexpected accidents and economic losses. To prevent these, prognostics and health management of rotating components attract much attention. The key for prognostics and health management of rotating components is to extract repetitive transients caused by early faults, which is further used to construct health indicators for describing rotating component degradation. In the past years, various theories and algorithms were developed to extract repetitive transients. One of the most famous theories is spectral kurtosis. Moreover, its fast realization is called fast Kurtogram. Even though several improvements have been made on spectral kurtosis and its fast realization, the fast Kurtogram is not an optimal filtering. In this paper, flexible Kurtogram is proposed. First, a flexible filtering framework based on multiple filter banks is constructed. Second, the objective functions of the flexible Kurtogram based on generalized spectral kurtosis are proposed. Third, optimization algorithms are introduced to maximize objective functions so as to make the flexible Kurtogram adaptive for the squared envelope with demodulation analysis. The results showed that the proposed flexible Kurtogram can be used to automatically find an optimal frequency band for demodulation analysis and it is better than the fast Kurtogram.