Papers by Keyword: Spectral Kurtosis

Abstract: Rolling element bearing is crucial for operation safety of modern mechanical equipment. Its poor damping and heavy load capacity usually makes it degrade faster than the shafts and the gear sets. In this paper, a recently proposed time-frequency (TF) representation tool, named the complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), is introduced for the TF representation of helicopter bearing degradation data. Totally three sets of sequentially collected accelerometer data were included for the comparison study, with the first two data sets showing no signs of degradation and the third data set corresponds to totally scored bearing rollers. Facilitated by CEEMDAN method, no clear distinction can be observed on the TF plane as for the strong background noises. Thus we previously highlight the bearing signatures using spectral kurtosis (SK), a forth order statistics which is sensitive for impulsive components. Its fast algorithm, called the kurtogram, implements the detection of considerably weak bearing components and suppresses other background components, e.g., gear mesh components. With the preprocessing by kurtogram, the feature information on the TF plane can be easily characterized for the third data set.

Abstract: Aiming at the problem that traditional demodulated resonance technology has the deficiency of difficulty to choose the parameters of band-pass filter, Kalman filter technology and fast spectral kurtosis were combined for fault feature extraction of rolling bearing. AR model was firstly built with gearbox original vibration signals, and then model order was ascertained with AIC formula, and finally model parameters were calculated with least-squares method. The original signals were pretreated by Kalman filter. Fast spectral kurtosis (FSK) was used to choose parameters of the best band-pass filter, and finally fault diagnosis was achieved by the energy operator demodulation spectrum analysis of band-pass filtered signal. The analysis result of engineering signals indicated that fault feature extraction method based on Kalman filter and fast spectral kurtosis can primely provide a new feature extraction method for rolling bearing’s week fault.

Abstract: This paper presents a detection algorithm for anomaly network traffic, which is based on spectral kurtosis analysis. Firstly, we turn network traffic into time-frequency signals at different scales. These time-frequency signals hold the more detailed nature corresponding to different scales. Secondly, the time-frequency signals at different scales are transformed into a series of new time signals by time-frequency analysis theory. These new time signals hold obvious narrowband nature and embody the local properties of network traffic. Thirdly, we calculate the spectral kurtosis values of the new time signals and then perform the feature extractions. As a result, the abnormal network traffic can be correctly identified. Simulation results show that our algorithm is feasible and promising.

Abstract: This paper proposed a new method of rolling element bearing (REB) fault diagnosis for metallurgical machinery. Mainly it stresses on the combination of spectral kurtosis (SK) and supports vector machine (SVM), using genetic algorithm (GA) to optimize the parameters of support vector machine at the same time. Thus, this study aims to integrate SK, GA and SVM in order to develop an intelligent REB fault detector for metallurgical machineries. Simulation study indicates that this method can effectively detect the REB faults with a high accuracy.

Abstract: Envelope analysis based on the combination of complex Morlet wavelet and Kurtogram have advantages of automatic calculation of the center frequency and bandwidth of required band-pass filter. However, there are some drawbacks in the traditional algorithm, which include that the filter bandwidth is not -3dB bandwidth and the analysis frequency band covered by the filter-banks are inconsistent at different levels. A new algorithm is introduced in this paper. Through it, both optimal center frequency and bandwidth of band-pass filter in the envelop analysis can be obtained adaptively. Meanwhile, it ensures that the filters in the filter-banks are overlapped at the point of -3dB bandwidth and the consistency of frequency band that the filter-banks covered.

Abstract: An improved envelope analysis approach based on spectral kurtosis (SK) and complex shifted Morlet wavelet in the diagnostics of local gear faults is proposed in this paper. For enhancing impact signals of gear faults, a prewhitting process is performed to the original signals by autoregressive (AR) models firstly in the approach. Then, the envelopes of signals are extracted by Morlet wavelet filter family and the optimal center frequency and bandwidth of the band-pass filters is determined by SK. In the improved approach, not only the optimal parameters of the band-pass filters in envelope analysis can be obtained adaptively by, but also the number of the required band-pass filters can be reduced dramatically. Simulation results verified the feasibility of the present scheme positively.

Abstract: Many machine faults, such as local defects in bearings and gears, manifest themselves in vibration signals as a series of impulsive events. Kurtosis is a measure of the impulsiveness of a signal, and spectral kurtosis (SK) gives an indication of how the kurtosis (of a bandpass filtered signal) varies with frequency. This not only gives an indication of the frequency bands to be processed, but can also be used to generate a filter to extract the most impulsive part of a signal. The first step in calculating SK is to perform a time/frequency decomposition of the signal, and then calculate the kurtosis for each frequency line. The paper compares the original STFT (short time Fourier transform) with wavelet analysis for the time/frequency decomposition, and for determining the optimum combination of centre frequency and bandwidth for maximizing the SK. The paper also describes how the SK can be enhanced by “prewhitening” the signal using an autoregressive (AR) model, this sometimes revealing an incipient fault at a much earlier stage.