关键词：
signals separation;adaptive line enhancer;colored noise;fault diagnosis
摘要：
The rub failures between rotating and stationary parts in rotating machinery may contribute to vibration signals measured on the machine as the colored noise component. But, for the colored noise mixed in vibration signals, ordinary FFT or autoregressive (AR) spectrum analysis methods can't provide sufficient and reasonable spectra estimates. In this paper, we exploited Adaptive Line Enhancer (ALE) technique to separate the colored noise from vibration signals in the time domain. The capability of the technique was evaluated using Simulink for some simulated vibration signals composed of three sinusoids and an additive colored noise. The results show that the technique is capable of separating accurately colored noise even at a lower level of colored noise, and tracking the level variation of colored noise in simulated signals. Further experimental tests for actual complex vibration signals will be conducted to validate the technique.
关键词：
nonlinear systems;fault diagnosis;differential algebra;characteristic set
摘要：
Differential algebra tools have been applied to study the identifiability of dynamic systems described by polynomial or rational equations, but have hardly ever been considered to check the diagnosability of nonlinear systems. The diagnosability here is given by the algebraic observability of the variable modelling the fault. In this note, the differential polynomials defining dynamical systems are considered as the generators of a differential ideal in a differential ring. The characteristic set of this ideal describes the same solution set of the original system. Its special structure allows to construct the exhaustive summary of the model used to test the diagnosability.
会议论文集名称：
Lecture Notes in Control and Information Sciences
摘要：
The ability to detect a new fault class can be a useful feature for an intelligent fault classification and diagnosis system. In this paper, we adopt two novelty detection methods, the support vector data description (SVDD) and the Parzen density estimation, to represent known fault class samples, and to detect new fault class samples. The experiments on real multi-class bearing fault data show that the SVDD can give both high identification rates for the prescribed ‘unknown’ fault samples and the known fault samples, which shows an advantage over the Parzen density estimation method in our experiments, via choosing the appropriate SVDD algorithm parameters.
关键词：
rolling element bearing;fault diagnosis;blind deconvolution;impulse signal
摘要：
A blind deconvolution algorithm was applied to extract or enhance the impulse impact signals in vibration signals of the localized defect bearings. The algorithm is based on maximizing the kurtosis value of the deconvolved signals. In the deconvolved signals collected from rolling bearings with inner race and outer race faults, the faulty impulse signals were adequately enhanced, and appeared regularly in the time domain. By taking two times the value of the standard deviation of the deconvolved signals as the threshold values, the "clean" faulty impulse line sequences were obtained using the threshold denoising. Computational results show that the mean impulse line repetition rate of each impulse sequence in the time.. domain matches the corresponding characteristic defect frequency of the bearing very well.
摘要：
A kernel Fisher discriminant (KFD) method is applied to the bearing fault diagnosis (i.e. classification of multiple fault classes). This paper deals with KFD for two multi-class fault recognition examples. One example is to recognize faults on different bearing elements; another is to recognize four different severities of the ball faults. The time-domain vibration signals of normal bearings, bearings with different faults have been used for feature extraction. The features are obtained from direct processing of the signal segments using simple preprocessing. The classification results demonstrate that KFD method is effective on the examples. Furthermore, in terms of classification performance, KFD method competes with support vector machines.
