Detection of weak transient signals based on wavelet packet transform and manifold learning for rolling element bearing fault diagnosis

Wang, Yi; Xu, Guanghua; Li, Lin; Jiang, Kuosheng

doi:10.1016/j.ymssp.2014.09.002

Cited by 208 publications

(117 citation statements)

References 43 publications

Supporting

Mentioning

117

Contrasting

Order By: Relevance

“…Numerous techniques have been developed in the time domain [2], the frequency domain [3,4], and timefrequency domain [5,6]. Artificial intelligent techniques, fuzzy inference [7], neurofuzzy [8], and ART-Kohonen neural network (ART-KNN) [9], for instance, are introduced to enhance fault diagnosis.…”

Section: Introductionmentioning

confidence: 99%

“…However, the feature of rolling bearing early fault is very weak and is often buried in the strong background noises. Signal modulation effect and noise are two major barriers in incipient defect detection for bearing fault diagnosis [5]. The success of the envelope analysis technique highly depends on the selection of the center frequency and bandwidth of the band-pass filter used for demodulation [12].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A Novel Approach of Impulsive Signal Extraction for Early Fault Detection of Rolling Element Bearing

Lin

Shangfei

et al. 2017

Shock and Vibration

View full text Add to dashboard Cite

The fault signals of rolling element bearing are often characterized by the presence of periodic impulses, which are modulated high-frequency harmonic components. The features of early fault in rolling bearing are very weak, which are often masked by background noise. The impulsiveness of the vibration signal has affected the identification of characteristic frequency for the early fault detection of the bearing. In this paper, a novel approach based on morphological operators is presented for impulsive signal extraction of early fault in rolling element bearing. The combination Top-Hat (CTH) is proposed to extract the impulsive signal and enhance the impulsiveness of the bearing fault signal, and the envelope analysis is applied to reveal the fault-related signatures. The impulsive extraction performance of the proposed CTH is compared with that of finite impulse response filter (FIR) by analyzing the simulated bearing fault signals, and the result indicates that the CTH is more effective in extracting impulsive signals. The method is evaluated using real fault signals from defective bearings with early rolling element fault and early fault located on the outer race. The results show that the proposed method is able to enhance the impulsiveness of early bearing fault signals.

show abstract

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

A Novel Approach of Impulsive Signal Extraction for Early Fault Detection of Rolling Element Bearing

Lin

Shangfei

et al. 2017

Shock and Vibration

View full text Add to dashboard Cite

show abstract

“…The data processing method simplifies the computational expense and benefits the improvement of the generation performance. Some typical feature extraction methods, such as wavelet packet transform (WPT) [7][8][9][10], empirical mode decomposition (EMD) [11], time-domain statistical features (TDSF) [12,13] and independent component analysis (ICA) [14][15][16][17] have been proved to be equivalent to a large-scale matrix factorization problem (i.e., there may be still some irrelevant or redundant noise in the extracted features) [18]. In order to resolve this problem, a feature selection method could be employed to wipe off irrelevant and redundant information so that the dimension of extracted feature is reduced.…”

Section: Introductionmentioning

confidence: 99%

Representational Learning for Fault Diagnosis of Wind Turbine Equipment: A Multi-Layered Extreme Learning Machines Approach

2016

View full text Add to dashboard Cite

Reliable and quick response fault diagnosis is crucial for the wind turbine generator system (WTGS) to avoid unplanned interruption and to reduce the maintenance cost. However, the conditional data generated from WTGS operating in a tough environment is always dynamical and high-dimensional. To address these challenges, we propose a new fault diagnosis scheme which is composed of multiple extreme learning machines (ELM) in a hierarchical structure, where a forwarding list of ELM layers is concatenated and each of them is processed independently for its corresponding role. The framework enables both representational feature learning and fault classification. The multi-layered ELM based representational learning covers functions including data preprocessing, feature extraction and dimension reduction. An ELM based autoencoder is trained to generate a hidden layer output weight matrix, which is then used to transform the input dataset into a new feature representation. Compared with the traditional feature extraction methods which may empirically wipe off some "insignificant' feature information that in fact conveys certain undiscovered important knowledge, the introduced representational learning method could overcome the loss of information content. The computed output weight matrix projects the high dimensional input vector into a compressed and orthogonally weighted distribution. The last single layer of ELM is applied for fault classification. Unlike the greedy layer wise learning method adopted in back propagation based deep learning (DL), the proposed framework does not need iterative fine-tuning of parameters. To evaluate its experimental performance, comparison tests are carried out on a wind turbine generator simulator. The results show that the proposed diagnostic framework achieves the best performance among the compared approaches in terms of accuracy and efficiency in multiple faults detection of wind turbines.

show abstract

“…Traditional methods for the extraction of signals under Gaussian strong noise can be divided into three categories: time domain, frequency domain, and time-frequency domain methods [6]. Time-frequency domain methods are often appropriate for analyzing the non-stationary signal, which needs to choose the appropriate parameters; for example, the wavelet basis and decomposing level have to be chosen when using discrete wavelet transform (DWT) [5].…”

Section: Introductionmentioning

confidence: 99%

Development of a Method for Selection of Effective Singular Values in Bearing Fault Signal De-Noising

Gao

Wang

et al. 2016

Applied Sciences

View full text Add to dashboard Cite

Singular value decomposition (SVD) is a widely used and powerful tool for signal extraction under noise. Noise attenuation relies on the selection of the effective singular value because these values are significant features of the useful signal. Traditional methods of selecting effective singular values (or selecting the useful components to rebuild the faulty signal) consist of seeking the maximum peak of the differential spectrum of singular values. However, owing to the small number of selected effective singular values, these methods lead to excessive de-noised effects. In order to get a more appropriate number of effective singular values, which preserves the components of the original signal as much as possible, this paper used a difference curvature spectrum of incremental singular entropy to determine the number of effective singular values. Then the position was found where the difference of two peaks in the spectrum declines in an infinitely large degree for the first time, and this position was regarded as the boundary of singular values between noise and a useful signal. The experimental results showed that the modified methods could accurately extract the non-stationary bearing faulty signal under real background noise.

show abstract

Detection of weak transient signals based on wavelet packet transform and manifold learning for rolling element bearing fault diagnosis

Cited by 208 publications

References 43 publications

A Novel Approach of Impulsive Signal Extraction for Early Fault Detection of Rolling Element Bearing

A Novel Approach of Impulsive Signal Extraction for Early Fault Detection of Rolling Element Bearing

Representational Learning for Fault Diagnosis of Wind Turbine Equipment: A Multi-Layered Extreme Learning Machines Approach

Development of a Method for Selection of Effective Singular Values in Bearing Fault Signal De-Noising

Contact Info

Product

Resources

About