KPCA Denoising and its Application in Machinery Fault Diagnosis

2011 ◽  
Vol 103 ◽  
pp. 274-278 ◽  
Author(s):  
Ling Li Jiang ◽  
Zong Qun Deng ◽  
Si Wen Tang
Keyword(s):  
Time Series ◽  
Principal Component ◽  
Vibration Signal ◽  
Rolling Bearing ◽  
Kernel Principal Component Analysis ◽  
Reconstruction Method ◽  
Denoising Method ◽  
One Dimensional ◽  
Multidimensional Time Series ◽  
Machinery Fault Diagnosis

This paper proposes a kernel principal component analysis (KPCA)-based denoising method for removing the noise from vibration signal. Firstly, one-dimensional time series is expanded to multidimensional time series by the phase space reconstruction method. Then, KPCA is performed on the multidimensional time series. The first kernel principal component is the denoised signal. A rolling bearing denoising example verify the effectiveness of the proposed method

scholarly journals An Integrated Cumulative Transformation and Feature Fusion Approach for Bearing Degradation Prognostics

2018 ◽  
Vol 2018 ◽  
pp. 1-15 ◽  
Author(s):  
Lixiang Duan ◽  
Fei Zhao ◽  
Jinjiang Wang ◽  
Ning Wang ◽  
Jiwang Zhang
Keyword(s):  
Feature Fusion ◽  
Model Performance ◽  
Principal Component ◽  
Vibration Signal ◽  
Degradation Process ◽  
Rolling Bearing ◽  
Kernel Principal Component Analysis ◽  
Machine Model ◽  
Monotonic Trend ◽  
Learning Machine

Aimed at degradation prognostics of a rolling bearing, this paper proposed a novel cumulative transformation algorithm for data processing and a feature fusion technique for bearing degradation assessment. First, a cumulative transformation is presented to map the original features extracted from a vibration signal to their respective cumulative forms. The technique not only makes the extracted features show a monotonic trend but also reduces the fluctuation; such properties are more propitious to reflect the bearing degradation trend. Then, a new degradation index system is constructed, which fuses multidimensional cumulative features by kernel principal component analysis (KPCA). Finally, an extreme learning machine model based on phase space reconstruction is proposed to predict the degradation trend. The model performance is experimentally validated with a whole-life experiment of a rolling bearing. The results prove that the proposed method reflects the bearing degradation process clearly and achieves a good balance between model accuracy and complexity.

Bearing Condition Recognition Based on Kernel Principal Component Analysis and Genetic Programming

2013 ◽  
Vol 397-400 ◽  
pp. 1282-1285 ◽  
Author(s):  
Wen Bin Liu ◽  
Yu Xin He ◽  
Hua Qing Wang ◽  
Jian Feng Yang
Keyword(s):  
Principal Component Analysis ◽  
Genetic Programming ◽  
Principal Component ◽  
Vibration Signal ◽  
Rolling Bearing ◽  
Component Analysis ◽  
Kernel Principal Component Analysis ◽  
Initial Symptom ◽  
The Time Domain ◽  
Condition Recognition

In order to extract the fault feature validity in early fault diagnosis, method based on kernel principal component analysis and genetic programming (GP) is presented. The time domain features of the vibration signal are extracted and the initial symptom parameters (SP) are constructed. Then the combination to the initial SPs is carried on to optimize and build composite characteristics by GP. Through kernel principal component analysis (KPCA), the nonlinear principal component of the original characteristics is produced. Finally, the nonlinear principal components are selected as the feature subspace to classify the conditions of rolling bearing. Meanwhile, the within-class and among-class distance is introduced to compare and analyze the bearing condition recognition effect by using KPCA and GP plus KPCA separately. Experimental results show that the features extracted by kernel principal component analysis and genetic programming perform better ability in identifying the working states of the rolling bearing.

Local Geometric Projection-Based Noise Reduction for Vibration Signal Analysis in Rolling Bearings

2008 ◽  
Author(s):  
Ruqiang Yan ◽  
Robert X. Gao ◽  
Kang B. Lee ◽  
Steven E. Fick
Keyword(s):  
Time Series ◽  
Phase Space ◽  
Noise Reduction ◽  
Signal Analysis ◽  
Signal To Noise Ratio ◽  
Multifractal Spectrum ◽  
Vibration Signal ◽  
Rolling Bearings ◽  
One Dimensional ◽  
Vibration Signal Analysis

This paper presents a noise reduction technique for vibration signal analysis in rolling bearings, based on local geometric projection (LGP). LGP is a non-linear filtering technique that reconstructs one dimensional time series in a high-dimensional phase space using time-delayed coordinates, based on the Takens embedding theorem. From the neighborhood of each point in the phase space, where a neighbor is defined as a local subspace of the whole phase space, the best subspace to which the point will be orthogonally projected is identified. Since the signal subspace is formed by the most significant eigen-directions of the neighborhood, while the less significant ones define the noise subspace, the noise can be reduced by converting the points onto the subspace spanned by those significant eigen-directions back to a new, one-dimensional time series. Improvement on signal-to-noise ratio enabled by LGP is first evaluated using a chaotic system and an analytically formulated synthetic signal. Then analysis of bearing vibration signals is carried out as a case study. The LGP-based technique is shown to be effective in reducing noise and enhancing extraction of weak, defect-related features, as manifested by the multifractal spectrum from the signal.

Signal Denoising Method Based on KICA by Noise Components

2013 ◽  
Vol 329 ◽  
pp. 269-273
Author(s):  
Ling Li Jiang ◽  
Ping Li ◽  
Bo Zeng
Keyword(s):  
Independent Component Analysis ◽  
Signal Analysis ◽  
Vibration Signal ◽  
Independent Component ◽  
Signal Denoising ◽  
Denoising Method ◽  
One Dimensional ◽  
Kernel Independent Component Analysis ◽  
Multidimensional Signal

Denoising is an essential part of fault signal analysis. This paper proposes a kernel independent component analysis (KICA)-based denoising method for removing the noise from vibration signal. By introducing noise components of the observed signal, one-dimensional observed signal is extended to multi-dimensional signal. Then performing KICA on multidimensional signal, the noise in the observed signal consistent with the introduced noise will be removed that achieve the purpose of denosing. The effectiveness of the proposed method is demonstrated by the case study.

scholarly journals Signal Denoising Method Based on Adaptive Redundant Second-Generation Wavelet for Rotating Machinery Fault Diagnosis

2016 ◽  
Vol 2016 ◽  
pp. 1-10 ◽  
Author(s):  
Na Lu ◽  
Guangtao Zhang ◽  
Yuanchu Cheng ◽  
Diyi Chen
Keyword(s):  
Fault Diagnosis ◽  
Second Generation ◽  
Vibration Signal ◽  
Rotating Machinery ◽  
Support Vector ◽  
Signal Denoising ◽  
Denoising Method ◽  
Machine Method ◽  
Second Generation Wavelet ◽  
Machinery Fault Diagnosis

Vibration signal of rotating machinery is often submerged in a large amount of noise, leading to the decrease of fault diagnosis accuracy. In order to improve the denoising effect of the vibration signal, an adaptive redundant second-generation wavelet (ARSGW) denoising method is proposed. In this method, a new index for denoising result evaluation (IDRE) is constructed first. Then, the maximum value of IDRE and the genetic algorithm are taken as the optimization objective and the optimization algorithm, respectively, to search for the optimal parameters of the ARSGW. The obtained optimal redundant second-generation wavelet (RSGW) is used for vibration signal denoising. After that, features are extracted from the denoised signal and then input into the support vector machine method for fault recognition. The application result indicates that the proposed ARSGW denoising method can effectively improve the accuracy of rotating machinery fault diagnosis.

Application of Balanced Multi-Wavelet in the Fault Diagnosis of Rolling Bearing

2012 ◽  
Vol 226-228 ◽  
pp. 210-215
Author(s):  
Sui Zheng Zhang ◽  
Jian Yu Zhang ◽  
Yang Yang
Keyword(s):  
Fault Diagnosis ◽  
Vibration Signal ◽  
Rolling Bearing ◽  
Vanishing Moments ◽  
One Dimensional ◽  
Bearing Fault ◽  
Bearing Fault Diagnosis ◽  
Band Index ◽  
Different Characteristics ◽  
Better Than

Multi-wavelet has many excellent properties that single wavelet cannot satisfy simultaneously, such as symmetry, orthogonality, compact support and high vanishing moments etc. It contains several scaling functions and wavelet functions, which can make it match different characteristics of analyzed signal. Therefore, it is always used in bearing fault diagnosis. However, multi-wavelet is multi-dimensional and vibration signal is one-dimensional, so the 1-D vibration signal should be preprocessed before being decomposed with multi-wavelet. It means that the initial data need to be converted to r-dimensional data, and then is input to a tower algorithm. If preprocessing is done, multi-wavelet properties will be destroyed. Due to balanced multi-wavelet has unique properties, the preprocessing can be omitted. In this paper, a balanced multi-wavelet called CL4BAL is designed through balancing original CL4 multi-wavelet and is applied in the vibration signal processing. Comparing the frequency band index after decomposition and reconstruction of CL4BAL and CL4 multi-wavelet, it can be proved that CL4BAL is much better than that of CL4 multi-wavelet in bearing fault diagnosis.

Rolling bearing fault diagnosis of PSO–LSSVM based on CEEMD entropy fusion

2020 ◽  
Vol 44 (3) ◽  
pp. 405-418
Author(s):  
Shuzhi Gao ◽  
Tianchi Li ◽  
Yimin Zhang
Keyword(s):  
Fault Diagnosis ◽  
Roller Bearing ◽  
Vibration Signal ◽  
Rolling Bearing ◽  
Fault Classification ◽  
Kernel Principal Component Analysis ◽  
Permutation Entropy ◽  
Support Vector ◽  
Bearing Fault ◽  
Bearing Fault Diagnosis

Taking aim at the nonstationary nonlinearity of the rolling bearing vibration signal, a rolling bearing fault diagnosis method based on the entropy fusion feature of complementary ensemble empirical mode decomposition (CEEMD) is proposed in combination with information fusion theory. First, CEEMD of the vibration signal of the rolling bearing is performed. Then the signal is decomposed into the sum of several intrinsic mode functions (IMFs), and the singular entropy, energy entropy, and permutation entropy are obtained for the IMFs with fault features. Second, the feature extraction method of entropy fusion is proposed, and the three entropy data obtained are input into kernel principal component analysis (KPCA) for feature fusion and dimensionality reduction to obtain complementary features. Finally, the extracted features are imported into the particle swarm optimization (PSO) algorithm to optimize the least-squares support vector machine (LSSVM) for fault classification. Through experimental verification, the proposed method can be used for roller bearing fault diagnosis.

An Improved KPCA Method of Fault Detection Based on Wavelet Denoising

2011 ◽  
Vol 467-469 ◽  
pp. 1427-1432 ◽  
Author(s):  
Xiao Qiang Zhao ◽  
Zhan Ming Li
Keyword(s):  
Principal Component Analysis ◽  
Nonlinear Systems ◽  
Fault Detection ◽  
Principal Component ◽  
Wavelet Denoising ◽  
Kernel Principal Component Analysis ◽  
Random Disturbance ◽  
Kernel Matrix ◽  
Denoising Method ◽  
Simulation Results

For complicated nonlinear systems, the data inevitably have noise, random disturbance, Traditional kernel principal component analysis (KPCA) methods are very difficult to calculate the kernel matrix K for fault detection with large sample sets. So an improved KPCA method based on wavelet denoising is proposed. First, wavelet denoising method is used for data processing, then the improved KPCA method can reduce calculational complexity of fault detection. The proposed method is applied to the benchmark of Tennessee Eastman (TE) processes. The simulation results show that the proposed method can effectively improve the speed of fault detection.

Sign in / Sign up

Export Citation Format

Share Document