Tao Xinmin

A Novel Model of one-class Bearing Fault Detection using SVDD and Genetic Algorithm

In many bearing fault anomaly detection application, only positive (normal) samples are available for training purposes, other abnormal samples are difficult to be available. In order to solve these practical application problems, a novel model of one-class bearing fault detection based on SVDD and genetic algorithm is presented in this paper. The time domain statistics features are processed as inputs to SVDD for one-class (normal) recognition. Then SVDD is used to describe the normal data distribution characteristics with high data description ability. The SVDD is trained only with a subset of normal samples. This paper also analyzes the behavior of the classifier based on parameter selection and proposes a novel way based on genetic algorithm to determine the optimal threshold parameters. The hybrid one-class classification model of SVDD and genetic algorithm is determined to address the problem of difficultly collecting abnormal samples in bearing fault detection. Comparison of the performance of detection of SVDD with different kernel parameters is experimented. This hybrid approach is compared against other MLP detection techniques. The results show the relative effectiveness of the proposed classifiers in detection of the bearing condition with some concluding remarks.

Bearings Fault Diagnosis Based on HMM and Fractal Dimensions Spectrum

Fractional dimensions have wide applications in fault diagnosis fields as a nonlinear signal processing method. Especially, there are more correlation between the generalized fractal dimensions of bearings signals and fault activities. Generalized fractal dimensions spectrum is different with different bearings condition and faults. To solve the problems of detection rate decreasing due to the noise influence within certain fractal dimensions, factor of linear discrimination ability is employed as the indicator for optimizing fractal dimensions. The results indicate that the proposed approach can effectively remove the noise and improve the performance. Furthermore, In order to solve the problems of traditional classifications overfitting due to data unbalanced, the model based on HMM is proposed in this paper. HMM-based single fault detection, HMM-based single fault diagnosis models are also presented. More especially, we focus on analysis of the HMM-based single bearings fault diagnosis model in this paper. This proposed approach is compared against other approaches such as MLP detection techniques. The results show the relative effectiveness of the investigated classifiers in detection and diagnosis of the bearing condition with some concluding remarks.

A Novel Model of one-class Bearing Fault Detection using RNCS Algorithm based on HOS

A novel model of bearing fault detection based on improved real-valued negative clone selection algorithm (RNCS) is presented In this paper. In many bearing fault detection application, only positive (normal) samples are available for training purposes, Then RNCS is used to generate probabilistically a set of fault detectors that can detect any abnormalities(including faults and damages) in the behavior pattern of bearings. Faults occurring in machine elements are often related to non-linear effects which may lead to non-linearity in the machine vibration signature. HOS make it possible to analyze the structure of the output signal and to provide information related to the non-linearity within the system. The extracted HOS features matrix from original signals are transformed to SVD features which are used as inputs to RNCS for one-class (normal) recognition to address the problem of difficultly collecting abnormal samples in bearing fault detection. Comparison of the performance of detection of RNCS with different detectors numbers is experimented. This proposed approach is compared against other MTP detection techniques. The results show the relative effectiveness of the proposed classifiers in detection of the bearing condition with some concluding remarks.

One-class Bearing Fault Detection using Negative Clone Selection Algorithm

In order to solve the problems that in bearing fault detection application, only normal samples are available for training purposes, a one-class fault detection based on negative clone selection algorithm (NCSA) is investigated in this paper. NCSA with only normal samples for training is used to generate probabilistically a set of fault detectors that can detect any abnormalities in bearings. By incorporating the self-adaptive clone-mutation operator and the clone mature operator into conventional real-valued negative selection algorithm, the performance of convergence of the proposed approach is significantly improved and thus accuracy of detection is strongly enhanced. This paper analyzes the behavior of the classifier based on parameter selection and number of normal training samples. Furthermore, Comparison of the performance of detection of NCSA with different detectors numbers is also experimented. Finally, the proposed approach is compared against other detection techniques such as MLP (multi-layer perception), etc. the experiments demonstrate that the proposed approach outperforms other methods with some concluding remarks.

Bearing fault detection using SOM based on singular value spectrum

A novel fault diagnosis approach based on singular value spectrum of signals in phase-space is presented in this paper, which can effectively solve the problems of detection-time delay due to the requirement for a large number of samples to calculate those dynamic invariants, such as the largest Lyapunov exponent etc. In order to avoid the practical problems of difficultly obtaining abnormal samples in fault diagnosis applications, a diagnosis model based on SOM is proposed in this paper. The corresponding models are established according to different types of training samples to diagnose new test sample, which is determined by the samplepsilas distance to the current model. In experiments, comparison of different distance functions is conducted which indicates the Euclidean distance achieves the best performance. The performance of detectors based on singular value spectrum and singular value spectrum entropy is also compared. The results show the effectiveness of the investigated techniques. Finally, the proposed approach is compared against the detector based on the largest Lyapunov exponent and MLP techniques based on 10-order temporary signals. The results illustrate the proposed approach significantly outperforms other methods in terms of detection rate.

Bearings Fault Diagnosis based on GMM Model using Lyapunov Exponent Spectrum

The scheme of the bearings fault diagnosis based on Lyapunov exponent spectrum is investigated in this paper. During experiments, it is clearly observed that the largest Lyapunov exponent can effectively implement the bearing fault detection, however it fails to accurately separate the ball and outer vibration signals in the bearing fault diagnosis applications. In order to solve this problem, the two-dimensional Lyapunov exponents are exploited as features for subsequent classification tasks. Experiments show that the proposed approach can effectively remove the noises and improve significantly the performance. Furthermore, to deal with the problem of difficultly obtaining abnormal samples in fault diagnosis, a novel approach based on GMM and Bayesian classifier is proposed in this paper. The performances of detectors with two-dimensional Lyapunov exponents, the large Lyapunov exponent and Lyapunov exponent spectrum entropy as classification features are compared in experiments later. The results demonstrate the effectiveness and improvement of the proposed approach. Finally, a comparison with other methods such as MLP demonstrates its excellent performance with some concluding remarks.

One-class Bearings Fault Detection Model Based on Lyapunov Exponent Spectrum

In order to avoid the practical application problems, abnormal data insufficiency and unavailability, and solve the difficulty of calculating embed dimension and time tag before calculating Lyapunov exponent, an one-class Bearings fault detection model based on genetic algorithm and Lyapunov exponent spectrum is proposed in this paper. The normal training samples are used to decide reconstructed phase space. Then the signals with different conditions will be projected into RPS, the Lyapunov exponent is calculated which is classified as features. The optimum decision threshold values are determined by Genetic Algorithm. The results show that Lyapunov exponent for fault detection is more efficient than for fault diagnosis. In experiment, the performance of detector with largest Lyapunov exponent and Lyapunov exponent spectrum entropy is compared .The results evaluates the effectiveness of the proposed approach. This proposed approach is compared against MLP and other detection techniques. The results show the relative effectiveness of the proposed classifiers in detection of the bearing condition with some concluding remarks.