Papers by Author: Guang Bin Wang

Abstract: This thesis constructs the dynamical model of the imbalance-misalignment coupling faults and the finite element model of the rotor system which are supported by rolling element bearing. It analyses the impacts from the coupling faults to the system on the basis of nonlinear finite element method, dynamic theory and Newmark-beta numerical integration method. It also studies the influence of the unbalance, misalignment and coupling faults to the system by applying the dynamic response chart and time-frequency properties. The study shows that there exist unstable high and low harmonic components, the unbalanced signal overshadowed by misalignment. It also discovers that besides the working frequency, there also exist tow times frequency and other high doubling components on the response spectra with two times frequency for the most. All those study results provide some theoretical reference for the fault diagnosing of the rotor bearing system, the vibration control and the stability research.

Abstract: Characteristic signals in rotating machinery fault diagnosis with the issues of complex and difficult to deal with, while the use of non-linear manifold learning method can effectively extract low-dimensional manifold characteristics embedded in the high-dimensional non-linear data. It greatly maintains the overall geometric structure of the signals and improves the efficiency and reliability of the rotating machinery fault diagnosis. According to the development prospects of manifold learning, this paper describes four classical manifold learning methods and each advantages and disadvantages. It reviews the research status and application of fault diagnosis based on manifold learning, as well as future direction of researches in the field of manifold learning fault diagnosis.

Abstract: In acoustic emission (AE) detection technique, to avoid the serious noise disturbance in the fault diagnosis of rotary machine, a de-noising method based on adaptive wavelet correlation analysis to be applied to the AE signal is proposed. First, AE signals are decomposed by dyadic wavelet transform and at the same time the AE signal is divided into available coefficients and noise coefficients. Secondly, the available coefficients are reconstructed to restore the original real signal after de-noising process. Finally, the de-noising threshold is set by adaptive threshold method based on wavelet entropy. On the simulation of AE signal and the bearing fault measured AE signal using wavelet entropy correlation de-noising method, the traditional wavelet de-noising method and the traditional lifting wavelet de-noising method three kinds of de-noising methods are compared, the results show that the wavelet entropy correlation de-noising method can greatly improve the rolling bearing AE signal de-noising effect.

Abstract: In signal denoise method to nonlinear time series based on principle manifold learning, reduction target dimension is chosen at random, which cause low efficiency. Local low dimensional manifold is obtained by the eigenvalue decomposition to the covariance matrix, but covariance belongs to the second order statistics and cannot reflect the nonlinear essential structure of signal, these reduce denoise efficiency and effect. In order to solve these problem, a new denoise algorithm based on the higher order cumulant and local tangent space mean reconstruction is proposed in this reserch. First, the signal's intrinsic dimension is obtained as dimension of reduction targets by maximum likelihood estimation. And then making use of restraining character to colored noise of high order cumulan,covariance matrix is constructed by high order cumulant function instead of second order moment function. The data outside intrinsic dimension space will be regarded as noise signal to be eliminated. Finanly the process of global array by affine transformation will be replaced by mean reconstruction,the data after denoise may be obtained in the inverse process of the phase space reconstruction. The effectiveness of the algorithm is verified through the denoise experiment in fan vibration signal with noise.

Abstract: In the rolling process, serious deviation will cause product quality drop and rolling equipment fault. This reserch propose tail deviation’s predictive control method of the tandem rolling strip based on manifold learning. Based on real deviation data in the rolling production site,tail deviation patterns are divided according to deviation’s value. Using manifold learning method to deviation data in middle rolling stage , tail deviation pattern and scope are obtained. According to regression model between the control variable and deviation, predictive control strategy of the tandem rolling strip may be implemented. Experiment shows this method may control tail deviation in preconcerted permission range.

Abstract: In the rolling process, deviation is the phenomenon that the strap width direction's centerline deviates from rolling system setting centerline,serious deviation will cause product quality drop and rolling equipment fault. This paper has established the finite element model to the hot tandem rolling aluminum strap, analyzed the strap’s deviation rule under four kinds of incentives,obtained the neural network predictive model and the control policy of the tail deviation.The result to analyze a set of fact deviation data shows this method may control tail deviation in preconcerted permission range.

Abstract: In order to better identify the fault of bearing,one new fualt diagnosis method based on supervised Linear local tangent space alignment (SLLTSA) and support vector machine (SVM) is proposed..In this methd, the supervised learning is embedded into the linear local tangent space alignment algorithm,making full use of experience category information for fault feature extraction, and then using linear transformation matrix to fast process the new monitoring data, finally distinguishing fault status of the incremental data by nonlinear SVM algorithm. The experiment result for roller bearing fault diagnosis shows that SLLTSA-SVM method has better diagnosis effect than related unsupervised methods.

Abstract: To enhance the effect of fault diagnosis, a new fualt diagnosis method based on supervised incremental local tangent space alignment (SILTSA) and support vector machine (SVM) is proposed. The supervised learning approach is embedded into the incremental local tangent space alignment algorithm, to realize fault feature extraction and new data processing for equipment fault signal, and then correctly classify the faults by non-linear support vector machines. The experiment result for roller bearing fault diagnosis shows that SILTSA-SVM method has better diagnosis effect to related methods

Abstract: In the noise reduction algorithm based on manifold learning, phase space data may be distorted and reduction targets are chosen at random, it made efficiency and effect of noise reduction lower.To solve this problem, a improved noise reducation method (local tangent space mean reconstruction) was proposed.The process of global array by affine transformation will be replaced with mean reconstruction,and the intrinsic dimension was estimate as dimension of reduction targets by using maximum likehood estimation method, the data in addition to intrinsic dimension space will be eliminated.Noise reduction experiment to fan vibration signal with noise shows this method had better noise reduction effect.

Abstract: Locally linear embedding (LLE) algorithm is an unsupervised technique recently proposed for nonlinear dimension reduction. In this paper,LLE manifold learning algorithm is introduced into the field of equipment fault diagnosis firstly, a method of the fault diagnosis based on LLE_KFDA is proposed. By LLE algorithm, original sample data is directly mapped to its’ intrinsical dimension space,which data still keep primary nonlinear form. then via kernel fisher discriminant analysis(KFDA), the characteristics data in intrinsical dimension space are mapped into knernel high-dimensional linear space,and then different fault data are discriminated based on a criterion of between-class and insid-class deviatione ratio maximum. LLE_KFDA algorithm is based on original data, avoided from fall of pattern recognition ability which caused by inappropriate or blind choice of the feature parameters in the traditional fault diagnosis method.The experiment to fault diagnosis of rolling bearing shows this method can effectively identify the equipment fault pattern, diagnostic result is good.