Jaouher Ben Ali

Application of higher order spectral features and support vector machines for bearing faults classification

Condition monitoring and fault diagnosis of rolling element bearings timely and accurately are very important to ensure the reliability of rotating machinery. This paper presents a novel pattern classification approach for bearings diagnostics, which combines the higher order spectra analysis features and support vector machine classifier. The use of non-linear features motivated by the higher order spectra has been reported to be a promising approach to analyze the non-linear and non-Gaussian characteristics of the mechanical vibration signals. The vibration bi-spectrum (third order spectrum) patterns are extracted as the feature vectors presenting different bearing faults. The extracted bi-spectrum features are subjected to principal component analysis for dimensionality reduction. These principal components were fed to support vector machine to distinguish four kinds of bearing faults covering different levels of severity for each fault type, which were measured in the experimental test bench running under different working conditions. In order to find the optimal parameters for the multi-class support vector machine model, a grid-search method in combination with 10-fold cross-validation has been used. Based on the correct classification of bearing patterns in the test set, in each fold the performance measures are computed. The average of these performance measures is computed to report the overall performance of the support vector machine classifier. In addition, in fault detection problems, the performance of a detection algorithm usually depends on the trade-off between robustness and sensitivity. The sensitivity and robustness of the proposed method are explored by running a series of experiments. A receiver operating characteristic (ROC) curve made the results more convincing. The results indicated that the proposed method can reliably identify different fault patterns of rolling element bearings based on vibration signals.

The use of SESK as a trend parameter for localized bearing fault diagnosis in induction machines.

A critical work of bearing fault diagnosis is locating the optimum frequency band that contains faulty bearing signal, which is usually buried in the noise background. Now, envelope analysis is commonly used to obtain the bearing defect harmonics from the envelope signal spectrum analysis and has shown fine results in identifying incipient failures occurring in the different parts of a bearing. However, the main step in implementing envelope analysis is to determine a frequency band that contains faulty bearing signal component with the highest signal noise level. Conventionally, the choice of the band is made by manual spectrum comparison via identifying the resonance frequency where the largest change occurred. In this paper, we present a squared envelope based spectral kurtosis method to determine optimum envelope analysis parameters including the filtering band and center frequency through a short time Fourier transform. We have verified the potential of the spectral kurtosis diagnostic strategy in performance improvements for single-defect diagnosis using real laboratory-collected vibration data sets.

The use of spectral kurtosis as a trend parameter for bearing faults diagnosis

Vibration signals are widely used in the health monitoring of rolling element bearings. A critical work of the bearing fault diagnosis is locating the optimum frequency band that contains faulty bearing signal, which is usually buried in the noise background. Now, envelope analysis is commonly used to obtain the bearing defect harmonics from the envelope signal spectrum analysis and has shown fine results in identifying incipient failures occurring in the different parts of a bearing (inner race, outer race, cage, as well as balls). However, a main step in implementing envelope analysis is to determine a frequency band that contains faulty bearing signal component with highest signal noise level. Conventionally, the choice of the band is made by manual spectrum comparison via identifying the resonance frequency where the largest change occurred. In This paper, we present a spectral kurtosis based method to determine optimum envelope analysis parameters including the filtering band and centre frequency through a short time Fourier transform. In the literature, spectral kurtosis is mainly presented as a tool used to detect non-stationary components in a signal. The results show that the maximum amplitude of the kurtogram (ways to compute the spectral kurtosis) provides the optimal parameters for band pass filter which allows both small outer race fault and large inner race fault to be determined from optimized envelope spectrum.

Bi-spectrum based-EMD applied to the non-stationary vibration signals for bearing faults diagnosis

Empirical mode decomposition (EMD) has been widely applied to analyze vibration signals behavior for bearing failures detection. Vibration signals are almost always nonstationary since bearings are inherently dynamic (e.g., speed and load condition change over time). By using EMD, the complicated non-stationary vibration signal is decomposed into a number of stationary intrinsic mode functions (IMFs) based on the local characteristic time scale of the signal. Bi-spectrum, a third-order statistic, helps to identify phase coupling effects, the bi-spectrum is theoretically zero for Gaussian noise and its flat for non-Gaussian white noise, consequently the bi-spectrum analysis is insensitive to random noise, which are useful for detecting faults in induction machines. Utilizing the advantages of EMD and bi-spectrum, this article proposes a joint method for detecting such faults, called bi-spectrum based EMD (BSEMD). First, original vibration signals collected from accelerometers are decomposed by EMD and a set of intrinsic mode functions (IMFs) is produced. Then, the IMF signals are analyzed via bi-spectrum to detect outer race bearing defects. The procedure is illustrated with the experimental bearing vibration data. The experimental results show that BSEMD techniques can effectively diagnosis bearing failures.

The use of nonlinear future reduction techniques as a trend parameter for state of health estimation of lithium-ion batteries

Remaining Useful Life (RUL) prediction accurately is an imperative industrial challenge. In this sense, the monitoring of lithium-ion battery is very significant for planning repair work and minimizing unexpected electricity outage. As the RUL estimation is essentially a problem of pattern recognition, the most valuable feature extraction techniques and more accurate classifier are needed to obtain higher prognostic effectiveness. Consequently, this paper discusses the importance of non linear feature reduction techniques for more adequate prognosis feature data base. For more convenience, the isometric feature mapping technique (ISOMAP) is used to reduce some features extracted from lithium-ion batteries, with different health states, in both modes of charge and discharge. Experimental results show that non linear feature reduction techniques are very promising to provide some trend parameters for industrial prognostic.

Application of feature reduction techniques for automatic bearing degradation assessment

Bearings are important assets for most industrial applications. The non-destructive diagnosis of these elements needs an accurate and reliable acquisition of its dynamic vibration signals affected by noise and the other part of system such as gears, shafts, etc. Empirical mode decomposition is an advanced signal processing tool for bearing fault feature extraction. In this paper, empirical mode decomposition is used to decompose non-linear and non-stationary bearing vibration signals into several stationary intrinsic mode functions and the empirical mode decomposition energy entropy is computed for each intrinsic mode function. Moreover, principal component analysis and linear discriminant analysis are used for feature reduction. Based on the Fishers criterion, experimental results show that linear discriminant analysis features are highlighted compared to principal component analysis features and original empirical mode decomposition features for bearing fault diagnosis as type (inner race, outer race, rolling element) and severity (normal, degraded, faulting).

Wind turbine high-speed shaft bearing degradation analysis for run-to-failure testing using spectral kurtosis

Premature failures of wind turbine gearboxes increase the price of energy and affect their reliability. Most gearbox failures initiate in bearings. High-speed bearings and planetary bearings exhibit a high rate of premature failure. A critical work of bearing fault diagnosis is finding the optimum frequency band that covers faulty bearing signal, which is a challenging task in practice. The kurtogram is a high technic used to characterize non-stationarities hidden in a signal. Thus, allows responding to the given problem. It consists to determine the central frequency (resonance) and the appropriate bandwidth witch maximizes the kurtosis. This paper addresses a squared envelope based spectral kurtosis method diagnosis for skidding in high-speed shaft bearings. We have verified the potential of the spectral kurtosis diagnostic strategy in performance improvements for single-defect diagnosis using real measured data from a drive train wind turbine.

Bi-spectrum analysis of coupled harmonics and its application to rotor faults diagnosis

The paper aims to clarify the use of the bi-spectrum to detect non-linearity in time series. Further we show how patterns in the bi-spectrum are useful for identifying the frequency (or bi-frequency) components involved in the nonlinear interaction. The bi-spectrum, a third-order spectrum, has properties that lend themselves to the measurement of nonlinearities in systems. The properties of interest are insensitivity to Gaussian noise and ability to detect quadratic phase coupling. This paper considers the properties of a bi-spectrum estimate when applied to a system with quadratic nonlinearity excited by the superposition of harmonics in the presence of additive Gaussian noise. This is compared, using signal-to-noise ratios, to the power spectrum. Numerical examples were included to verify the results. The study aims to expand the domain of induction machines faults diagnosis. Therefore, to verify the theoretical development, an experimental test bed has been used in a steady-state condition.

State of health estimation of lithium-ion batteries based on regression techniques

Lithium-ion batteries present the energy source for many industrial applications such as telecommunication, robotic, informatics… Consequently, the monitoring of these assets is considered as primordial to minimize unexpected electricity outage and thereby to ensure the effective operation of the used machines. In this sense, this paper presents a comparison of various regression techniques for modeling batteries degradations. Indeed, a rich discussion is introduced to enhance the benefits of regression techniques and to highlight the challenges still untreated by the literature in this important area. This paper begins with a literature review with experimental and numerical applications of regression techniques and it ends by highlighting the new industrial challenges.

Artificial neural network for blood glucose level prediction

To prevent complications in diabetes, a fitted control of blood glucose level is needed. In fact, patients have to know the future value of blood glucose in order to bring blood glucose level as close to the normal as possible. Artificial Intelligence could be more efficient to perform the prediction that could not be easily defined by mathematical algorithms. In this study, we exploit Artificial Neural Network (ANN) for an accurate blood glucose level prediction of Type 1 Diabetes (T1D). To validate the proposed method, real Continuous Glucose Monitoring (CGM) data of 12 patients were investigated. For the entire dataset, the average of the Root Mean Square Error (RMSE) (mg/dL) is 6.43, the average of the Sum of Squares of the Glucose Prediction Error SSGPE is 5.09% and the average of the Relative Error (e) is 3.71. Experimental results have shown that the proposed model is able to predict blood glucose with minimum errors and can be used to detect hyperglycemia or hypoglycemia 15 minutes in advance.