Jean-Paul Dron

Improvement of the sensitivity of the scalar indicators (crest factor, kurtosis) using a de-noising method by spectral subtraction: application to the detection of defects in ball bearings

The aim of this article is to show the interest of spectral subtraction for the improvement of the sensitivity of scalar indicators (crest factor, kurtosis) within the application of conditional maintenance by vibratory analysis on ball bearings. The case of a bearing in good conditions of use is considered; the distribution of amplitudes in the signal is of Gaussian kind. When the bearing is damaged, the appearance of spallings comes to disturb this signal, thus modifying this distribution. This modification is due to the presence of periodical impulses produced each time a rolling element meets a discontinuity on its way. Nevertheless, the presence of background noise induced by random impulse excitations can have an influence on the values of these temporal indicators. The de-noising of these signals by spectral subtraction in different frequency bands allows to improve the sensitivity of these indicators and to increase the reliability of the diagnosis.

Early detection of fatigue damage on rolling element bearings using adapted wavelet

Among the advanced techniques of the predictive maintenance, the vibratory analysis proves to be very effective, in particular, for monitoring rotating components such as the bearings. Their damage creates cyclic efforts which are at the origin of the processing of vibratory measurements. This processing can be made by temporal methods, frequential methods, or by time-scale methods using the wavelets for 2 decades. The wavelet transform is a very effective processing, however, the difficulties of application and interpretation of the results slow down their employment. The determination of the parameters of the wavelets makes its use all the more difficult. Moreover, the use of these time-scale methods is very expensive in time computation. This paper proposes a wavelet adapted to the mechanical shock response of a structure with n degrees of freedom. In addition, we developed a procedure for analysis of signals by this wavelet which makes it possible to accelerate the process and to improve detection in the case of disturbed signals. This methodology is compared with the traditional time-scale methods and is implemented to detect defects of different sizes on outer rings and inner rings of ball bearings.

Method of De-Noising By Spectral Subtraction Applied to the Detection of Rolling Bearings Defects

In this paper we aim to show the significance of spectral subtraction for the improvement of the sensitivity of scalar indicators (crest factor, kurtosis) within the application of conditional maintenance by vibratory analysis on ball bearings. If we consider the case of a bearing in good condition of use, the distribution of the amplitudes in the signal is Gaussian. When the bearing is damaged, the appearance of spallings disturbs this signal, modifying this distribution. This modification goes through the presence of periodical impulses produced each time a rolling element meets a discontinuity on its way. Nevertheless, the presence of background noise induced by random impulse excitations can have an influence on the values of these temporal indicators. The de-noising of these signals by spectral subtraction in different frequency bands allows us to improve the sensitivity of these indicators and to increase the reliability of the diagnosis.

On the extraction of rules in the identification of bearing defects in rotating machinery using decision tree

A methodology for the extraction of expert rules in the identification of bearing defects in rotating machinery is presented. Data sets are collected from signals measured by piezoelectric accelerometer fixed on bearings of an experimental set-up. Temporal and frequential analyses are then conducted to determine statistical parameters (crest factor (CF), kurtosis, root mean square) and spectrums (Fast Fourier Transform, envelope spectrum). The decision tree is then constructed by applying C4.5 algorithm on the dataset, and thus expert rules are established. The efficiency and applicability of expert rules over rules resulting from human experiments in rotating machinery maintenance is shown throughout the present study.

High-resolution methods in vibratory analysis : application to ball bearing monitoring and production machine

Abstract This paper is concerned with the implementation of parametric spectrum analysis using a high-resolution technique for setting up a conditional maintenance program via vibration analysis on a forming press. To achieve this, the resolving power of signal-processing-based parametric techniques is illustrated using spectrum assessment computation. Processing of the experimental results enabled (i) various autoregressive (AR) spectrum analysis methods and especially Burg’s algorithm to be tested and (ii) conventional spectrum analysis techniques such as the correlogram to be compared with parametric methods in terms of detection level as well as for mechanical component fault monitoring, especially ball bearing defects. Among various possible models, the AR model was retained along with Burg’s algorithm and the Akaike information criterion. A detection and location methodology for faults likely to occur on rotating machinery was developed on the basis of the results that were obtained. The methodology, supplementing other analysis techniques, relies on the understanding of component spectrum behavior and various constraints, such as component access, spectral resolution of the industrial measuring device, and statistical properties of the power spectral density measurements of a random signal. The results show that parametric methods are particularly worthwhile in the early detection of component defects, especially when two characteristic frequencies are close to one another. However, the complexity of these techniques necessitates many precautions when they are implemented; consequently, they should not replace conventional methods, but supplement them.

Effect of cascade methods on vibration defects detection

Within the framework of monitoring rotating machines, vibration analysis remains an effective tool for fault detection. This analysis generally consists of measuring acceleration signals from critical and judiciously chosen points of a machine with the help of piezoelectric sensors. However, relevant information concerning the machine health can be masked by disturbances such as noise. The detection reliability will then be conditioned directly by the quality of the collected signal. Signal preprocessing methods, in particular denoising methods, can significantly improve the detection quality in terms of reliability. In this paper we aim to compare two methods of denoising based on signal spectral content analysis: discrete wavelet transform and empirical mode decomposition. A first study is carried out in order to optimize specific parameters related to each of the two methods, starting from experimental data obtained on degraded bearings. In fact for each parameter, one has to define conditions which allow the best detection of periodic pulses in vibration signals thanks to indicators such as kurtosis and crest factor. The second study consists of assessing the effectiveness of each denoising method on a vibration signal measured on a failed bearing. This signal is then disturbed by various noises simulated with variable levels. This study aims to show the effectiveness of each of these two methods on the early detection of impulse defects.

Low speed bearings fault detection and size estimation using instantaneous angular speed

The fault diagnosis and prognosis of low speed machines remains a difficult problem despite remarkable advances in the conditional monitoring domain. The Rolling-element bearing is a vital part of these machines and its failure is one of the main causes of machine breakdown. In order to have an efficient maintenance strategy, fault diagnosis of a bearing and time estimation of its remaining useful life is needed. However, conventional vibration analysis at very low speeds generally fails to detect vibrations issued from a faulty bearing due to its low energy, high and variable loading conditions and to the noisy environment generated by other mechanical components of low speed machines such as gearing systems. In this work, instantaneous angular speed (IAS)-based fault diagnosis is introduced in order to compensate for the shortcoming of conventional monitoring techniques since it is strictly synchronized to shaft rotation and much less dependent on the transfer path between the defect and the sensor contrary to vibration and acoustic emission analysis. At very low speeds and in the case of a seeded spall on the bearing’s race, the shaft IAS reveals the shaft dynamical behavior when the rolling element passes into the spall. It is proven that this behavior is different when entering the spall than when exiting. The determination of entrance and exit moments allows a precise fault size estimation which is a critical step for bearing prognosis. The proposed fault size estimation method is tested on different seeded spall widths at different low speeds. The results gave a satisfactory fault width estimation and show that IAS measurement is a promising tool for the health monitoring of very low speed machines.

Early Detection of Gear Failure by Vibration Analysis

Gear mechanisms are an important element in a variety of mechanical systems, such as industrial machinery and automotive. Health monitoring of rotating machines is important to avoid failure of the system in advance. Principally, this paper consists of two parts: in the first part, a gear dynamic model including localized tooth defect has been developed. The model consists of a spur gear pair, two inertias. The model incorporates the effects of time-varying mesh stiffness and damping, excitation due to gear errors. The results of a dynamic modeling of the gears transmission are calculated by using the Newmark integration scheme. The second part consists of signal processing of simulated and experimental signals using the wavelet transform. It is shown that the kurtosis of the vibration signal is a sensitive indicator of the existence of damage in the gear pair.

Inverse Approach to the Reconstruction and Quantification of Vibratory Sources

Vibratory analysis allows us to interpret the fundamental conditions of rotating machines. This interpretation is useful in the diagnosis of faults. In practice, the vibratory measurements made by sensors come from a mixture of vibratory sources corresponding to different machine components. Thus, it becomes hard to conduct state interpretation for each individual component. This fact leads us to consider the inverse problem, reconstruction of the vibratory sources based on measurements. However, the inversion is unfortunately unstable. To limit the instability, previous studies proposed use of a large number of sensors and the processing of measurements using regularization methods. However, regularization requires estimation of the regularization parameter, which is a hard task. In addition, regularization can remove some vibratory sources. Moreover, the sensor positions are “random”, but the stability is significantly influenced by the sensor positions and by the mixture properties described by modal analysis. This article proposes to link modal analysis and stability. This link allows us to reduce the number of sensors and to avoid regularization methods. Thus, two complementary approaches are developed to determine optimal sensor positions: A numerical approach with a numerical updated model and an experimental approach using modal analysis. Thanks to the use of these optimal positions, stability is improved and vibratory sources are correctly reconstructed.

Optimisation de la détection de defauts de roulements par débruitage des signaux par soustraction spectrale

The aim of this article is to show the interest of the spectral subtraction for the improvement of the sensitivity of the scalar indicators (crest factor, kurtosis) within the application of a conditional maintenance by vibratory analysis on ball bearings. The case of a bearing in good condition of use is considered; the distribution of the amplitudes in the signal is a Gaussian kind. When the bearing is damaged, the appearance of spallings comes to disturb this signal, modifying this distribution. This modification goes through the presence of periodical impulses produced each time a rolling element meets a discontinuity on its way. Nevertheless, the presence of background noise induced by random impulse excitations can have an influence on the values of these temporal indicators. The de-noising of these signals by spectral subtraction in different frequency bands allows one to improve the sensitivity of these indicators and to increase the reliability of the diagnosis.