The prefailure detection of faults in operating plant can effect major rewards in both safety and economy. A successful on-condition maintenance philosophy would pay great dividends particularly in the offshore oil industry where -until recently, only token methods have been employed. Many techniques are available for monitoring mechanical plant and several of these are considered in general terms. Industrial methods are subsequently evaluated on reciprocating compressor and rolling element bearing faults. Bearing fault analysis is considered in two stages. Initially, a series of vibration based techniques are evaluated on a large relatively noise free rotating machine. The techniques of greatest worth carrier spectra, autospectra, time signature analysis and statistical assessments - are then applied to bearings in the hostile environment of a reciprocating machine. It is shown that while discrete faults often produce predictable periodic vibrational patterns, a monitoring system aimed solely at such vibrational phenomena cannot be relied upon. To this end, a diagnostic system must encompass a series of techniques, including carrier spectrum, time signature and statistical analyses. A series of valve and piston faults in reciprocating machines are also studied. By using a number of monitoring techniques, a catalogue of fault characteristics is constructed, and the methods of greatest worth are high-lighted. It is noted that due to the complexities of a reciprocating machine, fault characteristics vary with load, and this must be borne in mind when interpreting the various parameter displays. No single technique can provide a complete cover for all compressor faults, and it is shown that those of greatest worth are acoustic emission, combined pressure and vibration plots, temperature and performance analysis. An indication of compressor temperature and internal cylinder pressure can greatly ease the detection and diagnostic process, and for the latter, bolt load determinations may be a valuable aid.

Abstract A powerful tool for bearing time series feature extraction and classification is introduced that is computationally inexpensive, easy to implement and suitable for real-time applications. In this paper the proposed technique is applied to two rolling element bearing time series classification problems and shown that in some cases no data pre-processing, artificial neural network or nearest neighbour approaches are required. From the results obtained it is clear that for the specific applications considered, the proposed method performed as well as or better than alternative approaches based on conventional feature extraction.

Rolling element bearings are one of the most significant elements and frequently-used components in mechanical systems. Bearing fault detection and diagnosis is important for preventing productivity loss and averting catastrophic failures of mechanical systems. In industrial applications, bearing life is often difficult to predict due to different application conditions, load and speed variations, as well as maintenance practices. Therefore, reliable fault detection is necessary to ensure productive and safe operations. Vibration analysis is the most widely used method for detection and diagnosis of bearing malfunctions. A measured vibration signal from a sensor is often contaminated by noise and vibration interference components. Over the years, many methods have been developed to reveal fault signatures, and remove noise and vibration interference components. Though many vibration based methods have been proposed in the literature, the high frequency resonance (HFR) technique is one of a very few methods have received certain industrial acceptance. However, the effectiveness of the HFR methods depends, to a great extent, on some parameters such as bandwidth and centre frequency of the fault excited resonance, and window length. Proper selection these parameters is often a knowledge-demanding and time-consuming process. In particular, the filter designed based on the improperly selected bandwidth and center frequency of the fault excited resonance can filter out the true fault information and mislead the detection/diagnosis decisions. In addition, even if these parameters can be selected properly at beginning of each process, they may become invalid in a time-varying environment after a certain period of time. Hence, they may have to be re-calculated and updated, which is again a time-consuming and error-prone process. This undermines the practical significance of the above methods for online monitoring of bearing conditions. To overcome the shortcomings of existing methods, the following four non-parametric and non-filtering methods are proposed: 1. An amplitude demodulation differentiation (ADD) method, 2. A calculus enhanced energy operator (CEEO) method, 3. A higher order analytic energy operator (HO_AEO) approach, and 4. A higher order energy operator fusion (HOEO_F) technique. The proposed methods have been evaluated using both simulated and experimental data.

Bearing failure is one of the foremost causes of breakdown in rotating machinery. Such failure can be catastrophic and can result in costly downtime. Bearing condition monitoring has thus played an important role in machine maintenance. In condition monitoring, the observed signal at a measurement point is often corrupted by extraneous noise during the transmission process. It is important to detect incipient faults in advance before catastrophic failure occurs. In condition monitoring, the early detection of incipient bearing signal is often made difficult due to its corruption by background vibration (noise). Numerous advanced signal processing techniques have been developed to detect defective bearing signals but with varying degree of success because they require a high Signal to Noise Ratio (SNR), and the fault components need to be larger than the background noise. Vibration analyses in the time and frequency domains are commonly used to detect machinery failure, but these methods require a relatively high SNR. Hence, it is essential to minimize the noise component in the observed signal before post processing is conducted. In this research, detection of failure in rolling element bearing faults by vibration analysis is investigated. The expected time intervals between the impacts of faulty bearing components signals are analysed using the blind deconvolution technique as a feature extraction technique to recover the source signal. Blind deconvolution refers to the process of learning the inverse of an unknown channel and applying it to the observed signal to recover the source signal of a damaged bearing. The estimation time period between the impacts is improved by using the technique and consequently provides a better approach to identify a damaged bearing. The procedure to obtain the optimum inverse equalizer filter is addressed to provide the filter parameters for the blind deconvolution process. The efficiency and robustness of the proposed algorithm is assessed initially using different kinds of corrupting noises. The result show that the proposed algorithm works well with simulated corrupting periodic noises. This research also shows that blind deconvolution behaves as a notch filter to remove the noise components. This research involves the application of blind deconvolution technique with optimum equalizer design for improving the SNR for the detection of damaged rolling element bearings. The filter length of the blind equalizer needs to be adjusted continuously due to different operating conditions, size and structure of the machines. To determine the optimum filter length a simulation test was conducted with a pre-recorded bearing signal (source) and corrupted with varying magnitude noise. From the output, the modified Crest Factor (CF) and Arithmetic Mean (AM) of the recovered signal can be plotted versus the filter length. The optimum filter length can be selected by observation when the plot converges close to the pre-determined source feature value. The filter length is selected based on the CF and AM plots, and these values are stored in a data training set for optimum determination of filter length using neural network. A pre-trained neural network is designed to train the behaviour of the system to target the optimum filter length. The performance of the blind deconvolution technique was assessed based on kurtosis values. The capability of blind deconvolution with optimum filter length developed from the simulation studies was further applied in a life bearing test rig. In this research, life time testing is also conducted to gauge the performance of the blind deconvolution technique in detecting a growing potential failure of a new bearing which is eventually run to failure. Results from unseeded new bearing tests are different, because seeded defects have certain defect characteristic frequencies which can be used to track a specific damaged frequency component. In this test, the test bearing was set to operate continuously until failures occurred. The proposed technique was then applied to monitor the condition of the test bearing and a trend of the bearing life was established. The results revealed the superiority of the technique in identifying the periodic components of the bearing before final break-down of the test bearing. The results show that the proposed technique with optimum filter length does improve the SNR of the deconvolved signal and can be used for automatic feature extraction and fault classification. This technique has potential for use in machine diagnostics.

Rolling element bearings are one of the most significant elements and frequently-used components in mechanical systems. Bearing fault detection and diagnosis is important for preventing productivity loss and averting catastrophic failures of mechanical systems. In industrial applications, bearing life is often difficult to predict due to different application conditions, load and speed variations, as well as maintenance practices. Therefore, reliable fault detection is necessary to ensure productive and safe operations. Vibration analysis is the most widely used method for detection and diagnosis of bearing malfunctions. A measured vibration signal from a sensor is often contaminated by noise and vibration interference components. Over the years, many methods have been developed to reveal fault signatures, and remove noise and vibration interference components. Though many vibration based methods have been proposed in the literature, the high frequency resonance (HFR) technique is one of a very few methods have received certain industrial acceptance. However, the effectiveness of the HFR methods depends, to a great extent, on some parameters such as bandwidth and centre frequency of the fault excited resonance, and window length. Proper selection these parameters is often a knowledge-demanding and time-consuming process. In particular, the filter designed based on the improperly selected bandwidth and center frequency of the fault excited resonance can filter out the true fault information and mislead the detection/diagnosis decisions. In addition, even if these parameters can be selected properly at beginning of each process, they may become invalid in a time-varying environment after a certain period of time. Hence, they may have to be re-calculated and updated, which is again a time-consuming and error-prone process. This undermines the practical significance of the above methods for online monitoring of bearing conditions. To overcome the shortcomings of existing methods, the following four non-parametric and non-filtering methods are proposed: 1. An amplitude demodulation differentiation (ADD) method, 2. A calculus enhanced energy operator (CEEO) method, 3. A higher order analytic energy operator (HO_AEO) approach, and 4. A higher order energy operator fusion (HOEO_F) technique. The proposed methods have been evaluated using both simulated and experimental data.

<p>Roller bearings are the common components used in the mechanical systems for mechanical processing and production. The running state of roller bearings often determines the machining accuracy and productivity on a manufacturing line. Roller bearing failure may lead to the shutdown of production lines, resulting in serious economic losses. Therefore, the research on roller bearing fault diagnosis has a great value. This thesis research first proposes a method of signal frequency spectral resampling to tackle the problem of bearing fault detection at different rotating speeds using a single speed dataset for training the network such as the one dimensional convolutional neural network (1D CNN). Second, this research work proposes a technique to connect the graph structures constructed from spectral components of the different bearing fault frequency bands into a sparse graph structure, so that the fault identification can be carried out effectively through a graph neural network in terms of the computation load and classification rate. Finally, the frequency spectral resampling method for feature extraction is validated using our self-collected datasets. The performance of the graph neural network with our proposed sparse graph structure is validated using the Case Western Reserve University (CWRU) dataset as well as our self-collected datasets. The results show that our proposed method achieves higher bearing fault classification accuracy than those recently proposed by other researchers using machine learning approaches and neural networks.</p>

Robust fault analysis (FA) including the diagnosis of faults and predicting their level of severity is necessary to optimise maintenance and improve reliability of Aircraft. Early diagnosis of faults that might occur in the supervised process renders it possible to perform important preventative actions. The proposed diagnostic models were validated in two experimental tests. The first test concerned a single localised and generalised roller element bearing fault in a permanent magnet brushless DC (PMBLDC) motor. Rolling element bearing defect is one of the main reasons for breakdown in electrical machines. Vibration and current are analysed under stationary and non-stationary load and speed conditions, for a variety of bearing fault severities, and for both local and global bearing faults. The second test examined the case of an unbalance rotor due to blade faults in a thruster, motor based on a permanent magnet brushed DC (PMBDC) motor. A variety of blade fault conditions were investigated, over a wide range of rotation speeds. The test used both discrete wavelet transform (DWT) to extract the useful features, and then feature reduction techniques to avoid redundant features. This reduces computation requirements and the time taken for classification by the application of an orthogonal fuzzy neighbourhood discriminant analysis (OFNDA) approach. The real time monitoring of motor operating conditions is an advanced technique that presents the real performance of the motor, so that the dynamic recurrent neural network (DRNN) proposed predicts the conditions of components and classifies the different faults under different operating conditions. The results obtained from real time simulation demonstrate the effectiveness and reliability of the proposed methodology in accurately classifying faults and predicting levels of fault severity.

Bearings are the essential components of modern rotating machines. Bearing faults can cause severe machine damages or even breakdowns. In recent years, artificial intelligence and deep learning have been successfully applied to fault detection. In this thesis, convolutional neural networks (CNN) are employed for bearing fault detection and classification. Computer simulations results demonstrate that the CNN based approach is advantageous over the conventional regression model, with an overall accuracy of 99.5%.

The objective of this research is to develop a bearing fault detection scheme for electric machines via stator current. A new method, called the stator current noise cancellation method, is proposed to separate bearing fault-related components in the stator current. This method is based on the concept of viewing all bearing-unrelated components as noise and defining the bearing detection problem as a low signal-to-noise ratio (SNR) problem. In this method, a noise cancellation algorithm based on Wiener filtering is employed to solve the problem. Furthermore, a statistical method is proposed to process the data of noise-cancelled stator current, which enables bearing conditions to be evaluated solely based on stator current measurements. A detailed theoretical analysis of the proposed methods is presented. Several online tests are also performed in this research to validate the proposed methods. It is shown in this work that a bearing fault can be detected by measuring the variation of the RMS of noise-cancelled stator current by using statistical methods such as the Statistical Process Control. In contrast to most existing current monitoring techniques, the detection methods proposed in this research are designed to detect generalized-roughness bearing faults. In addition, the information about machine parameters and bearing dimensions are not required in the implementation.

Le travail présenté dans ce mémoire a été effectué dans le cadre du projet FEDER ”Maintenance prédictive des éoliennes et maîtrise des impacts environnementaux”. Un des objectifs du projet a été de développer, dans le Poitou-Charentes, des compétences dans le domaine de l’éolien en lien avec les activités des laboratoires LIAS et LaSIE. Pour le LIAS, il a été décidé de lancer une nouvelle activité de recherche sur le diagnostic de défauts mécaniques. Le cadre du projet concernant l’éolien, les défauts localisés dans les multiplicateurs ont été privilégiés. Par ailleurs, nous avons restreint l’étude au régime stationnaire afin de simplifier l’apprentissage des différents phénomènes mis en jeu et des techniques de traitement du signal utilisées. Dans une première partie, nous avons étudié les signatures de défaut sur les signaux vibratoires. Cette phase a été facilitée par l’utilisation des données expérimentales mise à disposition par le Bearing Data Center de la Case Western Reserve - University de Cleveland. Parmi les méthodes de traitement de signal utilisées, nous avons opté pour l’analyse d’enveloppe mise en oeuvre dans les techniques de type Time Synchronous Analysis (TSA). A cette occasion, nous avons défini une procédure complète de détection de défaut que nous avons conservée tout au long de cette étude en appliquant une technique d’identification de type PNL qui nous a permis d’obtenir des résultats comparables à des méthodes haute résolution de type ESPRIT. Par la suite, nous nous sommes recentrés sur l’application éolienne en réalisant un banc d’essai original permettant d’émuler un défaut au niveau de l’accouplement de deux machines électriques. L’idée principale a été de recenser l’ensemble des signaux exploitables dans le cadre de la détection du défaut émulé et de fournir une classification entre les courants électriques, le couple mécanique et la vitesse des machines. Par ailleurs, un comparatif entre signaux mesurés et signaux estimés a été présenté. Il en ressort qu’il est possible d’obtenir un signal observé plus riche que la mesure directe en terme de composantes spectrales liées au défaut. Cette amélioration est rendue possible par une synthèse adéquate des gains d’observation qui a été obtenue après linéarisation de l’observateur étudié. En marge de l’application éolienne, le cas d’un moteur commandé vectoriellement a été abordé. L’idée a été d’exploiter les performances de la boucle de vitesse afin d’amplifier les composantes recherchées dans les courants électriques. L’ensemble de ces pistes de recherches a été testé en simulation et expérimentalement. / The work presented in this thesis was carried out under the FEDER project ”Maintenance prédictive des éoliennes et maîtrise des impacts environnementaux”. One of the project objectives was to develop, in Poitou-Charentes, expertise in the field of wind power in connection with the activities of LIAS and LaSIE laboratories. For LIAS, it was decided to launch a new research activity on the diagnosis of mechanical faults. The localized defects in gearbox were privileged. Furthermore, we restricted the study to the stationary system to simplify the learning of different phenomena involved and signal processing techniques. In the first part, we studied the fault signatures on the vibration signals. This phase was facilitated by the use of experimental data available from the Bearing Data Center of the Case Western Reserve - Cleveland University. Among the signal processing methods, we opted for envelope analysis implemented in the Synchronous Time Averaging (TSA). On this occasion, we defined a comprehensive fault detection procedure that we have maintained throughout this study by applying a NLP identification technique where we obtained similar results compared to high-resolution methods as ESPRIT. There after, we refocused on wind power applications by making an original test bench capable of emulating a fault in the coupling of two electrical machines. The main idea was to identify all usable signals in the context of emulated fault detection and to provide a classification between electric currents, mechanical torque and speed of the machines. Moreover, a comparison between measured signals and estimated ones was discussed. It shows that it is possible to get an observed signal richer than direct signal measurement in terms of spectral components related to the defect. This improvement is made possible by an appropriate synthesis of gains observer which was obtained after linearization of the studied observer. In the margin of wind application, the case of a motor controlled by vector was also discussed. The idea was to exploit the speed loop performance to amplify the fault components in electrical currents. All these researches have been tested in simulation and experimentally.