Search

Bearings are massively utilized in industries of nowadays due to their huge importance. Nevertheless, their defects can heavily affect the machines performance. Therefore, many researchers are working on bearing fault detection and classification; however, most of the works are carried out under constant speed conditions, while bearings usually operate under varying speed conditions making the task more challenging. In this paper, we propose a new method for bearing condition monitoring under time-varying speed that is able to detect the fault efficiently from the vibration signatures. First, the vibration signal is processed with the Empirical Wavelet Transform to extract the AM-FM modes. Next, time domain features are calculated from each mode. Then, the features’ set is reduced using the Cultural Clan-based optimization algorithm by removing the redundant and unimportant parameters that may mislead the classification. Finally, an ensemble learning algorithm “Random Forest” is used to train a model able to classify the fault based on the selected features. The proposed method was tested on a time-varying real dataset consisting of three different bearing health states: healthy, outer race defect, and inner race defect. The obtained results indicate the ability of our proposed method to handle the speed variability issue in bearing fault detection with high efficiency.

Bearing is a key component of satellite inertia actuators such as moment wheel assemblies (MWAs) and control moment gyros (CMGs), and its operating state is directly related to the performance and service life of satellites. However, because of the complexity of the vibration frequency components of satellite bearing assemblies and the small loading, normal running bearings normally present similar fault characteristics in long-term ground life experiments, which makes it difficult to judge the bearing fault status. There are various methods introduced for condition monitoring such as vibration analysis, temperature analysis, wear and debris analysis, image processing etc. Among this image analysis is found to be the most effective method for detection of machine faults. This paper proposes an automatic fault diagnosis method for bearings based on a DWT and CNN.

Detecting bearing faults in advance is critical for mechanical and electrical systems to prevent economic loss and safety hazards. As part of the recent interest in artificial intelligence, deep learning (DL)-based principles have gained much attention in intelligent fault diagnostics and have mainly been developed in a supervised manner. While these works have shown promising results, several technical setbacks are inherent in a supervised learning setting. Data imbalance is a critical problem as faulty data is scarce in many cases, data labeling is tedious, and unseen cases of faults cannot be detected in a supervised framework. Herein, a generative adversarial network (GAN) is proposed to achieve unsupervised bearing fault diagnostics by utilizing only the normal data. The proposed method first adopts the short-time Fourier transform (STFT) to convert the 1-D vibration signals into 2-D time-frequency representations to use as the input to our (DL) framework. Subsequently, a GAN-based latent mapping is constructed using only the normal data, and faulty signals are detected using an anomaly metric comprised of a discriminator error and an image reconstruction error. The performance of our method is verified using a classic rotating machinery dataset (Case Western Reserve bearing dataset), and the experimental results demonstrate that our method can not only detect the faults but can also cluster the faults in the latent space with high accuracy.

Rolling-element bearings are commonly used in rotary machinery. As a matter of fact, most machinery imperfections are related to bearing defects. Reliable bearing fault detection techniques are very useful in industries for predictive maintenance operations. Bearing fault detection still remains a very challenging task especially when defects occur on rotating bearing components because the fault-re-lated features could be nonstationary in nature. In this paper, the recent development of bearing fault detection and the challenges facing reliable bearing health condition monitoring will be discussed. Specifically, the paper will discuss the bearing characteristic frequency analysis, denoising to improve the signal-to-noise ratio, and advanced signal processing techniques for nonstationary signal analysis and bearing fault detection.

This article suggests that the correlation information, hidden in spatial configuration and temporal dynamic of frequencies, is an important indication for bearing health condition. To consider this information, we extend graph-modeling strategy, and introduce a bearing fault detection and diagnosis technique based on temporal-spatio graph. First, short-time periodogram is extracted from vibration signal, and, then, modeled by a temporal-spatio graph. In fault detection phase, the spectrum of temporal channel graph is used to map short-time periodogram to acquire the so-called graph-mapped spectrum (GMS). The principal frequency in resulting GMS is found highly related with the health condition of monitored bearing. Thus, any change of health condition can be detected by checking this principal frequency over time. Once a fault is detected, the spatio channel graph is fed to K-nearest neighbor classifier, coupled with a specific graph distance metric, for fault type identification. Comprehensive experiments on two benchmarking datasets along with theoretical interpretation demonstrate the superiority of proposed method over state of the arts. The proposed temporal-spatio graph provides a significant extension of existing spectrum analysis for fault detection and diagnosis.

The localized faults of bearings often produce a series of periodic impacts. However, how to extract these repetitive transients from the signals with strong noise interference is still a challenging problem. In this paper, a new method called collaborative double difference sparse regularization (CDDSR) is proposed for bearing fault detection. To be specific, the first and second-order difference matrices are integrated into sparse regularization term, and the differential sparsity and denoising effect of fault signal are enhanced by collaboration. According to the time domain impulse characteristics of fault signal, the sparsity of signal itself will also be considered. Based on the majorization-minimization (MM) algorithm, the objective optimization model can be solved quickly. Furthermore, the selection of regularization parameters is deeply studied, and an adaptive parameter selection strategy is given. The performance of CDDSR is verified through simulation analysis and two experimental cases, and the ability of CDDSR to extract fault features is further evaluated by quantitative index. Results demonstrate its superiority in eliminating noise interference and extracting periodic impulses in comparison to other state-of-the-art methods.

Bearings play a vital role in all rotating machinery, and their failure is one of the significant causes of machine breakdown leading to a profound loss of safety and property. Therefore, the failure of rolling element bearings should be detected early while the machine fault is small. This paper presents the model that detects bearing failures using the continuous wavelet transform and classifies them using a switchable normalization-based convolutional neural network (SN-CNN). State-of-the-art accuracy was achieved with the proposed model using the Case Western Reserve University (CWRU) bearing dataset, which serves as the primary dataset for validating various algorithms for bearing failure detection. Batch normalization techniques were also employed and compared to the proposed model. The spectrogram images were also used as input for further comparison. Using switchable normalization, the proposed model achieved the testing accuracy in between 99.44% and 100% for different batch sizes and datasets.

Inner race fault in bearing suspension is relatively the common fault in induction motors coupled with a gearbox, their detection is feasible by vibration monitoring of characteristic bearing frequencies. However, vibration signals have numerous drawbacks like signal background noise due to external excitation motion, sensitivity due to the installation position and their invasive measurement nature. For this reason, it is necessary to apply an extremely efficient method known as stator current signal analysis which offers significant savings and implementation advantages over traditional vibration monitoring. This paper represents a mathematical model for electromechanical systems and for rolling-element bearing faults to study the influence of mechanical defects on electrical variables (stator current). The novelty in this work involves three contributions: modelling of rolling bearing faults by external forces applied on the electromechanical system; Physical representation of rolling bearing fault allowing the modeling of the studied system functionality and, the influence of mechanical fault (inner race) in the electrical variables (stator current). Simulation results at the end of this paper demonstrate the effectiveness of the proposed mathematical model to detect gearbox’ bearing fault based on the electrical stator current signal with high sensitivity using fast Kurtogram approach.

Bearings are vital part of rotary machines. A failure of bearing has a negative impact on schedules, production operation and even human casualties. Therefore, in prior achieving fault detection and diagnosis (FDD) of bearing is ensuring the safety and reliable operation of rotating machinery systems. However, there are some challenges of the industrial FDD problems. First, there are unbalanced samples because industrial faults rarely occur. Conse-quently, the labeled data which can refer to failure information are limited in the industry and data augmentation methods are critical pre-processing be-fore training data driven models. Second, due to many learnable parameters in model and data of long sequence, both lead to time delay for FDD. There-fore, this paper proposes various data preprocessing methods and Light-Convolutional Neural Network (LCNN).

This paper proposes a method for calculating a health indicator (HI) for low-speed axial rolling element bearing (REB) health assessment by utilizing the latent representation obtained by variational inference using Variational Autoencoders (VAEs), trained on each speed reference in the dataset. Further, versatility is added by conditioning on the speed, extending the VAE to a conditional VAE (CVAE), thereby incorporating all speeds in a single model. Within the framework, the coefficients of autoregressive (AR) models are used as features. The dimensionality reduction inherent in the proposed method lowers the need of expert knowledge to design good condition indicators. Moreover, the suggested methodology allows for setting the probability of false alarms when encoding new data points to the latent variable space using the trained model. The effectiveness of the proposed method is validated based on two different datasets: from a workshop test of an offshore drilling machine and from an in-house test rig for axial bearings. In both datasets, the HI is exceeding the warning and alarm levels with a probability of false alarm (PFA) of 10 -6 , and the method is most effective at lower shaft speeds. publishedVersion

Wheelset bearings are crucial mechanical components of high-speed trains. Wheelset-bearing fault detection is of great significance to ensure the safety of high-speed train service. Convolution sparse representations (CSRs) provide an excellent framework for extracting impulse responses induced by bearing faults. However, the performance of CSR on extracting impulse responses is fairly sensitive to inappropriate selection of method-related parameters, and a convolution model for representing the impulse responses has not been discussed. In view of these two unsolved problems, a convolutional representation model of the impulse response series is developed. A novel fault detection method, named adaptive CSR (ACSR), is then proposed based on combinations of CSR and methods for estimating three parameters related to CSR. Finally, the effectiveness of the proposed ACSR method is validated via simulation, bench testing, and a real-life running test employing a high-speed train.

Motor bearings are one of the most critical components in rotating machinery. Envelope demodulation analysis has been widely used to demodulate bearing vibration signals to extract bearing defect frequency components but one of the main challenges is to accurately locate the major fault-induced frequency band with a high signal-to-noise ratio (SNR) for demodulation. Hence, an enhanced fault detection method combining the maximal overlap discrete wavelet packet transform (MODWPT) and the Teager energy adaptive spectral kurtosis (TEASK) denoising algorithms is proposed for identifying the weak periodic impulses. The Teager energy power spectrum (TEPS) defines the sparse representation of the filtered signals of the MODWPT in the frequency domain via the Teager energy operator (TEO), the TEASK helps determine the most informative frequency band for demodulation. The methodology is compared in terms of performance with the fast Kurtogram and the Autogram methods. The simulation and practical application examples have shown that the proposed MODWPT-TEASK method outperforms the above two methods in diagnosing defects of motor bearings.

Induction motor is widely used in industrial applications. The research paper presents the diagnosis of the induction motor bearing faults using Discrete Wavelet Transforms and Partitioning Around Medoids algorithm methods. The experimental test rig was developed to obtain data of the bearings on healthy or damaged conditions. Several mother-level wavelets are tried in order to get the best performance to find bearing faults. The wavelet transform results are used as an input of the Partitioning Around Medoids algorithm to cluster the bearing condition. The results showed that the methods proposed could provide an accurate diagnosis of the bearing condition.

Bearings are the major components in rotary machinery and very used in the industry. The time for bearing failures identification before interrupting operation or affecting product quality is the basis for most predictive maintenance programs. Taking readings, keeping history of failures and evaluating these results in the operation of rotating equipment on a regular basis, allows to detect possible failures before they become catastrophic. In this way, the damages or defects that are detected before a failure occurs, reduce the repair costs and the time that a rotating machine will be inactive. The bearing failures can generate losses due to machine downtime, unwanted vibration, noise and damage of other components, but if they are detected in time, repair costs and downtime are minimal. This article shows in detail the different detection and classification techniques most used to identify bearing failures such as vibration analysis, artificial neural networks (i.e ANN), convolutional neural networks (i.e CNN) and support vector machine (i.e SVM) and the relevant features of each detection technique.

A wheelset bearing is one of the crucial mechanical parts in a high-speed train. Its fault detection is of great significance, as it ensures the safety of a high-speed train service. To resolve the boundary changeability of the sidebands and false segments caused by measured noises and vibrational interferences, a novel automatic fault detection method of the wheelset bearing, named AFDMLEWT, is proposed based on the proposed multi-level empirical wavelet transform (MLEWT) and the developed segment selection method, named SSESKDB. MLEWT provides a complete framework for extracting the sidebands with different bandwidths distributed throughout different frequency positions. A novel index for evaluating the fault information capacity contained in different decomposition bandwidths, named ESKDB, is proposed. The segment selection method based on ESKDB, named SSESKDB, is proposed to select those segments of MLEWT containing the fruitful fault information in the full frequency range of wheelset bearing fault detection. Finally, the proposed method is then validated using two simulated signals and bench tests.

Conventional order tracking, which relies on a reference signal, is a common tool for rotary machinery fault diagnosis under speed fluctuation working conditions. However, it is inconvenient to install a speed sensor under certain circumstances. In this paper, we present a two-stage method to detect variable speed rolling bearing faults without a tachometer. In the first stage, the spline-kernelled chirplet transform (SCT) is applied to calculate the time-frequency distribution and extract the instantaneous rotation frequency. In the second stage, angle synchronous averaging is employed to resample the raw signal and detect fault features. The comparison investigations between SCT and other time-frequency analysis methods are performed by a numerical signal with a nonlinear instantaneous frequency. The experimental studies of variable speed rolling bearings with the outer and inner race faults are further performed. The results demonstrate that the proposed method can well detect rolling bearing faults without a key-phase signal from strong background noise under varying speed conditions.

Bearing problems are by far the biggest cause of induction motor failures in the industry. Since induction machines are used heavily by the industry, their unexpected failure may disturb the production process. Motor condition monitoring is employed widely to avoid such unexpected failures. The data that can be obtained from induction machines are non-stationary by nature since the loading may vary during their operation. Wavelet packet decomposition seems to better handle non-stationary nature of induction machines, the use of this method in monitoring applications is limited, since the computational complexity is higher than other methods. In this work four-band wavelet packet decomposition of motor vibration data is proposed to reduce the computational complexity without compromising accuracy. The proposed method is very suitable for parallel computational processing by its nature, and as a result it is predicted that the calculation time will be shortened further if field-progammable gate array is used in design.

Vibration signal analysis is an important means for bearing fault detection/diagnosis and bearings often operate under time-varying rotational speed conditions. This data article contains vibration datasets collected from bearings with different health conditions under different time-varying speed conditions. The health conditions of the bearing include healthy, faulty with an inner race defect, and faulty with an outer race defect. The operating rotational speed conditions for the dataset include increasing speed, decreasing speed, increasing then decreasing speed, and decreasing then increasing speed. Mendeley Data, http://dx.doi.org/10.17632/v43hmbwxpm.1 .

Preventive maintenance of modern electric rotating machinery (RM) is critical for ensuring reliable operation, preventing unpredicted breakdowns and avoiding costly repairs. Recently many studies investigated machine learning monitoring methods especially based on Deep Learning networks focusing mostly on detecting bearing faults; however, none of them addressed bearing fault severity classification for early fault diagnosis with high enough accuracy. 1D Convolutional Neural Networks (CNNs) have indeed achieved good performance for detecting RM bearing faults from raw vibration and current signals but did not classify fault severity. Furthermore, recent studies have demonstrated the limitation in terms of learning capability of conventional CNNs attributed to the basic underlying linear neuron model. Recently, Operational Neural Networks (ONNs) were proposed to enhance the learning capability of CNN by introducing non-linear neuron models and further heterogeneity in the network configuration. In this study, we propose 1D Self-organized ONNs (Self-ONNs) with generative neurons for bearing fault severity classification and providing continuous condition monitoring. Experimental results over the benchmark NSF/IMS bearing vibration dataset using both x-and y-axis vibration signals for inner race and rolling element faults demonstrate that the proposed 1D Self-ONNs achieve significant performance gap against the state-of-the-art (1D CNNs) with similar computational complexity. publishedVersion

With the rapid development of science and technology, the production facilities are also growing advanced. An intelligent production facility is the outcome of smart systems used within a factory. The smart factories yield more production; thus, the faults in the machinery are prompt to increase when they are operated on a daily basis and for almost all applications. Different deep learning-based methods have been used and implemented in detecting and diagnosing bearing faults using raw vibration data. To detect and analyze the machinery bearing faults, we have proposed a deep learning-based convolutional neural network, which uses the 2-D image representation of 1-D raw vibration data from the Case Western Reserve University (CWRU) bearing dataset as input. With the use of the data augmentation technique for increasing training data, the proposed model has achieved 99.38% accuracy. The proposed method is computationally less expensive and simple than most of the complex algorithms used for detecting and diagnosing the bearing faults.

Bearings are one of the most essential parts of rotary machines. The failure of bearings can lead to significant financial loss as well as personal casualties. Therefore, bearing defect diagnosis is a very important research project. Recently, a lot of bearing defect diagnosis studies using deep learning methods have been conducted. However, there are some challenges to be addressed. In a real working condition, there is always much more normal data than fault data, so a data imbalance problem exists. To address this situation, data augmentation method which generates more training data from the original data, was used. This method was done by applying a geometric transformation so that the class label did not be changed. Therefore, in this paper, we compared the results of using and without data augmentation technique through 1-D CNN and 2-D CNN deep learning algorithm that are effective on time series data analysis and pattern recognition. Finally, we obtained better results when using data augmentation technique.

Bearings play an important role in current industrial production. The fault of bearings can cause economic losses and even casualties. Therefore, it is very important for industrial production to be able to diagnose bearing faults. This paper proposes a bearing fault detection method based on convolutional neural network: the original bearing vibration data collected by the sensor is input into the convolutional neural network, and the calculation of the convolutional neural network can obtain the fault type of the bearing. In this paper, after building the model, the number of neurons in the fully connected layer is optimized and calculated on the data set of the rolling bearing data center of the Western Reserve University. The accuracy rate is as high as 99.7%.

International audience

A bearing fault diagnosis approach based on spectral kurtosis and empirical mode decomposition (EMD) is proposed. EMD is a signal decomposition technique, which can adaptively separate a number of intrinsic mode functions (IMFs) from the vibration signal according to the architectural characteristics of the data. The spectral kurtosis parameter takes as signal impulsive indicator. Firstly, EMD is utilized to process the sampling vibration signal. And then spectral kurtosis is calculated to select the optimal intrinsic mode functions, so as to suppress the noise and highlight the transient impact feature. Finally, the envelope spectrum is computed and the fault characteristic is recognized. The experimental results show that the proposed approach can identify bearing defects effectively and provide a reliable method for gearbox fault monitoring and diagnosis.

Bearings are vital part of rotary machines. A failure of bearing has a negative impact on schedules, production operation and even human casualties. Therefore, in prior achieving fault detection and diagnosis (FDD) of bearing is ensuring the safety and reliable operation of rotating machinery systems. However, there are some challenges of the industrial FDD problems. Since according to a literature review, more than half of the broken machines are caused by bearing fault. Therefore, one of the important thing is time delay should be reduced for FDD. However, due to many learnable parameters in model and data of long sequence, both lead to time delay for FDD. Therefore, this paper proposes a deep Light Convolutional Neural Network (LCNN) using one dimensional convolution neural network for FDD.

Faults in bearing are major issues for the failure of Induction Motor (IM). In case a fault occurs, it is a common practice to shut down the motor as early as possible in order to avoid catastrophic damages. However, shut down at inconvenient times results in economic losses and consumption of substantial time. Hence, timely detection of the bearing health is very important before its complete breakdown. Most popular technique used for detecting bearing faults is vibration-based technique, therefore in present work, a vibration-based simple approach is proposed to diagnose the inner and outer race faults in bearing because it is simpler and can provide early information about the progressing of malfunctions and also it doesn’t require detailed technical specifications of bearing. The proposed approach based on number of trials for inner and outer race have been performed in Jindal Steel and Power Limited, Raigarh, India. The experiment results proclaim that results are in agreement with the proposed concept and are capable of identifying the bearing faults.

Nowadays, induction machines are widely used in various sectors, their robustness allows them to operate even in fault conditions; this can lead to permanent damage to the machine and economic losses as a result of machine replacement and/or production downtime. This reflects the importance of monitoring methods or strategies to detect imminent failures. Bearing Failure can be detected by treating signals such as sound, vibration and/or current. This work is based on the comparison of the following methods: Spectral Analysis and 2 types of Goodness-of-Fit Tests; the database is obtained in the laboratory using real and controlled damage; these methods are performed using Motor Current Signature Analysis. The strategies mentioned above, as well as the comparison of their accuracy, are validated experimentally.

The behaviour of vibrations is widely used for non-intrusive inspection and health monitoring of bearings. However, automated methods, intended for predicting the bearing status, greatly depend on ...

Timely and accurate bearing fault detection and diagnosis is important for reliable and safe operation of industrial systems. In this study, performance of a generic real-time induction bearing fault diagnosis system employing compact adaptive 1D Convolutional Neural Network (CNN) classifier is extensively studied. In the literature, although many studies have developed highly accurate algorithms for detecting bearing faults, their results have generally been limited to relatively small train/test data sets. As opposed to conventional intelligent fault diagnosis systems that usually encapsulate feature extraction, feature selection and classification as distinct blocks, the proposed system takes directly raw time-series sensor data as input and it can efficiently learn optimal features with the proper training. The main advantages of the 1D CNN based approach are 1) its compact architecture configuration (rather than the complex deep architectures) which performs only 1D convolutions making it suitable for real-time fault detection and monitoring, 2) its cost effective and practical real-time hardware implementation, 3) its ability to work without any pre-determined transformation (such as FFT or DWT), hand-crafted feature extraction and feature selection, and 4) its capability to provide efficient training of the classifier with limited size of training data set and limited number of BP iterations. Effectiveness and feasibility of the 1D CNN based fault diagnosis method is validated by applying it to two commonly used benchmark real vibration data sets and comparing the results with the other competing intelligent fault diagnosis methods. Scopus

The excellent features of bearing vibration signal are gainful to make accurate diagnosis results on the type of bearing damage. In this paper, a signal analysis technique based on wavelet packet t...

The concepts of signal smoothness and signal similarity are very common in digital signal processing. In this paper, both concepts have been explored in the ball bearing failure diagnostics, where the bearing fault, bearing inner race and outer race faults were analysed. It was found out that the signal smoothness and the signal similarity estimates might reduce computational tolerance in the vibro-acoustic signal analysis. On this basis, numerous experiments have been carried out to demonstrate usefulness of the above-mentioned estimates (as useful parameters of signal features) in the signal quality analysis.

In order to realize single fault detection (SFD) from the multi-fault coupling bearing data and further research on the multi-fault situation of bearings, this paper proposes a method based on features self-extraction of a Sparse Auto-Encoder (SAE) and results fusion of improved Dempster&ndash, Shafer evidence theory (D&ndash, S). Multi-fault signal compression features of bearings were extracted by SAE on multiple vibration sensors&rsquo, data. Data sets were constructed by the extracted compression features to train the Support Vector Machine (SVM) according to the rule of single fault detection (R-SFD) this paper proposed. Fault detection results were obtained by the improved D&ndash, S evidence theory, which was implemented via correcting the 0 factor in the Basic Probability Assignment (BPA) and modifying the evidence weight by Pearson Correlation Coefficient (PCC). Extensive evaluations of the proposed method on the experiment platform datasets showed that the proposed method could realize single fault detection from multi-fault bearings. Fault detection accuracy increases as the output feature dimension of SAE increases, when the feature dimension reached 200, the average detection accuracy of the three sensors for bearing inner, outer, and ball faults achieved 87.36%, 87.86% and 84.46%, respectively. The three types&rsquo, fault detection accuracy&mdash, reached to 99.12%, 99.33% and 98.46% by the improved Dempster&ndash, Shafer evidence theory (IDS) to fuse the sensors&rsquo, results&mdash, is respectively 0.38%, 2.06% and 0.76% higher than the traditional D&ndash, S evidence theory. That indicated the effectiveness of improving the D&ndash, S evidence theory by evidence weight calculation of PCC.

This study proposes a new method for fault diagnosis in ball bearings based on wavelet transform and principal component analysis (PCA) of the acquired vibration signals. The signals collected are pre-processed using a wavelet transform to decompose the signals into low (approximated) and high (detailed) frequency part where the high-frequency part are needed for fault diagnosis purposes. Eleven potential statistical features are then extracted from the high-frequency part coming from different bearing fault signals and those from healthy bearings as well. Four types of signals are proposed, they are outer race fault, inner race fault, ball fault and no-fault signals. The PCA is used to linearly transform and reduce multidimensional data resulted from statistical extraction down to a few dimensions for more straightforward analysis. Six principal components retaining more than 95% significance level are used for bearing fault detection and classification. By combining the wavelet transform, statistical features extraction and PCA, the proposed method successfully detected and classified fault types without knowledge of a bearing fault frequencies and analysis from experienced users.

In order to promote the development of China's heavy industry economy and improve industrial production efficiency, a bearing fault detection method based on ARM is proposed. Firstly, the component standard deviation of the original features of motor bearings is obtained by combining with the ARM bearing fault detection algorithm, and the fault parameters are compared according to the classification standard deviation, so that the gears can be eliminated effectively. In order to ensure the accuracy and efficiency of motor bearing fault detection, the bearing fault detection process is optimized based on ARM bearing fault detection algorithm. Finally, the experiment proves that the fault detection method based on ARM is feasible, and the precision of motor bearing fault detection method is obviously improved compared with the traditional method.

Although promising results of high-performance AI algorithms have been reported in recent predictive maintenance researches, most of the existing studies merely deal with AI-only solutions and do not consider the interaction between humans and AI. In this study, we explicitly focus on the benefits of interactions where a human inspector is assisted by AI solutions. A case study is conducted for predictive maintenance of wind farms, where endoscopic images were used for bearing fault detection. The experiment consisted of 54 technical inspectors and 2301 images collected over 138 wind turbines, and each inspector was shown images and asked to identify bearing faults in the absence and presence of AI-assistance. The results showed that AI-assistance had a statistically significant impact on improving the technical inspector’s specificity and time efficiency. The level of improvement was dependent on the level of expertise, where the generalist group showed greater improvements in specificity and time efficiency (24.6% and 25.3%, respectively) when compared with the specialist group (4.7% and 6.4%, respectively). Both groups responded positively on the reuse intention and usefulness of AI-assistance, and the change in cognitive load was not statistically significant.

In the present study, a new bearing fault detection and recognition methodology is proposed based on complementary ensemble empirical mode decomposition method (CEEMD) and a newly developed weighted multiscale entropy method. The need for this methodology is felt due to the inability of the existing multiscale entropy methods in correctly identifying the nature of the signal, particularly in the initial scales. The implication of this drawback is strongly perceived in the experimental analysis in the present work. Vibration signals acquired from test machines/working machines have a substantial presence of noise which severely affects the consistency and reliability of the extracted features. Therefore, for effective implementation and comparing the efficiency of the proposed methods, the original signal is firstly processed with CEEMD. The processing of the signal includes its decomposition into several modes thereafter reconstructing a new signal from the modes chosen through Hurst exponent threshold analysis. From the reconstructed signals, the faulty feature vectors are extracted by the weighted multiscale entropy methods. The capabilities of the proposed method are intensively tested through simulation and experimental analysis. From the analysis of simulated signals, it is demonstrated that the drawback prevailing in the established entropy methods have strongly been mitigated by the newly developed weighted entropy methods. On the experimental front, an impressive improvement is observed by the proposed methods both qualitatively (in indicating the faulty system from the normal system) and quantitatively (in recognizing the fault type and severity by the support vector machine classifier). Apart from the analysis of vibration signals, the versatility of the proposed method is also verified on the acoustic signals acquired under similar experimental conditions.

Enhancing repetitive transients is a key to bearing fault detection. Blind deconvolution aims to recover impulsive impacts from repetitive transients and their mixture. Nowadays, based on different optimization criteria, such as kurtosis, entropy, generalized Lp/Lq norm, etc., various deconvolution methods have been proposed for recovering repetitive impacts. Due to the non-convex of these optimization criteria, local optimal solutions to these optimization problems may be obtained frequently. In this paper, inspired by the fast realization of spectral kurtosis, e.g. the fast kurtogram, an optimized minimum generalized Lp/Lq deconvolution (OMGD) is proposed and investigated. The main idea of the proposed method utilizes the filters designed in the fast kurtogram to provide proper initializations for the minimum generalized Lp/Lq deconvolution (MGD). Hence, multiple local optimal solutions can be iteratively obtained. Then, the minimum of these local optimal solutions is used as a precise estimate to approximate the globally optimal solution of MGD. Results show that the proposed method has better deconvolution performance than MGD and fast kurtogram for enhancement of repetitive impacts caused by localized bearing faults.

Robust identification of bearing health states is closely linked to timely condition monitoring and downtime reducing for rotating machinery. Although many proposed algorithms achieve extraordinary performances on feature extraction, uncertainty still remains for the bearing fault identification. To address this problem, this paper introduces a two-stage framework for bearing fault detection using graph similarity evaluation. The recognition stage is used to identify the operation state (fault or not) based on an improved graph-based method according to the sampled vibration signal for each spindle turn. The feature extraction stage, on the other hand, is implemented to extract the fault characters from the time-domain signals. The results indicate that the proposed method achieves 100% identification accuracy in bearing fault detection even with phase shifts. This work therefore provides a powerful tool for bearing faults detection and is broadly applicable to a variety of engineering applications and experimental conditions.

Existing bearing fault diagnosis approaches utilize signal processing techniques in combination with vibration measurements. The common vibration measurement method, using accelerometer, has disadvantages such as position-constrains, high costs and external interferences. Meanwhile, the commonly used modelling methods of motor are not able to analyse both mechanic and electromagnetism simultaneously. For detecting the fault-related frequencies in motor current, a mathematical model is developed in this paper based on modified winding function approach (MWFA). The simulation results show that the fault-related characteristic frequencies can be exactly found in the spectrum results of motor current. The proposed method is proven to be feasible.

The support vector machine (SVM) does not have a fixed parameter selection method and the manual selection of parameters is difficult to determine the validity, which affects the accuracy of recognition. simultaneously, The existing coarse-grained approach cannot effectively analyze the high-frequency components of time series. In view of the shortcomings of the above method, we put forward a new technique of rolling bearings for fault detection, which combines wavelet packet dispersion entropy (WPDE) and artificial fish swarm algorithm (AFSA) optimize support vector machines (AFSA-SVM). First of all, wavelet packet is devoted to decompose the original vibration signal into components of different frequency bands. Secondly, the dispersion entropy (DE) are calculated for each of the obtained frequency band components to acquire more comprehensive and complete fault information. Afterward, Input feature samples into the SVM model for training, and AFSA is used to optimize the parameters of SVM to obtain the optimal value so as to establish the best classification model. Finally, the prepared test set is input into AFSA-SVM for fault classification. The achievement of bearing detection experiments show that this approach can accurately and quickly identify fault types.