Search

Bearing degradation is the primary cause of electrical machine failures, making reliable condition monitoring essential to prevent breakdowns. This paper presents a novel hybrid model for the detection of multiple faults in bearings, combining Long Short-Term Memory (LSTM) networks with random forest (RF) classifiers, further enhanced by the Grey Wolf Optimization (GWO) algorithm. The proposed approach is structured in three stages: first, time and frequency domain features are manually extracted from vibration signals; second, these features are processed by a dual-layer LSTM network, which is specifically designed to capture complex temporal relationships within the data; finally, the GWO algorithm is employed to optimize feature selection from the LSTM outputs, feeding the most relevant features into the RF classifier for fault classification. The model was rigorously evaluated using a dataset comprising six distinct bearing health conditions: healthy, outer race fault, ball fault, inner race fault, compounded fault, and generalized degradation. The hybrid LSTM-RF-GWO model achieved a remarkable classification accuracy of 98.97%, significantly outperforming standalone models such as LSTM (93.56%) and RF (98.44%). Furthermore, the inclusion of GWO led to an additional accuracy improvement of 0.39% compared to the hybrid LSTM-RF model without optimization. Other performance metrics, including precision, kappa coefficient, false negative rate (FNR), and false positive rate (FPR), were also improved, with precision reaching 99.28% and the kappa coefficient achieving 99.13%. The FNR and FPR were reduced to 0.0071 and 0.0015, respectively, underscoring the model's effectiveness in minimizing misclassifications. The experimental results demonstrate that the proposed hybrid LSTM-RF-GWO framework not only enhances fault detection accuracy but also provides a robust solution for distinguishing between closely related fault conditions, making it a valuable tool for predictive maintenance in industrial applications. [ABSTRACT FROM AUTHOR]

This study aims to detect faults in wheelset bearings by analyzing vibration-sound fusion data, proposing a novel method based on Grey Wolf Optimizer (GWO) and Support Vector Machine (SVM). Wheelset bearings play a vital role in transportation. However, malfunctions in the bearing might result in extensive periods of inactivity and maintenance, disrupting supply chains, increasing operational costs, and causing delays that affect both businesses and consumers. Fast fault identification is crucial for minimizing maintenance expenses. In this paper, we proposed a new integration of GWO for optimizing SVM hyperparameters, specifically tailored for handling sound-vibration signals in fault detection. We have developed a new fault detection method that efficiently processes fusion data and performs rapid analysis and prediction within 0.0027 milliseconds per data segment, achieving a test accuracy of 98.3%. Compared to the SVM and neural network models built in MATLAB, the proposed method demonstrates superior detection performance. Overall, the GWO-SVM-based method proposed in this study shows significant advantages in fault detection of wheelset bearing vibrations, providing an efficient and reliable solution that is expected to reduce maintenance costs and improve the operational efficiency and reliability of equipment. [ABSTRACT FROM AUTHOR]

The primary goal of the paper is to monitor the health of the spindle in machine tools to ensure optimal performance and reduce costly downtimes. Spindle health monitoring is essential to detect wear and cracks in spindle bearings, which can be challenging due to their gradual development and hidden locations. The proposed approach combines physics-based modeling and data-driven techniques to monitor spindle health effectively. In Part I and Part II of the paper, mathematical models of bearing faults and spindle imbalance are integrated into the digital model of the spindle. This allows for simulating the operation of the spindle both with and without faults. The integration of fault models enables the generation of vibrations at sensor locations along the spindle shaft. The generated vibration data from the physics-based model are used to train a recurrent neural network-based (RNN) fault detection algorithm. The RNN learns from the labeled vibration spectra to identify different fault conditions. Bayesian optimization is used to automatically tune the hyperparameters governing the accuracy and efficiency of the learning models during the training process. The RNN classifiers are further fine-tuned using a small set of experimentally collected data for the generalization of the model on real-world data. Once the RNN classifier is trained, it can distinguish between different types of damage and identify their specific locations on the spindle assembly. The proposed algorithms achieved an accuracy of 98.43% on experimental data sets that were not used in training the network. [ABSTRACT FROM AUTHOR]

Abstract Bearing degradation is the primary cause of electrical machine failures, making reliable condition monitoring essential to prevent breakdowns. This paper presents a novel hybrid model for the detection of multiple faults in bearings, combining Long Short-Term Memory (LSTM) networks with random forest (RF) classifiers, further enhanced by the Grey Wolf Optimization (GWO) algorithm. The proposed approach is structured in three stages: first, time and frequency domain features are manually extracted from vibration signals; second, these features are processed by a dual-layer LSTM network, which is specifically designed to capture complex temporal relationships within the data; finally, the GWO algorithm is employed to optimize feature selection from the LSTM outputs, feeding the most relevant features into the RF classifier for fault classification. The model was rigorously evaluated using a dataset comprising six distinct bearing health conditions: healthy, outer race fault, ball fault, inner race fault, compounded fault, and generalized degradation. The hybrid LSTM-RF-GWO model achieved a remarkable classification accuracy of 98.97%, significantly outperforming standalone models such as LSTM (93.56%) and RF (98.44%). Furthermore, the inclusion of GWO led to an additional accuracy improvement of 0.39% compared to the hybrid LSTM-RF model without optimization. Other performance metrics, including precision, kappa coefficient, false negative rate (FNR), and false positive rate (FPR), were also improved, with precision reaching 99.28% and the kappa coefficient achieving 99.13%. The FNR and FPR were reduced to 0.0071 and 0.0015, respectively, underscoring the model’s effectiveness in minimizing misclassifications. The experimental results demonstrate that the proposed hybrid LSTM-RF-GWO framework not only enhances fault detection accuracy but also provides a robust solution for distinguishing between closely related fault conditions, making it a valuable tool for predictive maintenance in industrial applications.

An effective feature vector generation approach is herein presented based on the cointegration concept and the signal processing methods in order to improve varying speed bearing troubleshooting. Initially, each vibration signal is decomposed into its intrinsic mode functions (IMFs) using ensemble empirical mode decomposition method. Then, the cointegration relations are extracted by applying the Johansen trace test method to the obtained components. In order to form the feature matrix, the cointegration relations are analyzed by wavelet packet transform and the time-domain features are calculated for the wavelet packet coefficients. Consequently, by using a hybrid feature selection approach based on compensation distance evaluation technique, support vector machine, and binary particle swarm optimization algorithm, the optimal hybrid features were identified. The experimental results show that the optimal features are composed of the first cointegration relation, the third level of wavelet packet coefficients, the third IMF, Teager–Kaiser energy, energy, and root mean square. Also, the number of optimal features obtained from the first cointegration relation was less than other relationships. The results reveal that when the optimal feature set is computed for the first cointegration relationship, the identification accuracy of the proposed approach increases considerably. The result analysis also shows the efficiency of the proposed technique in detecting the condition of the bearing in time-varying rotational speed. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Zhejiang University: Science A is the property of Springer Nature and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Bearings are critical components of industrial equipment and have a significant impact on the safety of industrial physical systems. Their failure may lead to equipment shutdown and accidents, posing a significant risk to production safety. However, it is difficult to obtain a large amount of bearing fault data in practice, which makes the problem of small sample size a major challenge for bearing fault detection. In addition, some methods may overlook important features in bearing vibration signals, leading to insufficient detection capabilities. To address the challenges in bearing fault detection, this paper proposed a few sample learning methods based on the multidimensional convolution and attention mechanism. First, a multichannel preprocessing method was designed to more effectively utilize the information in the bearing vibration signal. Second, by extracting multidimensional features and enhancing the attention to important features through multidimensional convolution operations and attention mechanisms, the feature extraction ability of the network was improved. Furthermore, nonlinear mapping of feature vectors into the metric space to calculate distance can better measure the similarity between samples, thereby improving the accuracy of bearing fault detection and providing important guarantees for the safe operation of industrial systems. Extensive experiments have shown that the proposed method has good fault detection performance under small sample conditions, which is beneficial for reducing machine downtime and economic losses.

Abstract The research and application of bearing fault diagnosis techniques are crucial for enhancing equipment reliability, extending bearing lifespan, and reducing maintenance expenses. Nevertheless, most existing methods encounter challenges in discriminating between signals from machines operating under normal and faulty conditions, leading to unstable detection results. To tackle this issue, the present study proposes a novel approach for bearing fault detection based on graph neural networks and ensemble learning. Our key contribution is a novel stochasticity-based compositional method that transforms Euclidean-structured data into a graph format for processing by graph neural networks, with feature fusion and a newly proposed ensemble learning strategy for outlier detection specifically designed for bearing fault diagnosis. This approach marks a significant advancement in accurately identifying bearing faults, highlighting our study's pivotal role in enhancing diagnostic methodologies.

Rotary machines commonly use rolling element bearings to support rotation of the shafts. Most machine performance imperfections are related to bearing defects. Thus, reliable bearing condition monitoring systems are critically needed in industries to provide early warning of bearing fault so as to prevent machine performance degradation and reduce maintenance costs. The objective of this paper is to develop a smart monitoring system for real-time bearing fault detection and diagnostics. Firstly, a smart sensor-based data acquisition (DAQ) system is developed for wireless vibration signal collection. Secondly, a modified variational mode decomposition (MVMD) technique is proposed for nonstationary signal analysis and bearing fault detection. The proposed MVMD technique has several processing steps: (1) the signal is decomposed into a series of intrinsic mode functions (IMFs); (2) a correlation kurtosis method is suggested to choose the most representative IMFs and construct the analytical signal; (3) envelope spectrum analysis is performed to identify the representative features and to predict bearing fault. The effectiveness of the developed smart sensor DAQ system and the proposed MVMD technique is examined by systematic experimental tests. [ABSTRACT FROM AUTHOR]

Traditional fault detection methods in acoustic signal feature extraction of rolling bearings often make the signal denoising process complex due to low signal-to-noise ratio and weak fault features, making this method difficult to meet real-time requirements. Therefore, a fault detection model based on Fast-Renoriented SIFT feature extraction is proposed, which can quickly extract a large number of features from the original signal without the need for noise reduction processing and can effectively improve the efficiency and accuracy of fault detection. At the same time, to adapt to the fault detection of rolling bearings under multiple working conditions, this study also proposes an adaptive extended word bag model that combines local kurtosis and local 2-dimensional information entropy features, improving the adaptability and flexibility of the new model. It obtained a 100% overall recognition rate and a fault detection time of no more than 0.5 seconds in a 5-fold cross-validation experiment, verifying the excellent recognition accuracy, stability, and operational efficiency of the detection model. Its recognition accuracy in the multi-working condition rolling bearing fault detection experiment was above 97%, which was improved by about 21.25% compared to the traditional word bag model and had significant advantages in fault recognition accuracy and computational efficiency. [ABSTRACT FROM AUTHOR]

Wind turbines are a crucial part of renewable energy generation, and their reliable and efficient operation is paramount in ensuring clean energy availability. However, the bearings in wind turbines are subjected to high stress and loads, resulting in faults that can lead to costly downtime and repairs. Fault detection in real time is critical to minimize downtime and reduce maintenance costs. In this work, a simple neural network model was designed and implemented on a Raspberry Pi for the real-time detection of wind turbine bearing faults. The model was trained to accurately identify complex patterns in raw sensor data of healthy and faulty bearings. By splitting the data into smaller segments, the model can quickly analyze each segment and generate predictions at high speed. Additionally, simplified algorithms were developed to analyze the segments with minimum latency. The proposed system can efficiently process the sensor data and performs rapid analysis and prediction within 0.06 milliseconds per data segment. The experimental results demonstrate that the model achieves a 99.8% accuracy in detecting wind turbine bearing faults within milliseconds of their occurrence. The model's ability to generate real-time predictions and to provide an overall assessment of the bearing's health can significantly reduce maintenance costs and increase the availability and efficiency of wind turbines. [ABSTRACT FROM AUTHOR]

The research and application of bearing fault diagnosis techniques are crucial for enhancing equipment reliability, extending bearing lifespan, and reducing maintenance expenses. Nevertheless, most existing methods encounter challenges in discriminating between signals from machines operating under normal and faulty conditions, leading to unstable detection results. To tackle this issue, the present study proposes a novel approach for bearing fault detection based on graph neural networks and ensemble learning. Our key contribution is a novel stochasticity-based compositional method that transforms Euclidean-structured data into a graph format for processing by graph neural networks, with feature fusion and a newly proposed ensemble learning strategy for outlier detection specifically designed for bearing fault diagnosis. This approach marks a significant advancement in accurately identifying bearing faults, highlighting our study's pivotal role in enhancing diagnostic methodologies. [ABSTRACT FROM AUTHOR]

A novel data acquisition system for condition monitoring and predictive maintenance of mechanical parts, machinery, and industrial plants is presented. Current commercial solutions rely on an analog architecture and a star topology, in which all transducers are connected to a centralized acquisition unit. Usually this requires long shielded cables, which are sensitive to electromagnetic disturbances, always present in industrial environments. The proposed solution makes use of a digital bus implemented on an Unshielded Twisted Pair to connect one or more Acquisition Nodes to a data storage system (e.g., a laptop or an industrial computer). The wiring is simplified, cabling cost is reduced, high disturbance rejection is obtained, at the same time ensuring synchronization between all signals, mandatory for the computation of the most advanced diagnostic metrics. The performance and effectiveness of the developed system are proved in comparison with a top-quality, laboratory-grade commercial solution. A 10-days experiment was performed on a radial bearing mounted on a bearing test bench, by employing both systems side-by-side. Early-stage damage identification will be demonstrated with the described solution, despite costing a fraction and offering numerous advantages for industrial applications with respect to products currently available on the market.

The growth of global consumption has motivated important applications of deep learning to smart manufacturing and machine health monitoring. In particular, analyzing vibration data offers great potential to extract meaningful insights into predictive maintenance by the detection of bearing faults. Deep learning can be a powerful method to predict these mechanical failures; however, they lack generalizability to new tasks or datasets and require expensive, labeled mechanical data. We address this by presenting a novel self-supervised pretraining and fine-tuning framework based on transformer models. In particular, we investigate different tokenization and data augmentation strategies to reach state-of-the-art accuracies using transformer models. Furthermore, we demonstrate self-supervised masked pretraining for vibration signals and its application to low-data regimes, task adaptation, and dataset adaptation. Pretraining is able to improve performance on scarce, unseen training samples, as well as when fine-tuning on fault classes outside of the pretraining distribution. Furthermore, pretrained transformers are shown to be able to generalize to a different dataset in a few-shot manner. This introduces a new paradigm where models can be pretrained on unlabeled data from different bearings, faults, and machinery and quickly deployed to new, data-scarce applications to suit specific manufacturing needs.

Bearing is one of the key components widely used in mechanical equipment. Due to overload, fatigue, wear, corrosion and other reasons, bearings are easily damaged during machine operation. Therefore, the monitoring and analysis of the bearing state is very important. It can find the early weak fault of the bearing and prevent the loss caused by the fault. This paper proposes a long-term and short-term network combining the lightweight convolutional block attention module (CBAM-LSTM). In the field of bearing fault detection, the experimental results show that the CBAM-LSTM method can accurately identify a variety of mechanical bearing faults with an accuracy of 99,13 7 %.

This study aims to detect faults in wheelset bearings by analyzing vibration-sound fusion data, proposing a novel method based on Grey Wolf Optimizer (GWO) and Support Vector Machine (SVM). Wheelset bearings play a vital role in transportation. However, malfunctions in the bearing might result in extensive periods of inactivity and maintenance, disrupting supply chains, increasing operational costs, and causing delays that affect both businesses and consumers. Fast fault identification is crucial for minimizing maintenance expenses. In this paper, we proposed a new integration of GWO for optimizing SVM hyperparameters, specifically tailored for handling sound-vibration signals in fault detection. We have developed a new fault detection method that efficiently processes fusion data and performs rapid analysis and prediction within 0.0027 milliseconds per data segment, achieving a test accuracy of 98.3%. Compared to the SVM and neural network models built in MATLAB, the proposed method demonstrates superior detection performance. Overall, the GWO-SVM-based method proposed in this study shows significant advantages in fault detection of wheelset bearing vibrations, providing an efficient and reliable solution that is expected to reduce maintenance costs and improve the operational efficiency and reliability of equipment.

As regards bearing fault detection of induction motors using stator current, anything unrelated to the faults is the "noise". The noise attenuation leads to an improved estimation of the bearing fault signal. The gist of this study is the integration and synchronization of the three residues obtained from the current noise cancelation methods: Time Shifting and Linear Prediction. This procedure results in a richer bearing fault signal than any of the individual residues through cross-correlation. The developed method efficiently synchronizes and integrates the current residues so that fault characteristic frequencies can be more conspicuous in spectral analysis. Both simulation and experimental results attest to the merit and effectiveness of the developed method in detecting both outer and inner race bearing faults. [ABSTRACT FROM AUTHOR]

This study proposes a framework for the automated hyperparameter optimization of a bearing fault detection pipeline for Permanent Magnet Synchronous Motors (PMSMs) without the need for external sensors. An Automated Machine Learning (AutoML) pipeline search is performed through genetic optimization to reduce human-induced bias due to inappropriate parameterizations. A search space is defined, which includes general methods of signal processing and manipulation as well as methods tailored to the respective task and domain. The proposed framework is evaluated on the bearing fault detection use case under real-world conditions. Considerations on the generalization of the deployed fault detection pipelines are also considered. Likewise, attention was paid to experimental studies for evaluations of the robustness of the fault detection pipeline to variations of the motors working condition parameters between the training and test domain. [ABSTRACT FROM AUTHOR]

Time-frequency analysis is recognized as a dynamic tool to analyze the nonstationary signal. The synchrosqueezing transform is usually applied as a post-processing method to further improve the readability of the time-frequency representation. Synchrosqueezing transform is related to the reassignment method and can be performed in two directions, namely time direction and frequency direction. Frequency-direction reassignment helps to squeeze the slowly changing ridge. However, the time-direction reassignment is efficient to process the signal with rapid variation in instantaneous frequency. Thus, there exists a conflict in most of the time-frequency analysis methods while dealing with a signal containing both of these two components. In this study, a new method called demodulated time-direction synchrosqueezing transform is introduced, which is not only capable of achieving a higher compact TFR but also allow reconstructing the mode. In order to explain demodulated time-direction synchrosqueezing transform, a signal model is established in frequency domain. Then, a demodulated procedure is implemented to eliminate time-frequency analysis diffusion. Finally, time-direction reassignment is carried out to further enhance the energy concentration of the time-frequency analysis. The proposed demodulated time-direction synchrosqueezing transform method is evaluated by both simulation and experimental research. The results reveal that the performance of demodulated time-direction synchrosqueezing transform is better than the conventional time-frequency analysis methods, and it can be applied to the fault diagnosis in a machine. [ABSTRACT FROM AUTHOR]

Rotary machines commonly use rolling element bearings to support rotation of the shafts. Most machine performance imperfections are related to bearing defects. Thus, reliable bearing condition monitoring systems are critically needed in industries to provide early warning of bearing fault so as to prevent machine performance degradation and reduce maintenance costs. The objective of this paper is to develop a smart monitoring system for real-time bearing fault detection and diagnostics. Firstly, a smart sensor-based data acquisition (DAQ) system is developed for wireless vibration signal collection. Secondly, a modified variational mode decomposition (MVMD) technique is proposed for nonstationary signal analysis and bearing fault detection. The proposed MVMD technique has several processing steps: (1) the signal is decomposed into a series of intrinsic mode functions (IMFs); (2) a correlation kurtosis method is suggested to choose the most representative IMFs and construct the analytical signal; (3) envelope spectrum analysis is performed to identify the representative features and to predict bearing fault. The effectiveness of the developed smart sensor DAQ system and the proposed MVMD technique is examined by systematic experimental tests.

Wind turbines are a crucial part of renewable energy generation, and their reliable and efficient operation is paramount in ensuring clean energy availability. However, the bearings in wind turbines are subjected to high stress and loads, resulting in faults that can lead to costly downtime and repairs. Fault detection in real time is critical to minimize downtime and reduce maintenance costs. In this work, a simple neural network model was designed and implemented on a Raspberry Pi for the real-time detection of wind turbine bearing faults. The model was trained to accurately identify complex patterns in raw sensor data of healthy and faulty bearings. By splitting the data into smaller segments, the model can quickly analyze each segment and generate predictions at high speed. Additionally, simplified algorithms were developed to analyze the segments with minimum latency. The proposed system can efficiently process the sensor data and performs rapid analysis and prediction within 0.06 milliseconds per data segment. The experimental results demonstrate that the model achieves a 99.8% accuracy in detecting wind turbine bearing faults within milliseconds of their occurrence. The model’s ability to generate real-time predictions and to provide an overall assessment of the bearing’s health can significantly reduce maintenance costs and increase the availability and efficiency of wind turbines.

In an effort to overcome the problem that the traditional stochastic resonance system cannot adjust the structural parameters adaptively in bearing fault-signal detection, this article proposes an adaptive-parameter bearing fault-detection method. First of all, the four strategies of Sobol sequence initialization, exponential convergence factor, adaptive position update, and Cauchy–Gaussian hybrid variation are used to improve the basic grey wolf optimization algorithm, which effectively improves the optimization performance of the algorithm. Then, based on the multistable stochastic resonance model, the structure parameters of the multistable stochastic resonance are optimized through improving the grey wolf algorithm, so as to enhance the fault signal and realize the effective detection of the bearing fault signal. Finally, the proposed bearing fault-detection method is used to analyze and diagnose two open-source bearing data sets, and comparative experiments are conducted with the optimization results of other improved algorithms. Meanwhile, the method proposed in this paper is used to diagnose the fault of the bearing in the lifting device of a single-crystal furnace. The experimental results show that the fault frequency of the inner ring of the first bearing data set diagnosed using the proposed method was 158 Hz, and the fault frequency of the outer ring of the second bearing data set diagnosed using the proposed method was 162 Hz. The fault-diagnosis results of the two bearings were equal to the results derived from the theory. Compared with the optimization results of other improved algorithms, the proposed method has a faster convergence speed and a higher output signal-to-noise ratio. At the same time, the fault frequency of the bearing of the lifting device of the single-crystal furnace was effectively diagnosed as 35 Hz, and the bearing fault signal was effectively detected. [ABSTRACT FROM AUTHOR]

Bearing fault is the most likely to occur in mechanical fault, and stochastic resonance (SR), as a noise enhanced signal processing tool, can find mechanical faults as early as possible, so as to avoid larger problems. However, most of the existing research methods are based on the first-order Langevin equation. According to the previous studies of many scholars, the weak signal detection ability of the second-order system is better than that of the first-order system, and the coupled system also has better performance due to the addition of the control system. So, in order to detect the fault signal more easily, a second-order coupled tristable stochastic resonance system (SCTSR) based on the adaptive genetic algorithm (AGA) is proposed, it is an improvement on improving the first-order coupled tristable stochastic resonance system (FCTSR). First, based on the fourth-order Runge–Kutta algorithm (F-RK), the performances of monostable, bistable and tristable control systems to SCTSR are compared, it is verified that the monostable system has the best performance as SCTSR's control system. Secondly, the equivalent potential function of SCTSR is derived, and the influences of each system parameters on it are researched. The output signal-to-noise ratio gain (SNRG) is chosen as a measure to verify that SCTSR's performance is better than that of FCTSR, and the influences of parameters on SNRG are discussed. SCTSR and FCTSR are used to detect low-, high- and multi-frequency cosine signals combined with AGA. The simulation results are compared with the wavelet transform method, which proves the performance superiority of SR, and also prove that SCTSR is easier to detect weak signals and has a stronger de-noising ability. Finally, SCTSR and FCTSR are applied in bearing fault detection under Gaussian white noise and trichotomous noise. The results also prove that SCTSR can get larger peaks and SNRG, and it is easier to detect fault signals. This proves that SCTSR's performance is superior that of other methods in bearing fault detection, and has better engineering application value. [ABSTRACT FROM AUTHOR]

Renewable energy sources are a growing branch of industry. One such source is wind farms, which have significantly increased their number over recent years. Alongside the increased number of turbines, maintenance problems are growing. There is a need for newer and less intrusive predictive maintenance methods. About 40% of all turbine failures are due to bearing failure. This paper presents a modified neural direct classifier method using raw accelerometer measurements as input. This proprietary platform allows for better damage prediction results than convolutional networks in vibration spectrum image analysis. It operates in real time and without signal processing methods converting the signal to a time–frequency spectrogram. Image processing methods can extract features from a set of preset features and based on their importance. The proposed method is not based on feature extraction from image data but on automatically finding a set of features from raw tabular data. This fact significantly reduces the computational cost of detection and improves the failure detection accuracy compared to the classical methods. The model achieved a precision of 99.32% on the validation set, and 96.3% during bench testing. These results were an improvement over the method that classifies time–frequency spectrograms of 97.76% for the validation set and 90.8% for the real-world tests, respectively. [ABSTRACT FROM AUTHOR]

Rotating machine health monitoring is critical for system safety, cost savings, and increased reliability. The need for a simple and accurate fault diagnosis method has led to the development of various monitoring techniques. They incorporate vibration, motor’s current signature, and acoustic emission signals analysis in condition monitoring. So, based on using vibration signal analysis, a test rig was built for bearing fault identification. The test rig replicates and investigates various bearing problems, such as those found in the inner and outer races. An accelerometer, type ADXL335, was interfaced to a data acquisition device (DAQ USB-6215) for collecting vibration signals under various operating circumstances. In addition, a load cell was embedded with the test rig, interfaced with a digital panel meter, and used for recording the applied load on the bearings. The time-domain signal analysis technique was used after acquiring vibration signals at various bearing health states. Then, the time-domain signal was converted to the frequency domain using the fast Fourier transform, and the result was analyzed to investigate the generated fault frequencies. Finally, the obtained frequencies were compared with the theoretical values extracted from the theoretical equations, and the method proved its effectiveness in detecting the fault generated.

Machine learning (ML) based bearing fault detection is an emerging application of Artificial Intelligence (AI) that has proven its utility in effectively classifying various faults for timely measures. There are myriad studies dedicated to the effective classification of bearing faults under different conditions and experimental settings. In this study, we proposed a weighted voting ensemble (WVE) of three low-computation custom-designed convolutional neural networks (CNNs) to classify bearing faults at 48 KHz. Some of the recent studies have exploited 1-d time-series signals and time-frequency based 2-d transformations for bearing fault classification. However, 1-d signals lack contextual information and higher-dimensional interpretations whereas time-frequency based transformations provide a more appropriate, visually perceivable and explainable representation of the time and frequency changes. Therefore in this study, a scalogram based representation of the signals is leveraged for classification using the CNN. Furthermore, the class imbalance is a significant challenge that affects the modelling behavior and possibly create biases. This study provides a novel density and distance hybrid over-sampling approach namely Density-Aware SMOTE(DA-SMOTE) built upon the SMOTE methodology for a more refined representation of synthetic samples within the minority class distribution. The experimentation procedures were carried out before and after the oversampling and it was observed that the balanced dataset acquired much better accuracy then the imbalanced dataset. This is evident by the fact that the highest validation accuracy for the proposed ensemble method (WVCNN) reached at 0-HP and 1-HP reached 99.28% and 99.13% while for the over-sampled dataset the accuracy soared to 99.71% and 99.87% for 0 and 1-HP respectively. The performance was evaluated for other metrics apart from the accuracy to assess the model’s performance in terms of chance occurrences and the class wise performance.

In stochastic resonance (SR) weak signal detection, there is no literature currently report on the study and comparison of one-dimensional (1D) and 2D pinning potential worldwide which has potential research necessity. A one (ODPPBSR) and a 2D pinning potential bistable SR (TDPPBSR) are proposed. The expressions for MFPT, SPD and SNR are derived based on adiabatic approximation theory. To investigate the correctness of the theoretical results, numerical simulations are carried out with the Runge–Kutta algorithm and the genetic algorithm (GA) is used to optimize the system. The system has exceptional ability to restore signal periodicity and amplitude amplification at low frequency, high frequency and multi-frequency. The two systems are compared of the detection capabilities on weak signals through wavelet transform denoising and applied to the 6205-2RS JEM SKF and HRB 6205-2Z for bearing fault detection. The experimental results show that the 2D system is superior to the 1D system. [ABSTRACT FROM AUTHOR]

In the decomposition process of variational mode extraction (VME), it is hard to choose the approximate center frequency and the weighting coefficient reasonably. To address this issue, this paper aims to present a new bearing fault diagnosis scheme integrating VME with particle swarm optimization (PSO). Firstly, a new index combining correlation coefficient, L-kurtosis and information entropy is constructed. Then, the PSO is employed to optimize the inside VME parameters by combining the new index as the fitness function. Finally, the desired mode is analyzed by envelope demodulation to identify the fault characteristics. The effectiveness of the approach is validated using the experimental data sets collected from bearings with damaged outer race. A comparison of the PSO-VME method using the weighted kurtosis (KCI) as the fitness function highlights the superiority of the new index. Furthermore, comparisons with the PSO-VMD method and fast kurtogram method further validate the efficiency and accuracy of the proposed method. [ABSTRACT FROM AUTHOR]

A novel combined underdamped bistable stochastic resonance (CUBSR) is proposed in this paper. Under the noise-free condition, the output amplitude is used as a measurement index of classical bistable stochastic resonance (CBSR) and CUBSR, which demonstrate CUBSR does not have output saturation characteristics and has a more prominent signal enhancement capability. Then, the expressions of mean first-pass time (MFPT), steady-state probability density (SPD) and signal-to-noise ratio (SNR) are derived. Combined with the fourth-order Runge–Kutta algorithm and genetic algorithm (GA) for numerical simulations, the comparison of the theoretical derivation and numerical simulation of CUBSR can be verified. Then, the two systems are applied to the engineering application of bearing fault diagnosis. Finally, the multi-scale noise-modulated SR method based on wavelet packet transform is studied to overcome the limitation of traditional parameter modulation and to achieve SR detection at multiple frequencies. Simulation analysis and bearing fault diagnosis show that the method can effectively detect the multi-frequency weak signal submerged in noise, resulting in a significant enhancement in signal amplitude. [ABSTRACT FROM AUTHOR]

The stator current in an induction motor contains a large amount of information, which is unrelated to bearing faults. This information is considered as the noise component for the detection of bearing faults. When there is noise information in the current signal, it can affect the detection of motor bearing faults and lead to the possibility of false alarms. Therefore, to accomplish an effective bearing fault detection, all or some of these noise components must be properly eliminated. This paper proposes the use of fractional linear prediction (FLP) as a noise elimination method in bearing fault diagnosis, which makes these noise components the predictable components and this bearing fault information the unpredictable components. The basis of the FLP is to eliminate noise components in the current signal by predicting predictable components through linear prediction theory and optimal prediction order. Meanwhile, this paper adopts the FLP model with limited memory samples. After determining the optimal number of memories, only the fractional derivative order parameter needs to be optimized, which greatly reduces the computational complexity and difficulty in parameter optimization. In addition, this paper uses spectral analysis of the current signals through experimental simulation to compare the FLP method with the linear prediction (LP) method and the time-shifting (TS) method that have been successfully applied to bearing fault diagnosis. Based on the analysis results, the FLP method can better extract fault features and achieve better bearing fault diagnosis results, verifying the effectiveness and superiority of the FLP method in the field of bearing fault diagnosis. Additionally, the predictive performance of thevFLP and LP was compared based on experimental data, verifying the advantages of the FLP method in predictive performance, indicating that this method has a better noise cancellation effect.

In these last ten years, the use of wind energy has become a strategic alternative in several countries; the increase in reliability of these systems has become a crucial issue. This is explained by the economic, human or environmental) losses that could generate even if only for a moment, their malfunctioning or their shutdown. In this context, many research methods have been developed to monitor these machines during their operation time. But the problem with these methods was that they did not take into account the criticality, the type of fault, or the degradation degree of the components to trigger an alert, so that the real-time monitoring and the decisions taken were distorted. In this work, four types of defects encountered in wind turbines are studied. Three models of the fuzzy logic system are compared on the severity of the faults in order to know which one is the most efficient. However, an estimation of the fault parameters by the Fast-Estimation of Signal Parameters via Rotational Invariant Techniques (Fast-ESPRIT) algorithm followed by an identification of each fault type by the Classification Algorithm of Fault Harmonics (CAFH) algorithm are first performed. The obtained results show the possibility of monitoring the severity of faults in electric induction machines using the Tsukamoto model in real time. The simulations are performed by using MATLAB software and the obtained results demonstrate the feasibility of such a system.

Weak signal detection under strong noise is a common problem in many engineering fields. The research on the detection theory and method of stochastic resonance (SR) has very important theoretical significance and application value for the realization of early weak fault diagnosis. In order to further enhance the weak signal processing capability of SR, an improved novel composite multistable potential well model is proposed by combining the tristable model and the Woods–Saxon model. The switching mechanism of the novel model constructed with the fusion of the tristable model and the Woods–Saxon model between different steady states is studied, the output response performance of SR system with the novel composite multistable model is analyzed. The adaptive synchronization optimization method of multiple system parameters adopts the differential brainstorming algorithm to realize the adaptive selection of multiple parameters. Simulation experiments are carried out on single and multiple low-frequency periodic signals and single and multiple high-frequency periodic signals under the Gaussian noise environment, simulation results indicate that the novel composite multistable SR system performs better. On the basis of this model, the composite multistable SR system is applied to the fault detection of rolling bearings, which has a good detection effect. [ABSTRACT FROM AUTHOR]

In an effort to overcome the problem that the traditional stochastic resonance system cannot adjust the structural parameters adaptively in bearing fault-signal detection, this article proposes an adaptive-parameter bearing fault-detection method. First of all, the four strategies of Sobol sequence initialization, exponential convergence factor, adaptive position update, and Cauchy–Gaussian hybrid variation are used to improve the basic grey wolf optimization algorithm, which effectively improves the optimization performance of the algorithm. Then, based on the multistable stochastic resonance model, the structure parameters of the multistable stochastic resonance are optimized through improving the grey wolf algorithm, so as to enhance the fault signal and realize the effective detection of the bearing fault signal. Finally, the proposed bearing fault-detection method is used to analyze and diagnose two open-source bearing data sets, and comparative experiments are conducted with the optimization results of other improved algorithms. Meanwhile, the method proposed in this paper is used to diagnose the fault of the bearing in the lifting device of a single-crystal furnace. The experimental results show that the fault frequency of the inner ring of the first bearing data set diagnosed using the proposed method was 158 Hz, and the fault frequency of the outer ring of the second bearing data set diagnosed using the proposed method was 162 Hz. The fault-diagnosis results of the two bearings were equal to the results derived from the theory. Compared with the optimization results of other improved algorithms, the proposed method has a faster convergence speed and a higher output signal-to-noise ratio. At the same time, the fault frequency of the bearing of the lifting device of the single-crystal furnace was effectively diagnosed as 35 Hz, and the bearing fault signal was effectively detected.

Bearings are vital components of rotating machines that are prone to unexpected faults. Therefore, bearing fault diagnosis and condition monitoring are essential for reducing operational costs and downtime in numerous industries. In various production conditions, bearings can be operated under a range of loads and speeds, which causes different vibration patterns associated with each fault type. Normal data are ample as systems usually work in desired conditions. On the other hand, fault data are rare, and in many conditions, there are no data recorded for the fault classes. Accessing fault data is crucial for developing data-driven fault diagnosis tools that can improve both the performance and safety of operations. To this end, a novel algorithm based on conditional generative adversarial networks (CGANs) was introduced. Trained on the normal and fault data on actual fault conditions, this algorithm generates fault data from normal data of target conditions. The proposed method was validated on a real-world bearing dataset, and fault data were generated for different conditions. Several state-of-the-art classifiers and visualization models were implemented to evaluate the quality of the synthesized data. The results demonstrate the efficacy of the proposed algorithm. [ABSTRACT FROM AUTHOR]

It is always an important and challenging issue to achieve an effective fault diagnosis in rotating machinery in industries. In recent years, deep learning proved to be a high-accuracy and reliable method for data-based fault detection. However, the training of deep learning algorithms requires a large number of real data, which is generally expensive and time-consuming. To cope with this, we proposed a Resnet classifier with model-based data augmentation, which is applied for bearing fault detection. To this end, a dynamic model was first established to describe the bearing system by adjusting model parameters, such as speed, load, fault size, and the different fault types. Large amounts of data under various operation conditions can then be generated. The training dataset was constructed by the simulated data, which was then applied to train the Resnet classifier. In addition, in order to reduce the gap between the simulation data and the real data, the envelop signals were used instead of the original signals in the training process. Finally, the effectiveness of the proposed method was demonstrated by the real bearing experimental data. It is remarkable that the application of the proposed method can be further extended to other mechatronic systems with a deterministic dynamic model. [ABSTRACT FROM AUTHOR]

The early fault detection in the rotary electrical machines,such as induction motors (IMs), has been growing in modern industry. IMs have been widely used in industrial applications due to its easy installation, reliability, and low cost. However, the increasing usage of IMs also increases the need for timely maintenance in order to ensure their operation and a longer service life. This article proposes a new diagnosis methodology based on maximal overlap discrete wavelet transform and a lightweight 1-D convolutional neural network (CNN) architecture, in order to detect mechanical and electrical faults and their combination, in adjustable speed drive (ASD)-powered IMs. Specifically, single and combined faults were studied from the next: Outer raceway bearing (mechanical), turn-to-turn short-circuit, and phase-to-ground short circuit (electrical). The presented study was developed using current signals acquired from stators of IMs of 1 hp. The current signals are measured at powered conditions introduced by a power grid with a constant frequency at 60 Hz, and an ASD at three different frequencies. The proposed diagnostic methodology reaches more than 99% of accuracy. [ABSTRACT FROM AUTHOR]

Renewable energy sources are a growing branch of industry. One such source is wind farms, which have significantly increased their number over recent years. Alongside the increased number of turbines, maintenance problems are growing. There is a need for newer and less intrusive predictive maintenance methods. About 40% of all turbine failures are due to bearing failure. This paper presents a modified neural direct classifier method using raw accelerometer measurements as input. This proprietary platform allows for better damage prediction results than convolutional networks in vibration spectrum image analysis. It operates in real time and without signal processing methods converting the signal to a time–frequency spectrogram. Image processing methods can extract features from a set of preset features and based on their importance. The proposed method is not based on feature extraction from image data but on automatically finding a set of features from raw tabular data. This fact significantly reduces the computational cost of detection and improves the failure detection accuracy compared to the classical methods. The model achieved a precision of 99.32% on the validation set, and 96.3% during bench testing. These results were an improvement over the method that classifies time–frequency spectrograms of 97.76% for the validation set and 90.8% for the real-world tests, respectively.

Industrial machinery often breakdowns due to faults in rolling bearing. Bearing diagnosis plays a vital role in condition monitoring of machinery. Operating conditions and working environment of bearings make them prone to single or multiple faults. In this research, signals from both healthy and faulty bearings are extracted and decomposed into empirical modes. By analyzing different empirical modes from 8 derived empirical modes for healthy and faulty bearings under different fault sizes, the first mode has the most information to classify bearing condition. From the first empirical mode eight features in time domain were calculated for various bearing conditions like healthy, rolling element fault, outer and inner race fault. The feature ex- traction of vibration signal based on Empirical Mode Decomposition (EMD) is extensively explored and applied in diagnosis of fault in rolling bearings. This paper presents mathematical analysis for selection of valid Intrinsic Mode Functions (IMFs) of EMD. These chosen features are trained and classified using different classifiers. Among them K-star classifier is most reliable to categorize the bearing defects. [ABSTRACT FROM AUTHOR]

Purpose: In order to improve the system's output signal-to-noise ratio(SNR) and solve the output saturation problem, a novel time-delay feedback piecewise tri-stable stochastic resonance system (DFPTSR) is proposed. Methods: Combining the time-delay feedback term can improve the output of the system and the piecewise tri-stable stochastic resonance system can overcome the advantages of output saturation, a novel piecewise tri-stable potential function is proposed and applied to the timedelay feedback system. The effective potential function Ueff(x), the steady-state probability density (SPD) and mean first passage time (MFPT) are derived and the parameters' effects on them are discussed. Then by adiabatic approximation theory, SNR is derived and compared with the classical bi-stable stochastic resonance system(CBSR). Finally, the fourth-order Runge-Kutta method is utilized to carry out the weak signal detection and bearing fault detection in two scenarios, and compared with time-delay feedback classical tristable stochastic resonance system(DFCTSR) and piecewise tri-stable stochastic resonance system(PTSR). Results: In this work, it shows that the output SNR of DFPTSR is larger than that of CBSR theoretically. Simulation results show that the output SNR and amplitude of DFPTSR are better than those of DFCTSR and PTSR in weak signal detection and bearing fault detection. Proved the time-delay feedback term can enhance the system's output and the piecewise system has advantages to overcome output saturation, so a system that combines the two advantages will have better performance. Conclusions: Both theoretical analysis and numerical simulation show that DFPTSR has much better anti-noise performance, output amplitude and SNR. Weak signal detection and bearing fault detection results show that the fault frequency can be more accurately identified by DFPTSR and the characteristic signal's energy can be enhanced more, indicating good theoretical significance and practical engineering applications value. [ABSTRACT FROM AUTHOR]

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.

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.

Today's deep learning strategies require ever-increasing computational efforts and demand for very large amounts of labelled data. Providing such expensive resources for machine diagnosis is highly challenging. Transfer learning recently emerged as a valuable approach to address these issues. Thus, the knowledge learned by deep architectures in different scenarios can be reused for the purpose of machine diagnosis, minimizing data collecting efforts. Existing research provides evidence that networks pre-trained for image recognition can classify machine vibrations in the time-frequency domain by means of transfer learning. So far, however, there has been little discussion about the potentials included in networks pre-trained for sound recognition, which are inherently suited for time-frequency tasks. This work argues that deep architectures trained for music recognition and sound detection can perform machine diagnosis. The YAMNet convolutional network was designed to serve extremely efficient mobile applications for sound detection, and it was originally trained on millions of data extracted from YouTube clips. That framework is employed to detect bearing faults for the CWRU dataset. It is shown that transferring knowledge from sound and music recognition to bearing fault detection is successful. The maximum accuracy is achieved using a few hundred data for fine-tuning the fault diagnosis model. [ABSTRACT FROM AUTHOR]