No file available
To read the file of this research,
you can request a copy directly from the authors.
Remaining useful life prediction of rolling bearings based on refined composite multi-scale attention entropy and dispersion entropy
Preprints and early-stage research may not have been peer reviewed yet.
Abstract
Remaining useful life (RUL) prediction based on vibration signals is crucial for ensuring the safe operation and effective health management of rotating machinery. Existing studies often extract health indicators (HI) from time domain and frequency domain features to analyze complex vibration signals, but these features may not accurately capture the degradation process. In this study, we propose a degradation feature extraction method called Fusion of Multi-Modal Multi-Scale Entropy (FMME), which utilizes multi-modal Refined Composite Multi-scale Attention Entropy (RCMATE) and Fluctuation Dispersion Entropy (RCMFDE), to solve the problem that the existing degradation features cannot accurately reflect the degradation process. Firstly, the Empirical Mode Decomposition (EMD) is employed to decompose the dual-channel vibration signals of bearings into multiple modals. The main modals are then selected for further analysis. The subsequent step involves the extraction of RCMATE and RCMFDE from each modal, followed by wavelet denoising. Next, a novel metric is proposed to evaluate the quality of degradation features. The attention entropy and dispersion entropy of the optimal scales under different modals are fused using Laplacian Eigenmap (LE) to obtain the health indicators. Finally, RUL prediction is performed through the similarity of health indicators between fault samples and bearings to be predicted. Experimental results demonstrate that the proposed method yields favorable outcomes across diverse operating conditions.
ResearchGate has not been able to resolve any citations for this publication.
The remaining useful life (RUL) prediction of rolling bearings based on vibration signals has attracted widespread attention. It is not satisfactory to adopt information theory (such as information entropy) to realize RUL prediction for complex vibration signals. Recent research has used more deep learning methods based on the automatic extraction of feature information to replace traditional methods (such as information theory or signal processing) to obtain higher prediction accuracy. Convolutional neural networks (CNNs) based on multi-scale information extraction have demonstrated promising effectiveness. However, the existing multi-scale methods significantly increase the number of model parameters and lack efficient learning mechanisms to distinguish the importance of different scale information. To deal with the issue, the authors of this paper developed a novel feature reuse multi-scale attention residual network (FRMARNet) for the RUL prediction of rolling bearings. Firstly, a cross-channel maximum pooling layer was designed to automatically select the more important information. Secondly, a lightweight feature reuse multi-scale attention unit was developed to extract the multi-scale degradation information in the vibration signals and recalibrate the multi-scale information. Then, end-to-end mapping between the vibration signal and the RUL was established. Finally, extensive experiments were used to demonstrate that the proposed FRMARNet model can improve prediction accuracy while reducing the number of model parameters, and it outperformed other state-of-the-art methods.
Due to the importance of bearings in modern machinery, the prediction of the remaining useful life (RUL) of rolling bearings has been widely studied. When predicting the RUL of rolling bearings in engineering practice, the RUL is usually predicted based on historical data, and as the historical data increases, the prediction results should be more accurate. However, the existing methods usually have the shortcomings of low prediction accuracy, large cumulative error and failure to dynamically give prediction results with the increase of historical data, which are not suitable for engineering practice.To address the above problems, a novel RUL prediction method is proposed. The proposed method consists of 3 parts: First, the multi-scale entropy-based feature – namely “average multi-scale morphological gradient power spectral information entropy (AMMGPSIE)”– from the rolling bearings as the Health indicator (HI) is extracted to ensure all the fault-related information is well-included; Then, the HI is processed with the enhanced Hodrick Prescott trend-filtering with boundary lines (HPTF-BL) to ensure good performance and small fluctuation on the HI; Finally, the deterioration curve is predicted using an LSTM neural network and the improved Particle Filter algorithm that we proposed. The proposed method is validated using the experimental bearing degradation dataset and the casing data of a centrifugal pump bearing from an actual industrial site. Comparing the results with other recent RUL prediction methods, the proposed method achieved state-of-the-art feasibility and effectiveness, conform to the needs of practical application of the project.
When rotating machinery fails, the consequent vibration signal contains rich fault feature information. However, the vibration signal bears the characteristics of nonlinearity and nonstationarity, and is easily disturbed by noise, thus it may be difficult to accurately extract hidden fault features. To extract effective fault features from the collected vibration signals and improve the diagnostic accuracy of weak faults, a novel method for fault diagnosis of rotating machinery is proposed. The new method is based on Fast Iterative Filtering (FIF) and Parameter Adaptive Refined Composite Multiscale Fluctuation-based Dispersion Entropy (PARCMFDE). Firstly, the collected original vibration signal is decomposed by FIF to obtain a series of intrinsic mode functions (IMFs), and the IMFs with a large correlation coefficient are selected for reconstruction. Then, a PARCMFDE is proposed for fault feature extraction, where its embedding dimension and class number are determined by Genetic Algorithm (GA). Finally, the extracted fault features are input into Fuzzy C-Means (FCM) to classify different states of rotating machinery. The experimental results show that the proposed method can accurately extract weak fault features and realize reliable fault diagnosis of rotating machinery.
The gear fault signal has some defects such as nonstationary nonlinearity. In order to increase the operating life of the gear, the gear operation is monitored. A gear fault diagnosis method based on variational mode decomposition (VMD) sample entropy and discrete Hopfield neural network (DHNN) is proposed. Firstly, the optimal VMD decomposition number is selected by the instantaneous frequency mean value. Then, the sample entropy value of each intrinsic mode function (IMF) is extracted to form the gear feature vectors. The gear feature vectors are coded and used as the memory prototype and memory starting point of DHNN, respectively. Finally, the coding vector is input into DHNN to realize fault pattern recognition. The newly defined coding rules have a significant impact on the accuracy of gear fault diagnosis. Driven by self-associative memory, the coding of gear fault is accurately classified by DHNN. The superiority of the VMD-DHNN method in gear fault diagnosis is verified by comparing with an advanced signal processing algorithm. The results show that the accuracy based on VMD sample entropy and DHNN is 91.67% of the gear fault diagnosis method. The experimental results show that the VMD method is better than the complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) and empirical mode decomposition (EMD), and the effect of it in the diagnosis of gear fault diagnosis is emphasized.
Classification of interbeat interval time-series which fluctuates in an irregular and complex manner is very challenging. Typically, entropy methods are employed to quantify the complexity of the time-series for classifying. Traditional entropy methods focus on the frequency distribution of all the observations in a time-series. This requires a relatively long time-series with at least a couple of thousands of data points, which limits their usages in practical applications. The methods are also sensitive to the parameter settings. In this paper, we propose a conceptually new approach called attention entropy, which pays attention only to the key observations. Instead of counting the frequency of all observations, it analyzes the frequency distribution of the intervals between the key observations in a time-series. The advantages of the attention entropy are that it does not need any parameter to tune, is robust to the time-series length, and requires only linear time to compute. Experiments show that it outperforms fourteen state-of-the-art entropy methods evaluated by real-world datasets. It achieves average classification accuracy of AUC=0.71 while the second-best method, multiscale entropy, achieves AUC=0.62 when classifying four groups of people with a time-series length of 100.
Fluctuation-based dispersion entropy (FDispEn) is a new approach to estimate the dynamical variability of the fluctuations of signals. It is based on Shannon entropy and fluctuation-based dispersion patterns. To quantify the physiological dynamics over multiple time scales, multiscale FDispEn (MFDE) is developed in this article. MFDE is robust to the presence of baseline wanders or trends in the data. We evaluate MFDE, compared with popular multiscale sample entropy (MSE), multiscale fuzzy entropy (MFE), and the recently introduced multiscale dispersion entropy (MDE), on selected synthetic data and five neurological diseases’ datasets: 1) focal and non-focal electroencephalograms (EEGs); 2) walking stride interval signals for young, elderly, and Parkinson’s subjects; 3) stride interval fluctuations for Huntington’s disease and amyotrophic lateral sclerosis; 4) EEGs for controls and Alzheimer’s disease patients; and 5) eye movement data for Parkinson’s disease and ataxia. MFDE avoids the problem of undefined MSE values and, compared with MFE and MSE, leads to more stable entropy values over the scale factors for white and pink noises. Overall, MFDE is the fastest and most consistent method for the discrimination of different states of neurological data, especially where the mean value of a time series considerably changes along the signal (e.g., eye movement data). This study shows that MFDE is a relevant new metric to gain further insights into the dynamics of neurological diseases recordings. The Matlab codes for MFDE and its refined composite form are available in Xplore.
A new fault feature extraction method for rolling element bearing is put forward in this paper based on the adaptive local iterative filtering (ALIF) algorithm and the modified fuzzy entropy. Due to the bearing vibration signals’ non-stationary and nonlinear characteristics, the ALIF method, which is a new approach for the analysis of the non-stationary signals, is used to decompose the original vibration signals into a series of mode components. Fuzzy entropy (FuzzyEn) is a nonlinear dynamic parameter for measuring the signals’ complexity. However, it only emphasizes the signals’ local characteristics while neglecting its global fluctuation. Considering the global fluctuation of bearing vibration signals will change with the bearing working condition varying, we modified the FuzzyEn. The modified FuzzyEn (MFuzzyEn) of the first few modes obtained by the ALIF is utilized to form the fault feature vectors. Subsequently, the corresponding feature vectors are input into the multi-class SVM classifier to accomplish the bearing fault identification automatically. The experimental analysis demonstrates that the presented ALIF-MFuzzyEn-SVM approach can effectively recognize the different fault categories and different levels of bearing fault severity.
The characteristics of the early fault signal of the rolling bearing are weak and this leads to difficulties in feature extraction. In order to diagnose and identify the fault feature from the bearing vibration signal, an adaptive local iterative filter decomposition method based on permutation entropy is proposed in this paper. As a new time-frequency analysis method, the adaptive local iterative filtering overcomes two main problems of mode decomposition, comparing traditional methods: modal aliasing and the number of components is uncertain. However, there are still some problems in adaptive local iterative filtering, mainly the selection of threshold parameters and the number of components. In this paper, an improved adaptive local iterative filtering algorithm based on particle swarm optimization and permutation entropy is proposed. Firstly, particle swarm optimization is applied to select threshold parameters and the number of components in ALIF. Then, permutation entropy is used to evaluate the mode components we desire. In order to verify the effectiveness of the proposed method, the numerical simulation and experimental data of bearing failure are analyzed.
Motor bearing is subjected to the joint effects of much more loads, transmissions and shocks that cause bearing fault and machinery breakdown. Vibration signal analysis method is the most popular technique that is used to monitor and diagnose the fault of motor bearing. However, the application of vibration signal analysis method for motor bearing is very limited in engineering practice. In this paper, on the basis of comparing fault feature extraction by using empirical wavelet transform(EWT) and Hilbert transform with the theoretical calculation, a new motor bearing fault diagnosis method based on integrating EWT, fuzzy entropy and support vector machine(SVM), called EWTFSFD is proposed. In the proposed method, a novel signal processing method called EWT is used to decompose vibration signal into multiple components in order to extract a series of amplitude modulated-frequency modulated(AM-FM) components with supporting Fourier spectrum under an orthogonal basis. Then fuzzy entropy is utilized to measure the complexity of vibration signal, reflect complexity changes of intrinsic oscillation, and compute the fuzzy entropy values of AM-FM components, which are regarded as the inputs of SVM model to train and construct a SVM classifier for fulfilling fault pattern recognition. Finally, the effectiveness of the proposed method is validated by using the simulated signal and real motor bearing vibration signals. The experiment results show that the EWT outperforms empirical mode decomposition for decomposing the signal into multiple components, and the proposed EWTFSFD method can accurately and effectively achieve the fault diagnosis of motor bearing. OAPA
One of the most powerful tools to assess the dynamical characteristics of time series is entropy. Sample entropy (SE), though powerful, is not fast enough, especially for long signals. Permutation entropy (PE), as a broadly used irregularity indicator, considers only the order of the amplitude values and hence some information regarding the amplitudes may be discarded. To tackle these problems, we introduce a new method, termed dispersion entropy (DE), to quantify the regularity of time series. We gain insight into the dependency of DE on several straightforward signal-processing concepts via a set of synthetic time series. The results show that DE, unlike PE, can detect the noise bandwidth and simultaneous frequency and amplitude change. We also employ DE to three publicly available real datasets. The simulations on real-valued signals show that the DE method considerably outperforms PE to discriminate different groups of each dataset. In addition, the computation time of DE is significantly less than that of SE and PE.
This paper describes the three methodologies used by CALCE in their winning entry for the IEEE 2012 PHM Data Challenge competition. An experimental data set from seventeen ball bearings was provided by the FEMTO-ST Institute. The data set consisted of data from six bearings for algorithm training and data from eleven bearings for testing. The authors developed prognostic algorithms based on the data from the training bearings to estimate the remaining useful life of the test bearings. Three methodologies are presented in this paper. Result accuracies of the winning methodology are presented.
This paper presents the Professional-category winning algorithm of bearing Remaining Useful Life (RUL) prediction for the 2012 IEEE PHM challenge problem. The algorithm consists of extraction of bearing characteristic frequency features with envelop analysis, fault detection with PCA, and two RUL prediction strategies to address the scenarios when the bearing faults have and have not been detected. The paper will go through various aspects to investigate the challenge problem, synthesize modeling strategies, and summarize the lessons learned from this bearing life prediction case study.
This paper deals with the presentation of an experimental
platform called PRONOSTIA, which enables testing,
verifying and validating methods related to bearing health
assessment, diagnostic and prognostic. The choice of bearings
is justified by the fact that most of failures of rotating machines
are related to these components. Therefore, bearings can be
considered as critical as their failure significantly decreases
availability and security of machines.
The main objective of PRONOSTIA is to provide real data related
to accelerated degradation of bearings performed under constant
and/or variable operating conditions, which are online controlled.
The operating conditions are characterized by two sensors:
a rotating speed sensor and a force sensor. In PRONOSTIA
platform, the bearing’s health monitoring is ensured by gathering
online two types of signals: temperature and vibration (horizontal
and vertical accelerometers). Furthermore, the data are recorded
with a specific sampling frequency which allows catching all the
frequency spectrum of the bearing during its whole degradation.
Finally, the monitoring data provided by the sensors can be used
for further processing in order to extract relevant features and
continuously assess the health condition of the bearing.
During the PHM conference, a “IEEE PHM 2012 Prognostic
Challenge” is organized. For this purpose, a web link to the
degradation data is provided to the competitors to allow them
testing and verifying their prognostic methods. The results of
each method can then be evaluated regarding its capability
to accurately estimate the remaining useful life of the tested
bearings.
Accurate remaining useful life (RUL) prediction for rolling bearings encounters many challenges such as complex degradation processes, varying working conditions, and insufficient run-to-failure data. Transfer learning (TL), one paradigm of artificial intelligence technology, has demonstrated its powerful performance and great effectiveness for such challenges. As a result, many TL-based solutions have been widely developed and extensively studied for rolling bearing RUL prediction. Admittedly, several review articles have been published on RUL prediction. Nevertheless, the majority of these articles only concentrated on deep learning-based RUL prediction methods, and a review article that systematically overviews the status of TL-based RUL prediction has not been published. Therefore, it is urgent and significant to thoroughly summarize the academic publications and industrial applications related to TL-based RUL prediction, and present its potential challenges and future research directions. With such goals, the problem definitions of TL-based RUL prediction, the general procedure of RUL prediction, and typical TL-based RUL prediction algorithms are first introduced to help researchers quickly overview the state-of-the-art approaches and recent developments. Thereafter, relevant TL-based RUL prediction solutions are comprehensively discussed from the perspectives of three industrial scenarios, providing suggestions to researchers and engineers for selecting appropriate solutions in practical industrial applications. Finally, the key challenges and future trends in RUL prediction are presented to conclude this paper. We hope that this review of TL-based RUL prediction for rolling bearings can contribute to a better understanding of intelligent prognostic technology and will inspire researchers to extend their work on RUL prediction.
In industrial production, effectively predicting the remaining useful life (RUL) of cutting tools can avoid overuse or underuse, which is of great significance for ensuring the processing quality of products and reducing enterprises’ production costs. This paper proposes a new method for RUL prediction of cutting tools based on robust empirical mode decomposition (REMD) and capsule bidirectional long short-term memory (Capsule-BiLSTM) network to improve accuracy. On one hand, new state features are extracted using REMD as the input of the deep learning network. On the other hand, a Capsule-BiLSTM network structure is designed to achieve RUL prediction of cutting tools by connecting the four layers. Finally, the effectiveness of the proposed method is verified by a series of cutting tool life tests. Comparison with some mainstream methods indicates that the proposed method has more advantages in RUL prediction of cutting tools with the average accuracy reaching up to 93.97%.
The dynamic model can provide high capabilities at accurate simulation for rolling element bearings (REBs), but to simulate the continuous dynamic response over the whole life time is difficult. This paper proposes a hybrid method based on a data-driven and dynamic model to obtain continuous parameters of dynamic model, which can be used to predict the remaining useful life (RUL) of REBs. First, four dynamic models of the REB with the variable parameters of roughness, crack growth rate, crack length and crack depth are established to describe the degradation process of REB with three stages. In particular, the self-healing phenomenon in bearing fault stage is analysed, and the three models are used to describe the trend of fluctuating degradation process in this stage. The dynamic model parameters are updated for the degradation process of REBs with the interacting multiple-model particle filtering, which taken root mean square (RMS) as the observation. Then, the continuously growing crack length is used as the index to predict the RUL. The proposed method realises the continuity of dynamic model in the whole life time, and the fluctuating RMS is transformed into a continuously growing parameter for prediction to avoid the prediction difficulty caused by the decrease in RMS in the degradation process. Finally, The bearing datasets of Xi’an Jiaotong University are used to verify the effectiveness of the proposed method. The comparative test results show that the proposed method can predict the results more accurately without sample training and obtain a clearer degradation mechanism.
The prediction of the remaining useful life (RUL) of rolling bearings is an important means to ensure the rotating machinery's safe operation. At present, most of the proposed methods use direct prediction based on bearing vibration signals, which not only have low prediction accuracy but also time-consuming. This paper proposes a staged prediction method, and the regularized learning machine (RELM) based on the proposed sensitive degradation feature is applied to predict RUL of the bearing with high accuracy and speed. Firstly, the relative root mean square value (RRMS) is used to divide the degradation stages of rolling bearings. Secondly, the RRMS indicator is used for multi-step time series prediction in the normal phase of the bearing. Thirdly, in the bearing's degradation stage, the Pearson Correlation Coefficient (PCC) Combined Entropy Weight Method (EWM) feature selection criterion is proposed to predict the RUL of the rolling bearing. Finally, the sensitive degradation feature of the bearing vibration signals is input into RELM to predict the RUL. The bearing data sets of PHM Challenging 2012 are used to verify the effectiveness of the proposed method. Three comparative experiments have been verified to prove the accuracy and rapidity of the proposed method in time series forecasting.
Feature Extraction Algorithms (FEAs) aim to address the curse of dimensionality that makes machine learning algorithms incompetent. Our study conceptually and empirically explores the most representative FEAs. First, we review the theoretical background of many FEAs from different categories (linear vs. nonlinear, supervised vs. unsupervised, random projection-based vs. manifold-based), present their algorithms, and conduct a conceptual comparison of these methods. Secondly, for three challenging binary and multi-class datasets, we determine the optimal sets of new features and assess the quality of the various transformed feature spaces in terms of statistical significance and power analysis, and the FEA efficacy in terms of classification accuracy and speed.
Link of the paper: https://authors.elsevier.com/a/1cc%7E%7E5buwWjZZw
With advanced measurement technologies and signal analytics algorithms developed rapidly, the past decades have witnessed large amount of successful breakthroughs and applications in the field of intelligent fault diagnosis (IFD). However, the historical IFD methods have difficulties for compound fault diagnosis, when labeled data cannot be collected in advance for new or extreme working conditions. Facing with such challenges, a Deep Adversarial Capsule Network (DACN) is proposed to embed multi-domain generalization into the intelligent compound fault diagnosis, so that the DACN not only can intelligently decouple the compound fault into multiple single faults for industrial equipment (IE) but also can be generalized from certain working conditions to another new. First, a DACN including feature extractor, decoupling classifier and multi-domain classifier is constructed for feature learning, fault decoupling and unsupervised multi-domain adaptation, respectively. Second, adversarial training is introduced into the DACN in training stage via a gradient reversal layer that can build the connection between the feature extractor and domain classifier, which can force the DACN to learn the domain-invariance features. Finally, the DACN is trained using the single-faults data collected under multiple working conditions, and then applied to monitor the health condition of IE under new working conditions. The cross-validation experiments have been implemented on an automobile transmission (AT), which illustrates that the DACN obtains an optimum performance with the highest average accuracy of 97.65% for compound fault diagnosis of IE under multi-domain generalization task, and outperforms other related methods.
In order to improve the prediction accuracy of performance degradation trends of rolling bearings, a method based on the joint approximative diagonalization of eigen-matrices (JADE) and particle swarm optimization support vector machine (PSO-SVM) was proposed. Firstly, the features of the time-domain, frequency-domain, and time-frequency-domain eigenvalues of the vibration signal corresponding to the entire life cycle of the rolling bearing are extracted, and the performance degradation parameters are initially selected by using the monotonicity parameter. Then, a fusion feature that can effectively represent the performance degradation is obtained by using the JADE method. Finally, the prediction model based on PSO-SVM is constructed to predict the performance degradation trend. By comparing with the prediction results obtained by other classical methods, it can be proved that this method can accurately predict the performance degradation trend and the remaining useful life (RUL) of rolling bearings under small sample sizes, and has considerable application potentials.
Accurate and reliable remaining useful life (RUL) assessment result provides decision-makers valuable information to take suitable maintenance strategy to maximize the equipment usage and avoid costly failure. The conventional RUL prediction methods include model-based and data-driven. However, with the rapid development of modern industries, the physical model is becoming less capable of describing sophisticated systems, and the traditional data-driven methods have limited ability to learn sophisticated features. To overcome these problems, a multi-scale deep convolutional neural network (MS-DCNN) which have powerful feature extraction capability due to its multi-scale structure is proposed in this paper. This network constructs a direct relationship between Condition Monitoring (CM) data and ground-RUL without using any prior information. The MS-DCNN has three multi-scale blocks (MS-BLOCKs), where three different sizes of convolution operations are put on each block in parallel. This structure improves the network’s ability to learn complex features by extracting features of different scales. The developed algorithm includes three stages: data pre-processing, model training, and RUL prediction. After the min–max normalization pre-processing, the data is sent to the MS-DCNN network for parameter training directly, and the associated RUL value can be estimated base on the learned representations. Regularization helps to improve prediction accuracy and alleviate the overfitting problem. We evaluate the method on the available modular aero-propulsion system simulation data (C-MAPSS dataset) from NASA. The results show that the proposed method achieves good prognostics performance compared with other network architectures and state-of-the-art methods. RUL prediction result is obtained precisely without increasing the calculation burden.
In practical industrial applications, the working conditions of machinery are changing with long-term operation, and the health status declining with the degradation of crucial components. When the working condition changes, prior diagnosis models cannot be generalized from one condition to another. To solve this challenging issue, a robust weight-shared capsule network (WSCN) is introduced for intelligent fault diagnosis of machinery under varying working conditions. First, taking raw accelerometer signals as inputs, 1D convolutional neural network (1D CNN) is constructed to extract discriminative characteristics. Second, various capsule layers based on multi-stacked weight-shared capsules is developed to enhance the generalization performance for further fault classification. Finally, margin loss function as well as agreement-based dynamic routing algorithm are employed to optimize the WSCN. Two diagnosis cases are carried out to demonstrate the generalization performance of the WSCN which obtains higher accuracy under varying working conditions than that of other state-of-the-art methods.
Rolling bearing is a key component in rotation machine, whose remaining useful life (RUL) prediction is an essential issue of constructing condition-based maintenance (CBM) system. However, recent data-driven approaches for bearing RUL prediction still require prior knowledge to extract features, construct health indicate (HI) and set up threshold, which is inefficient in the big data era. In this paper, a pure data-driven method for bearing RUL prediction with little prior knowledge is proposed. This method includes three steps, i.e., features extraction, HI prediction and RUL calculation. In the first step, five band-pass energy values of frequency spectrum are extracted as features. Then, a recurrent neural network based on encoder–decoder framework with attention mechanism is proposed to predict HI values, which are designed closely related with the RUL values in this paper. Finally, the final RUL value can be obtained via linear regression. Experiments carried out on the dataset from PRONOSTIA and comparison with other novel approaches demonstrate that the proposed method achieves a better performance.
Nowadays, Deep Learning is the most attractive research trend in the area of Machine Learning. With the ability of learning features from raw data by deep architectures with many layers of non-linear data processing units, Deep Learning has become a promising tool for intelligent bearing fault diagnosis. This survey paper intends to provide a systematic review of Deep Learning based bearing fault diagnosis. The three popular Deep Learning algorithms for bearing fault diagnosis including Autoencoder, Restricted Boltzmann Machine, and Convolutional Neural Network are briefly introduced. And their applications are reviewed through publications and research works on the area of bearing fault diagnosis. Further applications and challenges in this research area are also discussed.
There is a constant call for reduction of operational and maintenance costs of induction motors (IMs). These costs can be significantly reduced if the health of the system is monitored regularly. This allows for early detection of the degeneration of the motor health, alleviating a proactive response, minimizing unscheduled downtime, and unexpected breakdowns. The condition based monitoring has become an important task for engineers and researchers mainly in industrial applications such as railways, oil extracting mills, industrial drives, agriculture, mining industry etc. Owing to the demand and influence of condition monitoring and fault diagnosis in IMs and keeping in mind the prerequisite for future research, this paper presents the state of the art review describing different type of IM faults and their diagnostic schemes. Several monitoring techniques available for fault diagnosis of IM have been identified and represented. The utilization of non-invasive techniques for data acquisition in automatic timely scheduling of the maintenance and predicting failure aspects of dynamic machines holds a great scope in future.
In data-driven prognostic methods, prediction accuracy of bearing remaining useful life (RUL) mainly depends on the performance of bearing health indicators, which are usually fused from some statistical features extracted from vibration signals. However, many existing bearing health indicators have the following two shortcomings: (1) many statistical features do not have equal contribution to construction of health indicators since the ranges of these statistical features are different; (2) it is difficult to determine a failure threshold since health indicators of different machines are generally different at a failure time. To overcome these drawbacks, a recurrent neural network based health indicator (RNN-HI) for RUL prediction of bearings is proposed in this paper. Firstly, six related-similarity features are proposed to be combined with eight classical time-frequency features so as to form an original feature set. Then, with monotonicity and correlation metrics, the most sensitive features are selected from the original feature set. Finally, these selected features are fed into a recurrent neural network to construct the RNN-HI. The performance of the RNN-HI is verified by two bearing data sets collected from experiments and an industrial field. The results show that the RNN-HI obtains fairly high monotonicity and correlation values and it is beneficial to bearing RUL prediction. In addition, it is experimentally demonstrated that the proposed RNN-HI is able to achieve better performance than a self organization map based method.
The sensitivity of various features that are characteristics of machine performance may vary significantly under different working conditions. Thus it is critical to devise a systematic feature extraction (FE) approach that provides a useful and automatic guidance on using the most effective features for machine performance prediction without human intervention. This paper proposes a locality preserving projections (LPP)-based FE approach. Different from principal component analysis (PCA) that aims to discover the global structure of the Euclidean space, LPP is capable to discover local structure of the data manifold. This may enable LPP to find more meaningful low-dimensional information hidden in the high-dimensional observations compared with PCA. The effectiveness of the proposed approach for bearing defect and severity classification is evaluated experimentally on bearing test-beds. Furthermore, a novel health assessment indication, Gaussian mixture model (GMM)-based negative log likelihood probability (NLLP) is developed to provide a comprehensible indication for quantifying bearing performance degradation. The proposed approach has shown to provide better performance than using regular features (e.g., root mean square (RMS)). The experimental results indicate potential applications of LPP-based FE and GMM as effective tools for bearing performance degradation assessment.
A method of bearing remaining useful life estimation based on convolutional long short-term memory neural network
- J Wang
- S Yang
- Y Liu
- G Wen
Wang, J., S. Yang, Y. Liu, and G. Wen. A method of bearing
remaining useful life estimation based on convolutional long
short-term memory neural network. Journal of Mechanical
Engineering, Vol. 57, No. 21, 2021, pp. 88-95.