Search

Abstract Deep learning (DL) is progressively popular as a viable alternative to traditional signal processing (SP) based methods for fault diagnosis. However, the lack of explainability makes DL-based fault diagnosis methods difficult to be trusted and understood by industrial users. In addition, the extraction of weak fault features from signals with heavy noise is imperative in industrial applications. To address these limitations, inspired by the Filterbank-Feature-Decision methodology, we propose a new Signal Processing Informed Neural Network (SPINN) framework by embedding SP knowledge into the DL model. As one of the practical implementations for SPINN, a denoising fault-aware wavelet network (DFAWNet) is developed, which consists of fused wavelet convolution (FWConv), dynamic hard thresholding (DHT), index-based soft filtering (ISF), and a classifier. Taking advantage of wavelet transform, FWConv extracts multiscale features while learning wavelet scales and selecting important wavelet bases automatically; DHT dynamically eliminates noise-related components via point-wise hard thresholding; inspired by index-based filtering, ISF optimizes and selects optimal filters for diagnostic feature extraction. It's worth noting that SPINN may be readily applied to different deep learning networks by simply adding filterbank and feature modules in front. Experiments results demonstrate a significant diagnostic performance improvement over other explainable or denoising deep learning networks. The corresponding code is available at https://github.com/albertszg/DFAWnet .

Bearings are one of the critical components of rotating machinery, and their failure can cause catastrophic consequences. In this regard, previous studies have proposed a variety of intelligent diagnosis methods. Most existing bearing fault diagnosis methods implicitly assume that the training and test sets are from the same distribution. However, in real scenarios, bearings have been working in complex and changeable working environments for a long time. The data during their working processes and the data used for model training cannot meet this condition. This paper proposes an improved adversarial transfer network for fault diagnosis under variable working conditions. Specifically, this paper combines an adversarial transfer network with a short-time Fourier transform to obtain satisfactory results with the lighter network. Then, this paper employs a channel attention module to enhance feature fusion. Moreover, this paper designs a novel domain discrepancy hybrid metric loss to improve model transfer learning performance. Finally, this paper verifies the method’s effectiveness on three datasets, including dual-rotor, a Case Western Reserve University dataset and the Ottawa dataset. The proposed method achieves average accuracy, surpassing other methods, and shows better domain alignment capabilities.

To extract the global temporal correlations and local features together to enhance the accuracy for fault diagnosis, this paper proposes an effective convolutional Transformer (ECT), which can learn the global temporal correlations using Transformer and local features with convolution at the same time. The proposed method designs a multi-stage hierarchical structure of Transformer, which utilizes convolutional tokenization to distill dominating sequence features from raw vibration signals while increasing the dimension of token embedding across stages at the same time as that in CNNs. The spatial-reduction attention (SRA) and the linear dimension reduction projections are introduced respectively to Transformer at different stages to reduce the resource consumption of the model. Finally, the proposed method utilizes a sequence pooling strategy on the output of Transformer to eliminate the requirement of the class token and make the model accurate for classification. The specially designed structure makes the model flexible and effective for planetary gearbox fault diagnosis. Experiments performed on planetary gearbox fault simulators indicate that the ECT method has significant effectiveness and high accuracy compared with the state-of-the-art methods for planetary gearbox fault diagnosis.

Abstract As an integrated application of modern information technologies and artificial intelligence, Prognostic and Health Management (PHM) is important for machine health monitoring. Prediction of tool wear is one of the symbolic applications of PHM technology in modern manufacturing systems and industry. In this paper, a multi-scale Convolutional Gated Recurrent Unit network (MCGRU) is proposed to address raw sensory data for tool wear prediction. At the bottom of MCGRU, six parallel and independent branches with different kernel sizes are designed to form a multi-scale convolutional neural network, which augments the adaptability to features of different time scales. These features of different scales extracted from raw data are then fed into a Deep Gated Recurrent Unit network to capture long-term dependencies and learn significant representations. At the top of the MCGRU, a fully connected layer and a regression layer are built for cutting tool wear prediction. Two case studies are performed to verify the capability and effectiveness of the proposed MCGRU network and results show that MCGRU outperforms several state-of-the-art baseline models.

Abstract Prognostics and Health Management (PHM), including monitoring, diagnosis, prognosis, and health management, occupies an increasingly important position in reducing costly breakdowns and avoiding catastrophic accidents in modern industry. With the development of artificial intelligence (AI), especially deep learning (DL) approaches, the application of AI-enabled methods to monitor, diagnose and predict potential equipment malfunctions has gone through tremendous progress with verified success in both academia and industry. However, there is still a gap to cover monitoring, diagnosis, and prognosis based on AI-enabled methods, simultaneously, and the importance of an open source community, including open source datasets and codes, has not been fully emphasized. To fill this gap, this paper provides a systematic overview of the current development, common technologies, open source datasets, codes, and challenges of AI-enabled PHM methods from three aspects of monitoring, diagnosis, and prognosis.

Abstract Gear fault diagnosis technologies have received rapid development and been effectively implemented in many engineering applications. However, the various working conditions would degrade the diagnostic performance and make gear fault diagnosis (GFD) more and more challenging. In this paper, a novel model parameter transfer (NMPT) is proposed to boost the performance of GFD under varying working conditions. Based on the previous transfer strategy that controls empirical risk of source domain, this method further integrates the superiorities of multi-task learning with the idea of transfer learning (TL) to acquire transferable knowledge by minimizing the discrepancies of separating hyperplanes between one specific working condition (target domain) and another (source domain), and then transferring both commonality and specialty parameters over tasks to make use of source domain samples to assist target GFD task when sufficient labeled samples from target domain are unavailable. For NMPT implementation, insufficient target domain features and abundant source domain features with supervised information are fed into NMPT model to train a robust classifier for target GFD task. Related experiments prove that NMPT is expected to be a valuable technology to boost practical GFD performance under various working conditions. The proposed methods provides a transfer learning-based framework to handle the problem of insufficient training samples in target task caused by variable operation conditions.

Abstract Most gear fault diagnosis (GFD) approaches suffer from inefficiency when facing with multiple varying working conditions at the same time. In this paper, a non-negative matrix factorization (NMF)-theoretic co-clustering strategy is proposed specially to classify more than one task at the same time using the high dimension matrix, aiming to offer a fast multi-tasking solution. The short-time Fourier transform (STFT) is first used to obtain the time-frequency features from the gear vibration signal. Then, the optimal clustering numbers are estimated using the Bayesian information criterion (BIC) theory, which possesses the simultaneous assessment capability, compared with traditional validity indexes. Subsequently, the classical/modified NMF-based co-clustering methods are carried out to obtain the classification results in both row and column tasks. Finally, the parameters involved in BIC and NMF algorithms are determined using the gradient ascent (GA) strategy in order to achieve reliable diagnostic results. The Spectra Quest’s Drivetrain Dynamics Simulator gear data sets were analyzed to verify the effectiveness of the proposed approach.

Rotating blade is one of the most important components for turbomachinery. Blade crack is one of the most common and dangerous failure modes for rotating blade. Therefore, the fault mechanism and feature extraction of blade crack are vital for the safety assurance of turbomachinery. This study is aimed at the nonlinear dynamic model of rotating blade with transverse crack and the prior feature extraction of blade crack faults based on the vibration responses. First and foremost, a high-fidelity breathing crack model (HFBCM) for rotating blade is proposed on the basis of criterion for stress states at crack section. Since HFBCM is physically deduced from the perspective of energy dissipation and the coupling between centrifugal stress and bending stress is considered, the physical interpretability and the accuracy of the crack model are enhanced comparing with conventional models. The validity of the proposed HFBCM is verified through the comparison study among HFBCM, conventional crack models, and finite element-based contact crack model (FECCM). It is suggested that HFBCM behaves best among the analytical models and matches well with FECCM. With the proposed HFBCM, the nonlinear vibration responses are investigated, and four types of blade crack detection indicators for rotating blade and their quantification method are presented. The numerical study manifests that all these indicators can well characterize the occurrence and severity of crack faults for rotating blade. It is indicated that these indicators can serve as the crack-monitoring indexes.

With the rapid technological development and production requirement, machines and equipment in modern industry, such as advanced manufacturing, transportation, aerospace, and civil infrastructure, become increasingly functional and complex. Machine fault diagnosis plays a significant role for the productivity, reliability, and safety of industrial systems. In the recent decade, data-driven solutions have become more effective and promising for fault diagnosis of complex machines due to increasing data availability and processing capacity. Artificial intelligence (AI) techniques, especially deep learning approaches, are the most powerful tools in achieving accurate fault diagnosis of complex systems. However, AI-based fault diagnosis still faces great challenges in feasibility and reliability for real applications, including the lack of fault condition data, varying working conditions, insufficient generalization capability of AI models, the black-box nature of most AI methods, etc. This special section focuses on advanced and innovative solutions to address a broad view of problems about AI-based machine fault diagnosis. Six out of 18 submissions included in this special section summarize some of the research and applications in this field.

This paper presents an improved cross recurrence quantitative analysis (CRQA) method for bearing degradation evaluation. It extracts CRQA variable from bearings' vibration signals to establish the degradation trend. Due to the fact that traditional process for calculating the hyperparameters of CRQA requires a large amount of time to reconstruct phase space, this step is replaced by calculating divergence rate of each signal sample to cut down the time consumption. After establishing the degradation trend based on the extracted variable, nonlinear auto-regressive neural network (NARNN) is developed to predict future degradation trend. Furthermore, a new failure detection method is investigated to indicate the first failure point during degradation tracking. The method applies temperature signals as auxiliary information to calculate the adaptive threshold and classify extracted features into different health status. The experiments on bearing vibration signals have verified that the improved CRQA method can reduce time consumption by more than 90%. In addition, defect characteristic frequencies extracted using wavelet analysis have validated the detection accuracy.

This paper aims at proposing an abnormality detection framework for electrocardiogram (ECG) signals, which owns unbalance distribution among different classes and gaining high accuracy in rhythm/morphology abnormalities classification. The proposed framework is composed of two models: data augmentation model and classification model. In this framework, data augmentation model is designed to recast a class-balanced training dataset by generating artificial data of minor class. The outputs of augmentation model are transferred into classification model. The classification model is designed to identify abnormalities accurately after training using both the experimental and generated datasets. Data augmentation model is supported by auxiliary classifier generative adversarial network (ACGAN). We construct Generator and Discriminator of the ACGAN by stacking multiple 1-dimensional convolutional layers with small kernel size. Dropout function and batch normalization are added to prevent gradients vanish and speed up convergence. In order to evaluate the performance of augmentation model, a set of quantitative indicators are introduced to verify the quality of generated ECG signals. We establish classification model based on stacked residual network parallel connected with long short-term memory (LSTM) network. The experimental study is conducted for single heartbeat detection and consecutive heartbeat detection. The results based on standard benchmark, MIT-BIH, and competition database provided by 2018 China physiological signal challenge (CPSC) have verified the proposed framework can achieve high performance in robustness and accuracy for class-imbalanced dataset.

Multiline spectral vibration (or noise) is very common in rotating machinery like motor, gearbox, compressor, propeller, etc. Active reshaping of these vibrations is one of the most important branches for active vibration control, which may find application in fields of sound quality control, psychoacoustics, military camouflage, etc. The traditional filtered-reference LMS is the most popular structure for multiline spectral reshaping control. However, with the increase of spectral lines, vibration sources and observation points (feedback channels), the traditional structure will have extremely heavy computational complexity, especially for the system whose dynamic of secondary path is complex, since it requires two long filters for every single spectral line and secondary path to adjust the reference signals. In this study, we propose a fast-multiline spectrum reshaping algorithm for active vibration control. We first construct an equivalent system of the traditional multiline spectrum reshaping algorithm by introducing two extra control branches, which is able to improve the convergence property of the system. Then, we generate a group of pre-phase-scheduled reference signals using the phase information at certain frequencies of the secondary paths instead of using filters. Afterwards, we conduct a theoretical analysis of computational complexity of multiline spectrum reshaping and fast multiline spectrum reshaping algorithms. Finally, we carried out two case studies based on the FEMilS test-bed to verify the effectiveness of the proposed algorithm. The results show that the proposed algorithm performs an accurate residual vibration control with uniformed convergence speed and reduced computational complexity.

Premature ventricular contraction (PVC) is one of the common ventricular arrhythmias, which may cause stroke or sudden cardiac death. Automatic long-term electrocardiogram (ECG) analysis algorithms could provide diagnosis suggestion and even early warning for physicians. However, they are mutually exclusive in terms of robustness, generalization and low complexity. In this study, a novel PVC recognition algorithm that combines deep learning-based heartbeat template clusterer and expert system-based heartbeat classifier is proposed. A long short-term memory-based auto-encoder (LSTM-AE) network was used to extract features from ECG heartbeats for K-means clustering. Thus, the templates were constructed and determined based on clustering results. Finally, the PVC heartbeats were recognized based on a combination of multiple rules, including template matching and rhythm characteristics. Three quantitative parameters, sensitivity (Se), positive predictive value (P+) and accuracy (ACC), were used to evaluate the performances of the proposed method on the MIT-BIH Arrhythmia database and the St. Petersburg Institute of Cardiological Technics database. Se on the two test databases was 87.51% and 87.92%, respectively; P+ was 92.47% and 93.18%, respectively; and ACC was 98.63% and 97.89%, respectively. The PVC scores on the third China Physiological Signal Challenge 2020 training set and hidden test set were 36,256 and 46,706, respectively, which could rank first in the open-source codes. The results showed that the combination strategy of expert system and deep learning can provide new insights for robust and generalized PVC identification from long-term single-lead ECG recordings.

Shock response spectrum (SRS) test is a kind of vibration test that uses the principle of equivalent damage to simulate complex shock and vibration environment for evaluating product fragility. However, for a certain SRS, neither an analytical nor a unique inverse time-domain shock waveform (TSW) exists, making the SRS test a very challenging problem. Synthesis of a TSW using a group of wavelets with different frequencies, amplitudes, and phases is considered to be a very promising method. However, it is challenging to find an optimal group of wavelets since there are hundreds of optimization variables and objective functions. In this paper, a novel intelligent parameter matching method for TSW synthesis is proposed for the SRS test. A variable-weight method has been introduced to combine the effects of hundreds objective functions so that the optimization problem can be solved by using the genetic algorithm. The frequency response function of the shaking system has been taken into consideration in the calculation of the objective function, so the synthesized TSW can be applied directly to the SRS test. In the optimization process, normalized control factors have been formulated for the optimization variables so that they can be tuned in a reasonable small range to avoid irrationality and accelerating convergence in searching process. The effectiveness of the proposed method has been verified experimentally in two active vibration control systems, in which one contains a 500 N minishaker and the other contains a 1-ton shaking table. It can be seen from the experimental results that the proposed method can accurately synthesize the input TSW for the shaking system, where the output TSW and SRS can accurately meet the time- and frequency-domain specification.

Ocular contamination of EEG data is an important and very common problem in the diagnosis of neurobiological events. An effective approach is proposed in this paper to remove ocular artifacts from the raw EEG recording. First, it conducts the blind source separation on the raw EEG recording by the stationary subspace analysis, which can concentrate artifacts in fewer components than the representative blind source separation methods. Next, to recover the neural information that has leaked into the artifactual components, the adaptive signal decomposition technique EMD is applied to denoise the components. Finally, the artifact-only components are projected back to be subtracted from EEG signals to get the clean EEG data. The experimental results on both the artificially contaminated EEG data and publicly available real EEG data have demonstrated the effectiveness of the proposed method, in particular for the cases where limited number of electrodes are used for the recording, as well as when the artifact contaminated signal is highly non-stationary and the underlying sources cannot be assumed to be independent or uncorrelated.

This paper presents an improved gearbox fault diagnosis approach by integrating complementary ensemble empirical mode decomposition (CEEMD) with permutation entropy (PE). The presented approach identifies faults appearing in a gearbox system based on PE values calculated from selected intrinsic mode functions (IMFs) of vibration signals decomposed by CEEMD. Specifically, CEEMD is first used to decompose vibration signals characterizing various defect severities into a series of IMFs. Then, filtered vibration signals are obtained from appropriate selection of IMFs, and correlation coefficients between the filtered signal and each IMF are used as the basis for useful IMFs selection. Subsequently, PE values of those selected IMFs are utilized as input features to a support vector machine (SVM) classifier for characterizing the defect severity of a gearbox. Case study conducted on a gearbox system indicates the effectiveness of the proposed approach for identifying the gearbox faults.

In this research, a semi-analytical model of the adaptive piezoelectric metamaterial, built upon continuum mechanics characterization, was formulated and analyzed to reveal the fundamental features of bandgap with respect to unit-cell parameters under transverse wave. A new mechanism to broaden the bandgap width, was then introduced through geometric cavity synthesis. It was demonstrated that the cavities incorporated into the host structure of the piezoelectric metamaterial can increase the electro-mechanical coupling of the system, which effectively yields broadened bandgap width. Case studies were performed to demonstrate the enhanced performance of the new design, as well as the tunability. Compared with the conventional piezoelectric metamaterial, the metamaterial with cavity synthesis can increase the bandgap width from 45 Hz to 126.7 Hz.

This paper presents an enhanced rolling bearing fault diagnosis approach, based on optimized wavelet packet transform (WPT) assisted with quantitative wavelet function selection. Mutual information is utilized as a quantitative measure to select the most suitable wavelet function for the WPT-based vibration analysis. Energy features from coefficients of an optimal set of orthogonal wavelet subspaces which resulted from the WPT-based vibration analysis are input to different classifiers. The fault states of the rolling bearings can then be identified. Experiment studies conducted on a rolling bearing test system have verified the effectiveness of the proposed approach for rolling bearing fault diagnosis.

The imbalance between limited organ supply and huge potential need has hindered the development of organ-graft techniques. In this paper a low-cost hypothermic machine perfusion (HMP) device is designed and implemented to maintain suitable preservation surroundings and extend the survival life of isolated organs. Four necessary elements (the machine perfusion, the physiological parameter monitoring, the thermostatic control and the oxygenation apparatus) involved in this HMP device are introduced. Especially during the thermostatic control process, a modified Bayes estimation, which introduces the concept of improvement factor, is realized to recognize and reduce the possible measurement errors resulting from sensor faults and noise interference. Also, a fuzzy-PID controller contributes to improve the accuracy and reduces the computational load using the DSP. Our experiments indicate that the reliability of the instrument meets the design requirements, thus being appealing for potential clinical preservation applications.

In modern manufacturing systems and industries, more and more research efforts have been made in developing effective machine health monitoring systems. Among various machine health monitoring approaches, data-driven methods are gaining in popularity due to the development of advanced sensing and data analytic techniques. However, considering the noise, varying length and irregular sampling behind sensory data, this kind of sequential data cannot be fed into classification and regression models directly. Therefore, previous work focuses on feature extraction/fusion methods requiring expensive human labor and high quality expert knowledge. With the development of deep learning methods in the last few years, which redefine representation learning from raw data, a deep neural network structure named Convolutional Bi-directional Long Short-Term Memory networks (CBLSTM) has been designed here to address raw sensory data. CBLSTM firstly uses CNN to extract local features that are robust and informative from the sequential input. Then, bi-directional LSTM is introduced to encode temporal information. Long Short-Term Memory networks(LSTMs) are able to capture long-term dependencies and model sequential data, and the bi-directional structure enables the capture of past and future contexts. Stacked, fully-connected layers and the linear regression layer are built on top of bi-directional LSTMs to predict the target value. Here, a real-life tool wear test is introduced, and our proposed CBLSTM is able to predict the actual tool wear based on raw sensory data. The experimental results have shown that our model is able to outperform several state-of-the-art baseline methods.

Milling vibration is one of the most serious factors affecting machining quality and precision. In this paper a novel hybrid error criterion-based frequency-domain LMS active control method is constructed and used for vibration suppression of milling processes by piezoelectric actuators and sensors, in which only one Fast Fourier Transform (FFT) is used and no Inverse Fast Fourier Transform (IFFT) is involved. The correction formulas are derived by a steepest descent procedure and the control parameters are analyzed and optimized. Then, a novel hybrid error criterion is constructed to improve the adaptability, reliability and anti-interference ability of the constructed control algorithm. Finally, based on piezoelectric actuators and acceleration sensors, a simulation of a spindle and a milling process experiment are presented to verify the proposed method. Besides, a protection program is added in the control flow to enhance the reliability of the control method in applications. The simulation and experiment results indicate that the proposed method is an effective and reliable way for on-line vibration suppression, and the machining quality can be obviously improved.

This paper presents an experimental investigation on virtual texture force perception, where the Just Noticeable Difference (JND) method is used to evaluate the effects of texture spatial period and human exploratory speed on texture force perception. Two experiments have been conducted in this study. The first experiment focused on JNDs of texture force under different spatial periods (0.055–90mm) without limiting subjects' exploratory speeds. It was found that JNDs of texture force were between 6.20% and 9.80%. The spontaneous exploratory speed was highly affected by spatial period. It increased and then became stable as the spatial period increased. The second experiment assessed the effect of spatial period and human exploratory speed on JNDs of texture force. The spatial period had a significant effect on JNDs of texture force, which decreased as spatial period increased in the range of 2.9–14.9mm. Although the subject's exploratory speed has no noticeable effect on JNDs of texture force, the effect of interaction between spatial period and exploratory speed was significant. The experiment result indicated that when texture spatial period is small, low exploratory speed can be chosen to get small JND of texture force.