Machinery Diagnosis Research Articles

Fault Diagnosis of Rolling Bearings Based on Variational Mode Decomposition and Genetic Algorithm-Optimized Wavelet Threshold Denoising

Fault diagnosis of rolling bearings can be a serious challenge, as rolling bearings often work under complex conditions and their vibration signals are typically nonlinear and nonstationary. This paper proposes a novel approach to diagnosing faults of rolling bearings based on variational mode decomposition (VMD) and genetic algorithm-optimized wavelet threshold denoising. First, VMD was used to decompose the vibration signals of faulty rolling bearings into a series of band-limited intrinsic mode functions (BLIMFs). During the decomposition, the parameters of VMD were selected by Kullback–Leibler (K–L) divergence. Then, the effective BLIMFs were determined by the analysis of their correlation coefficients and variance contributions. Finally, genetic algorithm-optimized wavelet threshold denoising was proposed to optimize the selection of important parameters, and the optimized threshold function used not only ensures the continuity of the threshold function but also avoids the fixed deviation of the soft threshold. The validity and superiority of the proposed approach were verified by theoretical calculations, numerical simulations and application studies. The results indicate that the proposed approach is promising in fault diagnosis of rotary machinery.

Weak fault diagnosis of machinery using Laplacian eigenmaps and parallel sparse filtering

Machinery fault diagnosis is crucial for maintenance cost reduction and accident prevention. Vibration signal monitoring is an effective and feasible method for machinery fault diagnosis. However, extraction of the fault-related periodic impulses from weakly monitoring signals is basic but difficul . In this paper, a new weak feature extraction model using Laplacian eigenmaps and parallel sparse filtering (LE-PSF) is presented for mechanical weak fault diagnosis. Specifically, the weak vibration signal is measured from the machinery pedestal. Then, LE is used to extract principal components of the overlapped signal segments, and PSF is employed for weak feature extraction from the principal components. Finally, the extracted features are inputted to softmax regression for fault classification. A simulation study and two experimental cases are employed to testify the effect of the LE-PSF model. Experimental performances show that the LE-PSF can not only achieve accurate fault classification but also is superior to other traditional methods.

A transformer model with enhanced feature learning and its application in rotating machinery diagnosis

Deep learning has become the prevailing trend of intelligent fault diagnosis for rotating machines. Compared to early-stage methods, deep learning methods use automatic feature extraction instead of manual feature design. However, conventional intelligent diagnosis models are trapped by a dilemma that simple models are unable to tackle difficult cases, while complicated models are likely to over-parameterize. In this paper, a transformer-based model, Periodic Representations for Transformers (PRT) is proposed. PRT uses a dense-overlapping split strategy to enhance the feature learning inside sequence patches. Combined with the inherent capability of capturing long range dependencies of transformer, and the further information extraction of class-attention, PRT has excellent feature extraction abilities and could capture characteristic features directly from raw vibration signals. Moreover, PRT adopts a two-stage positional encoding method to encode position information both among and inside patches, which could adapt to different input lengths. A novel inference method to use larger inference sample sizes is further proposed to improve the performance of PRT. The effectiveness of PRT is verified on two datasets, where it achieves comparable and even better accuracies than the benchmark and state-of-the-art methods. PRT has the least FLOPs among the best performing models and could be further improved by the inference strategy, reaching an accuracy near 100%.

Semi-supervised fault diagnosis of machinery using LPS-DGAT under speed fluctuation and extremely low labeled rates

Recent research in semi-supervised fault diagnosis of machinery based on graph neural networks (GNNs) still has some problems, such as insufficient label information mining, static feature extraction of neighbor nodes, and relatively ideal diagnosis scenarios. In engineering practice, machinery often runs under speed fluctuation such as start-stop process, and labeling samples becomes increasingly expensive. To deal with the above challenges, a new semi-supervised fault diagnosis method called label propagation strategy and dynamic graph attention network (LPS-DGAT) is proposed in this paper. The designed LPS can take full advantage of the label co-dependency between samples, so as to realize the full utilization of the limited label information. The constructed DGAT by dynamic attention can effectively extract feature information of the different neighbor nodes under speed fluctuation. The proposed method is used to analyze the vibration signals of bearing and gear under speed fluctuation, and the comparison results show that even in the extreme situations where the labeled rates are no more than 1%, the proposed method can still accurately extract discriminative features and diagnose different fault modes, which is better than other GNNs.

A Multisource Dense Adaptation Adversarial Network for Fault Diagnosis of Machinery

Deep learning theory has made great progress in the field of intelligent fault diagnosis, and the development of domain adaptation has greatly promoted fault diagnosis under polytropic working conditions (PWC). Extensive studies have been conducted to solve the problem of fault diagnosis under PWC. However, the existing fault diagnosis methods based on domain adaptation have the following shortcomings. First, multisource information fusion is rarely considered. Second, the utilization of inherent labels is also insufficient in classification problems. To deal with the above problem, a novel multisource dense adaptation adversarial network is proposed, which leverages multisensor vibration information and classification label information. Specifically, the frequency spectrum of multisensor data is first employed to make full use of fault information. Afterwards, the dense convolution and fusion convolution blocks are used for deep feature extraction and fusion. Finally, a joint loss function is reconstructed under the framework of unsupervised learning, which considers the distribution differences of the features and the label information simultaneously. The experimental results from various working conditions, including still distant working conditions, all demonstrate that the proposed method can achieve state-of-the-art performances, which has shown great promise as an intelligent fault diagnosis method.

Robust fault diagnosis of rolling bearing via phase space reconstruction of intrinsic mode functions and neural network under various operating conditions

Rolling bearings are important components in mechanical, civil, and aerospace engineering. The practical working conditions of rolling bearings are complex; hence, fault diagnosis of rolling bearings under various operating conditions is very challenging. This paper proposes a novel approach to fault diagnosis of rotary machinery using phase space reconstruction (PSR) of intrinsic mode functions (IMFs) and neural network under various operating conditions. Complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) is employed to decompose vibration signal of rotary component into IMFs denoting high-to-low instantaneous frequencies adaptively. PSR constructs one-dimensional IMFs to high-dimensional IMFs, which helps reveal the underlying nonlinear geometric topology via the reconstructed inherent and hidden dynamical characteristics of the one-dimensional vibration signal. To explore intrinsic dynamical properties, interquartile range (IQR) of Euclidean distance (ED) values of high-dimensional IMFs are extracted as condition indicators and used as input of back propagation (BP) neural network to fulfill fault identification of rolling bearings. The effectiveness and superiority of the proposed approach have been validated by theoretical derivations, numerical simulations and experimental data. The results show that the proposed approach is promising in fault diagnosis of rotary machinery under various operating conditions.

RADIS: A real-time anomaly detection intelligent system for fault diagnosis of marine machinery

By enhancing data accessibility, the implementation of data-driven models has been made possible to empower strategies in relation to O&M activities. Such models have been extensively applied to perform anomaly detection tasks, with the express purpose of detecting data patterns that deviate significantly from normal operational behaviour. Due to its preeminent importance in the maritime industry to adequately identify the behaviour of marine systems, the Real-time Anomaly Detection Intelligent System (RADIS) framework, constituted by a Long Short-Term Memory-based Variational Autoencoder in tandem with multi-level Otsu’s thresholding, is proposed. RADIS aims to address the current gaps identified within the maritime industry in relation to data-driven model applications for enabling smart maintenance. To assess the performance of such a framework, a case study on a total of 14 parameters obtained from sensors installed on a diesel generator of a tanker ship is introduced to highlight the implementation of RADIS. Results demonstrated the capability of RADIS to be part of a diagnostic analytics tool that will promote the implementation of smart maintenance within the maritime industry, as RADIS detected an average of 92.5% of anomalous instances in the presented case study.

Transferable graph features-driven cross-domain rotating machinery fault diagnosis

Graph data has been integrated into transfer learning-based cross-domain rotating machinery diagnosis for reducing domain discrepancy. Sample relationships, representing the correlations between data distribution and the sample label, have been destroyed in the graph construction process, resulting in transferable knowledge loss. To fully mine and retain the transferable knowledge, a transferable graph features-driven cross-domain rotating machinery fault diagnosis approach is proposed. An improved graph construction strategy is designed to establish the mapping between labels and nodes. Graphs with similar structure for source- and target-domain samples are constructed to preserve sample relationships under data distribution discrepancy. Domain adaptation is introduced to the graph convolutional network for reducing learned graph feature discrepancy. Case studies, including two cross-load and one cross-machine transfer diagnosis tasks, are conducted for effectiveness verification. Experimental results show that it can effectively learn transferable graph features to eliminate the cross-domain discrepancy.

An ensemble and shared selective adversarial network for partial domain fault diagnosis of machinery

Recently, some adversarial transfer learning (TL) approaches have been developed for addressing the partial domain adaptation (DA) problems in machinery fault diagnosis. However, these existing methods generally follow the partial DA framework of multiple sub-domain discriminators, which causes the overly complex model in dealing with many source classes. Moreover, the classifier mostly based on a single intelligent model with limited generalization, it may predict the wrong class-probability for unlabeled samples, which will cause the negative transfer when used in the model training. In response to the challenges, an ensemble and shared selective adversarial network (ES-SAN) is proposed in this paper. In this network, by introducing a correlation layer to correlate both class and domain informations for each sample, a single-intelligent model based shared module is constructed, which can transform between the classifier and the discriminator capable of multi sub-domain discrimination, so as to form a simplified partial DA framework. Moreover, multiple shared modules based on different intelligent models are integrated into an ensemble module, which can output reliable probability weights to promote positive transfer. Experimental investigations on two diagnosis datasets demonstrate that the proposed ES-SAN outperforms the existing methods in the partial DA diagnosis.

Task-incremental broad learning system for multi-component intelligent fault diagnosis of machinery

Broad learning system (BLS) is widely used in intelligent fault diagnosis (IFD) since its high computation efficiency and incremental learning ability. However, its applicability is limited to the single-task learning scenario in which only one diagnosis task needs to be learned. Generally, a machinery system contains multiple critical components that need to be diagnosed. The fault data of different components will be collected at different times for model training. It is essentially a task-incremental learning scenario in which the diagnosis model needs to learn a series of diagnosis task for different components at different times. Existing BLS methods cannot meet this requirement due to the catastrophic forgetting issue. Therefore, this paper proposes a task-incremental broad learning system (TiBLS) for multi-component IFD. The TiBLS is developed as a multi-head configuration with a series of BLS blocks to learn different diagnosis tasks sequentially. Then, the catastrophic forgetting is prevented via parameter isolation. Finally, the structure-incremental learning ability is developed for the TiBLS to enhance the diagnosis performance of each task without retraining. In this way, the TiBLS will gain more and more functions to diagnose different components over time. The experiment validation is implemented on a simulated machinery system including three critical components. The diagnosis accuracies of the three tasks are 94.75%, 93.02%, and 92.00%, respectively. The training times of the three tasks are 22.2 s, 34.5 s, and 7.6 s, respectively. The satisfying results demonstrate that the TiBLS is an effective and efficient method for multi-component IFD.

Convolution enabled transformer via random contrastive regularization for rotating machinery diagnosis under time-varying working conditions

Mechanical equipment such as wind turbines often operates under time-varying working conditions (TVWC). The vibration signals collected from their key rotating components, such as bearings and gears, are often affected by TVWC. In this situation, eliminating the influence of working conditions is the key to effectively implementing end-to-end intelligent diagnosis of rotating machinery (RM). This paper first introduces the transformer encoder architecture to construct a diagnostic model. Then an improved self-attentive module based on a depth-separable convolution operation is built to encode better condition-independent depth diagnostic features and reduce the number of model parameters. Thus, a convolution enabled Transformer (Con-eT) is constructed as a deep encoder for the diagnostic model. Subsequently, a random contrastive regularization (RCR) method inspired by recently proposed contrastive learning is proposed to incentivize the model to learn working-condition-independent features and to improve the model's working condition generalization performance. A publicly available dataset of bearings at time-varying speeds is adopted to verify the effectiveness of the proposed method. Meanwhile, a specific experiment with a wind turbine gearbox designed explicitly for time-varying speed conditions further demonstrates the advantage of the proposed method over the comparative method in terms of the model’s working condition generalization.

Rotating Machinery Fault Diagnosis Using HHO-RSSD and RCGmvMAAPE

Rotating machinery has played an enormous role in industrial production, and its stable operation is related to whether production can proceed smoothly. At present, multichannel entropy-based methods are usually be adopted to analyze multichannel vibration signals. However, the collected signal may only have one channel in the actual situation. At this time, analyzing only a single channel signal cannot effectively utilize the advantages of multivariate analysis. For this reason, this paper presents a novel multivariate analysis approach and applies it to the fault diagnosis of machinery. Firstly, the parameter-optimized resonance sparse decomposition (RSSD) algorithm is adopted to decompose the single-channel vibration signal into high and low resonance components. Then, the two components are regarded as dual-channel vibration signals and input into the refined composite generalized multivariate multiscale amplitude aware permutation entropy (RCGmvMAAPE) method to gain fault features. Eventually, the features are input to the deep belief network (DBN) classifier to perform fault judgment. The experiments of rotating machinery are carried to verify the effectiveness of the developed approach. The results display that the proposed fault diagnosis method can achieve the classification accuracy of 100% and 98% when only a single-channel vibration signal is used, which is better than the fault diagnosis method based on a multichannel vibration signal and enjoys strong stability.

An improved multi-channel graph convolutional network and its applications for rotating machinery diagnosis

Different from most of deep learning-based rotating machinery diagnosis methods, graph convolutional network-based method can effectively mine relationship between nodes in the graph by feature aggregation and transformation. But the performance is limited to graph quality. Currently, edge connections of the graph are often established by calculating the feature similarity of single sensor data. To further improve graph quality, an improved multi-channel graph convolutional network (iMCGCN) for rotating machinery diagnosis is proposed in this paper. Multi-sensor data are used to construct graphs, where corresponding undirected k-nearest neighbor graphs (UK-NNGs) are constructed for each sensor data. A parallel graph data processing framework is designed to extract graph features from the constructed UK-NNGs. Then, an iMCGCN is constructed to learn graph features and achieve multi-channel feature fusion. Case studies are implemented to verify effectiveness of the proposed iMCGCN in learning health features for fault diagnosis.

Conditional Contrastive Domain Generalization for Fault Diagnosis

Data-driven fault diagnosis plays a key role in stability and reliability of operations in modern industries. Recently, deep learning has achieved remarkable performance in fault classification tasks. However, in reality, the model can be deployed under highly varying working environments. As a result, the model trained under a certain working environment (i.e., certain distribution) can fail to generalize well on data from different working environments (i.e., different distributions). The naive approach of training a new model for each new working environment would be infeasible in practice. To address this issue, we propose a novel conditional contrastive domain generalization (CCDG) approach for fault diagnosis of rolling machinery, which is able to capture shareable class information and learn environment-independent representation among data collected from different environments (also known as domains). Specifically, our CCDG attempts to maximize the mutual information of similar classes across different domains while minimizing mutual information among different classes, such that it can learn domain-independent class representation that can be transferable to new unseen domains. Our proposed approach significantly outperforms state-of-the-art methods on two real-world fault diagnosis datasets with an average improvement of 7.75% and 2.60%, respectively. The promising performance of our proposed CCDG on new unseen target domain contributes toward more practical data-driven approaches that can work under challenging real-world environments.

Dual-Weight Consistency-Induced Partial Domain Adaptation Network for Intelligent Fault Diagnosis of Machinery

Domain adaptation-based methods have been broadly developed for cross-domain fault diagnosis of machinery, in which the target and source label spaces are identical. However, a more challenging and practical diagnostic scenario in which the source label space subsumes target one is remaining to be solved. To address this issue, an innovative two-stage dual-weight consistency-induced partial domain adaptation (DCPDA) network, consisting of two feature extractors, a classifier, a domain discriminator, a dual-weight consistency-induced reweighting (DCR) module, and a Wasserstein distance based-domain adaptation (WDA) module, is proposed. In the pre-trained stage, a well-trained model on the source domain is acquired. In the fine-tuned stage, the DCR, which conducts a dual-weight consistency-induced weighting strategy, obtains a bi-level compound weight from both class-level and sample-level weights. Embedded with the bi-level weight, the WDA is encouraged to transfer the source knowledge with mutual agreement from double-level weights across domains. Accordingly, the DCPDA is capable of alleviating negative transfer dramatically and mapping the source diagnostic knowledge into the target domain selectively. Extensive experiments indicate the validity and advantage of the proposed DCPDA.

Automated Model Generation for Machinery Fault Diagnosis Based on Reinforcement Learning and Neural Architecture Search

In recent years, deep learning (DL)-based fault diagnosis methods have demonstrated significant success in various domains due to their high accuracy. Similar to other data-driven techniques, DL models are application-specific and strongly depend on the data which they are developed upon. Different DL models need to be designed for different tasks. In addition, manual-tuning of the DL structures and associated parameters as part of the model design is a trial-and-error optimization process, which is time-consuming and sensitive to changes. To address these limitations, this article presents a reinforcement learning (RL) and neural architecture search (NAS)-based automatic modeling framework (RL-NAS) for fault diagnosis of machinery. The RL-NAS method can adaptively and automatically design high-accuracy network models according to the intended diagnosis tasks. A weight-sharing mechanism has been developed to alleviate the bottleneck of NAS where the computation time increases exponentially with the expansion of the search space and the deepening of the model structure. To evaluate its effectiveness, efficiency, and reproducibility, the proposed framework is validated on five datasets from different institutes, spanning the applications of rolling element bearings, gears, and ball screws. The results show that for all five applications, the RL-NAS method successfully searched out diagnostics models with over 97% accuracy within a short time, from 40 s to 30 min depending on the complexity of the diagnosis task, showing good performance of the proposed method.

An Impedance Sensor for Distinguishing Multi-Contaminants in Hydraulic Oil of Offshore Machinery.

The cleanliness of hydraulic oil can reflect the service life of the oil and the wear state of hydraulic machinery. An impedance sensor is proposed to distinguish multi-contaminants in hydraulic oil. The impedance sensor has two detection modes: the inductance-resistance mode is used to detect metal debris, and the capacitance mode is used to distinguish water droplets and air bubbles. By adding a built-in silicon steel strip and an external silicon steel strip with high magnetic permeability, the distribution area, strength, and uniformity of the magnetic field are enhanced to improve the detection sensitivity under inductance and resistance parameters. In addition, the silicon steel strips are used as electrode plates to introduce capacitance parameter detection. The experimental results show that the resistance detection method based on coil successfully improves the detection ability for non-ferromagnetic metal debris. The impedance sensor for distinguishing multi-contaminants in hydraulic oil can provide technical support for fault diagnosis of offshore hydraulic machinery.

Cycle-consistent Adversarial Adaptation Network and its application to machine fault diagnosis

Driven by industrial big data and intelligent manufacturing, deep learning approaches have flourished and yielded impressive achievements in the community of machine fault diagnosis. Nevertheless, current diagnosis models trained on a specific dataset seldom work well on other datasets due to the domain discrepancy. Recently, adversarial domain adaptation-based approaches have become an emerging and compelling tool to address this issue. Nonetheless, existing methods still have a shortcoming since they cannot guarantee sufficient feature similarity between the source domain and the target domain after adaptation, resulting in unguaranteed performance. To this end, a Cycle-consistent Adversarial Adaptation Network (CAAN) is advanced to realize more effective fault diagnosis of machinery. In CAAN, specifically, an adversarial game formed by the feature extractor and the domain discriminator is constructed to induce transferable feature learning. Meanwhile, the feature translators and discriminators between source and target domains are further designed to build a more comprehensive cycle-consistent generative adversarial constrain, which can more reliably ensure domain-invariant and class-separate characteristics of learned features. Extensive experiments constructed on three datasets from different mechanical systems demonstrate the effectiveness and superiority of CAAN.

Cantilever Type Accelerometer Based on a Mirror-Terminated Hetero-Core Optical Fiber

This paper describes a novel cantilever type accelerometer using a mirror-terminated hetero-core optical fiber for fault diagnosis of machinery. Hetero-core optical fiber sensors are suitable for long-term measurement in on-site fields with attractive advantages of robustness to environmental disturbances such as electromagnetic interference, chemical corrosion, and temperature fluctuation. Moreover, the proposed sensor contained a mirror termination on the hetero-core fiber in order to form a compact and cantilever-shaped structure. It was experimentally found that the mirror-terminated hetero-core fiber worked as Mach-Zehnder interferometry having interference fringes in transmission spectrum, in which linear changes in peak intensities enabled to detect the deflection of fiber beam using a cost-effective LED/PD measurement instrument. For the vibration response characteristics, the sensor reproduced the waveforms of input vibrations to optical loss changes when sinusoidal vibration was applied. Furthermore, the sensitivities were stable in the range of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$5.4\times10^{3}$ </tex-math></inline-formula> to <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$2.3\times10^{2}$ </tex-math></inline-formula> dB/g for the frequency ranges less than 620–130 Hz, when setting a fiber beam length by 6–13 mm. Additionally, finite element simulations indicated that the frequency response characteristics were well correlated with the vibration characteristics of the cantilever-shaped fiber beam, which implied that the sensor performance can be optimized with respects to the sensitivity and the measurable frequency range by changing the fiber beam length.

Feature-Based Multi-Class Classification and Novelty Detection for Fault Diagnosis of Industrial Machinery

Given the strategic role that maintenance assumes in achieving profitability and competitiveness, many industries are dedicating many efforts and resources to improve their maintenance approaches. The concept of the Smart Factory and the possibility of highly connected plants enable the collection of massive data that allow equipment to be monitored continuously and real-time feedback on their health status. The main issue met by industries is the lack of data corresponding to faulty conditions, due to environmental and safety issues that failed machinery might cause, besides the production loss and product quality issues. In this paper, a complete and easy-to-implement procedure for streaming fault diagnosis and novelty detection, using different Machine Learning techniques, is applied to an industrial machinery sub-system. The paper aims to offer useful guidelines to practitioners to choose the best solution for their systems, including a model hyperparameter optimization technique that supports the choice of the best model. Results indicate that the methodology is easy, fast, and accurate. Few training data guarantee a high accuracy and a high generalization ability of the classification models, while the integration of a classifier and an anomaly detector reduces the number of false alarms and the computational time.