Fault Diagnosis Performance Research Articles

Compound fault diagnosis for industrial robots based on dual-transformer networks

The accurate diagnosis of the compound fault of industrial robots can be highly beneficial to maintenance management. In the actual noisy working environment of industrial robots, the mixed and feeble failure features are easy to be overwhelmed, which poses a major challenge for the industrial robot compound fault diagnosis. Meanwhile, in the existing studies, a large-size deep learning model is the guarantee of decent denoising and fault diagnosis performance. However, this demands expensive computational costs and large data samples, which are not always available. In order to address both challenges, in this study, an integrated approach that contains two compact Transformer networks is proposed to achieve accurate compound fault diagnosis for industrial robots. In this approach, the feedback current signals collected from a six-axis industrial robot are first transformed into time-frequency image representation via continuous wavelet transformation (CWT). Secondly, a novel deep learning algorithm called compact Uformer is proposed to denoise the time-frequency image. Subsequently, the denoised time-frequency images are fed into compact convolutional Transformer (CCT) for compound fault diagnosis. An experimental study based on a real-world industrial robot compound fault dataset was conducted. The experimental results reveal that the proposed method can achieve satisfactory compound fault diagnosis accuracy based on the data collected from the noisy environment in comparison with the state-of-the-art algorithms.

Noise-robust machinery fault diagnosis based on self-attention mechanism in wavelet domain

Machinery fault diagnosis in noisy scenes is important and challenging to ensure the safety and stability of machinery. Existing machinery fault diagnosis methods mostly handle noisy signals in the time or frequency domains. However, these methods lack fully consideration of the noise characteristics in both domains, resulting in unstable performance. This paper proposes an anti-noise Wavelet based Self-attention Network for machinery fault diagnosis, named Wavelet-SANet. Our approach combines the frequency-oriented fusion modules and the Transformer modules to restrain noise in both frequency and time domains. Firstly, the multiple frequency-oriented fusion modules exploit a self-attention mechanism to weight different frequency features of noisy signals in the wavelet domain for filtering the persistent noise. Furthermore, the Transformer modules are designed to suppress the remaining scattered impulse noise in the time domain. The experimental results on two public bearing datasets demonstrate the promising performance for machinery fault diagnosis. Experimental results on CWRU and SEU datasets indicate the superior performance of Wavelet-SANet on different SNRs.

A two‐stage transformer fault diagnosis method based multi‐filter interactive feature selection integrated adaptive sparrow algorithm optimised support vector machine

AbstractThe scarcity of samples and disunity of feature inputs hinder the enhancement of transformer fault diagnosis performance, and there are mutual influences between model construction and feature selection, which cannot only consider a single process. Therefore, this study proposes a novel two‐stage transformer fault diagnosis strategy, which includes a multi‐filter interactive feature selection method (MIFS) constructed, and a diagnosis model ASSA‐SVM based on the adaptive sparrow algorithm (ASSA) optimised support vector machine (SVM). Firstly, the proposed MIFS incorporates ReliefF and mRMR to establish a comprehensive criterion ReliefF‐mRMR for feature importance ranking, and then performs dimension‐by‐dimension input classifier interaction selection based on the ranking results to obtain the optimal feature subset. Secondly, ASSA was proposed to optimise the kernel parameters of SVM. A two‐stage integration model MIFS‐ASSA‐SVM was developed. Finally, Experiments were conducted using real fault data, and the diagnostic performance of different feature inputs, optimisation algorithms and classifiers were compared. The results show that the proposed method performs well on parameter optimisation, can dynamically and interactively select feature subsets with few dimensions and good generalisation performance, its overall diagnosis accuracy reached 92.47%, and the diagnosis performance of each fault type has good performance in multiple evaluation metrics.

Deep Negative Correlation Multisource Domains Adaptation Network for Machinery Fault Diagnosis Under Different Working Conditions

Machinery fault diagnosis is crucial to ensure production safety. Deep learning has been widely investigated in intelligent fault diagnosis. However, the mechanical equipment generally runs in a complex and polytropic condition. It is challenging to perform machinery fault diagnosis under polytopic conditions. In this article, a novel multisource domain adaptation method is proposed for machinery fault diagnosis under different working conditions. First, a deep negative correlation multisource domains adaptation network (DNC-MDAN) is proposed, where the information from multisource domains is transferred into target domain based on the multisource feature alignment and adversarial learning. Second, in order to make full of the multisource domains information, an ensemble classifier corresponding to multisource domains is developed, where a DNC learning is performed in the ensemble classifier. Third, an end-to-end feature generator, discrete cosine convolutional block, is proposed in DNC-MDAN, where discrete cosine transform is embedded in the network smoothly for noise reduction. In order to validate the effectiveness of DNC-MDAN for machinery fault diagnosis, three cases are considered in this article. In comparison with other transfer learning methods, DNC-MDAN has a better fault diagnosis performance.

Stepwise Adaptive Convolutional Network for Fault Diagnosis of High-Speed Train Bogie Under Variant Running Speeds

Fault diagnosis of high-speed train (HST) bogie plays an indispensable role in guaranteeing the safety and stability of the daily operation of HST. Up to now, the well-established fault diagnosis models of HST bogie, which are usually trained by supervised learning mechanism, require that the training and testing data satisfy the same probability distribution. Consequently, those methods could lose their efficacy when the probability distribution of the testing data is changed owing to the variation of HST running speed and the labeled data corresponding to the changed speed is absent. To address this, a stepwise adaptive convolutional network (SACN) is proposed to learn the domain-invariant features of vibration data in different speed domains. Moreover, a novel in-and-out stepwise transfer method is designed to deal with the scenario of continuous change in vehicle speed. The experimental tests of the proposed method are conducted using the dataset acquired by SIMPACK via the HST model CRH380 A, mainly addressing the single and compound failure classification problem of the three key components of HST bogie, namely, the air spring, antiyaw damper, and lateral damper. Overall, the proposed SACN achieves an average accuracy of 96.1%, demonstrating the remarkable performance of domain adaptation in performing fault diagnosis of HST bogie under variant running speeds.

A lightweight model for train bearing fault diagnosis based on multiscale attentional feature fusion

As one of the key components of a train, the running gear bearing has the highest fault rate, and its health condition is very important for the safe operation of the train. Therefore, how to quickly and accurately diagnose the health condition of the train running gear bearings under strong noise and variable working conditions has become one of the core contents of the intelligent operation and maintenance strategy. To meet these requirements, a lightweight convolutional neural network based on multiscale attentional feature fusion (MA-LCNN) is proposed in this paper, which takes the inverted residual network as the main structure. Firstly, a multiscale attention module (MA) was designed to extract fault feature information. Secondly, by embedding MAs in different locations, the ability of the MA-LCNN to extract fault feature information is greatly improved. Finally, an ablation experiment and noise resistance experiment are performed. The recognition accuracy scores of the MA-LCNN for cases 2 and 3 are 99.70% and 99.83%, respectively. The results show that the proposed attention module has better learning ability and stability compared to the contrast modules. The MA-LCNN demonstrates better fault diagnosis performance than contrast models under different noise environments and variable working conditions.

Vibration Image Representations for Fault Diagnosis of Rotating Machines: A Review

Rotating machine vibration signals typically represent a large collection of responses from various sources in a machine, along with some background noise. This makes it challenging to precisely utilise the collected vibration signals for machine fault diagnosis. Much of the research in this area has focused on computing certain features of the original vibration signal in the time domain, frequency domain, and time–frequency domain, which can sufficiently describe the signal in essence. Yet, computing useful features from noisy fault signals, including measurement errors, needs expert prior knowledge and human labour. The past two decades have seen rapid developments in the application of feature-learning or representation-learning techniques that can automatically learn representations of time series vibration datasets to address this problem. These include supervised learning techniques with known data classes and unsupervised learning or clustering techniques with data classes or class boundaries that are not obtainable. More recent developments in the field of computer vision have led to a renewed interest in transforming the 1D time series vibration signal into a 2D image, which can often offer discriminative descriptions of vibration signals. Several forms of features can be learned from the vibration images, including shape, colour, texture, pixel intensity, etc. Given its high performance in fault diagnosis, the image representation of vibration signals is receiving growing attention from researchers. In this paper, we review the works associated with vibration image representation-based fault detection and diagnosis for rotating machines in order to chart the progress in this field. We present the first comprehensive survey of this topic by summarising and categorising existing vibration image representation techniques based on their characteristics and the processing domain of the vibration signal. In addition, we also analyse the application of these techniques in rotating machine fault detection and classification. Finally, we briefly outline future research directions based on the reviewed works.

Deep Scattering Spectrum Germaneness for Fault Detection and Diagnosis for Component-Level Prognostics and Health Management (PHM)

Most methodologies for fault detection and diagnosis in prognostics and health management (PHM) systems use machine learning (ML) or deep learning (DL), in which either some features are extracted beforehand (in the case of typical ML approaches) or the filters are used to extract features autonomously (in the case of DL) to perform the critical classification task. In particular, in the fault detection and diagnosis of industrial robots where the primary sources of information are electric current, vibration, or acoustic emissions signals that are rich in information in both the temporal and frequency domains, techniques capable of extracting meaningful information from non-stationary frequency-domain signals with the ability to map the signals into their constituent components with compressed information are required. This has the potential to minimise the complexity and size of traditional ML- and DL-based frameworks. The deep scattering spectrum (DSS) is one of the approaches that use the Wavelet Transform (WT) analogy for separating and extracting information embedded in a signal's various temporal and frequency domains. Therefore, the primary focus of this work is the investigation of the efficacy and applicability of the DSS's feature domain relative to fault detection and diagnosis for the mechanical components of industrial robots. For this, multiple industrial robots with distinct mechanical faults were studied. Data were collected from these robots under different fault conditions and an approach was developed for classifying the faults using DSS's low-variance features extracted from input signals. The presented approach was implemented on the practical test benches and demonstrated satisfactory performance in fault detection and diagnosis for simple and complex classification problems with a classification accuracy of 99.7% and 88.1%, respectively. The results suggest that, similarly to other ML techniques, the DSS offers significant potential in addressing fault classification challenges, especially for cases where the data are in the form of signals.

Performance evaluation of IEC 60891:2021 procedures for correcting I–V curves of photovoltaic modules under healthy and faulty conditions

AbstractCorrection of PV modules' current–voltage characteristics (I–V curves) is essential before they can be used for performance analysis and fault diagnosis under real‐life conditions. IEC 60891 (version 2021) has updated Procedure 2 and proposed a new correction Procedure 4 compared to the 2009 version. This study aims to analyze the performance of these new procedures applied to I–V curves of faulty PV modules. The work is based on an mc‐Si PV module considering healthy and four common fault conditions with varying fault severity. The irradiance and temperature measured in the field are used to generate I–V curves. The correction procedures of IEC 60891 (version 2021) based on a single curve (Procedures 1, 2, and 4) are evaluated. Environmental factors such as measurement season and irradiation level on the correction performance are studied. The results show that Procedure 2 is relatively better with the relative root‐mean‐square error of the curve current as 2.6% compared to Procedure 1 (2.8%) and Procedure 4 (4.8%). Experimental tests using real I–V curves also show that Procedure 2 exhibits better robustness of correction. The new Procedure 4, whose correction coefficients are determined dynamically, performs poorly under partial shading and short‐circuit bypass conditions. However, it achieves similar or better performance than Procedures 1 and 2 under degraded conditions, where the PV fault is generally not easy to detect, making the I–V correction more necessary. It is, therefore, a promising alternative correction procedure when it is difficult to determine the correction coefficients in advance. Finally, the pros and cons of the procedures are discussed with the suggestion of correction procedures under different conditions. The challenges and prospects are also provided.

The fault frequency priors fusion deep learning framework with application to fault diagnosis of offshore wind turbines

In fault diagnosis, deep learning plays an important role, but still lacks good interpretability. To address this issue, we develop a novel fault frequency priors fusion deep learning (FFP-DL) framework by introducing fault frequency priors into deep learning. The FFP-DL framework contains two branches: fault frequency priors learning branch (FFPLB) and self-learning branch (SLB). We then propose a pre-training algorithm which can shorten the overall training time especially for training multiple models simultaneously. To illustrate its efficacy, we take convolutional neural network (CNN) as the specific deep learning model in the FFP-DL framework (FFP-CNN), and apply the FFP-CNN model to a private offshore wind turbines (OWTs) data. The experimental results show that the FFP fusion does help improve the performance of fault diagnosis in terms of accuracy and Marco-F1-score and provide good interpretability to the diagnosis results with the distinguished feature of predicted FFP. With the training data reduction, the performance of the FFP-CNN model does not deteriorate quickly, which implies that this framework is also suitable for less data. In addition, the result reveals the fact that the pre-training algorithm does reduce convergence epochs, which will help the FFP-CNN model train faster during the training process.

Feature-based statistical process monitoring for pressure swing adsorption processes

Pressure swing adsorption (PSA) is a widely used technology to separate a gas product from impurities in a variety of fields. Due to the complexity of PSA operations, process and instrument faults can occur at different parts and/or steps of the process. Thus, effective process monitoring is critical for ensuring efficient and safe operations of PSA systems. However, multi-bed PSA processes present several major challenges to process monitoring. First, a PSA process is operated in a periodic or cyclic fashion and never reaches a steady state; Second, the duration of different operation cycles is dynamically controlled in response to various disturbances, which results in a wide range of normal operation trajectories. Third, there is limited data for process monitoring, and bed pressure is usually the only measured variable for process monitoring. These key characteristics of the PSA operation make process monitoring, especially early fault detection, significantly more challenging than that for a continuous process operated at a steady state. To address these challenges, we propose a feature-based statistical process monitoring (SPM) framework for PSA processes, namely feature space monitoring (FSM). Through feature engineering and feature selection, we show that FSM can naturally handle the key challenges in PSA process monitoring and achieve early detection of subtle faults from a wide range of normal operating conditions. The performance of FSM is compared to the conventional SPM methods using both simulated and real faults from an industrial PSA process. The results demonstrate FSM’s superior performance in fault detection and fault diagnosis compared to the traditional SPM methods. In particular, the robust monitoring performance from FSM is achieved without any data preprocessing, trajectory alignment or synchronization required by the conventional SPM methods.

Fault Diagnosis Method Based on Two-Stage GAN for Data Imbalance

In the field of fault diagnosis, a large number of sensors are used to reflect the working status of equipment with the increasing complexity of mechanical systems. However, collecting enough fault monitoring data, which leads to the widespread problem of data imbalance and continues effect on the improvement of fault diagnosis accuracy, is difficult for some devices in the actual work process. To improve the performance of fault diagnosis under imbalance, first, this article proposes a generative model based on generative adversarial networks (GANs), the first stage of which uses time-series GAN to overcome the problem, wherein traditional GAN cannot effectively extract the time–series information of sequence data, and the second stage incorporates the characteristics of supervised learning of auxiliary classifier GAN to solve the problem, wherein it cannot generate multimode samples. Second, this article also studies the fault diagnosis method based on two-stage GAN (2S GAN) and applies the classification model combining continuous wavelet transform and ResNet18 to the field of mechanical fault diagnosis. The method includes the data preprocessing stage, data augmentation, and model training stages. Finally, this method is validated by two datasets, and the results show that the absolute increments of the accuracy of this method are improved by 5.70%, 6.34%, 9.08%, and 16.35% in case 1 under four imbalance ratios and by 3.36%, 4.65%, 15.93%, and 26.34% in case 2. Thus, it provides an effective solution to the fault diagnosis problem in the data imbalance state.

A Fault Diagnosis Method of Rolling Bearing Based on Attention Entropy and Adaptive Deep Kernel Extreme Learning Machine

To address the difficulty of early fault diagnosis of rolling bearings, this paper proposes a rolling bearing diagnosis method by combining the attention entropy and adaptive deep kernel extreme learning machine (ADKELM). Firstly, the wavelet threshold denoising method is employed to eliminate the noise in the vibration signal. Then, the empirical wavelet transform (EWT) is utilized to decompose the denoised signal, and extract the attention entropy of the intrinsic mode function (IMF) as the feature vector. Next, the hyperparameters of the deep kernel extreme learning machine (DKELM) are optimized using the marine predators algorithm (MPA), so as to achieve the adaptive changes in the DKELM parameters. By analyzing the fault diagnosis performances of the ADKELM model with different kernel functions and hidden layers, the optimal ADKELM model is determined. Compared with conventional machine learning models such as extreme learning machine (ELM), back propagation neural network (BPNN) and probabilistic neural network (PNN), the high efficiency of the method proposed in this paper is verified.

Comparative study on deep transfer learning strategies for cross-system and cross-operation-condition building energy systems fault diagnosis

Timely and accurate fault diagnosis (FD) in building energy systems (BESs) can promote energy efficiency and sustainable development. Especially the heating, ventilating, and air-conditioning (HVAC) systems are diverse and operate under complex and variable operation conditions. System and operation differences lead to great differences in operational data which causes poor adaptability of data-driven FD models that are developed using data from a single HVAC system or limited operation condition. To improve diagnostic performance across different HVAC systems and operation conditions, this study proposes high-adaptability FD models using three deep transfer learning (DTL) strategies including network-based fine-tuning (FT), mapping-based domain-adaptive neural network (DaNN) and adversarial-based domain adversarial neural network (DANN). The effectiveness of DTL-based FD is validated by fault datasets of two typical BESs: one is a 703-kW screw chiller while the other is the 316-kW centrifugal chiller from ASHRAE RP-1043. Two types of TL scenarios (cross-system and cross-operation-condition fault diagnosis) are set up consisting of eight TL tasks. For DTL strategies, both FD performance and transferability are evaluated using metrics like accuracy and accuracy improvement degree (AID). Results indicate that FT obtains 93% FD accuracy averagely for all tasks of the two TL scenarios considered, which is an average 55% AID compared with the non-transfer benchmark model convolutional neural network (CNN). Further, the impacts of source and target data volumes, and TL tasks are analyzed. For cross-operation-condition scenario, DTL-based FD accuracy grows with the increase of target data volume. For cross-system scenario, FT still show high FD performance with less training data. The reason why FT outperforms DANN and DaNN is explained by visualizing classification scatterplots of the last NN layers. Practical application issues of the DTL-based FD strategy for building energy systems are discussed at last.

Binomial adversarial representation learning for machinery fault feature extraction and diagnosis

The quality of feature extraction is a key factor in determining the performance of machinery fault diagnosis. The feature extraction of conventional deep learning-based methods has the disadvantages of uncontrollability and low quality. To overcome these disadvantages, we propose a binomial adversarial representation learning (BARL) method. Considering that the binomial distribution can make the representations have smaller intra-class distances (ICDs) and larger cross-class distances (CCDs), the adversarial learning mechanism and autoencoder are combined to learn representations obeying binomial distributions from raw monitoring signals to extract key features containing machine health information. Two case studies on rolling element bearings and gearboxes are carried out to validate the performance of the proposed method. The results show that the diagnosis accuracy of BARL outperformed comparison methods, especially under the condition of insufficient training data. According to quantitative and qualitative analysis, the representations learned by BARL were proven to have smaller ICDs and larger CCDs than the comparison methods, which illustrates the effectiveness of the proposed method.

Intelligent rolling bearing imbalanced fault diagnosis based on Mel-Frequency Cepstrum Coefficient and Convolutional Neural Networks

To improve the bearing fault diagnosis performance under the condition of data distribution shift, an intelligent diagnosis method based on MFCC (Mel-Frequency Cepstrum Coefficient) and MECNN (Convolutional Neural Networks optimized by Mode Normalization (MN) and Efficient Channel Attention (ECA)) is proposed. Firstly, Mel filters are adopted to extract the feature of different frequency bands of vibration signal, and by the feature enhancement of Cepstrum Lifting Technique, the final 2D MFCC is obtained. Secondly, MN is applied to reduce the internal covariant shift caused by the data distribution discrepancy, and improve the generalization ability. ECA is adopted to enhance the fault feature and improve anti-interference ability. Finally, experiments under data distribution shift have been carried out, and an average accuracy of 99.72% was obtained under the data imbalance, and 99.50% was obtained under the operating condition change. Compared with the existing methods, the proposed has higher accuracy and better domain adaptability.

A novel complex network community clustering method for fault diagnosis

The complex network, as a method for the analysis of nonlinear and non-stationary signals, overcomes the shortcomings of traditional time-frequency analysis methods and proves its effectiveness in mechanical fault diagnosis. Community clustering, a type of complex network, has made great progress in recent years. However, the existing community clustering algorithms have disadvantages in that they lack significant global extreme value and huge search spaces. Therefore, a Fast Newman algorithm based on reliability judgment is proposed. Starting from the community structure characteristics of the complex network, with the fault sample as a network node, the relationship between the samples as a connected edge and a complex network model of fault data is established. Clusters in troubleshooting are transformed into community structure discovery in the network. Firstly, the initial division of the community is obtained by measuring the distance between the samples. Then, the modularity index of the network is used as a standard function of the community division, and the bottom-up community merger is performed. The local edge density index is used for reliability determination before each combination to achieve global optimization, and the network block structure is the most obvious. Finally, with all the data merged into one community, the optimal division of the community structure is obtained, while accurate community clustering and fault diagnosis is realized. The benchmark graphs for testing community detection (Lancichinetti–Fortunato–Radicchi benchmark standard test network, LFR) and different fault data of rolling bearings under multiple operating conditions are applied to verify the effectiveness of this method; the results prove that the modified Fast Newman algorithm has better clustering effects and a higher accuracy rate than the original algorithm. Compared with K-means clustering and fuzzy clustering, the modified Fast Newman algorithm achieves higher performance in fault diagnosis of rolling bearings under multiple operating conditions.

Deep discriminative transfer learning network for cross-machine fault diagnosis

Many domain adaptation methods have been presented to deal with the distribution alignment and knowledge transfer between the target domain and the source domain. However, most of them only pay attention to marginal distribution alignment while neglecting the discriminative feature learning in two domains. Thus, they still cannot satisfy the diagnosis requirement in some cases. To enhance the distribution alignment and match the marginal distributions as well as conditional distributions of two domains, an improved joint distribution adaptation (IJDA) mechanism is proposed. In IJDA, to enhance domain confusion, maximum mean discrepancy and CORrelation Alignment (CORAL) are combined as a new distribution discrepancy metric. Furthermore, an improved conditional distribution alignment mechanism is constructed. To contribute to feature learning and learn more separable features, a new I-Softmax loss that can be optimized like the original Softmax loss and possesses a stronger classification ability is proposed. With the IJDA mechanism and I-Softmax loss, the deep discriminative transfer learning network (DDTLN) is built to implement fault transfer diagnosis. Under the unlabeled target-domain samples, the experimental results on six cross-machine diagnostic tasks verify that the proposed DDTLN has a higher performance of transfer fault diagnosis than other typical domain adaptation methods.

An attention-enhanced multi-modal deep learning algorithm for robotic compound fault diagnosis

Compound fault diagnosis plays a critical role in lowering the maintenance time and cost of industrial robots. With the advance of deep learning and industrial big data, a compound fault diagnosis model can be established through a data-driven approach. However, current methods mainly focus on the single fault diagnosis of assets, which cannot achieve satisfactory performance for compound fault diagnosis. This study proposes a compound fault diagnosis algorithm for an industrial robot based on multi-modal feature extraction and fusion. Firstly, the multi-head self-attention enhanced convolution neural network module and long short-term memory network module are adopted to learn the fault-related features from different perspectives simultaneously. The local and global features extracted by the aforementioned modules are then fused for subsequent compound fault classification. An experimental study was implemented based on real-world robotic sensor data. The experimental results indicated that the proposed multi-modal algorithm shows merits in compound fault diagnosis in comparison with other state-of-the-art methods.

Multi-channel data fusion and intelligent fault diagnosis based on deep learning

In complex manufacturing systems, multi-channel sensor data are usually recorded for fault detection and diagnosis. Most existing multi-channel data processing methods adopt tensor analysis technology, which cannot effectively describe the temporal and spatial structures of the multi-channel data. The obstacles in multi-channel data analysis are the temporal correlation between the time-series data of the single-channel and the spatial correlation between different channels. In this paper, a novel deep convolutional neural network model is proposed for multi-channel data fusion and intelligent fault diagnosis. First, features of the multi-channel data are extracted from two scales. The extracted features are then fused through a multi-layer neural network. Finally, a classifier of fault modes is established by using the improved Softmax function. The fault diagnosis performance of the proposed model is evaluated and compared with other common methods in both the simulation studies and real-world case studies. Results show that the proposed methodology has superior fault diagnosis performance for multi-channel data.