Focus Issue on Novel Practical Predictive Maintenance Measurement Solutions for Industrial Systems

Failures to equipment such as milling machines and inertial navigation systems (INSs) can affect their normal operation, resulting in economic losses and personal injury in severe cases. Therefore, fault detection is of great importance. Belief rule base (BRB) is an expert system that plays an important role in fault detection. The traditional BRB has some problems in the explosion of the number of combination rules, the process of model inference, and the process of parameter optimization. To better deal with the above problems, this paper proposes a complex system fault detection method based on an interval-valued BRB fault detection interval-valued (FDIV) and provides the construction and inference process of the method. In the method construction, the form of interval value and disjunction rules are introduced to solve the problem of the number explosion of combination rules, the indicator reliability is added to improve the accuracy of the method, and a new calculation method of rule availability is proposed. In the inference process, twice fusions are made based on evidence reasoning (ER) analysis algorithm and ER rule algorithm respectively to deal with the interval uncertainties. Moreover, the proposed FDIV method is optimized by the projection covariance matrix adaptive evolutionary strategy algorithm projection covariance matrix adaptive evolutionary strategy (P-CMA-ES). Finally, the effectiveness of the proposed method was verified through the research on milling fault detection and the experimental verification of INS fault detection. The superiority of the model was also confirmed through comparative experiments.

This study proposes an efficient rolling bearing fault diagnosis model of a hybrid neural network with a lightweight attention mechanism. Firstly, to achieve the low complexity of deep learning (DL) computation, data reduction and denoising are performed by sparse convolutional network (principal component analysis and improved complete ensemble empirical modal decomposition of adaptive noise), then processed data is imported to the hybrid neural network model with convolutional block attention module. The bi-directional long short-term memory and sparse convolutional networks are used in the backbone of the model. A lightweight, generalized attention mechanism is introduced to the last layer of the model for enhancing feature learning, which can further improve the diagnostic accuracy and efficiency. Compared with existing DL fault diagnosis models, In simulating the most realistic cross-conditions and cross-platform conditions, which leads to the random nature of fault generation and makes model diagnosis more complex, the proposed method still maintains less running time and excellent diagnostic accuracy. Finally, the experimental results fully prove that the model has reliable robust and efficient, and it achieves the best balance of diagnostic accuracy and diagnostic efficiency of the hybrid DL model.

In this paper, a hybrid convolutional neural network (CNN)-bidirectional gated recurrent unit (BiGRU) model is integrated with the bootstrap method to endow the deep learning (DL) based prognostic method with the quantification capability of the prognostic intervals. The proposed hybrid method contains three parts: (I) The complete ensemble empirical mode decomposition with adaptive noise and principal component analysis and the CNN-BiGRU are utilized to automatically construct the health indicator (HI). (II) 3σ criterion is employed to detect the first predicting time based on the HIs of rolling bearings. (III) The bootstrap method is imposed to endow the proposed DL method with the quantification capability of the prognostic intervals. The experimental validation is carried out on the XJTU-SY bearing dataset and the proposed method outperforms the other four methods in the majority of cases. In addition, the proposed method not only comprehensively considers the fault prognosis error caused by model parameters and noise, but also considers the prediction error caused by different combinations of features on the model.

At present, pipeline inspection gauge (PIG) is the main means to ensure the reliability of oil and gas transportation pipelines. Due to the unstable friction between the PIG and the pipeline, the failure and blockage of the PIG often occurs. In this paper, the theoretical modeling and experimental measurement of fluctuating friction on PIG are studied. Firstly, considering the coupling relationship between cylindrical shell and conical shell, the contact radial force model between PIG and pipeline is derived. Based on which, the nonlinear frictional model of PIG is proposed, and the parameters are measured by several experiments. Then an experimental platform is developed to measure the fluctuating friction force of PIG in actual operation, and verify the accuracy of the proposed model. The motion characteristic analysis and dynamics research are carried out, and the PIG creeping phenomenon caused by fluctuating friction is explained. Finally, based on the proposed nonlinear friction model, the influence of parameters is analyzed. The conclusions based on the proposed model and the experiments results can provide technical supports for practical engineering application.

Rolling bearings play a crucial role as components in mechanical equipment. Malfunctioning rolling bearings can disrupt the normal operation of the equipment and pose safety hazards. Traditional deep learning-based methods for diagnosing faults in rolling bearings present several issues, such as insufficient feature information of fault samples, high model complexity and low accuracy. To overcome these challenges, this paper introduces an intelligent approach for rolling bearing fault diagnosis using intrinsic feature extraction and convolutional block attention module (CBAM)-enhanced InceptionNet. In our researches, variational mode decomposition (VMD) is adopted to decompose the original signal into multiple band-limited intrinsic mode functions (BLIMFs). In the decomposition process of VMD, the number of decomposition layers k is determined by center frequency method and the optimal BLIMF is chosen based on minimum envelope entropy. Subsequently, the continuous wavelet transform is employed to transform the optimal BLIMFs into time-frequency images. Finally, the obtained time-frequency images are fed into the proposed CBAM-enhanced InceptionNet for fault state diagnosis. Experiments on two different datasets prove that the method has stable and reliable accuracy. Comparative experiments have demonstrated that this method can reduce network model parameters and improve diagnosis efficiency while achieving high accuracy.

Data-driven machine learning (ML) for rolling bearing remaining useful life (RUL) prediction is a promising method in condition-based maintenance. However, due to the uncertainty of optimal hyperparameter tuning of the ML model, it is very difficult for a data-driven method to accurately predict the RUL of rolling bearings. Aiming to address this problem, this paper proposes a hybrid model-based on continuous wavelet transform (CWT), convolutional neural network (CNN), Bayesian network and long short-term memory network for estimating the remaining usage of rolling bearings lifetime. Firstly, the one-dimensional vibration signal of a bearing is divided into six segments and then it is converted into the corresponding two-dimensional time-frequency feature images via CWT. Secondly, the two-dimensional images are input into the two-dimensional CNN for deep feature extraction in order to obtain a series of one-dimensional feature vectors. Finally, it is input into a Bayesian-optimized long short-term memory model to obtain a prediction of the RUL of the bearing. The effectiveness of the proposed method is verified using bearing data. The verification results show that the proposed method has better prediction accuracy than the other two compared prediction methods, which indicates that the proposed method can effectively extract the bearing fault features and accurately predict the RUL of rolling bearings.

Accurate partial discharge (PD) measurement is critical to ensure the stable operation of transformers. The ultrasonic method is a low-cost, safe, and reliable technology that is widely available and provides real-time monitoring capability. The PD ultrasonic signals propagation is complex and severely attenuated in the transformer, which greatly affects the measurement accuracy of the sensor. In order to improve the accurate monitoring of PD in complicated transformer environments, an optimization monitoring method based on sub-scene detection and quantitative analysis and evaluation is proposed in this paper. Firstly, to address this concern, a sub-scene monitoring method is designed and explores the optimal monitoring points separately. In addition, establish the partition model of an oil-immersed power transformer, and compare the ultrasonic wave propagation characteristics and sound pressure attenuation characteristics of different monitoring points. Then, analyzed by wavelet transform algorithm and Pearson correlation coefficient to determine the best monitoring point location for each scene. Finally, we further tested the proposed method through extensive experiments based on simulations, testbed, and trial deployment. The experimental results have demonstrated the feasibility and accuracy of the proposed method in transformer PD monitoring under complicated environments.

Data-driven intelligent methods have demonstrated their effectiveness in the area of fault diagnosis. However, most existing studies are based on the assumption that the distributions of normal and faulty samples are balanced during the diagnostic process. This assumption significantly decreases the application range of a diagnostic model as the samples in most real-world scenarios are highly unbalanced. To cope with the limitations caused by unbalanced data, this paper proposed an original dual-view network (DVN). Firstly, an interactive graph modeling strategy is introduced for relationship information modeling of multi-sensor data. Meanwhile, the graph convolution operation is used as the baseline for feature extraction of the constructed interactive graph to mine for fault representations. Secondly, an original dual-view classifier consisting of a binary classifier and a multi-class classifier is proposed, which divides fault diagnosis into two stages. Specifically, in the first stage, the binary classifier performs the binary inference from the view of fault detection. In the second stage, the multi-class classifier performs the full-state inference from the view of fine-grained fault classification. Then, based on the dual-view classifier, a weight activation module is designed to alleviate training bias toward majority classes by sample-level re-weighting. Finally, the diagnosis results can be obtained according to the output of the multi-class classifier. Fault diagnosis experiments using two different datasets with varying data unbalance ratios were conducted to validate the effectiveness of the proposed method. The superiority of the proposed DVN is verified through comparisons with state-of-the-art methods. The effectiveness of the DVN is further validated through ablation studies with some ablative models. The DVN code is available at: https://github.com/CQU-ZixuChen/DualViewNetwork.

Complexity measures typically represented by entropy are capable of detecting and characterizing underlying dynamic changes in a system, and they have been considerably studied for machine condition monitoring and fault diagnosis. Various entropies have been developed based on Shannon entropy to meet actual demands. Nevertheless, currently existing research works about complexity measures mainly focus on experimental studies, and their theoretical studies are still ongoing and not fully explored. In previous studies, it was theoretically and experimentally proved that two complexity measures including correlation dimension and approximate entropy have a 'bilateral reduction' effect. Since sample entropy and fuzzy entropy are two more advanced complexity measures that were developed based on the concept of correlation dimension and approximate entropy, this paper continues conducting theoretical and experimental investigations on sample entropy and fuzzy entropy and exploring their theoretical properties to enrich the domain of complexity measure analysis and its applications to machine condition monitoring. Specifically, this paper theoretically proves and verifies that sample entropy and fuzzy entropy still have a similar 'bilateral reduction' effect with correlation dimension and approximate entropy, and they are indeed complexity measures. The relationships between sample entropy, fuzzy entropy, and their key parameters during their calculation are numerically and experimentally studied. Bearing and gear run-to-failure datasets are used to investigate the effectiveness of sample entropy and fuzzy entropy for bearing and gear condition monitoring, and experimental results of sample entropy and fuzzy entropy are well-matched with the theoretical 'bilateral reduction' effect of sample entropy and fuzzy entropy. Overall, this paper will provide a guideline for correct uses of sample entropy and fuzzy entropy for engineering applications, especially for machine condition monitoring.

Rolling bearings play a vital role in ensuring the safe operation of rotating machinery. However, in many application scenarios, the collected data has a low signal-to-noise ratio and the samples with faults are rare, which affects the generalization capability of the model, making it impossible to achieve accurate diagnosis. To solve this problem, the selection of time-frequency (TF) maps was considered in this paper through reinforcement learning. The TF maps are built by four classical TF characterization methods such as short-time Fourier transform and synchro squeezing transform. And the match-reinforcement learning time frequency selection (MRLTFS) fault diagnosis model is proposed to extract the fault-related features. Experiments show that the proposed MRLTFS method is superior to existing methods in robustness, generalization and feature selection capability.

Intelligent fault diagnosis of rotating equipment is increasingly reliant on algorithms that are driven by big data. By contrast, signal processing was once widely utilized for fault diagnosis in machinery as a classical tool for signal analysis due to its capability to investigate the fault-related mechanism and almost no demand on the number of data samples. This investigation was motivated by the notion that signal processing and data-driven algorithms are combined to exploit their respective characteristics and strengths. Furthermore, in engineering practice, numerous complex factors such as time-variable operating conditions of equipment, non-stationary properties of signals, and extremely limited samples available for model training, can make it difficult to learn discriminative features from input data, thereby diminishing the diagnostic accuracy. In this paper, a novel framework of vibration amplitude normalization (VAN) enhanced fault diagnosis is proposed. Firstly, after dissects deeply the effects of the time-varying speed conditions on vibration signal and its characteristics, VAN technique is proposed for non-stationary signal processing to obtain the approximate stationary signal, so as to facilitate the subsequent state characteristics mining from the vibration signal. Then, two VAN enhanced fault diagnosis methods—i.e. signal amplitude normalization integrated with shallow learning by cascade and VAN integrated with deep learning by embedding—are developed to capture discriminative features from approximate stationary signal for fault diagnosis under conditions of variable speed and extremely limited samples. Finally, the feasibility and effectiveness of the proposed methods are verified using actual vibration datasets measured on test rig and in-site wind turbines. The number of samples required to achieve the same diagnostic accuracy is reduced by an average of 60%, demonstrating the superiority.

Maintenance is a critical aspect of complex products through entire life cycle, often requiring coordination of production planning and available resources, while previous studies appear to have rarely addressed. With this in mind, this paper presents a prescriptive maintenance framework based on digital twins (DTs) for reducing operational risk and maintenance costs of complex equipment clusters. Virtual entities are firstly constructed for each single asset in multiple dimensions, which use real-time or historical sensing data collected from the physical entities to predict the corresponding remaining useful life (RUL). Then such RUL information is incorporated into a stochastic programming model with chance constraints to enable dynamic decision making. In particular, a risk-based optimization model is formulated to take full account of the physical distances between facilities and production gaps. Further, a dual-sense pyramidal transformer model is proposed to sense important details of data in both time and space while capturing temporal dependencies at different scales. Compared to existing data-driven approaches, the proposed DT-based alternative achieves dynamic real-time interaction between physical and virtual units driven by both models and data, while virtual verification based on high-fidelity models ensures high reliability of maintenance decisions, which has also been validated in an aero-engine maintenance case study.

To address the shortcomings of existing bearing remaining useful life (RUL) prediction process such as low accuracy and reliance on expert experience for parameter estimation, this paper proposes a bearing RUL prediction method combining relevance vector (RV) machine (RVM) and hybrid degradation model. The bearing degradation characteristics are extracted from the acquired vibration acceleration signals, the time-varying 3σ criterion is then used to determine the bearing first predicting time, and the sequence from initial failure time point to the inspection time is regressed by differential kernel parameter RVM to obtain the different sparse RVs. A mixed degenerate model combined single exponential, weighted double exponential, and polynomial is used to fit the sparse RVs to obtain the fitted curve clusters. The similarity based on bidirectional Hausdorff distance is used to select the best degradation curve, and to extrapolate the best degradation curve to the failure threshold. The experimental results indicate that the proposed method overcomes the widespread drawbacks of monotonicity and trend bias in model-based methods, and has better prediction efficiency than the conventional exponential models.

Fault diagnosis of asynchronous motors has become a pressing need in the metallurgical industry. Due to the complex structure of asynchronous motors, fault types and fault characteristics are diverse, with strong nonlinear relationships between them, which leads to the difficulty of fault diagnosis. To efficiently and accurately diagnose various motor faults, we propose a fault diagnosis method based on an optimal deep bidirectional long short-term memory neural network. First, the three-phase current, multidimensional vibrational signal, and acoustic signal of the asynchronous motor are collected and construct diverse and robust data sample set to enhance the generalization ability of the model. Next, a modified 3D logistics-sine complex chaotic map (3D LSCCM) is constructed to improve the global and local search capabilities of the pigeon swarm optimization algorithm (PIO). Then, we construct a deep bidirectional long short-term memory network (Bid-LSTM) with attention mechanism to mine high-value fault characteristic information. Meanwhile, the optimal hyper-parameters of the deep ABid-LSTM are explored using the modified PIO to improve the training performance of the model. Finally, the fault data samples of asynchronous motor are induced to train and test the proposed framework. By fusing diverse data samples, the proposed method outperforms conventional deep Bid-LSTM and achieves fault diagnosis accuracy of 99.13%. It provides a novel diagnostic strategy for motor fault diagnosis.

Remaining useful life (RUL) prediction plays an important role in prognostics and health management (PHM) and can significantly enhance equipment reliability and safety in various engineering applications. Accurate RUL prediction enables proactive maintenance planning, helping prevent potential hazards and economic losses caused by equipment failures. Recently, while deep learning-based methods have swept the RUL prediction field, most existing methods still have difficulties in simultaneously extracting multiscale global and local degradation feature information from raw multi-sensor monitoring data. To address these issues, we propose a novel multiscale global and local self-attention-based network (MGLSN) for RUL prediction. MGLSN consists of global and local feature extraction subnetworks based on self-attention, which work in parallel to simultaneously extract the global and local degradation features of equipment and can adaptively focus on more important parts. While the global network captures long-term dependencies between time steps, the local network focuses on modeling local temporal dynamics. The design of parallel feature extraction can avoid the mutual influence of information from global and local aspects. Moreover, MGLSN adopts a multiscale feature extraction design (multiscale self-attention and convolution) to capture the global and local degradation patterns at different scales, which can be combined to better reflect the degradation trend. Experiments on the widely used Commercial Modular Aero-Propulsion System Simulation (CMAPSS), New CMAPSS (N-CMAPSS), and International Conference on Prognostics and Health Management 2008 challenge datasets provided by NASA show that MGLSN significantly outperforms state-of-the-art RUL prediction methods and has great application prospects in the field of PHM.

Federal learning based on parameter sharing under the assumption that the data obey independent identical distribution (IID has already achieved good results in areas such as fault diagnosis. Data collected by the decentralized devices often do not obey IID. However, when faced with the scenario of client data obeying Non-IID distribution, its diagnostic accuracy is usually weak. Based on this, we did an investigation on the mechanism causing this phenomenon and found that it was attributed to the weight shift of the network. Therefore, based on the stimulus response principle, we investigated the network similarity of federal clients under different data distributions and explain the reasons for the weight shift. Firstly, it was pointed out that there are differences in the regions where the network is activated when performing different classification tasks. Then, similarity metric federal learning (FedSiM) was proposed based on the principle that there are differences between the activated regions. Finally, experiments were designed on the Case Western Reserve University bearing failure dataset for different degrees of IID cases. The results show that FedSiM can improve the diagnostic accuracy by 15.8 percentage points in the case of Non-IID, and a few shared FedSiM methods to further improve the accuracy were also given.

The aim of this study is to develop a contour extraction method for the investigation of the temporal and spatial characteristics of cavitation evolution in high-speed inducers. Cavitation experiments were conducted, and corresponding snapshots were captured using a high-speed camera subsequently. The raw images were processed while using threshold value filtering techniques to extract only relevant bubble information. Proper orthogonal decomposition was employed to explore the evolving characteristics in a quantitative manner. Results show that the cavitation evolution process is an unsteady process when viewed from the side. In high net positive suction head (NPSH) environments, multiple small-scale spatial modals contribute significantly to the effective proportion of the cavitation structure. The attached sheet cavitation exists in the blade edge which is the main embodiment of cavitation area evolvement. Under low NPSH conditions, the primary flow feature can be reflected by a low-mode flow pattern, associated with the inducer's rotation effect. The evolving regulation of cavitation area in developed stage is closely submitted to the trigonometric function. And the exhibited spatial coefficient exposes the hidden reflux vortex feature. The proposed image separation techniques enable the quantification of morphological features, which can provide technical details for the digital design of inducers.

Imperfect data, such as data scarcity and imbalance, have a negative impact on intelligent fault diagnosis. Generative adversarial networks (GANs) have proven to be a potential method for augmenting data. However, the overall quality of data generated by GANs still requires improvement. In this study, a multiscale feature-fusion GAN was proposed for generating high-dimensional signals. A strategy that generates signals with dimensionality higher than the desired dimensionality was proposed. The generated signals were resampled to improve the phase diversity of the generated signals. To enhance the feature extraction capability of the GAN for high-dimensional signals, a multiscale feature extraction structure was designed. The integration of multiscale feature extraction and fusion was achieved without a significant increase in the computational burden. To enhance the amplitude diversity of the generated signals, a reconstruction network was designed that directly constrained the spatial distribution of the generated signals. Experimental results show that the model has advantages in terms of the similarity and diversity of the generated signals. The effectiveness of the model in fault diagnosis was verified using two motor datasets, where the fault diagnosis model using the sample set augmented by the proposed model obtained a 36.32% improvement in accuracy compared with using the original sample set. In addition, in comparison experiments, the model achieved a higher diagnostic accuracy improvement of 28.87% compared to four other published models.

Ultrasonic in-line inspection is one of the most widely adopted nondestructive testing methods for defect detection in pipelines. In practical industrial scenes, pipeline specification varies, sensor failure occurs frequently, and the number of pipeline defect samples is scarce. How to detect defects without false detection caused by invalid data and with limited labeled samples is a challenging problem in this area. An ultrasonic in-line inspection data processing method considering invalid data caused by sensor failure is proposed in this paper to enhance the accuracy of the defect and its profile detection. Firstly, the multi-channel data is aggregated according to sampling time. The data dimension is reduced, and the accuracy of the invalid waveform detection adopting isolation forest arithmetic is improved. The methods are adopted for the invalid waveforms replacement with two-dimensional cubic spline interpolation using adjacent sensors. Secondly, a natural breakpoint method is adopted to locate the echo peaks. Residual wall thickness is evaluated by calculating the time difference between the echo peaks. A fast pseudo-colorization method based on sliding windows and a morphological image processing method are proposed to detect defects and their profiles efficiently utilizing the residual wall thickness. Finally, practical in-line inspection data is utilized to evaluate the performance. The experiment results illustrate that the detection accuracy is enhanced on different sizes of pipelines without requiring labeled samples.

Due to the complex changes in physicochemical properties of lithium-ion batteries during the process from degradation to failure, it is difficult for methods based on physical or data-driven models to fully characterize this nonlinear process, and existing methods that hybridize physical and data-driven models suffer from ambiguous hybridization, which results in the vast majority of existing methods for predicting the remaining useful life (RUL) of lithium-ion batteries suffering from a lack of accuracy and robustness. In this study, a novel hybrid approach based on empirical modeling and data-driven techniques is proposed for predicting the RUL of lithium-ion batteries. To better capture its complexity, stochasticity, and state transition, and improve the modeling accuracy and RUL prediction precision, Gamma stochasticity and state-space modeling are used to empirically model the complex Li-ion battery degradation process. Moreover, the expectation maximization (EM) method of particle filtering (PF) was used to estimate the hidden parameters of the empirical model, and the estimated parameters were corrected using an optimized support vector regression (SVR) method to enhance the generalization performance and robustness of the data-driven model. The results show that the gamma state-space model is effective in capturing the inherent stochastic properties of the battery degradation and the proposed hybrid method outperforms the existing prediction methods in RUL prediction. The experiments show that the sparrow search algorithm (SSA) optimized SVR is considered to be the most effective correction method for the estimated parameters, while the new EM-PF-SSA-SVR hybrid method provides better performance for state assessment and RUL prediction of lithium-ion batteries. It is indicated that the proposed EM-PF-SSA-SVR method with Gamma stochastic process has hybrid validity and superior performance with equal performance and less parameter computation relative to the existing state-of-the-art deep learning RUL prediction methods.

In industrial scenarios, machinery and equipment operate in environments with many uncontrollable factors. These factors have a significant impact on vibration signals. For instance, differences in rotational speeds can make the collected data inconsistent. As a result, the tags collected from industrial environments are untrustworthy. It is impossible to know how many types of vibration signals are included in the dataset. In this paper, we propose a self-supervised pre-training method for bearing fault diagnosis called contrastive learning via swapped cluster assignments (CLSCA). Technically, CLSCA is based on momentum contrast (MoCo) and uses an unsupervised clustering algorithm to replace the classification of positive and negative samples in MoCo. By autonomously extracting representations from unlabelled datasets and then generating pseudo-labels using a clustering algorithm, CLSCA is able to learn in a self-supervised manner. When the value of the clustering center is in the appropriate range, the classification accuracy of CLSCA is improved by about $4\%$ compared to MoCo.

The hydropower generator unit (HGU) is a complex nonlinear dynamic system, and how to utilize the monitoring data to achieve the performance assessment of the HGU is a challenge at present. In this paper, based on the linear transmissibility function and the nonlinear output frequency response function, the nonlinear transmissibility function (NTF) method is proposed to characterize the variation rule of the HGU performance, and the effectiveness of the proposed method is confirmed by the monitoring data from a hydropower station located in Guangxi, China. Firstly, the feature parameters related to the operation state of the unit are extracted from the monitoring data of the HGU. Secondly, the HGU system is divided into three subsystems according to the installation position of the sensors collecting these feature parameters, and the NTFs of them are estimated from multiple time scales. Finally, the performance of the HGU is comprehensively assessed utilizing the Fe indicator and the correlation transmissibility damage indicator. The analysis results demonstrate that the NTF method can reveal the performance variation trend and detect the early fault of the HGU, which is of great significance to the stable operation of the HGU.

Online anomaly detection (AD) of analog signals plays an important role in equipment fault diagnosis and predictive maintenance. However, the signal often deviates slightly from those seen previously in the early stage of equipment failure, and the anomaly is invisible to the human eye. This kind of anomaly belongs to the typical contextual anomaly. Whether this anomaly can be effectively detected determines whether the failure of the equipment can be detected in the early stage, which is of great significance for safety in production. This study aimed to propose an online AD method for the analog signals of the quasi-sine waveform class. The sample similarity in the sliding window was evaluated using a sample trend rather than sample amplitude deviation to detect anomalies based on the principle that the trend of the quasi-sinusoid waveform signal in the adjacent space was similar. Compared with the traditional method, the proposed method was sensitive to contextual anomalies and did not need a complete sample data set for model training. The proposed method was finally validated by three data sets with good results.

The diagnosis of faults in rolling bearings plays a critical role in monitoring the condition and maintaining the performance of rotating machinery, while also preventing major accidents. In this article, a new approach to diagnosing faults in rolling bearings is proposed, using wavelet packet decomposition (WPD) for features extraction and the chaotic sparrow search optimization algorithms (CSSOAs) to optimize the parameters of a deep belief network (DBN). Firstly, the WPD method is used for the decomposition of vibration signals in rolling bearings, which are decomposed into three layers, and reconstruction is performed on the nodes of the last layer based on the decomposition. Furthermore, the energy characteristics of the reconstructed nodes are then utilized as inputs to DBN, and the CSSOA is employed to optimize the hyperparameters of DBN. Ultimately, a fault diagnosis model combining WPD with optimizing parameters is presented. This model is validated on bearing datasets from Case Western Reserve University (CWRU) and Jiangnan University (JNU). Experimental results indicate that the average accuracy achieved when modeling with WPD-CSSOA-DBN on the CWRU dataset is $98.24\%$, with a root mean square error of 0.0713. On the JNU bearing dataset, the modeling achieves an average accuracy of $95.15\%$ with a root mean square error of 0.1018. Compared to other methods, this approach demonstrates stronger feature extraction capabilities and outstanding rolling bearing fault diagnosis abilities.

A novel operational risk assessment method for hydroelectric generating units (HGUs) is presented in this article. First, a multi-head spatio-temporal attention gated network (MSTAGN) is proposed to establish an operation risk benchmark model for HGUs to reveal the intricate relationship between performance and its multiple influencing factors. In particular, MSTAGN learns complex interaction relationships among multiple influencing factors in both temporal and spatial dimensions and automatically extracts important features. Then, a nonlinear mapping function is constructed to extract the deviation of the current measured performance parameters from the predicted baseline performance parameters as the operation risk degree. On this basis, an adaptive fuzzy clustering algorithm is proposed to achieve a clear classification of the operating risk level for HGUs. The proposed method is applied in a HGU in Sichuan province, China. The results of comparative experiments demonstrate its viability and efficacy.

Given the importance of accelerator devices, major developed and emerging countries around the world have invested substantial human and material resources into the development, construction, and operation of more advanced large-scale accelerator devices. High-precision and high-efficiency alignment of the control network is crucial for the stable operation of accelerators. This study proposes a method for measuring tunnel control networks using total station angle observation. By taking advantage of the high precision of angle observation with a total station, we use the measured results of the designed and deployed control points measured by the laser tracker to simulate the total station angle observations, combined with some length scale values in the control network measured by the laser range finder. With the angle observations and several known distances, the point coordinates can be calculated by angle intersection methods. Only angle observations are used from total stations, while different length scale reference conditions are added to constrain the control network. This is done by calculating the point accuracy of the China Spallation Neutron Source circular accelerator. The overall plane point accuracy can reach 66 μm. By using this method, personnel will spend less time on the site and are at lower risk of radiation exposure, for example by automating measurements at night. This method can be used at nighttime for measurements, avoiding the need for daytime work and shielding employees from severe radiation exposure. By employing this method, efficiency can be increased twice. For on-site measurement, it enables monitoring and automated measurement during operation, as well as providing reference for the installation and measurement of tunnel control networks for large scale accelerators such as the Circular Electron Positron Collider.

As a pivotal element within the drive system of mechanical equipment, the remaining useful life (RUL) of rolling bearings not only dictates the lifespan of the equipment's drive system but also the overall machine. An inaccurate prediction of the RUL of rolling bearings could hinder the formulation of maintenance strategies and lead to a chain of failures stemming from bearing malfunction, culminating in potentially catastrophic accidents. This paper designs a novel temporal convolutional network-multi-head self-attention (TCN-MSA) model for predicting the RUL of rolling bearings. This model considers the intricate non-linearity and complexity of mechanical equipment systems. It captures long-term dependencies using the causally inflated convolutional structure within the temporal convolutional network (TCN) and simultaneously extracts features from the frequency domain signal. Subsequently, by employing the multi-head self-attention (MSA) mechanism, the model discerns the significance of different features throughout the degradation process of rolling bearings by analyzing global information. The final prediction for rolling bearings' RUL has been successfully attained. To underline the excellence of the method presented in this paper, a comparative analysis was performed with existing methods, such as convolutional neural network, gate recurrent unit, and TCN. The results highlight that the model designed in this paper surpasses other existing methods in predicting the RUL of rolling bearings, demonstrating superior prediction accuracy and robust generalization capability.

Planetary gearbox operates under complex working conditions involving high speed, heavy load, and corrosion. When the planetary gearbox is in tight spaces, it is difficult to measure its signal by conventional methods. In this case, acoustic sensors can measure signal with the noncontact method. This paper proposes a vibro-acoustic fault diagnosis method with respect to planetary gearbox. The method addresses challenges related to weak vibro-acoustic signal, difficulty in extracting fault features, and low diagnostic accuracy and efficiency. Firstly, vibro-acoustic signal is captured by a unidirectional microphone. Next, intrinsic wavelet analysis extracts intrinsic features of the planetary gears. The band-limited intrinsic mode functions (BLIMFs) of the acoustic signal are obtained by optimized variational mode decomposition, and the BLIMFs are then transformed into time-frequency map features. Finally, these time-frequency map features are utilized as the inputs for Ghost module and Efficient channel attention module (GE)-improved EfficientNet model, namely GE-EfficientNet model, to achieve fault diagnosis of planetary gearbox. The superiority of the proposed method is verified by the experimental results which show that the diagnostic accuracy of GE-EfficientNet reached 100%, and the floating-point operations and parameter numbers are only 5.1 G and 0.4 MB, respectively. The results demonstrate that the proposed vibro-acoustic fault diagnosis method achieves good diagnostic accuracy and efficiency.

Similarity has been extensively utilized to measure the degree of conflicts between evidences in multisource information fusion. The existent works, however, assumed that the contribution of each focal element's belief to the similarity measure is the same, and the influence of the weights of focal element's belief is not considered, which is unreasonable. This article proposes a new Gaussian kernel similarity approach to measure the similarity between evidences. The proposed Gaussian kernel similarity coefficient can effectively take account of the weights of focal element's beliefs. In addition, it possesses some preferable properties, such as, bounded, consistent, and symmetrical. A multisource information fusion method based on the Gaussian kernel similarity coefficient is, therefore, investigated. The developed method mainly contains three steps: (1) the Gaussian kernel similarity coefficient, as a connection, is leveraged to calculate the weight of evidences based on the weight of focal element's beliefs; (2) the initial evidences are, thereby, modified based on the weight of evidence via the weight-average method; and (3) the final multisource information fusion can be achieved by the Dempster's combination rule using the modified evidences. An illustrative example with single-element subset and an application with multi-element subset are presented, and it is verified that the proposed method is effective in dealing with conflicting evidences.

In practical underground production environments, low light intensity and the presence of dust can disrupt the recognition of coal and gangue. To advance the separation of coal and gangue in underground settings, this paper introduces a coal–gangue recognition approach (DG Module (DGM) + YOLOX-PSB) tailored for conditions of low light intensity and dust. To address the impact of dusty conditions, a preprocessing module (DGM) is proposed. This module employs a dark channel prior dehazing algorithm to mitigate the impact of fog on coal–gangue images. Subsequent steps include white balancing, bilateral filtering, and gamma correction to alleviate noise and distortion issues arising from the dehazing algorithm. To counteract potential drawbacks of the DGM and enhance target recognition accuracy, a polarized self-attention mechanism is integrated during the feature extraction stage to prioritize edge information of coal–gangue targets. By combining the attributes of the weighted bidirectional feature pyramid network, multiple layers of coal–gangue features are efficiently fused to achieve precise identification of coal–gangue targets. Experimental results using a custom dataset demonstrate that the enhanced algorithm outperforms YOLOv3, YOLOv5, YOLOv7-Tiny, and YOLOX, achieving a recognition accuracy of 97.6%, a frames per second rate of 99 and a good smoke concentration robustness. The proposed DGM + YOLOX-PSB serves as a valuable reference for accurate coal and gangue identification in conditions of low light intensity and dust within underground environments.

Pointer meters are widely used in modern industries, such as petrochemical applications, substations, and nuclear power plants. To overcome the reading errors and inaccurate measurements due to uneven or fluctuating illumination in practical applications, this paper proposes an improved UNet++ network for recognizing pointer meter readings. First, the scale invariant feature transform feature-matching algorithm is used to adjust the captured tilted meter images to a symmetrical and upright shape. Then, the UNet++ network is used to segment the scale and pointer regions in the dashboard to eliminate background interference. Furthermore, part of the convolution in the UNet++ network is replaced with dilated convolution with different expansion rates to expand the perceptual field during network training. In the UNet++ network jump connection, the attention mechanism module is also introduced in the path to enhance the region's features to be segmented and suppress the parts of the non-segmented area. A hybrid loss function is used for the network model training to prevent the imbalance of the segmented region share. Finally, the distance method is used to read the gauge representation. Experiments were conducted to compare the performance of the proposed method with that of the original UNet++ network in terms of feasibility and precision. The experimental results showed that the recognition reading accuracy was significantly improved by the enhanced network, with the accuracy, sensitivity, and specificity reaching 98.65%, 84.33%, and 99.38%, respectively. Furthermore, when using the improved UNet++ network for numerical reading, the average relative error was only 0.122%, indicating its robustness in a natural environment.

Batch processes are often characterized by multiphase and different batch durations, which vary from phase to phase presenting multiple local neighborhood features. In this paper, a sequential phase division-multiway sparse weighted neighborhood preserving embedding method is proposed for monitoring batch processes more sensitively. First, batches with uneven durations are synchronized, and the phases are automatically determined in chronological order. Secondly, the nearest neighbors are computed at each phase and the optimal sparse representation (SR) is obtained based on the nearest neighbors. This improves the robustness of the algorithm to noise and outliers, and solves the problem of computational difficulties associated with global SR based. Thirdly, the distance values of the neighbor elements are considered to fully extract the neighbor structure when the optimal SR is calculated. Finally, after dimension reduction, T² and squared prediction error statistics are established in feature space and residual space respectively for fault detection. The effectiveness of the method is verified by a multiphase numerical simulation example and the penicillin fermentation process.

Quality inspection of injection molding products with intricate three-dimensional structures and diffuse reflection characteristics is a very important procedure in industrial production. However, the current inspection process for these products still heavily relies on visual inspection, which introduces various issues including low efficiency, and missing or false detection. While previous studies have utilized deep-learning methods in conjunction with specific optical sensors and imaging systems to detect defects, the intricate structure of injection molding products and the small magnitude of defects pose significant challenges in defect detection. To address these challenges, this paper proposes an inspection system based on Michelson interferometer capable of detecting and characterizing defects of injection molding products. Notably, by utilizing the light intensity modulation and an improved image differencing approach, this inspection system is capable of detecting defects with a magnitude as small as 0.1 mm and achieving a remarkable detection accuracy exceeding 93% on self-made datasets without utilizing phase information. The effectiveness of our method is validated by comparison with mainstream deep-learning-based defect detection methods and visual inspection methods.

As a vital constituent of rotating machinery, rolling bearings assume a pivotal function in ensuring the stable operation of equipment. Recently, deep learning (DL)-based methods have been able to diagnose bearing faults accurately. However, in practical applications, the severe data imbalance caused by the limited availability of fault data compared to the abundance of healthy data poses challenges to the effective training of DL models, leading to a decrease in diagnostic accuracy. In this paper, a bearing fault diagnosis method with the improved residual Unet diffusion model (IResUnet-DM) based on a data generation strategy is proposed to solve the extreme data imbalance. Initially, a deep feature extraction network named improved residual Unet is built, which effectively enhances the information learning ability from vibration signals of the Unet network by one-dimensional residual block and self-attention block. Furthermore, the IResUnet-DM is constructed, which generates vibration signals under extreme data imbalance based on a probability model. The variational bound on the negative log-likelihood of the distribution of generated data was optimized to make the generated data similar to the real data distribution. Finally, wide deep convolutional neural network and one-dimensional ResNet classification networks were used for fault identification to verify the validity and generalization of the IResUnet-DM. Experiment results at different data imbalance rates on two bearing datasets demonstrate that the proposed method can effectively improve fault diagnosis accuracy under extreme data imbalances and outperform the comparison method.

This paper presents a fault diagnosis framework of rolling bearings based on the adaptive dynamic activation convolutional capsule network (CN). The CN is first used to vectorize and mine the spatial information of features aiming at extracting more comprehensive spatial location features. Then, the feature extraction layer of the CN is improved to extract deeper features and reduce the number of parameters. The dynamic activation function is then introduced to extract features better than the steady-state activation function, which can self-adapt the activation features and capture variable feature information. Finally, real rolling bearing data sets are used to verify the superiority and effectiveness of the proposed method with the assistance of comparisons with existing fault diagnosis methods. The results confirmed that the proposed framework has better performance in terms of accuracy and generalization.

Identification of motor vibration signals is one of the important tasks in motor fault diagnosis and predictive maintenance, and wavelet time–frequency diagram is a commonly used signal analysis method to extract the frequency and time characteristics of signals. In this paper, a method based on local sensitive hashing (LSH)-Swin transformer network is proposed for identifying the wavelet time–frequency diagrams of motor vibration signals to analyze the fault types. The traditional Swin transformer model converges slowly due to the smoothing of the attention distribution when dealing with data with sparse features, while the method proposed in this paper reduces the smoothing of the computed attention and enables the network to learn the key features better by introducing locally-sensitive hash attention in the network model, dividing the sequences in the input attention into multiple hash buckets, calculating the attention weights of only some of the vectors with a high degree of hash similarity, and by sampling discrete samples with the use of the Gumbel Softmax. The experimental results show that the method proposed in this paper has better recognition accuracy and higher computational efficiency compared with the traditional network when processing wavelet time–frequency maps of motor vibration signals, and its validation accuracy reaches 99.7%, the number of parameters also has a decrease of about 13%, and the training network to reach converged epochs is also faster. The method in this paper can provide an effective solution for the analysis and processing of motor vibration signals, and has certain application value in practical engineering.

The periodic transient shocks triggered by damages in rolling bearings are frequently overshadowed by disruptive elements such as noise and harmonics. Therefore, the extraction of fault characteristics from these disturbances to identify the health status of the bearing is crucial for fault diagnosis. This study presents a novel approach, the period analysis dictionary weighted sparse representation classification (PAD-WSRC) method, designed specifically for rolling bearings. The proposed approach incorporates Bi-damped wavelet as the dictionary wavelet atom, while accounting for the pulse characteristics induced by faults and leveraging prior knowledge of periodicity. A weighted sparse representation (SR) method was also designed, which calculates the weighted SR coefficients of samples, amplifying the local features of samples while addressing the impact of time-shift bias. In addition, the bearing fault type is identified using a mutual correlation classification criterion based on sparse approximation. Our PAD-WSRC strategy has demonstrated its effectiveness in classifying the health status of bearings across three datasets, achieving recognition accuracies of 99.75%, 99.69% and 99.17%, respectively. Comparisons with several traditional methods further underscore the effectiveness and superiority of our proposed method in classifying rolling bearing faults.

A method based on laser tracing multi-station measurement technology is proposed in this paper. The method identifies the robotic kinematic parameters and compensates for the absolute positioning errors of industrial robots to improve the absolute positioning accuracy further. The position coordinates of industrial robots are typically measured using laser tracking devices. In this study, the measurement accuracy of an industrial robot is further enhanced using laser tracer multi-station measurement technology. Additionally, the least absolute shrinkage and selection operator (LASSO) algorithm was used to identify the robotic kinematic parameters. Compared with the commonly used least squares algorithm, the LASSO algorithm improved the parameter identification accuracy and the compensation effect on absolute positioning errors. A position error model was established based on the parameters of the modified Denavit–Hartenberg model of an industrial robot. Using the LASSO algorithm, the robotic kinematic parameters were accurately identified, and the original data in the controller were replaced to compensate for the geometric errors of the industrial robot. In the compensation experiments, after implementing the geometric error compensation, the average absolute positioning error of the industrial robot decreased by 41.15%, demonstrating a significant improvement in the absolute positioning accuracy.

Bearings are one of the core components of rotating machine machinery. Monitoring their health status can ensure the safe and stable operation of rotating machine equipment. The limited nature of bearing fault samples makes it difficult to meet the demand for sufficient samples based on deep learning methods. Therefore, how to solve the problem of small- samples is the key to achieving intelligent fault diagnosis. In bearing failures based on vibration signals, the complex operating environment causes the vibration signals to inevitably mix with noise. The mixing of fault signature features and noise intensifies the strong spatial coupling of different types of fault features. In addition, diagnosing bearing failures under different loads is challenging because of the complex working conditions of bearings. Given the above problems, a small sample-bearing fault diagnosis method based on a high and low-frequency layered algorithm (HLFLA) and a novel Zernike moment feature attention convolutional neural network (ZMFA-CNN) is proposed. First, the proposed HLFLA converts one-dimensional time series signals into two-dimensional signals distributed rectangularly according to different frequency bands, and is used to simplify network feature screening, reduce the impact of noise, and retain adjacent signal constraint information. In addition, a new ZMFA-CNN is proposed to further extract multi-order moment features and attention weights, and can significantly improve the model generalization ability without increasing model parameters. At the same time, it is combined with FilterResponseNorm2d and thresholded linear unit to further improve model performance. Finally, sufficient experiments verified that the algorithm proposed in this paper can solve the above problems and has excellent transfer generalization ability and noise robustness. In addition, the experimental results of applying the algorithm proposed in this article to gas turbine main bearing fault diagnosis prove the reliability of the algorithm proposed in this article.

The acoustic-based approach is a prevalent way for non-contact fault diagnosis on gas-insulated switch-gear (GIS). GIS always works under different voltages causing great diversity in acoustic frequency. However, based on the frequency principle, neural networks always focus on a specific frequency, which challenges robust fault detection on GIS. This paper introduces a novel multi-stage training method to improve the robustness of fault detection on GIS. The proposed method consists of three components: a multi-channel based frequency regressor (MCBFR), an audio spectrogram transformer auto-encoder (AST-AE), and a feature interaction module (FIM). MCBFR and AST-AE are optimised to extract specific features from acoustics during the pre-training stage. The FIM fuses components extracted by MCBFR and AST when training the model that can indicate the final result. Also, we apply a multi-stage training strategy during the training stage to reduce the cost of potential model retraining. The efficacy of the proposed method was validated using experimental data from a real GIS, and it shows competitive performance in fault detection compared to existing methods.

To address the adverse effects of nonlinearity and dynamic time-varying in complex chemical processes on the accuracy of the soft sensor model, a local-semi-supervised ensemble learning for soft sensor modeling (local semi-supervised-selective ensemble learning-long short term memory, LS-SEL-LSTM) method is proposed in this article. Firstly, a hierarchical clustering method incorporating spatiotemporal criteria is proposed to reduce the influence of nonlinearity in global model prediction accuracy. The method considers the dynamic time-varying characteristics of temporal data and generates multiple local datasets. Then, to address the issue of multi-rate between auxiliary variables and dominant variables, a semi-supervised weight fusion mechanism based on temporal correlation is proposed, which effectively utilizes auxiliary variables to reconstruct local semi-supervised datasets and establishes local soft sensing models using LSTM. Concurrently, the parameters of the established model were optimized using the flower pollination algorithm. Subsequently, a selective ensemble learning method based on sub-model prediction accuracy and an adaptive combination weight calculation method for sub-models were proposed to improve the prediction accuracy. Finally, the effectiveness of the proposed method was verified through the actual dataset of the sulfur recovery process. The results indicate that LS-SEL-LSTM performs well in handling complex chemical processes with nonlinear and dynamic time-varying characteristics.

Rotating machinery (e.g. rolling bearings and gearboxes) is usually operated in high-risk and vulnerable environments such as time-varying loads and poor lubrication. Timely assessment of the operational status of rotating machinery is crucial to prevent damage caused by potential failure and shutdown, which significantly enhances the reliability of mechanical systems, prolongs the service life of critical components in rotating machinery, and minimizes unnecessary maintenance costs. In this regard, in this paper, a novel approach named self-attention mechanism combining time convolutional network with soft thresholding algorithm (SAM-TCN-ST) is proposed for fault intelligent recognition of rotating machinery. Specifically, the vibration signals are transformed into time-frequency graphs with distinct features utilizing the continuous wavelet transform, and then the proposed SAM-TCN-ST algorithm is employed for capturing essential data characteristics and classification performance. Eventually, datasets from rolling bearings and gearboxes are used to verify the accuracy and effectiveness of the proposed method compared with state-of-the-art benchmark networks such as pure TCN, convolutional neural networks and long short-term memory models. Experimental results demonstrate that the recognition accuracy rate of the proposed SAM-TCN-ST is higher than that obtained from the benchmark methods. This research presents an intelligent and viable solution for achieving real-time monitoring of the status and detecting faults in rotating machinery, thereby expectedly enhancing the reliability of mechanical systems. Consequently, the proposed SAM-TCN-ST algorithm holds significant potential for applications in prognostic and health management practices related to rotating machinery.