Defect inversion is a critical quantitative assessment technique to evaluate the integrity of energy transportation systems. In practice, non-stationary states and frequent situation switches in real-world industrial processes make it difficult to acquire ideal sensor signals with precise annotations, there by compromising accurate defect inversion. Simulation systems is a viable solution. However, the dramatic inter-domain disparity between simulated and real system poses challenges to existing studies. Moreover, intra-domain divergence within the real energy transportation systems is also unexplored. To address the above issues, an intra- and inter-domain interactive learning method is proposed for simulation-to-real (sim2real) defect inversion. Specifically, reality-aware feature stylization is proposed which diversifies simulated signals with real style variants, so as to alleviate the domain gap initially. Then, bi-level inter-domain alignment is proposed to not only penalize hard-to-align samples in abstract feature space, but also innovatively exploits intrinsic geometric relations of sensor signals to guide inter-domain gaps. Next, physical-informed intra-domain alignment is proposed where a similarity matrix is constructed to pursue the consistency between sensor signals and physical knowledge, and it is also served as pseudo-labels to constrain inter-domain adaptation, so that inter- and intra- domain adaptation interact with united strength to enhance information exchange. Finally, the proposed method is systematically validated on energy transportation experimental platform and competitive results are obtained, which shows its promise in real-world industrial processes.

Incremental learning is a critical yet challenging problem in automation engineering, especially across heterogeneous domains. Existing incremental learning methods utilize mixup and mosaic techniques to replay previous knowledge. However, the saliency discrepancy between replayed samples is often ignored, leading to suboptimal accuracy. To address this problem, we propose a novel replay-based incremental learning framework that replays previous defect samples based on mixup and mosaic. To dynamically assign samples to different image augmentation techniques, a saliency score calculation mechanism is proposed by using posterior probability and spatial location. To eliminate the step noise brought by mixup, we propose a Gaussian operator-improved mixup to smoothly merge previous foregrounds with current images. Additionally, a domain-attentive distillation loss is proposed to further alleviate forgetting. Experiments on three scenarios verify the improved accuracy of the proposed method, achieving improvements of 2.4% mAP on weld defect detection, 2.7% mAP on steel defect detection and 4.0% mAP on no-service rail defect detection, respectively.

This article explores predictor-based fractional-order sliding mode load frequency control for interconnected power systems, accounting for input delay. First, a predictor-based method is developed to deal with the input delay. By designing a predictor to accurately predict the future state, the delayed control input can be replaced by the delay-free control input. Then, a novel fractional-order sliding mode controller is designed, where the predictor is used rather than the system state, reducing the dependence on directly measurable system states and enhancing the dynamic response of the system. Furthermore, a disturbance observer is designed to estimate the disturbance, allowing the controller to correspondingly compensate for it, and the robustness of the system is improved. Finally, three cases are conducted to show the validity of the presented method.

Magnetic flux leakage (MFL) detection has been widely used in the pipeline integrity evaluation. However, the MFL signals will be distorted due to the speed effect, which reduces the accuracy of the defect size inversion severely. To solve above problems, a multiple information-guided collaborative network (MIG-CN) is proposed to invert defect sizes accurately for the distorted pipeline MFL signals caused by the speed effect. First, a multiview feature fusion method is proposed, which enables MIG-CN to predict the defect sizes of specific types accurately by fusing the features of specific views, where the mechanism features are injected into the network to fully exploit the physical information of the MFL signals, improving the inversion accuracy. Second, a physics-informed constrained network is proposed, in which a speed magnetic dipole model is developed to impose constraints from real physical laws with the speed effect, thus enhancing the physical plausibility of the inversion results. Third, a dynamic-static collaborative decision strategy is proposed, in which a two-point constrained quantile probability modelling is proposed to model the inversion results of the undistorted MFL signals and obtain the prior probability distribution. The accuracy and robustness of the inversion results at different speeds are improved by minimizing the difference between the probability distribution of the inversion results of the distorted MFL signals and those of the undistorted. The above three components are jointly trained and optimized to improve the inversion accuracy under the influence of speed effect. Finally, the experiments on real pipeline networks verifies the effectiveness of MIG-CN.

Remarkable progress has been achieved in the detection and segmentation of the baseline; however, for high-level visual tasks in complex scenes (e.g., dense, occlusion, scale diversity, high background noise, etc.), existing frameworks often fail to provide satisfactory performance. To further improve the object recognition ability, this article introduces a leader-based multiexpert mechanism into the detection and segmentation tasks. In this work, we first design a leader-based attention learning layer to fully integrate multilevel features from the backbone network, which can effectively obtain global semantics and assign instructions to detection experts. Then, we propose multiple feature pyramids with dual fusion paths to replace the traditional single pipeline using semantic and spatial allocators. With this strategy, we can further establish deep supervision for multiple experts during training and sufficiently utilize the multiexpert detection results from leaders' assignments during reasoning, thereby comprehensively improving the performance of the model in complex scenarios. In the experiment, we established ablation studies and performance comparisons on COCO 2017 detection and segmentation tasks. Finally, we demonstrated the model's performance in three complex application scenarios (remote sensing, autonomous driving, and industrial fields), and the results showed our advantages.

Magnetic flux leakage (MFL) testing is an effective non-destructive testing (NDT) method for pipeline safety operation, and defect detection is one of the core issues in MFL signal processing. Currently, MFL defect detection is tough due to insufficient defect samples. To obtain more serviceable information, heterogeneous signals are collected in MFL process, while taking full advantage of heterogeneous signals is still a hard problem. In this paper, an end-to-end MFL defect detection architecture with insufficient training samples called heterogeneous signal fusion method (IHSF) is proposed. Firstly, heterogeneous signals from the same pipeline are collected by axial and radial sensors. Secondly, the features from fine-tuned and non-fine-tuned pre-trained models are fused, which increases the generality and adaptability of the features. Moreover, fine-tuning of a few parameters reduces the number of parameters during model training, which is more suitable for the insufficient sample training. Thirdly, a mutual supervision training strategy based on the features fusion of general and adaptive features is proposed to update the parameters of the proposed network, which establishes the latent relationship of heterogeneous signals. Finally, experiments on MFL defect detection are conducted, and IHSF is compared to the state-of-the-art methods. The results validate that the proposed method is effective.Note to Practitioners—The motivation of this paper is a hot signal processing issue on pipeline safety operation called defect detection with insufficient samples. Traditionally, the defect is usually detected by training a large number of defect samples, while defect samples are usually insufficient in practical MFL measurement. Moreover, conventional deep networks only use a single signal, which ignores information from multiple signals. Regarding the problems above, a heterogeneous signal mutual supervision method is proposed to replace the traditional fixed label mechanism. The mutual supervision method fully extracts the information in multiple signals by establishing hidden connections between defective heterogeneous signals. Furthermore, the fusion of the general features and adaptive features, which are extracted by the pre-trained model with non-fine-tuning and fine-tuning respectively, not only reduces the parameters in the network to be suitable for the insufficient sample problem, but also enriches the extracted features. The case on MFL signals demonstrates that the proposed method is more effective than the comparison methods.

Defect recognition optimizes industrial manufacturing processes by monitoring the health of equipment and structures. The success of current deep learning (DL)-based defect recognition methods relies on the assumption that the samples are label-noise free, yet the low-quality of industrial data inevitably introduces label noise, resulting in a substantial drop in recognition performance. To tackle this problem, this paper develops a parallel hybrid penalty network (PHL-Net). It solves two essential issues in defect recognition under label noise: low sample utilization and difficulty in recognizing the ambiguous sample in the decision boundary region, so that the industrial defect recognition performance can be greatly enhanced under label noise. PHL-Net consists of three stages: data selection (DS) for explicitly selecting clean and noisy data; parallel feature mining (PFM) for improving the sample utilization by label-driven mining and distance-driven mining; hybrid penalty learning (HPL) to improve decision reliability to label noise, thereby the ambiguous samples in the decision boundary region can be accurately recognized. In PHL-Net, PFM and HPL are co-optimized to facilitate data selection. In turn, reliable data selection prevents the model from overfitting label noise. Three groups of experimental results on real-world industrial datasets MPV-W and NEU-CLS have demonstrated the effectiveness of our PHL-Net over state-of-the-art alternatives. Note to Practitioners—Defect recognition plays a vital role in industrial production, ensuring product quality and efficiency. When training deep learning-based defect recognition models, having datasets with precise annotations is essential. Unfortunately, label errors, known as label noise, are common when annotating industrial datasets due to the need for domain expertise, subjective judgments, and confusing samples. This makes it challenging to create a large-scale and accurate industrial dataset. To address this issue, our goal is to develop a robust model that can achieve optimal recognition performance even on datasets with noisy labels. On this basis, this paper proposes a parallel hybrid penalty network (PHL-Net). It improves recognition performance under label noise by improving sample utilization and recognizing ambiguous samples near the decision boundary. In experiments, we utilize the noise transition matrix to introduce label noise for MPV-W and NEU-CLS. The results of the experiments demonstrate that our PHL-Net outperforms the comparison methods on industrial datasets, highlighting the practical significance of our approach. This provides inspiration and reference for industrial defect recognition under label noise.

This research investigates the problem of exponential tracking consensus for nonlinear switched multi-agent systems (NSMASs) under limited communication resources. To alleviate the communication load, edge-event-triggered strategies are applied to switched multi-agent systems for the first time. Due to the nonlinear switched dynamics in NSMASs, addressing the edge-event-triggered tracking consensus problem is more challenging compared to systems with linear or nonlinear fixed dynamics explored in previous studies. To tackle this challenge, the novel dynamic event-triggered rule for the leader and novel dynamic edge-event-triggered rules for edges are developed, both relying on model-dependent designs. Different triggering conditions are established for diverse system models to achieve enhanced triggering performance. Furthermore, the introduction of model-dependent dynamic auxiliary and edge-auxiliary variables contributes to a significant reduction in signal transmission frequency. Then, novel model-dependent edge-node-combined distributed dynamic event-triggered controllers are formed to ensure the global exponential consensus of NSMASs, with Zeno phenomenon explicitly excluded. Finally, the effectiveness, practicability and superiority of proposed methods are validated through two numerical examples and a practical physical simulation.

In industrial processes, the defect data is continuously accumulated over time, and new classes of defects can arise at any time. When a well-trained model is adapted to new classes, its performance in old classes will sharply decline. To solve the above issue, a novel incremental defect detection method named elastic heterogeneous distillation network (EHD-Net) is proposed. First, an elastic knowledge transfer method is proposed to selectively transfer the core features of the old classes, so that the old knowledge is retained and new knowledge is effectively learned at the feature level. Second, warping heterogeneous distillation learning is proposed for the first time. In warping heterogeneous distillation learning, a gradient warping layer is proposed to balance the optimization direction of gradients for both old and new classes, and the proposed heterogeneous distillation learning strategy can clarify the association and difference between new and old classes, so that the rapid learning of new knowledge and comprehensive transfer of old knowledge are guaranteed at the decision level. Finally, a well-trained evolutionary model is employed to achieve the detection of both old and new classes. The proposed method can effectively overcome the catastrophic forgetting of old classes and guarantee the independent learning of new classes. Experimental results on two industrial datasets show that EHD-Net outperforms existing advanced methods. Note to Practitioners—The motivation for this paper is an important and highly practical issue called class-incremental defect detection. Specifically, new classes of defects always appear in industrial scenarios over time and when a well-trained detection model adapts to the new classes, its performance in the old classes decreases dramatically. To solve the above issues, an elastic heterogeneous distillation network (EHD-Net) via elastic knowledge transfer and warping heterogeneous distillation learning is proposed. The proposed method not only effectively suppresses the catastrophic forgetting of old classes while realizing the fast learning of new classes, so that the model can simultaneously achieve the high-accuracy detection of both the new and old classes without the participation of old class samples. The validity of the proposed method is verified under two industrial datasets, which fully guarantees its practical application value.

Magnetic Flux Leakage(MFL) inspection is one of the most efficient non-destructive testing method used to detect anormalies of ferromagnetic materials which is widely used in many industry scenarios. Sizing of the defect from inspection signal is the key problem which is also a classic inverse problem. Reconstruction of corrosion liked defect with high degree of freedom(DOF) is a challenging problem in this area. A multi-agent based reconstruction algorithm(MAR) is proposed in this paper. Instead of designing an algorithm to build the entire defect, the defect region is divided into several segments. Each segment is reconstructed by its corresponding agent. The dimension of reward function is reduced number-of-agent times from the DOF of defect which makes the critic network efficient. Base on MAR, an uni-agent base reconstruction algorithm(UAR) is also given as another solution which requires less computing resources. How to divide the defect and corresponding signal to achieve better performance is also studied. Algorithms proposed in this paper solves the problem of inverse problem of MFL inspection with high DOF. The effectiveness of the algorithms proposed in this paper are validated with simulated data and practical inspection data. The results show that the algorithms proposed in this paper has good reconstruction accuracy with robustness when dealing with defects with high DOF.Note to Practitioners—MFL inspection is the widely used pipeline in-line inspection method. Currently, defects detected are usually quantified as simple cube defects. Since most of the practical defects on industrial pipelines are caused by corrosion which have complex shape, how to quantify defects with more details is not only a research topic but also practical demand. In this paper, MAR and UAR are proposed to reconstruct the depth with 99 degree of freedoms. In addition, the algorithms proposed can be extended to reconstruct the depth with much more higher degree of freedoms by just adopting more agents. The impact of wings data on reconstruction results is also discussed in this paper. The algorithms proposed in this paper focus on 2-D dimension which considers only the length and depth reconstruction problem. In further research, we will try to give solutions to 3-D dimension problem which considers length, width and depth in the same time.

The event-triggered tracking consensus of multi-agent systems (MASs) with model switching, topological switching and non-zero leader inputs is concerned in this paper. In previous event-triggered studies on MASs, only a single switching behavior is considered and triggering parameters remain unchanged during switching, which shows more conservatism. In sharp contrast to existing research, this paper deals with two switching behaviors simultaneously, i.e. model switching and topological switching. Accordingly, dual-switch-based dynamic event-triggered rules are proposed to ensure better triggering performance. Further, dual-switch-based event-triggered control protocols are devised for agents. By constructing dual-switch-based multiple Lyapunov functions and utilizing the average dwell time (ADT) method, sufficient conditions are given to ensure the tracking exponential consensus. Finally, two examples of practical mass-spring systems with variable masses are carried out to certify the effectiveness and superiority of our approaches. Note to Practitioners—In real-world MASs, model switching and topological switching often coexist. Considering the limited computational and communication resources, it is of practical significance to conduct event-triggered research on MASs with model switching and topological switching. However, in existing event-triggered research on MASs, triggering control protocols are designed only for one type of switching behavior. To fill the research gap, this paper delves into the event-triggered control of MASs under both model switching and topological switching, offering greater practicality for engineering applications. Meanwhile, the devised triggering protocol is less conservative, ensuring better triggering performance and saving more computing and communication resources. In the simulation, the proposed approach is applied to practical mass-spring control problems under variable masses and switching topologies, and its practical application value is illustrated.

Magnetic flux leakage (MFL) detector can analyze the state of pipelines and is widely used in pipeline inspection. Object detection of class extremely imbalanced MFL signals is a challenging and demanding task. It is difficult for existing deep learning-based detection methods to effectively detect a small number of component classes. To address the above issues, a novel masked multi-view bidirectional network (MMB-Net) is proposed in this paper. First, based on the problem of insufficient component samples, a mask-based data augmentation module is designed, where we develop a novel wavelet convolution blocking module and Gauss-like distribution mask strategy so as to expand component sample sets. Second, based on the problem that poor feature extraction of partial MFL signals leads to low accuracy of component detection, a learnable multi-view attention (LMA) module is designed to expand the perceptual field of the network and fully mine the features of MFL signals. Finally, a bidirectional attention module based on multiple classification layers (BAMC) is proposed to learn different decision boundaries, which maintains the accuracy of the defects while improving the accuracy of the components. Experimental results illustrate that the proposed network can effectively build a detection model under MFL signals and outperforms the state-of-the-art methods on mean precision (9.5%).Note to Practitioners—This article is motivated by the problem of extreme class imbalance in the MFL signals collected from pipelines, which seriously reduces the detection accuracy of defects and components in the MFL signals. In terms of the MFL signal intrinsic characteristics, a masked multi-view bidirectional network (MMB-Net) based on data dynamic adjustment of model parameters and bidirectional learning decision boundaries are constructed. Our method can extend the component samples and learn different decision boundaries to improve the detection accuracy of the component classes. The experimental results show that our proposed method is more advantageous than the state-of-the-art methods, and therefore, our proposed method has strong practical value.

Domain-adaptive object detection (DAOD) aims to generalize detectors trained in labeled source domains to unlabeled target domains by mitigating domain bias. Recent studies have confirmed that pretrained vision-language models (VLMs) are promising tools to enhance the generalizability of detectors. However, there exist paradigm discrepancies between single-domain detection in most existing works and DAOD tasks, which may hinder the fine-grained alignment of cross-domain visual-text features. In addition, some preliminary solutions to these discrepancies may potentially neglect relational reasoning in prompts and cross-modal information interactions, which are crucial for fine-grained alignment. To this end, this article explores fine-grained visual-text feature alignment in DAOD with prompt tuning and organizes a novel framework termed FGPro that contains three elaborated levels. First, at the prompt level, a learnable domain-adaptive prompt is organized and a prompt relation encoder is constructed to infer intertoken semantic relations in the prompt. At the model level, a bidirectional cross-modal attention is structured to fully interact visual and textual fine-grained information. In addition, we customize a prompt-guided cross-domain regularization strategy to inject domain-invariant and domain-specific information into prompts in a disentangled manner. The three designs effectively align the fine-grained visual-text features of the source-target domain to facilitate the capture of domain-aware information. Experiments on four cross-domain scenarios show that FGPro exhibits notable performance improvements over existing work (Cross-weather: +1.0% AP50; Simulation-to-real: +1.2% AP50; Cross-camera: +1.3% AP50; Industry: +2.8% AP50), validating the effectiveness of its fine-grained alignment.

Pipeline system, as a vital industrial infrastructure, inevitably suffers from corrosion due to long-term service, which poses great threat to safe and stable energy transport. Many intelligent quantitative assessment technologies have emerged to evaluate the pipeline corrosion condition. However, they heavily rely on the quality and quantity of corrosion data. For industrial scenarios, the acquisition of pipeline corrosion data is costly with precision equipment for excavation works, and the annotation task is also labor-intensive, which severely limits their applicability. To address these issues, a mechanism-aided bilevel knowledge transfer framework is proposed to achieve effective quantitative assessment of pipeline corrosion condition. First, a high-fidelity mechanism model is designed to simulate the pipeline corrosion process and generates sufficient simulation data to alleviate the dependence on real-field resources. Then, a bilevel distribution-aware idea is proposed to transfer knowledge from simulated to real corrosion data, which is our original effort to focus on both feature complexity and label finesse, so as to accommodate the real-world corrosion for sensible domain adaptation. Next, we propose the idea of divide-and-conquer based on variational integration embedding (VIE) to maximize model performance improvement with minimal expert workload, where unreliable corrosion data are asked to expert annotation, while others are utilized via pseudolabels obtained from VIE. In the experiments, the performance of our method based on both hardware test platform and practical application case is systematically validated, and the competitive results indicate that our method has great potential in intelligent pipeline systems.

Fault transfer diagnosis is a key technology to ensure the reliability and safety of industrial systems, the core of which is to identify the health status of the equipment among different working conditions with multiclassification methods. However, most of them are based on a closed-set assumption that the label space among different working conditions is consistent, which is hard to satisfy in a practical industrial environment as unknown faults would inevitably occur during operation, i.e., the open-set fault transfer diagnosis (OSFTD) problem. Moreover, during the transfer process, unnecessary source-specific knowledge tends to be adapted, which brings about biased diagnostics on both domain and category. Aiming at this issue, an OSFTD framework, coined as knowledge transfer and reinforcement based on biunbiased neural network (KTR-BUNN), is proposed. First, a domain-unbiased knowledge transfer subnet is proposed, including an uncertainty-aware fault transferability evaluator (FTE) that estimates the transferability of target-domain samples unbiasedly to guide distribution alignment of known faults and a triple-tier unknown fault separator (UFS) that takes transferability as the criterion to extrapolate unknown faults. Second, a class-unbiased knowledge reinforcement subnet is designed to promote the recognition of fault semantic features at the embedding space, where fault knowledge graphs (FKGs) are constructed to describe the relationships between fault types, and they are optimized by a contrastive fault correlation loss, so that fine-grained class-level fault features can be further aligned. The knowledge transfer and knowledge reinforcement mechanisms work jointly to facilitate the performance of OSFTD. Finally, extensive experimental results conducted on diverse diagnostic tasks illustrate the superiority of the proposed KTR-BUNN.

Magnetic flux leakage (MFL) testing and analysis is an effective non-destructive testing (NDT) method of pipeline health. The health status of pipelines is evaluated by analyzing MFL signals, especially the defect analysis. However, complex operating conditions and equipment deviations may cause deformation on the defects. Availability and integrity of the MFL defects are the keys to accurate signal processing. Due to the limited interpretability of abstract networks and insufficient feature extraction, it is hard to accurately recover the deformed defects. Considering the problems above, an MFL deformed defect recovery method with the guidance of physical features is proposed. The proposed method integrates a deep feature extraction model into a sparse autoencoder, whose parameters are guided by the combination of MSE loss and physics loss. First, the deep feature extraction model fully extracts the abstract features, which contains more information on deformed defects. Second, the sparse penalty factor in the sparse autoencoder reduces the feature redundancy of deformed defects. Third, the physical features are extracted to guide the MFL deformed defect recovery relying on the MFL mechanism for the first time, which enhances the robustness and accuracy of the network. Finally, several comparison experiments have been conducted on measured and simulated MFL defects. The results show that the proposed method is effective in deformed defect recovery.Note to Practitioners—The motivation of this paper is a practical problem of MFL deformed defect recovery. The deformation on MFL defects affects defect inversion and evaluation seriously, which causes great potential harm to pipeline safety transmission. In this paper, compared with traditional neural network-based methods, physical features based on MFL mechanism are extracted to guide the parameter update. In addition, a pre-trained model replaces regular convolutional layers to extract features sufficiently, while a sparse factor is applied to reduce redundant features. The proposed method not only utilizes deep features from deep networks, but also increases the guidance of physical features, so that the algorithm is more robust and accurate. The experiment results show that our proposed method is more effective than the comparative methods on MFL simulated and measured defects. In a word, the proposed method has strong theoretical research and practical value.

This article proposes a practical and generalizable object detector, termed feature extraction-fusion-prediction network (FEFP-Net) for real-world application scenarios. The existing object detection methods have recently achieved excellent performance, however they still face three major challenges for real-world applications, i.e., feature similarity between classes, object size variability, and inconsistent localization and classification predictions. In order to effectively alleviate the current difficulties, the FEFP-Net with three key components is proposed, and the improved detection accuracy is proved in various applications: 1) Extraction Phase: an adaptive fine-grained feature extraction network is proposed to capture features of interest from coarse to fine details, which effectively avoids misclassification due to feature similarity; 2) Fusion Phase: a bidirectional neighbor connection network is designed to identify objects with different sizes by aggregating multilevel features and 3) Prediction Phase: in order to improve the accuracy of object localization and classification, a task specific prediction network is presented, which sufficiently exploits both the spatial and channel information of features. Compared with the State-of-the-Art methods, we achieved competitive results in the MS-COCO dataset. Further, we demonstrated the performance of FEFP-Net in different application fields, such as medical imaging, industry, agriculture, transportation, and remote sensing. These comprehensive experiments indicate that FEFP-Net has satisfactory accuracy and generalizability as a basic object detector.

Process fault diagnosis is of great importance to ensure the safe and stable operation of industrial systems. Many existing deep-learning-based process fault diagnosis methods assume that the samples are sufficient and obey the same distribution; however, it is almost impossible to achieve in practical industrial applications due to changing working conditions and the high cost of acquiring fault samples, which leads to a prominent performance degradation. In essence, those methods do not fully exploit the intrinsic and relevant knowledge under different working conditions. To address the above issue, a knowledge-mining-based semantic-aware network (KMSA-Net) is proposed in this article. First, a self-correlation knowledge mining subnet is proposed, where unshared attention mechanism is designed to extract knowledge inherent in each working condition so that the discriminative features can be captured. Second, a cross-correlation knowledge mining subnet is proposed, where we develop a fault relational knowledge graph so as to explicitly constrain the local consistency between the source domain, target domain, and cross-domain. Third, a semantic-aware knowledge transfer subnet is designed to impose a semantic constraint during knowledge transfer by encouraging the output of KMSA-Net to be consistent and distinguishable. These three subnets are jointly trained and then applied for cross-domain industrial process fault diagnosis. Finally, benchmark simulated experiments and real-world application experiments are conducted, and the experimental results validate the effectiveness and superiority of the proposed method.

Pipeline magnetic flux leakage (MFL) detection is an efficient and energy-saving nondestructive testing (NDT) method. However, under the high-speed detector, MFL signals become distorted with the influence of velocity effects, which adversely affects the accuracy of defect size inversion. The essential cause is the distorted signal multiplicity by velocity effects. In response to this issue, a high-precision defect size inversion method is proposed for the first time, which is called knowledge-guided contrastive fusion network (KCF-Net). First, MFL and eddy current (EC) mechanisms are analyzed, which are concluded that the sensitivity of EC signals to speed is much lower than that to defect sizes, so that EC and MFL abstract features are mined to improve the sensitivity of defect sizes. Moreover, MFL mechanism representations are mined to supervise neural networks to enhance the interpretability of the network. MFL and EC knowledge including abstract features and mechanism representations is fused to highlight the disparities between undistorted and distorted signals and enrich available information. Then, joint decision-making is proposed to eliminate the instability of fusion knowledge and enhance the universality and effectiveness of defect inversion. Finally, the experiments prove the effectiveness of KCF-Net. The length, width, and depth inversion MAEs of measured signals reach 2.2676, 1.6185, and 0.5664, respectively.

Defect detection aims to locate and classify defects in images, which is a necessary yet challenging task in industrial product quality monitoring. The current anchor-based detectors have weak generalization performance due to their inability to consider numerous scale priors. Moreover, the basic networks lack the ability to dynamically capture and utilize multiscale feature representations, resulting in low accuracy in industrial defect detection. To counter these challenges, an efficient anchor-free detector with dynamic receptive field assignment (DRFA) and task alignment is proposed. First, a feature pyramid structure with DRFA is innovatively designed to sufficiently extract multiscale feature representation and flexibly adjust the receptive field to detect diverse defects. Second, a task decoupling prediction mechanism is proposed to improve localization and classification prediction capabilities by introducing feature reassembly and task-specific information enhancers. Next, an anchor-free-based deep supervision with task-aligned is presented to encourage both to make accurate and consistent predictions, thereby effectively improving the overall detection performance. Finally, three industrial defect datasets (NEU-DET, PCB, WELD) are employed for experiments. The results show that the proposed method achieves 5.3% higher average AP than other state-of-the-art detectors.

Automatic defect detection plays an important role in industrial production. Deep learning-based defect detection methods have achieved promising results. However, there are still two challenges in the current defect detection methods: 1) high-precision detection of weak defects is limited and 2) it is difficult for current defect detection methods to achieve satisfactory results dealing with strong background noise. This article proposes a dynamic weights-based wavelet attention neural network (DWWA-Net) to address these issues, which can enhance the feature representation of defects and simultaneously denoise the image, thereby improving the detection accuracy of weak defects and defects under strong background noise. First, wavelet neural networks and dynamic wavelet convolution networks (DWCNets) are presented, which can effectively filter background noise and improve model convergence. Second, a multiview attention module is designed, which can direct the network attention toward potential targets, thereby guaranteeing the accuracy for detecting weak defects. Finally, a feature feedback module is proposed, which can enhance the feature information of defects to further improve the weak defect detection accuracy. The DWWA-Net can be used for defect detection in multiple industrial fields. Experiment results illustrate that the proposed method outperforms the state-of-the-art methods (mean precision: GC10-DET: 6.0%; NEU: 4.3%). The code is made in https://github.com/781458112/DWWA.

The problem of limited samples with missing information is an open challenge in data-driven fault diagnosis. Existing work has limited application in this field, since the reconstructed missing samples participating in sample generation may hurt the quality of the generated samples. To address this issue, the joint modeling of sample reconstruction and sample generation is proposed. First, the differentiated evaluation and reconstruction strategies are designed, which make reconstructed samples more reasonable and realistic, so that they can be employed to participate in sample generation. Second, the adaptive fusion mechanism is presented to introduce the knowledge of actual fault samples into the laboratory simulation samples, by which the quality and diversity of generated samples are guaranteed. By doing so, limited samples with missing information are enhanced to enable reliable fault diagnosis modeling. The proposed method is applied to the actual industrial process and benchmark simulated process. The experimental results highlight the superiority of the proposed method.

Small sample size leads to low accuracy and poor generalization of industrial fault diagnosis modeling. Domain adaptation (DA) attempts to enhance small samples by transferring samples in other similar domains, but it has limited application in industrial fault diagnosis, since the differences in working conditions lead to large variations of fault samples. To address the above issues, this article proposes a heterogeneous sample enhancement network with lifelong learning (HSELL-Net). First, a heterogeneous DA subnet (HDA-subnet) is presented, in which the designed heterogeneous supporting domain ensures dimension alignment and the designed distribution jointly matching improves the performance of distribution matching; thus, fault samples from other working conditions can be employed to reliably enhance small samples. Second, a lifelong learning subnet (LL-subnet) is designed, in which the proposed Admixup and shared knowledge repository enable incremental samples to further enhance small samples without retraining the network. The two subnets are mutually embedded and reinforced to enhance the number and types of small samples; thus, the accuracy and generalization of fault diagnosis under industrial small samples are improved. Finally, benchmark simulated experiments and real-world application experiments are conducted to evaluate the proposed method. Experimental results show the HSELL-Net outperforms the existing works under industrial small samples.

Class-imbalanced weld defect recognition, which realizes defect recognition via learning features of class-imbalanced X-ray images, is an emerging but challenging task. Nevertheless, the existing studies on the class-imbalanced problem mainly focus on large-scale data, and it is difficult to extract high-quality features from the insufficient industrial data, resulting in weak recognition performance. To address the above issue, this article comprehensively learns features from the perspective of basic-class and cross-class, on this basis, a novel hybrid feature learning model for class-imbalanced weld defect recognition is proposed. First, an image acquisition method completed by photographing, scanning, and sampling is designed to collect the class-imbalanced X-ray images. Second, a hybrid feature learning model is proposed to learn the distinctive and effective features from acquired images, so that the class-imbalanced data are mapped to a balanced feature distribution. Third, with the distinguishable features learned by our hybrid feature learning model, an unbiased defect recognition model can be trained to recognize different types of defects. The practical weld data W-PPLN, W-MTL, and W-GDXray are adopted in the experiments, and the experimental results show that our method outperforms the state-of-the-art methods on the task of class-imbalanced weld defect recognition.

Deep learning methods have demonstrated promising performance in magnetic flux leakage (MFL) defect detection under adequate amounts of labeled samples. However, in industrial occasions, obtaining adequate amounts of labeled samples is time-consuming and expensive, and applying only limited labeled samples can lead to unsatisfactory defect detection accuracy. To address the above issues, a defect detection method named semisupervised circular teacher network (SSCT-Net) is proposed in this article. First, a parallel feature extraction network with hybrid attention is proposed in SSCT-Net so that the useful features of multiview MFL signals can be extracted simultaneously. Second, semisupervised circular learning is proposed for the first time. In semisupervised circular learning, a distinguishable feature embedding space is constructed, and two structurally identical deep networks cosupervise and collaborate through the proposed consistent circular strategy so that the decision bias of unlabeled samples can be reduced. Finally, the trained model is applied for defect detection in practice. The proposed method can establish a potential connection between multiview MFL signals and fully utilize labeled and unlabeled MFL signals. The experiments in simulations and real-world applications demonstrate that SSCT-Net can reach 92% detection accuracy with only 20% labeled samples, which is more effective than the state-of-the-art methods and leads to a promising practical utility of the proposed method.

In complex real-world industrial systems, few-shot fault diagnosis greatly challenges model-free methods. Interest in domain adaptation methods, which enriches the diversity of accessible samples by narrowing the distance between the source and target domain distributions, has grown. However, these approaches generally rely on specific domain pairs and numerous labeled source data, which are difficult conditions to satisfy in industrial scenarios with complex/changeable working conditions and limited fault samples. Herein, we creatively propose a distribution-agnostic few-shot framework by imitating brain awareness process in unseen tasks. Our framework can generate a learnable and interpretable paradigm to learn common similarities in task embedding space, alleviating the dependence on deep supervised training and reducing the time required for conducting credible exploration from scratch. In particular, we design an adaptation-aware nonconvex matrix optimization procedure for optimal deep adaptation features transport process updating. Experimental results validate the superiority of our framework in two industrial applications, magnetic flux leakage, and bearing datasets, showing it is feasible and promising.

To improve the validity of magnetic flux leakage (MFL) multisensor signals, anomaly detection has become a significant part of MFL signal processing. The anomalies in MFL are uncertain and have no prior information or labels. Therefore, the detection and location of the anomalies become a difficult issue. Regarding the abovementioned problem, we propose an unsupervised method called multisensor cycle-supervised convolutional neural network (CsCNN). The CsCNN is built including multiple CNNs with the same structure and a cycle-supervised part. The proposed model realizes unsupervised anomaly detection through multiple cycle-supervised CNNs for the first time. Moreover, the latent relationship between multisensor signals is established by CsCNN to take full use of multisensor information. Besides, a dynamic threshold is applied to detect anomalies. In the end, experiments on simulated signals and measured signals are conducted, and CsCNN is compared to the state-of-the-art methods. The results show that the proposed method is effective.

The small sample problem is an open challenge in industrial data-driven fault diagnosis. Domain adaptation (DA) has been successfully used to enhance small samples by transferring fault samples in other working conditions, but it suffers from low accuracy and poor generalization for modeling. To address this issue, in this article, a task-based bias DA network (TBDA-net) is proposed. First, an adaptive dimension alignment subnet is proposed, which overcomes the information loss of the target domain caused by the dimension alignment of different domains, thus the accuracy of modeling is improved. Second, a task-based distribution matching subnet is proposed, in which the correlation difference of domains is considered and the designed auxiliary branch is embedded in the matching network to protect the sample diversity of multisource domains, thus the generalization of modeling is improved. Finally, benchmark simulated experiments and real-world application experiments are conducted to evaluate the proposed method. Experiment results show the TBDA-net outperforms the existing methods for modeling under industrial small samples

Safe transmission of oil pipelines is one of the guarantees of national defense and environmental protection. Magnetic flux leakage (MFL) testing is critical to the safety inspection of in-service pipelines. In the detection process, the incompleteness of MFL data affects defect location and inversion severely. This article proposes an MFL data recovery method based on multifeature condition risk, which can effectively handle the block data gap problem. First, a preprocessing mechanism is proposed to determine defect boundaries and uniformly interpolate the raw data automatically. Second, a multifeature extraction method of MFL defect data is proposed, which takes full advantage of the complete data and reduces the impact of data gaps. Third, a novel data reconstruction method based on feature conditional risk is proposed, where prior information of MFL data is regressed, relying on the regression coefficient calculated by dynamic programming. Finally, comparison experiments on different sizes of data gaps, varying robustness, and average running time are conducted, respectively. The MFL data are derived from actual measurements. The results indicate that the proposed method is more robust, more efficient, and faster.Note to Practitioners—This article is motivated by the problem of data gaps in magnetic flux leakage (MFL) data in actual pipeline safety transmission measurements, which seriously affects defect inversion and assessment. In this article, in terms of the intrinsic characteristics for MFL data, we propose a new and effective approach by extracting the multifeature of defects and using conditional risk to recovery data gaps. Our method takes advantage of the no-missing part of MFL data and prior samples, which are obtained from practical measurements. The experiment results show that our proposed method is more advantageous than the comparison methods, and therefore, our proposed method has strong practical value. Our future work will aim at improving the versatility of the proposed method.

The lack of fault samples is a challenging issue for fault diagnosis in the industrial process. It is difficult for conventional fault diagnosis methods to achieve satisfactory results under small samples. This article proposes a general transfer framework with evolutionary capability to address the above issue. First, a general transfer framework is proposed, in which the transfer learning strategy is applied to guarantee the number and diversity of expanded samples and achieve accurate modeling. Second, the adaptive mixup (Admixup) method is presented, which can adaptively expand the fault samples and make the shared distribution smoother to guarantee the stability and accuracy of the fault diagnosis results. Finally, an optimized evolution strategy is designed, in which the transformation matrix is used as an evolutionary channel to reduce the fault diagnosis errors without retraining the framework as fault samples increase. The presented framework can utilize the generalization of small samples and the knowledge of various working condition samples to achieve accurate modeling. The proposed framework is applied to simulated and real industrial processes. Experiment results illustrate that the fault diagnosis model can be effectively established by the proposed framework under small samples, and the proposed framework is evolutionary capable.

Condition monitoring (CM) has been playing an important role in the operation and maintenance of industrial equipment. However, the changeable environments where some equipment, such as wind turbines (WTs) are installed, have put a negative effect on CM. To deal with this problem, this article proposes a CM method of WTs based on assembled multidimensional membership functions (MFs) using a fuzzy inference system. First, multidimensional MFs, which are formed by fusing environment factor into conventional MFs, are proposed to reduce the negative effect from changeable environments. Second, the input data are properly divided into segments and the segments are classified into four types to distinguish the effects of different data on CM. Different weights are assigned to each segment and the corresponding membership of each segment is calculated separately. Then, these memberships are assembled for fuzzy inference. Finally, based on the assembled multidimensional MFs, a new CM architecture is established. Four groups of experiments were carried out to evaluate the proposed method with the data collected from a wind farm in northern China. The experiments results show that the proposed method can not only detect anomalies at an early stage but also effectively decrease the false alarms and missing detections.

The limited fault information caused by small fault data samples is a major problem in wind turbine (WT) fault detection. This paper proposes a small-sample WT fault detection method with the synthetic fault data using generative adversarial nets (GANs). First, based on prior knowledge, a rough fault data generation process is developed to transform the normal data to the rough fault data. Second, a rough fault data refiner is developed by GANs to make the rough fault data more similar with the real fault data. Moreover, to make the generated data better suited to the WT conditions, GANs are improved in both the generative model and the discriminative model. Third, artificial intelligence (AI)-based WT fault detection models can be well trained by using only the generated data in the condition of small fault data sample. Finally, three groups of generated data evaluation experiments and four groups of WT fault detection comparative experiments are conducted using real WT data collected from a wind farm in northern China. The results indicate that the method proposed in this paper is effective.

In most current nondestructive testing systems, a magnetic flux leakage (MFL) method is widely used in various industry fields, where the structural integrity of specimens is of vital importance. The estimation of defect size in specimen from the MFL measurements is a key and difficult problem. The traditional methods have low precision because feature extraction procedure relies on prior knowledge and the ability of designer. Inspired by the idea of convolutional neural network (CNN), a novel visual transformation CNN (VT-CNN) is proposed in this paper to overcome the limitation of traditional method in a feature extraction procedure. By adding a visual transformation layer according to the characteristics of the MFL measurements, the VT-CNN can distinguish the defect feature with different sizes more accurately. Moreover, since the VT-CNN method is designed based on the deep learning theory, more industrial big data with accurate label should be used to train the network. Due to the difficulty of making real industrial big data, a novel mesher magnetic dipole model is designed to simulate this industrial process. A large simulated MFL measurement of irregular defects produced by this model can increase the number of training samples and improve the robustness of the network. Experiments to estimate natural corrosion defects on real industrial pipelines are performed to validate the proposed framework. The experimental results are illustrated in detail, which highlights the superiority of the proposed method in industrial applications.

A fuzzy min-max neural network based on data core (DCFMN) is proposed for pattern classification. A new membership function for classifying the neuron of DCFMN is defined in which the noise, the geometric center of the hyperbox, and the data core are considered. Instead of using the contraction process of the FMNN described by Simpson, a kind of overlapped neuron with new membership function based on the data core is proposed and added to neural network to represent the overlapping area of hyperboxes belonging to different classes. Furthermore, some algorithms of online learning and classification are presented according to the structure of DCFMN. DCFMN has strong robustness and high accuracy in classification taking onto account the effect of data core and noise. The performance of DCFMN is checked by some benchmark datasets and compared with some traditional fuzzy neural networks, such as the fuzzy min-max neural network (FMNN), the general FMNN, and the FMNN with compensatory neuron. Finally the pattern classification of a pipeline is evaluated using DCFMN and other classifiers. All the results indicate that the performance of DCFMN is excellent.