Local Contextual Information Research Articles

Temporal convolution long short-term memory network with multiple attention for remaining useful life prediction of rolling bearings

Abstract Remaining useful life (RUL) prediction of rolling bearings prevents sudden mechanical failures and reduces equipment maintenance costs. Due to the strong performance in time series forecasting tasks, the temporal convolutional network (TCN) has become a mainstream model for RUL prediction. However, existing TCN-based prediction models struggle to fully capture both long-term and global dependencies in complex data. To address these issues, a temporal convolutional long short-term memory network integrated with multi-head attention mechanism (TCLSTM-MA) is proposed to predict the rolling bearings’ RUL. Firstly, the time-domain and frequency-domain features are extracted from the acquired raw vibration signals to form a complete degradation feature. Secondly, we enhance the traditional TCN by combining it with LSTM and introducing a multi-head attention mechanism. This integration allows the model to effectively capture both global degradation information and local context information. Additionally, a time-weighted t-MSE loss function is employed throughout training to make the model focus more on data close to failure points. Finally, the trained TCLSTM-MA model is used for RUL prediction. Extensive experiments were conducted on two authoritative rolling bearing datasets and compared with other methods. The experimental results demonstrate that the proposed method exhibits good accuracy and generalization capability.

Severity grading of hypertensive retinopathy using hybrid deep learning architecture.

Hypertensive Retinopathy (HR) is a retinal manifestation resulting from persistently elevated blood pressure. Severity grading of HR is essential for patient risk stratification, effective management, progression monitoring, timely intervention, and minimizing the risk of vision impairment. Computer-aided diagnosis and artificial intelligence (AI) systems play vital roles in the diagnosis and grading of HR. Over the years, very limited research has been conducted for the grading of HR. Nevertheless, there are no publicly available datasets for HR grading. Moreover, one of the key challenges observed is high-class imbalance. To address these issues, in this paper, we develop "HRSG: Expert-Annotated Hypertensive Retinopathy Severity Grading" dataset, classifying HR severity into four distinct classes: normal, mild, moderate, and severe. Further, to enhance the grading performance on limited datasets, this paper introduces a novel hybrid architecture that combines the strengths of pretrained ResNet-50 via transfer learning, and a modified Vision Transformer (ViT) architecture enhanced with a combination of global self-attention and locality self-attention mechanisms. The locality self-attention addresses the common issue of a lack of inductive bias in ViT architecture. This architecture effectively captures both local and global contextual information, resulting in a robust and resilient classification model. To overcome class imbalance, Decouple Representation and Classifier (DRC) - based training approach is proposed. This method improves the model's ability to learn effective features while preserving the original dataset's distribution, leading to better diagnostic accuracy. Performance evaluation results show the competence of the proposed method in accurately grading the severity of HR. The proposed method achieved an average accuracy of 0.9688, sensitivity of 0.9435, specificity of 0.9766, F1-score of 0.9442, and precision of 0.9474. The comparative results indicate that the proposed method outperforms existing HR methods, state-of-the-art CNN models, and baseline pretrained ViT models. Additionally, we compared our method with a CNNViT model, which combines a shallow CNN architecture with 3 convolution blocks consisting of a convolution layer, a batch normalization layer, a max pooling layer, and lightweight ViT architecture, due to limited datasets. In comparison with the CNNViT, the proposed method achieved superior performance, demonstrating its effectiveness. The experimental results demonstrate the efficacy of the proposed method in accurately grading HR severity.

Point cloud graph for LiDAR segmentation

Point clouds have benefited autonomous driving applications. Thanks to advancements in CNN-based and Transformer-based architectures, projection and voxel representations for point clouds have become mainstream for 3D point cloud tasks. However, current feature backbones still face challenges in capturing local neighbor communication and contextual information due to the sparse and discrete point clouds. In this work, we explore an order reconstruction for discrete point clouds based on the LiDAR measurement principles and discuss the point graph representation. The point cloud graph provides geodesic information through directed edges. We integrate a graph branch based on GNN into a vanilla 3D CNN-based segmentation backbone while addressing the interaction between Euclidean and geodesic features. We examine the improvements in segmentation accuracy, semantic consistency, and receptive fields powered by the point cloud graph on 3D point cloud benchmarks. Our experimental results demonstrate that the proposed methods achieve the 3D CNNs-based state-of-the-art for LiDAR segmentation.

An Efficient Group Convolution and Feature Fusion Method for Weed Detection

Weed detection is a crucial step in achieving intelligent weeding for vegetables. Currently, research on vegetable weed detection technology is relatively limited, and existing detection methods still face challenges due to complex natural conditions, resulting in low detection accuracy and efficiency. This paper proposes the YOLOv8-EGC-Fusion (YEF) model, an enhancement based on the YOLOv8 model, to address these challenges. This model introduces plug-and-play modules: (1) The Efficient Group Convolution (EGC) module leverages convolution kernels of various sizes combined with group convolution techniques to significantly reduce computational cost. Integrating this EGC module with the C2f module creates the C2f-EGC module, strengthening the model’s capacity to grasp local contextual information. (2) The Group Context Anchor Attention (GCAA) module strengthens the model’s capacity to capture long-range contextual information, contributing to improved feature comprehension. (3) The GCAA-Fusion module effectively merges multi-scale features, addressing shallow feature loss and preserving critical information. Leveraging GCAA-Fusion and PAFPN, we developed an Adaptive Feature Fusion (AFF) feature pyramid structure that amplifies the model’s feature extraction capabilities. To ensure effective evaluation, we collected a diverse dataset of weed images from various vegetable fields. A series of comparative experiments was conducted to verify the detection effectiveness of the YEF model. The results show that the YEF model outperforms the original YOLOv8 model, Faster R-CNN, RetinaNet, TOOD, RTMDet, and YOLOv5 in detection performance. The detection metrics achieved by the YEF model are as follows: precision of 0.904, recall of 0.88, F1 score of 0.891, and mAP0.5 of 0.929. In conclusion, the YEF model demonstrates high detection accuracy for vegetable and weed identification, meeting the requirements for precise detection.

A Multi-Level SAR-Guided Contextual Attention Network for Satellite Images Cloud Removal

In the field of remote sensing, cloud cover severely reduces the quality of satellite observations of the earth. Due to the complete absence of information in cloud-covered regions, cloud removal with a single optical image is an ill-posed problem. Since the synthetic aperture radar (SAR) can effectively penetrate clouds, fusing SAR and optical remote sensing images will effectively alleviate this problem. However, existing SAR-based optical cloud removal methods fail to effectively leverage the global information provided by the SAR image, resulting in limited performance gains. In this paper, we introduce a novel cloud removal method named the Multi-Level SAR-Guided Contextual Attention Network (MSGCA-Net). MSGCA-Net is designed with a multi-level architecture that integrates a SAR-Guided Contextual Attention (SGCA) module to fuse the dependable global contextual information from SAR images with the local features of optical images effectively. In the module of SGCA, the SAR image provides reliable global contextual information and genuine structure of cloud-covered regions, while the optical image provides the local feature information. The proposed model can efficiently extract and fuse global and local contextual information in SAR and optical images. We trained and evaluated the performance of the model on both simulated and real-world datasets. Both qualitative and quantitative experimental evaluation demonstrated that the proposed method can yield high quality cloud-free images and outperform state-of-the-art cloud removal methods.

MCPA: multi-scale cross perceptron attention network for 2D medical image segmentation

The UNet architecture, based on convolutional neural networks (CNN), has demonstrated its remarkable performance in medical image analysis. However, it faces challenges in capturing long-range dependencies due to the limited receptive fields and inherent bias of convolutional operations. Recently, numerous transformer-based techniques have been incorporated into the UNet architecture to overcome this limitation by effectively capturing global feature correlations. However, the integration of the Transformer modules may result in the loss of local contextual information during the global feature fusion process. In this work, we propose a 2D medical image segmentation model called multi-scale cross perceptron attention network (MCPA). The MCPA consists of three main components: an encoder, a decoder, and a Cross Perceptron. The Cross Perceptron first captures the local correlations using multiple Multi-scale Cross Perceptron modules, facilitating the fusion of features across scales. The resulting multi-scale feature vectors are then spatially unfolded, concatenated, and fed through a Global Perceptron module to model global dependencies. Considering the high computational cost of using 3D neural network models, and the fact that many important clinical data can only be obtained in two dimensions, our MCPA focuses on 2D medical image segmentation. Furthermore, we introduce a progressive dual-branch structure (PDBS) to address the semantic segmentation of the image involving finer tissue structures. This structure gradually shifts the segmentation focus of MCPA network training from large-scale structural features to more sophisticated pixel-level features. We evaluate our proposed MCPA model on several publicly available medical image datasets from different tasks and devices, including the open large-scale dataset of CT (Synapse), MRI (ACDC), and widely used 2D medical imaging datasets captured by fundus camera (DRIVE, CHASE_DB1, HRF), and OCTA (ROSE). The experimental results show that our MCPA model achieves state-of-the-art performance.

Enhanced scene text recognition using deep learning based hybrid attention recognition network

<span lang="EN-US">The technique of automatically recognizing and transforming text that is present in pictures or scenes into machine-readable text is known as scene text recognition. It facilitates applications like content extraction, translation, and text analysis in real-world visual data by enabling computers to comprehend and extract textual information from images, videos, or documents. Scene text recognition is essential for many applications, such as language translation and content extraction from photographs. The hybrid attention recognition network (HARN), unique technology presented in this research, is intended to greatly improve efficiency and accuracy of text recognition in complicated scene situations. HARN makes use of cutting-edge elements including alignment-free sequence-to-sequence (AFS) module, creative attention mechanisms, and hybrid architecture that blends attention models with convolutional neural networks (CNNs). Thanks to its novel attention processes, HARN is capable of comprehending wide range of scene text components by capturing both local and global context information. Through faster network convergence, shorter training times, and better utilization of computing resources, the suggested technique raises bar for state-of-the-art. HARN’s versatility makes it a good choice for range of scene text recognition applications, including multilingual text analysis and data extraction. Extensive tests are conducted to assess the effectiveness of HARN approach and demonstrate it is ability to greatly influence real-world applications where accurate and efficient text recognition is essential.</span>

CSegNet: a hybrid transformer-CNN network for road crack image segmentation

Detecting cracks from optical images plays a crucial role in road maintenance but its good realisation has many challenges. Road cracks exhibit significant diversity and complexity in terms of shape, size and texture and road images may contain various types of noise and interference, such as lighting variations, shadows and different appearances, due to varying perspectives and scales. To address these challenges, a comprehensive road crack dataset called CRCrack has been constructed, which encompasses various crack characteristics. This study proposes a road crack segmentation network called CSegNet, which combines convolutional neural networks (CNNs) and transformers. The network adopts an encoder-decoder framework, namely DeepLabV3+. In the encoder, leveraging the transformers' flexibility in modelling long-term dependencies and the CNNs' ability to capture local contextual information through local receptive fields, weight sharing and spatial subsampling, a ResNeXt-Transformer (ResNeXTR) feature extraction module is designed as the backbone network to enhance the feature extraction capability for road crack images. To reduce the computational cost in the self-attention (SA) computation of the transformer, an average pooling layer is introduced to downsample the dimensions of the encoded features. In the decoder, to focus on the key information of road cracks under diverse environmental conditions and interferences, an efficient channel attention module (ECAM) and a spatial attention module (SAM) are combined to design an efficient convolutional block attention module (ECBAM) to optimise feature representation. Through comparative experiments on the CRCrack dataset, the results demonstrate that the proposed method outperforms classic networks such as U-Net and DeepLabV3+ in terms of intersection over union (IoU), Dice coefficient and area under the receiver operating characteristic (AUROC) curve evaluation metrics. It exhibits good adaptability to ground crack images from different sources, providing a basis for estimating the degree of road damage.

COSTA: A Multi-center TOF-MRA Dataset and A Style Self-Consistency Network for Cerebrovascular Segmentation.

Time-of-flight magnetic resonance angiography (TOF-MRA) is the least invasive and ionizing radiation-free approach for cerebrovascular imaging, but variations in imaging artifacts across different clinical centers and imaging vendors result in inter-site and inter-vendor heterogeneity, making its accurate and robust cerebrovascular segmentation challenging. Moreover, the limited availability and quality of annotated data pose further challenges for segmentation methods to generalize well to unseen datasets. In this paper, we construct the largest and most diverse TOF-MRA dataset (COSTA) from 8 individual imaging centers, with all the volumes manually annotated. Then we propose a novel network for cerebrovascular segmentation, namely CESAR, with the ability to tackle feature granularity and image style heterogeneity issues. Specifically, a coarse-to-fine architecture is implemented to refine cerebrovascular segmentation in an iterative manner. An automatic feature selection module is proposed to selectively fuse global long-range dependencies and local contextual information of cerebrovascular structures. A style self-consistency loss is then introduced to explicitly align diverse styles of TOF-MRA images to a standardized one. Extensive experimental results on the COSTA dataset demonstrate the effectiveness of our CESAR network against state-of-the-art methods. We have made 6 subsets of COSTA with the source code online available, in order to promote relevant research in the community.

Automatic detection and counting of wheat spike based on DMseg-Count

The automatic detection and counting of wheat spike images are of great significance for yield prediction and variety evaluation. Therefore, accurate and timely estimation of spike numbers is crucial for wheat production. However, in actual production, due to the susceptibility of wheat spike images to factors such as lighting conditions, shooting angles, occlusion, and overlap, the contour and features of wheat spike is unclear, which affects the accuracy of automatic detection and counting of wheat spike. In order to solve the above problems and further improve the accuracy of wheat spike counting, an improved wheat spike counting model DMseg-Count was proposed by enhancing local contextual supervision information based on existing target object counting model DM-Count. Firstly, wheat spike local segmentation branch was introduced to improve the network architecture of DM-Count, so as to extract the local contextual supervision information of wheat spike. Secondly, an element-by-element point multiplication mechanism was designed to fuse global and local contextual supervision information of wheat spike. Finally, the total loss function was constructed to optimize the model. The test results showed that the mean absolute error (MAE) and root mean square error (RMSE) of the proposed DMseg-Count model were 5.79 and 7.54, respectively, which were 9.76 and 10.91 higher than the standard distribution matching for crowd counting (DM-Count) model. Compared with other deep learning models, the proposed DMseg-Count model can detect wheat spike image in challenging situations, and has better computer vision processing capabilities and performance evaluation detection effect. In summary, the proposed DMseg-Count model can effectively detect wheat spike and has good counting performance, which provides a new method for automatic counting of wheat spike and yield prediction in complex field environments.

Real-time 3D temperature field reconstruction for aluminum alloy forging die using Swin Transformer integrated deep learning framework

Temperature field distribution in forging dies is crucial for quality control and defect prevention, particularly for aluminum alloys. Current methods are limited to discrete points or surface measurements, making real-time three-dimensional temperature field acquisition challenging. In this paper, a novel Swin Transformer-integrated deep learning framework is proposed for real-time 3D temperature field reconstruction of forging dies, pioneering the application of transformer architecture in physical field prediction. In this framework, numerical simulations are first conducted to provide ground truth and fundamental insights into the temperature evolution, and then limited sparse thermal sensors are utilized to offer corrected real-time input parameters. The model for 3D temperature field reconstruction is developed through the combination of Swin Transformers with the U-shaped encoder-decoder structure, which is trained and tested with various sensor configurations, initialization methods, and datasets, including actual experiments. The results demonstrate that the proposed Swin-UNETR model achieves 3D temperature field prediction with time cost of 0.98 s per frame, mean absolute error of 0.8658 °C, showing a 17.23 % improvement over the next best CNN-based model (ResUNet3D at 1.0461 °C), and a 4.63 % improvement over the next best machine learning model (LightGBM at 0.9078 °C), which can be attributed to the Swin Transformer’s ability to capture both local and global contextual information and shifted window mechanism. The proposed method holds significant implications for ensuring the forming quality of forgings and propelling the development of digital twin technology in forging processes.

ESL-YOLO: Small Object Detection with Effective Feature Enhancement and Spatial-Context-Guided Fusion Network for Remote Sensing

Improving the detection of small objects in remote sensing is essential for its extensive use in various applications. The diminutive size of these objects, coupled with the complex backgrounds in remote sensing images, complicates the detection process. Moreover, operations like downsampling during feature extraction can cause a significant loss of spatial information for small objects, adversely affecting detection accuracy. To tackle these issues, we propose ESL-YOLO, which incorporates feature enhancement, fusion, and a local attention pyramid. This model includes: (1) an innovative plug-and-play feature enhancement module that incorporates multi-scale local contextual information to bolster detection performance for small objects; (2) a spatial-context-guided multi-scale feature fusion framework that enables effective integration of shallow features, thereby minimizing spatial information loss; and (3) a local attention pyramid module aimed at mitigating background noise while highlighting small object characteristics. Evaluations on the publicly accessible remote sensing datasets AI-TOD and DOTAv1.5 indicate that ESL-YOLO significantly surpasses other contemporary object detection frameworks. In particular, ESL-YOLO enhances mean average precision mAP by 10% and 1.1% on the AI-TOD and DOTAv1.5 datasets, respectively, compared to YOLOv8s. This model is particularly adept at small object detection in remote sensing imagery and holds significant potential for practical applications.

EMPViT: Efficient multi-path vision transformer for security risks detection in power distribution network

To maintain the safe operation of power distribution network (PDN) equipment, it is important to accurately and promptly identify security risks. However, conventional drone-based object detection methods face challenges due to noise and similarity features in risk targets, as well as limited computing resources of unmanned aerial vehicles (UAVs). To address these challenges, an efficient embedding-based multi-path fusion architecture is proposed. This architecture uses a re-parameterized depthwise block to embed local context information at different scales, enhancing the extraction of tiny features while preserving inference speed. Additionally, a coordinated self-attention module is proposed to reduce computational complexity while maintaining the performance of global information. By fusing fine and coarse feature representations without requiring a lot of computation, this module efficiently learns from both local and global features from images. The goal is to create an efficient multi-path vision transformer (EMPViT) architecture that achieves a balance between accuracy and efficiency. The proposed EMPViT has been evaluated on two different drone image dataset, demonstrating better performance compared to other architectures. Specifically, the EMPViT-S improves the detection mAP by 1.2%, and the inference speed is improved to 1.24 times on average on Drone-PDN dataset. It has achieved the same performance improvement on VisDrone-DET2019 dataset, gaining detection performance by 1.3% and 1.2 times acceleration on average.

Combined query embroidery image retrieval based on enhanced CNN and blend transformer

Embroidery images carry rich historical information and are an important form of embroidery art. In the field of combination query image retrieval, how to efficiently retrieve the embroidery image information required by users has become a current research challenge. In recent years, convolutional neural networks (CNNs) have achieved significant success in image feature extraction, but they tend to focus on local information, making it easy to ignore global context information when processing such textured embroidery images. Therefore, we propose a combination query retrieval method for embroidery images. First, we propose Blend-Transformer, which introduces Group External Attention (GEA). GEA can integrate feature information from three different dimensions, effectively capturing the local and global context information of embroidery images. Second, we propose Enhanced CNN, which introduces Shuffle Attention (SA), regrouping the reference image features extracted by CNN and reaggregating them by channel to enhance the richness of embroidery image feature information. Through experiments on the TCE-S and ICR2020 standard datasets, we verify the excellent performance of the proposed algorithm in embroidery image retrieval. Our method fills the gap in embroidery image retrieval research and provides a new perspective for the protection of embroidery art.

Research on Land Use and Land Cover Information Extraction Methods for Remote Sensing Images Based on Improved Convolutional Neural Networks

To address the challenges that convolutional neural networks (CNNs) face in extracting small objects and handling class imbalance in remote sensing imagery, this paper proposes a novel spatial contextual information and multiscale feature fusion encoding–decoding network, SCIMF-Net. Firstly, SCIMF-Net employs an improved ResNeXt-101 deep backbone network, significantly enhancing the extraction capability of small object features. Next, a novel PMFF module is designed to effectively promote the fusion of features at different scales, deepening the model’s understanding of global and local spatial contextual information. Finally, introducing a weighted joint loss function improves the SCIMF-Net model’s performance in extracting LULC information under class imbalance conditions. Experimental results show that compared to other CNNs such as Res-FCN, U-Net, SE-U-Net, and U-Net++, SCIMF-Net improves PA by 0.68%, 0.54%, 1.61%, and 3.39%, respectively; MPA by 2.96%, 4.51%, 2.37%, and 3.45%, respectively; and MIOU by 3.27%, 4.89%, 4.2%, and 5.68%, respectively. Detailed comparisons of locally visualized LULC information extraction results indicate that SCIMF-Net can accurately extract information from imbalanced classes and small objects.

B-TBM: A Novel Deep Learning Model with Enhanced Loss Function for HAZOP Risk Classification Using Natural Language Statistical Laws

HAZOP is a paradigm of industrial safety, and the introduction of deep learning-based HAZOP text categorization marks the arrival of an intelligent era of safety analysis. However, existing risk analysis methods have limitations in processing complex texts and extracting deep risk features. To solve this problem, this paper proposes a novel HAZOP risk event classification model based on BERT, BiLSTM, and TextCNN. The complexity of HAZOP text is revealed by introducing statistical laws of natural language, such as Zipf’s law and Heaps’ law, and the outputs of different levels of BERT are further combined linearly to collaborate with BiLSTM and TextCNN to capture long-term dependency and local contextual information for a more accurate classification task. Meanwhile, an improved loss function is proposed to effectively solve the deficiencies of the traditional cross-entropy loss function in the mislabeling process and improve the generalization ability of the model. It is experimentally demonstrated that the accuracy of the model is improved by 3% to 4% compared to the traditional BERT model in the task of severity and possibility classification of HAZOP reports. This study not only improves the accuracy and efficiency of HAZOP risk analysis, but also provides new ideas and methods for the application of natural language processing in industrial safety.

A novel spatial pyramid-enhanced indoor visual positioning method

As a key application of Internet of Things (IoT) technology, visual localization plays an important role in everyday life. However, pedestrians in images can obstruct environmental features, negatively impacting the performance of visual localization systems. To address this issue, we propose a Spatial Pyramid-Enhanced MixVPR visual localization method (SPE-VL) that aims to enhance image feature descriptions through multi-scale spatial information, thereby mitigating the effects of pedestrian occlusion on localization accuracy. The SPE-VL method is divided into two main phases: sensor-based matching range constraint and image feature extraction and matching. In the matching range constraint phase, we propose a direction decision method based on a machine learning classifier that utilizes smartphone sensor data to restrict the direction of image matching, reducing the likelihood of mismatches. In the feature extraction and matching phase, we propose a Transformer-based feature cross-enhancement method that leverages local features and spatial contextual information to enhance features, improving both image retrieval accuracy and localization precision. Experimental results indicate that the SPE-VL method demonstrates higher localization accuracy and robustness compared to existing state-of-the-art methods, providing new insights and solutions for the application of visual localization in complex environments.

Identification of transmission line voltage sag sources based on multi‐location information convolutional transformer

AbstractConventional methods for identifying voltage sag sources are difficult to categorize accurately due to the complexity of transmission lines and the influence of noise. In order to solve the problem of difficulty in recognizing voltage sag sources of transmission lines under different locations, this paper proposes a new method for transmission line fault diagnosis based on modified wavelet denoising and multi‐location information convolution transformer. The improved wavelet denoising method proposed in this paper solves the problems of discontinuity and bias of the traditional wavelet denoising method, and is able to better reconstruct the original voltage signal in a strong noise environment. In addition, multi‐location information convolution transformer adopts a new model combining multi‐location information convolution and multi‐scale convolution transformer, which realizes the combination of global information and local context information capturing ability under multi‐location faults. This paper validates the method through simulation experiments and practical situations, and the results show that the method can well classify and identify the type and location of voltage sag sources in transmission lines.

RFNet: Multivariate long sequence time-series forecasting based on recurrent representation and feature enhancement

Multivariate time series exhibit complex patterns and structures involving interactions among multiple variables and long-term temporal dependencies, making multivariate long sequence time series forecasting (MLSTF) exceptionally challenging. Despite significant progress in Transformer-based methods in the MLSTF domain, many models still rely on stacked encoder–decoder architectures to capture complex time series patterns. This leads to increased computational complexity and overlooks spatial pattern information in multivariate time series, thereby limiting the model’s performance. To address these challenges, we propose RFNet, a lightweight model based on recurrent representation and feature enhancement. We partition the time series into fixed-size subsequences to retain local contextual temporal pattern information and cross-variable spatial pattern information. The recurrent representation module employs gate attention mechanisms and memory units to capture local information of the subsequences and obtain long-term correlation information of the input sequence by integrating information from different memory units. Meanwhile, we utilize a shared multi-layer perceptron (MLP) to capture global pattern information of the input sequence. The feature enhancement module explicitly extracts complex spatial patterns in the time series by transforming the input sequence. We validate the performance of RFNet on ten real-world datasets. The results demonstrate an improvement of approximately 55.3% over state-of-the-art MLSTF models, highlighting its significant advantage in addressing multivariate long sequence time series forecasting problems.

Anomaly detection in smart manufacturing: An Adaptive Adversarial Transformer-based model

In Industry 5.0, smart manufacturing brings additional intricacies and novel data processing challenges. Given the evolving nature of manufacturing processes and the inherent complexity of data, including noise and missing entries, achieving accurate anomaly detection becomes even more intricate. Conventional methods often miss nuanced anomalies, especially when dealing with high-dimensional, multivariate, non-stationary data. These data types are typical of smart manufacturing environments. Hence, many recent approaches have embraced deep learning to confront these challenges, making use of diverse attention mechanisms to acquire data representations. However, in manufacturing, where the dynamics of time series data change over time, methods relying solely on pointwise or pairwise representations often fall short. Thus, ensuring product quality and operational integrity calls for even more advanced methodologies. The deficiency lies in the capability of state-of-the-art models to effectively capture abnormal patterns while considering both local and global contextual information. This challenge is compounded by the rarity of anomalies, making it exceedingly challenging to establish substantial associations between individual abnormal points and the entire time series. To tackle these challenges, we introduce the “Adaptive Adversarial Transformer” as a novel deep learning technique that combines Transformer architecture with an anomaly attention mechanism and Adversarial Learning. Our Model effectively captures intricate temporal patterns, distinguishes normal and anomalous behaviors, and dynamically adjusts thresholds to align with the evolving dynamics of time-series data. Empirical validation on four benchmark datasets and three real-world manufacturing datasets demonstrates our model’s effectiveness compared to the state-of-the-art, as evidenced by the F1-Score.