检索结果-内蒙古大学图书馆

defect classification of composite materials using transfer learning methods

NONDESTRUCTIVE TESTING AND EVALUATION 2024年

作者： Gulsen, Abdulkadir Kolukisa, Burak Ozdemir, Ahmet Turan Bakir-Gungor, Burcu Gungor, Vehbi Cagri Abdullah Gul Univ Dept Comp Engn Kayseri Turkiye Erciyes Univ Dept Elect & Elect Engn Kayseri Turkiye Turkcell Dept Technol Istanbul Turkiye

Nowadays, composite materials have become prevalent across various sectors, particularly finding usage in large-scale applications such as spaceships, automobiles, and aircrafts. The accurate detection of the defects in these materials is crucial, yet traditional methods often rely on human inspection, which is susceptible to errors. Recent advancements in machine learning have enabled defect detection using ultrasonic non-destructive testing methods. This paper introduces a new dataset named UNDT, which is obtained from the scans of 60 different composite materials, generating a total of 1150 images depicting both defective and non-defective areas. Several transfer learning methods are applied on the newly introduced UNDT dataset as well as the publicly available USimgAIST ultrasonic dataset. Comparative performance assessments illustrate the significance of utilising the transfer learning approach for defect classification on ultrasonic inspection images. Furthermore, the research emphasises the substantial benefits of employing these transfer learning methods. Notably, the DenseNet121 and VGG19 models achieve the highest accuracy rates, with 98.8% and 98.6% on the UNDT and USimgAIST datasets, respectively.

关键词： Automated inspection defect classification non-destructive testing transfer learning ultrasonic testing

来源：评论

学校读者我要写书评

暂无评论

A knowledge distillation-based multi-scale relation-prototypical network for cross-domain few-shot defect classification

引用

JOURNAL OF INTELLIGENT MANUFACTURING 2024年第2期35卷 841-857页

作者： Zhao, Jiaqi Qian, Xiaolong Zhang, Yunzhou Shan, Dexing Liu, Xiaozheng Coleman, Sonya Kerr, Dermot Northeastern Univ Coll Informat Sci & Engn Shenyang Peoples R China Univ Ulster Intelligent Syst Res Ctr Londonderry North Ireland

Surface defect classification plays a very important role in industrial production and mechanical manufacturing. However, there are currently some challenges hindering its use. The first is the similarity of different defect samples makes classification a difficult task. Second, the lack of defect samples leads to poor accuracies when using deep learning methods. In this paper, we first design a novel backbone network, ResMSNet, which draws on the idea of multi-scale feature extraction for small discriminative regions in defect samples. Then, we introduce few-shot learning for defect classification and propose a Relation-Prototypical network (RPNet), which combines the characteristics of ProtoNet and RelationNet and provides classification by linking the prototypes distances and the nonlinear relation scores. Next, we consider a more realistic scenario where the base dataset for training the model and target defect dataset for applying the model are usually obtained from domains with large differences, called cross-domain few-shot learning. Hence, we further improve RPNet to KD-RPNet inspired by knowledge distillation methods. Through extensive comparative experiments and ablation experiments, we demonstrate that either our ResMSNet or RPNet proves its effectiveness and KD-RPNet outperforms other state-of-the-art approaches for few-shot defect classification.

关键词： Few-shot learning defect classification Multi-scale feature encoder Cross-domain Knowledge distillation

来源：评论

学校读者我要写书评

暂无评论

defect classification and Detection Using a Multitask Deep One-Class CNN

引用

IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING 2022年第3期19卷 1719-1730页

作者： Dong, Xinghui Taylor, Christopher J. Cootes, Tim F. Univ Manchester Ctr Imaging Sci Manchester M13 9PT Lancs England Ocean Univ China Sch Comp Sci & Technol Qingdao 266100 Peoples R China

defect classification and detection have been explored using convolutional neural networks (CNNs). Normally, a large set of training images containing defects and the associated annotation data are required by these approaches. However, such a large set of images is usually difficult to collect because defects are rare and annotation is time-consuming and expensive. To address these issues, we propose to use a multitask deep one-class CNN for defect classification. Compared with supervised classification methods, this CNN does not require abnormal images and annotated data for training. Specifically, we build a stacked encoder-decoder autoencoder for learning feature representation from normal images. The encoder is used as a feature extractor based on the hard sharing scheme of multitask learning. A one-class classification (OCC) objective learned as a hypersphere using minimum volume estimation is appended to it. Together the encoder and the OCC objective lead to a deep one-class classifier. To train both the autoencoder and one-class classifier end-to-end, a multitask loss function is built. Given an unknown sample, the distance between its feature representation and the center of the hypersphere is used as the anomaly score. Furthermore, defect detection is implemented using a moving-window scanning method on top of the deep one-class classifier. The proposed approach achieves better performance than its counterparts trained using a two-stage method. For defect detection, our approach achieves results almost as good as the supervised method even without using any annotated data. We attribute the promising results to the advantages of multitask learning.

关键词： Convolutional neural networks Feature extraction Training Inspection Anomaly detection Volume measurement Task analysis Autoencoder defect classification defect detection industrial inspection one-class classification (OCC) one-class convolutional neural networks (CNNs)

来源：评论

学校读者我要写书评

暂无评论

Multimodal Fabric defect classification Using Channel Switching and Multiscale Feature Fusion

引用

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT 2025年 74卷

作者： Li, Song Sun, Wei Liang, Qiaokang Sun, Jian Liu, Chongpei Hunan Univ Coll Elect & Informat Engn Natl Engn Res Ctr Robot Visual Percept & Control T Changsha Peoples R China Hunan Univ Coll Elect & Informat Engn Changsha 410082 Peoples R China Hunan Univ Coll Robot Changsha 410082 Peoples R China

Human perception relies on the important sensory modes of vision and touch, which provide complementary information about the surrounding environment. Robots can leverage this multimodal sensing capability. In this study, we propose a defect recognition and classification method based on multiscale and channel-switching fusion, integrating visual and tactile modalities to achieve superior performance. defect features are extracted by processing tactile data from GelSight sensors and captured visual images through deep neural networks. However, due to the heterogeneity between the two modalities and the diversity in defect sizes and shapes, commonly used methods including feature map superposition or concatenation operations increase computational costs and reduce model classification efficiency. To address this challenge, we designed a channel-switching module and a multiscale fusion module. The channel-switching module assigns different weights to each channel in the feature map to solve the problem of mutual exclusivity or unbalanced fusion in traditional methods, while the multiscale fusion module effectively processes features of different shapes by merging features from different scales. Results obtained on a newly collected fabric dataset show that the proposed defect classification method improves classification accuracy compared to learning from single-modal data, with significant improvement in complex backgrounds. The proposed fusion network framework achieves robust defect classification performance exceeding 90%.

关键词： Fabrics Visualization Feature extraction defect detection Sensors Cameras Tactile sensors Switches Surface treatment Accuracy Channel switching deep learning defect classification fabric defect detection visual-tactile fusion

来源：评论

学校读者我要写书评

暂无评论

A novel micro-defect classification system based on attention enhancement

引用

JOURNAL OF INTELLIGENT MANUFACTURING 2024年第2期35卷 703-726页

作者： Lin, Song He, Zhiyong Sun, Lining Harbin Inst Technol Harbin Peoples R China Soochow Univ Suzhou Peoples R China

A surface micro-defect is characterized by a small size and a susceptibility to noise. Micro-defect detection and classification is very challenging. This paper proposes a Micro-defect classification system based on attention enhancement (MDCS) for solving the detection and classification of micro-defects. We combine defect detection with defect classification in MDSC. Micro-defects classification can be better realized based on the auxiliary task of defect detection. In this system, the aim of attention formation in bionic vision is to guide the system to focus on the target by zooming in and out on micro-defects. To avoid noise interference, an attention module based on trilinear feature confluence has been incorporated. Last but not least, the enhancement process based on the attention map improves the classification ability of micro-defects. As part of comparative experiment, we analyzed data including 19,200 fabric images and 4,800 bamboo images. In the micro-defect classification experiment based on MDCS(ResNet-50), the accuracy of fabric data and bamboo data is 88.2% and 89.4% respectively. Compared with ResNet-50, the classification accuracy (64.8%, 67.7%) is improved by 23.4% and 21.7% respectively. In the object detection experiment of micro-defects based on MDCS (ResNet-50), the accuracy of fabric data and bamboo data is 65.1% mAP(s) and 63.3% mAP(s) respectively. Compared with HRDNet, the detection accuracy (59.6% mAP(s), 52.2% mAP(s)) is improved by 5.5% mAP(s) and 11.1% mAP(s) respectively. Experimental results demonstrate that the proposed system can counteract the interference caused by noise in small object detection, localize micro-defects accurately, and improve micro-defect classification accuracy significantly.

关键词： Convolutional neural network Vision attention Intelligent manufacturing defect classification Deep learning

来源：评论

学校读者我要写书评

暂无评论

Volumetric defect classification in Nano-resolution X-ray computed tomography images of laser powder bed fusion via deep learning

引用

JOURNAL OF MANUFACTURING PROCESSES 2024年 121卷 499-511页

作者： Vaghefi, Ehsan Hosseini, Seyedmehrab Azimi, Mohsen Shmatok, Andrii Zhao, Rong Prorok, Bart Mirkoohi, Elham Auburn Univ Dept Mech Engn 354 War Eagle Way Auburn AL 36849 USA Georgia Inst Technol Coll Comp 801 Atlantic Dr Atlanta GA 30332 USA Auburn Univ Dept Mat Engn 275 Wilmore Labs Auburn AL 36849 USA

Additively manufactured (AMed) components often contain volumetric defects that significantly impact mechanical and fatigue properties across various material systems. Nano -resolution X-ray computed tomography (XCT) provides precise imaging of these defects, however, XCT software struggles to differentiate between types of defects (e.g., lack of fusion, gas -entrapped pores, and keyhole), each with distinct effects on mechanical properties. To address this problem, we introduce VolDefSegNet , an automated framework based on convolutional neural networks (CNNs), for semantic segmentation of common volumetric defects in AMed components. VolDefSegNet incorporates deep learning techniques and leverages seven defect features for automated defect labeling, enhancing robustness and generalization. We trained VolDefSegNet on a comprehensive dataset of approximately 90,000 labeled defects and achieved superior performance metrics (F1 -score of 0.91 and IoU of 0.68) compared to alternative models. Additionally, we explored the impact of various hyperparameters on VolDefSegNet 's performance to identify optimal hyperparameters for the framework, highlighting the framework's potential in streamlining defect classification processes despite the challenges posed by XCT machine and software.

关键词： Laser powder bed fusion Volumetric defects defect classification CNN Deep learning

来源：评论

学校读者我要写书评

暂无评论

引用

PROCEEDINGS OF THE INSTITUTION OF MECHANICAL ENGINEERS PART E-JOURNAL OF PROCESS MECHANICAL ENGINEERING 2024年第6期238卷 2809-2820页

作者： Rajeshkanna, R. Meikandan, M. Daxayani, C. Kumar, P. Ganesh Hindusthan Coll Engn & Technol Dept Elect & Elect Engn Coimbatore Tamilnadu India Vel Tech Rangarajan Dr Sagunthala R&D Inst Sci & T Dept Mech Engn Chennai Tamilnadu India SRM Inst Sci & Technol Dept Mech Engn Chennai Tamil Nadu India SRM Inst Sci & Technol Dept Mech Engn Chennai 603203 Tamil Nadu India

During the production of energy using photovoltaic (PV) panels, solar cells may be affected by different environmental aspects, which cause many defects in the solar cells. Such defects should be identified and categorized to improve energy productivity. Most recent studies developed many deep learning models to categorize solar cell defects, whereas tiny defects were not categorized due to the blurry and overlapping edges. Hence, this paper develops an effective deep-learning model of cell partitioning and defect classification to satisfy the criteria for solar panel manufacturing monitoring. At first, the low-resolution solar cell images are enriched into super-resolved images using the improved super-resolved adversarial network. Then, those images are passed to the novel deep feature discrimination network, which extracts fault-related characteristics from the solar cell images. Those characteristics are further fed to the softmax classifier to categorize the defected or non-defected solar cell images. Moreover, a new mixture of error functions is used for fine-tuning and rapid convergence. Further, a detection bias and compensation in missed faults that limit the model accuracy are found to be unaffected by the categorization of defected/non-defected panels in a real-time manufacturing unit. Finally, assigning certain forecasting standards for various faults categories, the overall test analysis using solar cell images on 5 days exhibits that the presented model attains a maximum accuracy of 99.91% compared to the classical deep learning models. So, this model has a high probability to be used in a real manufacturing unit.

关键词： Solar panel defects cell partitioning deep learning defect classification super-resolved images softmax

来源：评论

学校读者我要写书评

暂无评论

Application of Machine Learning in Automotive Stud Weld defect classification

引用

IEEE SENSORS JOURNAL 2023年第20期23卷 25547-25556页

作者： Naddaf-Sh, Sadra Naddaf-Sh, M-Mahdi Dalton, Maxim Ramezani, Soodabeh Kashani, Amir R. Zargarzadeh, Hassan Lamar Univ Philip M Drayer Elect Engn Dept Beaumont TX 77710 USA Stanley Black & Decker Houston TX 77008 USA Stanley Black & Decker Washington DC 20015 USA

Arc stud welding (ASW) is a well-developed and automated manufacturing process commonly used to fasten studs to a workpiece. This process requires the implementation of measures to prevent low-quality welds that can lead to the disposal of an entire structure. In this article, a machine learning-based method is introduced that uses experimental data to detect and classify defective stud welds. The source of data is a series of experimental arc stud welds conducted by a precision robotic arm. Welding data log is recorded through various sensors to shape a multivariate dataset for ASW. Then, wavelet transforms (WTs) as well as synthetic signal generation are applied for feature extraction. For each set of extracted features, suitable models are designed, trained, and evaluated for automatic root cause classification. Next, the models and approaches are compared on synthetic measurement generation, feature extractions, and feature importance. Numerical results show that XGBoost, the best model, has an 85.41% F1-score on the test set. Additionally, the study reveals that voltage and lift play crucial roles in the root cause classification of arc stud weld. The proposed method not only offers promising results for defect detection in ASW but also provides outreach value for similar industrial processes.

关键词： Welding Feature extraction Discrete wavelet transforms Continuous wavelet transforms Sensors Wavelet analysis Voltage measurement Convolutional neural network (CNN) defect classification machine learning stud welding wavelet transform (WT)

来源：评论

学校读者我要写书评

暂无评论

Deep Learning Based Underground Sewer defect classification Using a Modified RegNet

引用

Computers, Materials & Continua 2023年第6期75卷 5451-5469页

作者： Yu Chen Sagar A.S.M.Sharifuzzaman Hangxiang Wang Yanfen Li L.Minh Dang Hyoung-Kyu Song Hyeonjoon Moon Department of Computer Science and Engineering Sejong UniversitySeoul05006Korea Department of Intelligent Mechatronics Engineering Sejong UniversitySeoul05006Korea Department of Information of Communication Engineering and Convergence Engineering for Intelligent DroneSejong UniversitySeoul05006Korea

The sewer system plays an important role in protecting rainfall and treating urban *** to the harsh internal environment and complex structure of the sewer,it is difficult to monitor the sewer *** are developing different methods,such as the Internet of Things and Artificial Intelligence,to monitor and detect the faults in the sewer *** learning is a promising artificial intelligence technology that can effectively identify and classify different sewer system ***,the existing deep learning based solution does not provide high accuracy prediction and the defect class considered for classification is very small,which can affect the robustness of the model in the constraint *** a result,this paper proposes a sewer condition monitoring framework based on deep learning,which can effectively detect and evaluate defects in sewer pipelines with high *** also introduce a large dataset of sewer defects with 20 different defect classes found in the sewer *** study modified the original RegNet model by modifying the squeeze excitation(SE)block and adding the dropout layer and Leaky Rectified Linear Units(LeakyReLU)activation function in the Block structure of RegNet *** study explored different deep learning methods such as RegNet,ResNet50,very deep convolutional networks(VGG),and GoogleNet to train on the sewer defect *** experimental results indicate that the proposed system framework based on the modified-RegNet(RegNet+)model achieves the highest accuracy of 99.5 compared with the commonly used deep learning *** proposed model provides a robust deep learning model that can effectively classify 20 different sewer defects and be utilized in real-world sewer condition monitoring applications.

关键词： Deep learning defect classification underground sewer computer vision convolutional neural network RegNet

来源：评论

学校读者我要写书评

暂无评论

Image defect classification of surface mount technology welding based on the improved ResNet model

JOURNAL OF ENGINEERING RESEARCH

引用

JOURNAL OF ENGINEERING RESEARCH 2024年第2期12卷 154-162页

作者： Zhang, Qiang Zhang, Kaiyun Pan, Kailin Huang, Wei Guilin Univ Elect Technol Sch Mech & Elect Engn Guilin 541004 Peoples R China Guilin Univ Elect Technol Engn Res Ctr Elect Informat Mat & Devices Minist Educ Guilin 541004 Peoples R China Ind & Commercial Bank China Ltd Software Dev Ctr Zhuhai 710000 Peoples R China

In mass production, welding flaw detection in existing surface mount technology (SMT) has certain constraints, including its high costs, heavy workloads, and time-consuming processes. However, image classification technology using computer vision demonstrates high detection speeds and considerably reduced detection costs in flaw detection. Nevertheless, the increased integration of chip components on printed circuit boards (PCBs) and reduced component sizes pose challenges for flaw detection technology. Therefore, in this paper, an SMT welding image flaw classification model-that is, the ResNet-34-ECA model-based on an improved ResNet model, is proposed. Initially, the dataset is amplified using data amplification methods, such as stochastic rotation, increased data diversity, and enhanced model robustness. The ResNet34 model is then optimized using the light quantization efficient channel attention (ECA) module, resulting in higher classification accuracy. The experimental data in this study were collected using automated optical inspection (AOI) equipment, following the manual creation and amplification of the dataset. The experimental results showed that the baseline model accuracy increased by 0.22 in the augmented dataset, reaching 97.2%. Moreover, the ResNet-34-ECA model proposed in this paper could realize the classification of SMT welding image defects successfully;the overall classification accuracy of the improved ResNet image classification model was 0.01 higher than that of the baseline model, reaching 98.2%. Consequently, the proposed model proves to be better than other models in defect classification on this dataset, providing an accurate classification of SMT welding image defects.

关键词： Attentional mechanisms Chip components defect classification Efficient channel attention (ECA) module ResNet Surface mount technology (SMT)

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：