Detection Of Manufacturing Defects Research Articles

A Methodology for Advanced Manufacturing Defect Detection through Self-Supervised Learning on X-ray Images

In industrial quality control, especially in the field of manufacturing defect detection, deep learning plays an increasingly critical role. However, the efficacy of these advanced models is often hindered by their need for large-scale, annotated datasets. Moreover, these datasets are mainly based on RGB images, which are very different from X-ray images. Addressing this limitation, our research proposes a methodology that incorporates domain-specific self-supervised pretraining techniques using X-ray imaging to improve defect detection capabilities in manufacturing products. We employ two pretraining approaches, SimSiam and SimMIM, to refine feature extraction from manufacturing images. The pretraining stage is carried out using an industrial dataset of 27,901 unlabeled X-ray images from a manufacturing production line. We analyze the performance of the pretraining against transfer-learning-based methods in a complex defect detection scenario using a Faster R-CNN model. We conduct evaluations on both a proprietary industrial dataset and the publicly available GDXray dataset. The findings reveal that models pretrained with domain-specific X-ray images consistently outperform those initialized with ImageNet weights. Notably, Swin Transformer models show superior results in scenarios rich in labeled data, whereas CNN backbones are more effective in limited-data environments. Moreover, we underscore the enhanced ability of the models pretrained with X-ray images in detecting critical defects, crucial for ensuring safety in industrial settings. Our study offers substantial evidence of the benefits of self-supervised learning in manufacturing defect detection, providing a solid foundation for further research and practical applications in industrial quality control.

Development of an Autonomous Component Testing System with Reliability Improvement Using Computer Vision and Machine Learning

This study evaluated computer vision-based models, including Histogram Analysis, Logistic Regression, Sift-SVM, and Deep learning models, in an autonomous testing system developed for smartphone camera modules. System performance was assessed in a practical factory setting with workers operating the system, and metrics such as processing time, sensitivity, specificity, accuracy, and defect rate were evaluated. Based on the results, the Sift-SVM model demonstrated the greatest potential for enhancing the reliability of the system with a processing time of 0.01578 seconds, a sensitivity of 99.811%, and a reduction in the failure rate to 1888 PPM. The study findings suggest that Sift-SVM has the potential to be practically applied in the industry, thus improving the speed and accuracy of automatic defect detection in manufacturing and reducing the defect rate.

Visual analytics and intelligent reasoning for smart manufacturing defect detection and judgement: A meta-learning approach with knowledge graph embedding case-based reasoning

Visual inspection is an important procedure in manufacturing industries to guarantee product quality and reduce costs of poor quality. Several challenges exist for manufacturing inspection automation, including the false positive and false negative trade-off, autonomous decision-making, and inspection algorithm selection. This paper addresses these challenges by proposing a two-level visual analytics and intelligent reasoning system, where machine vision identifies and locates potential defects and convolutional neural network models further classify them into acceptable and defective features. Symbolic reasoning techniques are applied for autonomous decision making on part-level judgement to enhance the system reasoning capability, including part judgement (accept, reject, rework) and rework operation assignment. Furthermore, a case-based reasoning approach for meta-learning of inspection algorithm selection is proposed, which uses knowledge graph embedding (TransR embedding) to embed cases into a two-dimensional space for case similarity estimation and case retrieval. A case of smart assembly inspection is presented to validate the feasibility of the proposed system.

Detection of Manufacturing Defects in Lithium-Ion Batteries-Analysis of the Potential of Computed Tomography Imaging

Realising an ideal lithium-ion battery (LIB) cell characterised by entirely homogeneous physical properties poses a significant, if not an impossible, challenge in LIB production. Even the slightest deviation in a process parameter in its production leads to inhomogeneities and causes a deviation in performance parameters of LIBs within the same batch. The greater the number and/or intensity of inhomogeneities, the more they need to be avoided. Severe inhomogeneities (defects), such as metal particle contamination, significantly impact the cell’s performance. Besides electrical measurements, image-based measurement methods can be used to identify defects and, thus, ensure the production quality and safety of LIBs. While the applicability of computed tomography (CT) as an image-based measurement method for detecting defects has been proven, the limitations of this method still need to be determined. In this study, a systematic analysis of the capabilities of CT imaging was conducted. A multilayer pouch cell without an electrolyte was reassembled with several defects on one of the middle anodes. To investigate the boundaries of CT, defects such as a partial and complete removal of the coating, a cut, or a kink, as well as particle contaminations of various sizes and materials (aluminium, copper, iron) were chosen. By comparing the CT images of the cell using laser scanning microscope images of the defective anode, it could be proven that all selected defects except the kink were detectable.

Giant enhancement of the Goos–Hänchen shift assisted by merging bound states in the continuum

Bound states in the continuum (BICs) have received considerable attention in the field of nanophotonics due to their highly confined resonance and high Q factor, which effectively eliminates radiation loss. Various periodic structures have been studied to achieve BICs, with photonic crystal slabs (PCSs) being a prominent example. In PCS, multiple BICs can be merged to strongly suppress out-of-plane-scattering losses caused by fabrication imperfections. In this paper, we investigate the impact of reflection-type merging BICs on the Goos–Hänchen shift (GH shift) and demonstrate a remarkable enhancement of the GH shift, exceeding five orders of wavelength. We show the dynamic changes of the GH shifts with the isolated, merging, and merged BICs, achieving positive and negative GH shifts in different angles of peak reflectance for the same frequency. Our research highlights that even minor fabrication imperfections can result in a significant change in the GH shift, which can serve as a means for detecting manufacturing defects. Furthermore, we propose an ultrasensitive environmental refractive index sensor based on the enhanced GH shift by an isolated BIC. Our study contributes to the understanding of BICs and their potential applications in nanophotonics, including advanced optical communication devices, nanodevice fabrication, and highly sensitive sensors.

A Robust Asymmetric Kernel Function for Bayesian Optimization, With Application to Image Defect Detection in Manufacturing Systems

Some response surface functions in complex engineering systems are usually highly nonlinear, unformed, and expensive to evaluate. To tackle this challenge, Bayesian optimization (BO), which conducts sequential design via a posterior distribution over the objective function, is a critical method used to find the global optimum of black-box functions. Kernel functions play an important role in shaping the posterior distribution of the estimated function. The widely used kernel function, e.g., radial basis function (RBF), is very vulnerable and susceptible to outliers; the existence of outliers is causing its Gaussian process (GP) surrogate model to be sporadic. In this article, we propose a robust kernel function, asymmetric elastic net radial basis function (AEN-RBF). Its validity as a kernel function and computational complexity are evaluated. When compared with the baseline RBF kernel, we prove theoretically that AEN-RBF can realize smaller mean squared prediction error under mild conditions. The proposed AEN-RBF kernel function can also realize faster convergence to the global optimum. We also show that the AEN-RBF kernel function is less sensitive to outliers, and hence improves the robustness of the corresponding BO with GPs. Through extensive evaluations carried out on synthetic and real-world optimization problems, we show that AEN-RBF outperforms the existing benchmark kernel functions. Note to Practitioners—Some industrial systems cannot be accurately represented by physical models. In this situation, data-driven black-box optimization is necessary for advancing the system automation and intelligence. BO is one of the widely used strategies for learning the global optimum of black-box functions. BO has been applied to robotics, anomaly detection, automatic learning algorithm configuration, reinforcement learning, and deep learning. This article proposes one new kernel function, named after AEN-RBF. The new kernel function will make BO with GPs more robust to outliers and lower the data quality barrier of model training. This article was motivated by the hyperparameter tuning problem of deep learning models for image defect detection in advanced manufacturing, but the method can be easily extended to other applications where kernel functions are needed. Our proposed method is verified by synthetic and real-world optimization problems.

Defect Detection in Transparent Printed Electronics Using Learning-Based Optical Inspection

Printed electronics (PE) is an emerging technology that provides attractive and complementary features compared to traditional wafer-scale silicon fabrication, such as flexible substrate and point-of-use ultralow-cost manufacturing. The low-cost manufacturing and larger feature sizes mandate reduced complexity in circuit size and also limited and transparent printing layers. This enables optical inspection for manufacturing defect detection, eliminating the need for electrical testing for gross defect detection. Therefore, the traditional problem of controllability and observability in logic testing can completely be alleviated. In this article, we present a learning-based method for optical inspection to detect defective transistors in transparent PE. The method leverages domain-specific as well as common inspection features extracted from optical images to detect defective transistors using supervised learning algorithms trained with real fabricated transistor images. The results show that the proposed method detects 95% of the defective transistors, which can significantly reduce the cost of the overall test flow.

Application of mathematical morphology operation with memristor-based computation-in-memory architecture for detecting manufacturing defects

Mathematical morphology operations are widely used in image processing such as defect analysis in semiconductor manufacturing and medical image analysis. These data-intensive applications have high requirements during hardware implementation that are challenging for conventional hardware platforms such as central processing units (CPUs) and graphics processing units (GPUs). Computation-in-memory (CIM) provides a possible solution for highly efficient morphology operations. In this study, we demonstrate the application of morphology operation with a novel memristor-based auto-detection architecture and demonstrate non-neuromorphic computation on a multi-array-based memristor system. Pixel-by-pixel logic computations with low parallelism are converted to parallel operations using memristors. Moreover, hardware-implemented computer-integrated manufacturing was used to experimentally demonstrate typical defect detection tasks in integrated circuit (IC) manufacturing and medical image analysis. In addition, we developed a new implementation scheme employing a four-layer network to realize small-object detection with high parallelism. The system benchmark based on the hardware measurement results showed significant improvement in the energy efficiency by approximately 358 times and 32 times more than when a CPU and GPU were employed, respectively, exhibiting the advantage of the proposed memristor-based morphology operation.

Autoencoder-based anomaly detection for surface defect inspection

In this paper, the unsupervised autoencoder learning for automated defect detection in manufacturing is evaluated, where only the defect-free samples are required for the model training. The loss function of a Convolutional Autoencoder (CAE) model only aims at minimizing the reconstruction errors, and makes the representative features widely spread. The proposed CAE in this study incorporates a regularization that improves the feature distribution of defect-free samples within a tight range. It makes the representative feature vectors of all training samples as close as possible to the mean feature vector so that a defect sample in the evaluation stage can generate a distinct distance from the trained center of defect-free samples. The proposed CAE model with regularizations has been tested on a variety of material surfaces, including textural and patterned surfaces in images. The experimental results reveal that the proposed CAE with regularizations significantly outperforms the conventional CAE for defect detection applications in the industry.

Design and implementation of convolutional neural network-based SVM technique for manufacturing defect detection

This paper introduces the design, implementation, training and testing of deep convolutional neural network (DCNN)-based support vector machines (SVMs). These DCNN-based SVMs are designed using software tools developed by the authors that enable them to construct, train, and test the DCNN-based SVMs and effectively facilitate vision-based inspection to detect different undesirable manufacturing defects. Two pretrained DCNNs are used for this purpose: the sssNet is developed by the authors and was trained using many actual and simple target images consisting of seven categories, and the standard AlexNet that was trained by a large number of images consisting of 1,000 categories. In this work, the pretrained sssNet and AlexNet are used as feature vector extractors in training and testing. The generated feature vectors are used as inputs to train SVMs for the final binary classification represented as accept (OK) or reject (NG) category.

Design and implementation of convolutional neural network-based SVM technique for manufacturing defect detection

Design and implementation of convolutional neural network-based SVM technique for manufacturing defect detection

Detecting Manufacturing Defects via Robust Registration Using Ridgelines of Industrial Products

Detecting Manufacturing Defects via Robust Registration Using Ridgelines of Industrial Products

CT Inspection of Cooled Turbine Blades

AbstractTo improve the engine efficiency by increasing the gas temperature, multi-layer cooling passages are applied in new designs of gas turbine blades. As a result, traditional non-destructive methods, which have been applied so far in the production control, became insufficient. The aim of this manuscript is to develop an inspection method for cooling passages of turbine blades, which would be helpful in detecting manufacturing defects of blades and their operational failures. GE v/tome/x/m 300 was applied to conduct dimensional control and check the interior of two types of turbine blades from turboshaft engines. The procedure for selecting X-ray parameters was suggested. The thickness of walls in the selected cross-section was measured with the accuracy of 0.01 mm, and the selected manufacturing defects of cooling passages were identified.

81‐1: Invited Paper: Data Augmentation for Applying Deep Learning to Display Manufacturing Defect Detection

Insufficient and imbalance data samples often prevent the development of accurate deep learning models for manufacturing defect detection. By applying data augmentation methods ‐ including VAE latent space oversampling and random data generation, and GAN multi‐modal complementary data generation, we overcome the dataset limitations and achieve Pass/No‐Pass accuracies of over 90%.

Deep Learning-Based Solar-Cell Manufacturing Defect Detection With Complementary Attention Network

The automatic defects detection for solar cell electroluminescence (EL) images is a challenging task, due to the similarity of defect features and complex background features. To address this problem, in this article a novel complementary attention network (CAN) is designed by connecting the novel channel-wise attention subnetwork with spatial attention subnetwork sequentially, which adaptively suppresses the background noise features and highlights the defect features simultaneously by employing the complementary advantage of the channel features and spatial position features. In CAN, the novel channel-wise attention subnetwork applies convolution operation to integrate the concatenated and discriminative output features extracted by global average pooling layer and global max pooling layer, which can make fully use of these informative features. Furthermore, a region proposal attention network (RPAN) is proposed by embedding CAN into region proposal network in faster R-CNN (convolution neutral network) to extract more refined defective region proposals, which is used to construct a novel end-to-end faster RPAN-CNN framework for detecting defects in raw EL image. Finally, some experimental results on a large-scale EL dataset including 3629 images, 2129 of which are defective, show that the proposed method performs much better than other methods in terms of defects classification and detection results in raw solar cell EL images.

A Perspective on Using Machine Learning in 3D Bioprinting.

Recently, three-dimensional (3D) printing technologies have been widely applied in industry and our daily lives. The term 3D bioprinting has been coined to describe 3D printing at the biomedical level. Machine learning is currently becoming increasingly active and has been used to improve 3D printing processes, such as process optimization, dimensional accuracy analysis, manufacturing defect detection, and material property prediction. However, few studies have been found to use machine learning in 3D bioprinting processes. In this paper, related machine learning methods used in 3D printing are briefly reviewed and a perspective on how machine learning can also benefit 3D bioprinting is discussed. We believe that machine learning can significantly affect the future development of 3D bioprinting and hope this paper can inspire some ideas on how machine learning can be used to improve 3D bioprinting.

Manufacturing Defect Detection in Ceiling Fans using Image processing

During manufacturing process of ceiling fans there may be possibility that any step of manufacturing process can be skipped or improper completion due to malfunctioning of the system. In manufacturing process of fans, the outer plate, stator, rotor, axle and other parts are manufactured and assembled. If the winding machine is not working properly and improper windings is done and no-one can acknowledge that winding machine is not working properly then the whole batch should be designed with defect and this will surely create negative impact on the production process of the industry. So this proposed work will overcome this problem. This paper guides to detect whether the windings are proper or not and detection is carried out by taking picture of armature-windings. If there is any problem in the windings then our system will generate an alert so that other armatures can be protected from failures. Manual results are not so accurate all the time but in manufacturing process we need high accuracy. For this skilled labor is required and for skilled labor we have to pay more, this automated system reduce the need of skilled labor. If fan manufacturing industries use this image processing system then this will be very helpful in increasing production of fans within the completion time with more accuracy

Experimental study of flexible injection to manufacture parts of strong curvature

Flexible injection (FI) is a new process for the manufacture of high performance composites, which consists of injecting a thermosetting resin through a fibrous reinforcement contained in the lower chamber of a double cavity mold. Resin is injected in the lower cavity, which is sealed by a membrane, and then a compaction fluid is injected in the upper chamber to compress the reinforcement. This new composite manufacturing technique, which allows a limited and controlled deformation of the flexible membrane during processing, was shown to be very effective in reducing filling times in the case of planar or slightly curved geometries. In the present study, flexible injection is applied to strongly curved parts, namely here a composite rectangular panel with two 90° corners. After setting up an experimental procedure to produce the stair-shaped components out of fiberglass and vinylester resin, longitudinal cross-sections of the parts are analyzed to assess the quality of the final product in both the flat and curved zones. This characterization method allows detecting manufacturing defects such as thickness gradients or resin-rich zones. Such defects are likely to induce geometrical deformations of the component and may decrease its mechanical performance. Therefore they ought to be minimized to improve the overall quality of the part. Modifications of the manufacturing procedure are proposed in this article to decrease the importance of process-induced defaults and improve the performance of the flexible injection. POLYM. COMPOS., 2011. © 2011 Society of Plastics Engineers

A novel manufacturing defect detection method using association rule mining techniques

In recent years, manufacturing processes have become more and more complex, and meeting high-yield target expectations and quickly identifying root-cause machinesets, the most likely sources of defective products, also become essential issues. In this paper, we first define the root-cause machineset identification problem of analyzing correlations between combinations of machines and the defective products. We then propose the Root-cause Machine Identifier (RMI) method using the technique of association rule mining to solve the problem efficiently and effectively. The experimental results of real datasets show that the actual root-cause machinesets are almost ranked in the top 10 by the proposed RMI method.

Sensitivity of the Electromagnetic Acoustical Technique for Testing Railway Rails by the Mirror-Shadow Method

A study of the sensitivity of the electromagnetic acoustical (EMA) technique for detecting manufacturing defects of railway rails by a mirror-shadow method (MSM) has been carried out. It has been found that the sensitivity of the MSM with the EMA technique using bulk shear ultrasonic waves is, in certain cases, considerably higher than the sensitivity of the traditional “wet” echo-pulse method or mirror-shadow method with the use of longitudinal waves. Therefore, in order to renovate or design new technical means of ultrasonic acceptance testing of rails in factories, it is advisable to use the mirror-shadow method of testing with bulk elastic shear waves polarized normally to the direction of rail rolling. This would make it possible to reduce the acceptance of intolerable internal flaws of manufacturing origin.