Article
Chemistry, Analytical
Eunmi Shin, Chang D. Yoo
Summary: The purpose of this study was to classify defect patterns on wafers with limited resources and time. The study introduced an efficient convolutional neural network model that achieved high performance while reducing resource usage and processing time. By classifying nine frequently found defect patterns, the experiment demonstrated that the model outperformed traditional models in terms of parameter usage, training speed, and inference speed, with an accuracy of 98% and an F1 score of 89.5%.
Article
Computer Science, Interdisciplinary Applications
Dong-Hee Lee, Eun-Su Kim, Seung-Hyun Choi, Young-Mok Bae, Jong-Bum Park, Young-Chan Oh, Kwang-Jae Kim
Summary: This study proposes a new method for defect pattern classification, which combines information on shape, size, location, and bin dimensions. The Bin2Vec method is used to determine the RGB code for each bin, and three levels of bin dimensions are defined through analysis and clustering of a large number of WBMs. Compared to existing taxonomies, this method can identify major defect patterns, new defect patterns, and non-critical defect patterns.
COMPUTERS IN INDUSTRY
(2023)
Article
Computer Science, Artificial Intelligence
Rui Wang, Songhao Wang, Ben Niu
Summary: This study proposes a shape prior guided method for the identification and separation of mixed-type defect patterns. The method deforms shape templates to match the patterns and achieves accurate results.
JOURNAL OF INTELLIGENT MANUFACTURING
(2023)
Article
Computer Science, Artificial Intelligence
Jaewoong Shim, Seokho Kang
Summary: Automation of wafer map pattern classification is a hot topic in semiconductor manufacturing for identifying root causes. The challenge lies in building a classification model due to the high cost of labeling wafer maps with defect categories. This study proposes a method using synthetic wafer maps generated from normal and single-defect maps to train a CNN for accurately classifying mixed-defect maps, and it outperforms existing methods according to experimental results using the MixedWM38 dataset.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Computer Science, Information Systems
Ho Sun Shon, Erdenebileg Batbaatar, Wan-Sup Cho, Seong Gon Choi
Summary: Visual defect inspection and classification are crucial steps in the manufacturing processes of the semiconductor and electronics industries. Deep learning techniques with data augmentation are commonly used for identifying wafer map defect patterns. This study proposes a DL-based method with automatic data augmentation, demonstrating its effectiveness through experimental results.
Article
Metallurgy & Metallurgical Engineering
Yu Nai-gong, Xu Qiao, Wang Hong-lu, Lin Jia
Summary: This paper proposes a WBM defect pattern inspection strategy based on the DenseNet deep learning model, with improvements made to the structure and training loss function. Additionally, a constrained mean filtering algorithm is introduced to filter noise grains. Using an entropy-based Monte Carlo dropout algorithm for model prediction, the uncertainty of the model decision can be quantified. Experimental results demonstrate that the improved DenseNet has better recognition ability compared to traditional algorithms in typical WBM defect patterns. Analyzing model uncertainty can effectively reduce the miss or false detection rate and help identify new patterns.
JOURNAL OF CENTRAL SOUTH UNIVERSITY
(2021)
Article
Engineering, Industrial
Sumin Kim, Heeyoung Kim
Summary: In this study, a new training algorithm called sample bootstrapping is proposed to address the issue of overfitting when training on mislabeled data in semiconductor manufacturing.
QUALITY ENGINEERING
(2023)
Article
Computer Science, Artificial Intelligence
Chia-Yu Hsu, Ju-Chien Chien
Summary: This study proposes an ensemble convolutional neural network (ECNN) framework for WBM pattern classification, which effectively identifies common WBM defect patterns and performs superior in terms of precision, recall, F1, and other conventional machine learning classifiers.
JOURNAL OF INTELLIGENT MANUFACTURING
(2022)
Article
Computer Science, Artificial Intelligence
Michal Koziarski
Summary: This paper proposes a unified framework for addressing the issues of oversampling and undersampling in imbalanced data, utilizing radial basis functions to preserve the original shape of class distributions and optimizing the positions of synthetic observations. Experimental results show that the proposed approach outperforms state-of-the-art resampling algorithms in handling imbalanced datasets.
PATTERN RECOGNITION
(2021)
Article
Chemistry, Multidisciplinary
Kunti Robiatul Mahmudah, Fatma Indriani, Yukiko Takemori-Sakai, Yasunori Iwata, Takashi Wada, Kenji Satou
Summary: Converting binary features into numerical ones can improve the performance of oversampling methods in classification tasks. Through experiments, it was observed that converting binary features into numerical features before applying oversampling methods resulted in maximum improvements of 35.11% in accuracy and 42.17% in F1-score.
APPLIED SCIENCES-BASEL
(2021)
Article
Automation & Control Systems
Youngjae Bae, Seokho Kang
Summary: In this paper, an improved training method for wafer map pattern classification is proposed, which trains a convolutional neural network using supervised contrastive learning. The experimental results show that the proposed method can enhance classification accuracy compared with existing methods, particularly when the training dataset is small.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Article
Engineering, Manufacturing
Hyungu Kahng, Seoung Bum Kim
Summary: This study introduces a self-supervised learning framework that utilizes unlabeled data to learn visual representations for efficient WBM defect pattern classification. By pre-training with noise-contrastive estimation on vast amounts of unlabeled data and fine-tuning on limited labeled data, the network significantly improves classification performance in cases of scarce labels.
IEEE TRANSACTIONS ON SEMICONDUCTOR MANUFACTURING
(2021)
Article
Chemistry, Multidisciplinary
Sangwoo Park, Cheolwoo You
Summary: In the semiconductor industry, achieving a high production yield is crucial. Wafer bin maps (WBMs) provide vital information for identifying anomalies in the manufacturing process. This study proposes a deep learning method using generative adversarial networks to improve the accuracy of defect pattern classification in the presence of imbalanced data.
APPLIED SCIENCES-BASEL
(2023)
Article
Computer Science, Interdisciplinary Applications
Wooksoo Shin, Hyungu Kahng, Seoung Bum Kim
Summary: This study proposes a method to use WBM data with only a single defect to train convolutional neural networks (CNNs) for classifying mixed-type defects. The method generates mixed-type defect patterns on the fly for model training, improving the classification performance compared to previous methods. The effectiveness and applicability of the proposed method are demonstrated through experiments on a real-world WBM benchmark.
COMPUTERS & INDUSTRIAL ENGINEERING
(2022)
Article
Computer Science, Artificial Intelligence
Jinjun Ren, Yuping Wang, Yiu-ming Cheung, Xiao-Zhi Gao, Xiaofang Guo
Summary: Class-imbalanced classification is a challenging problem where traditional classifiers exhibit bias towards majority classes and generate incorrect predictions. Existing algorithms struggle with the issue of class overlapping. This paper proposes a grouping scheme for minority class samples based on their possibilities of appearing in overlapping regions in the feature space. A new oversampling method is then proposed to generate samples far away from the overlapping region and rectify the decision boundary. An effective classification algorithm for imbalanced data is developed based on these techniques. Extensive experiments demonstrate the superiority of the proposed algorithm over seventeen benchmark algorithms, particularly on highly imbalanced datasets.
PATTERN RECOGNITION
(2023)
Article
Statistics & Probability
Dong-Hee Lee
COMMUNICATIONS IN STATISTICS-THEORY AND METHODS
(2020)
Article
Engineering, Multidisciplinary
In-Jun Jeong, Dong-Hee Lee
ENGINEERING OPTIMIZATION
(2020)
Article
Computer Science, Interdisciplinary Applications
Xianhui Yin, Zhanwen Niu, Zhen He, Zhaojun(Steven) Li, Donghee Lee
COMPUTERS & INDUSTRIAL ENGINEERING
(2020)
Article
Engineering, Industrial
Dong-Hee Lee, Jin-Kyung Yang, So-Hee Kim, Kwang-Jae Kim
QUALITY ENGINEERING
(2020)
Article
Engineering, Multidisciplinary
Dong-Hee Lee, So-Hee Kim, Jai-Hyun Byun
QUALITY AND RELIABILITY ENGINEERING INTERNATIONAL
(2020)
Article
Engineering, Multidisciplinary
Dong-Hee Lee, Jin-Kyung Yang, Kwang-Jae Kim
QUALITY AND RELIABILITY ENGINEERING INTERNATIONAL
(2020)
Article
Computer Science, Artificial Intelligence
Chang-Ho Lee, Dong-Hee Lee, Young-Mok Bae, Seung-Hyun Choi, Ki-Hun Kim, Kwang-Jae Kim
Summary: This study proposes a heuristic approach to derive the golden paths in a multistage manufacturing process by extracting superior machine sequence patterns and excluding inferior machine sequence patterns to produce products whose quality exceeds the desired level.
JOURNAL OF INTELLIGENT MANUFACTURING
(2022)
Article
Green & Sustainable Science & Technology
Do-Hyeon Ryu, Ryu-Hee Kim, Seung-Hyun Choi, Kwang-Jae Kim, Young Myoung Ko, Young-Jin Kim, Minseok Song, Dong Gu Choi
Article
Engineering, Industrial
Dong-Hee Lee, So-Hee Kim, Eun-Su Kim, Kwang-Jae Kim, Zhen He
Summary: This article introduces a new approach for multiresponse optimization using a classification and regression tree method and desirability functions, which can simultaneously optimize multiple responses. A case study of a steel manufacturing company demonstrates the effectiveness of the proposed method in obtaining an optimal region for simultaneously optimizing multiple responses.
QUALITY ENGINEERING
(2021)
Article
Engineering, Industrial
Chang-Ho Lee, Dong-Hee Lee, Seung-Hyun Choi, Kwang-Jae Kim
Summary: In a multistage manufacturing process, identical machines are utilized at each process stage, but with different operational performances which can impact the quality of the final product. This research proposes deriving golden paths (GPs) to ensure high-quality product production even as machine performance degrades over time. By introducing health indicators and discovering machine sequence patterns, the proposed approach aims to increase production of superior quality products and reduce re-production costs.
JOURNAL OF MANUFACTURING SYSTEMS
(2021)
Article
Automation & Control Systems
Jin-Soo Cho, Dong-Hee Lee, Gi-Jeong Seo, Duck-Bong Kim, Seung-Jun Shin
Summary: In this study, wire + arc additive manufacturing was used to perform welding experiments, and response surface models were fitted based on the measured geometry. The optimal process parameters were obtained using a desirability function method.
INTERNATIONAL JOURNAL OF ADVANCED MANUFACTURING TECHNOLOGY
(2022)
Article
Food Science & Technology
Dong-Hee Lee, Eun-Su Kim, Jin-Soo Cho, Jae-Hee Ryu, Byung-Seok Min
Summary: In the food industry, companies use image classification methods like CNNs to inspect X-ray images of food for foreign bodies. However, obtaining a large training dataset is challenging due to the time-consuming and labor-intensive manual labeling task. To address this, we propose an automatic labeling method that overlays additional X-ray images and identifies abnormal food items based on a predetermined threshold.
JOURNAL OF FOOD MEASUREMENT AND CHARACTERIZATION
(2022)
Article
Computer Science, Interdisciplinary Applications
Xiaolin Wang, Liyi Zhan, Yong Zhang, Teng Fei, Ming-Lang Tseng
Summary: This study proposes an environmental cold chain logistics distribution center location model to reduce transportation costs and carbon emissions. It also introduces a hybrid arithmetic whale optimization algorithm to overcome the limitations of the conventional algorithm.
COMPUTERS & INDUSTRIAL ENGINEERING
(2024)
Article
Computer Science, Interdisciplinary Applications
Hong-yu Liu, Shou-feng Ji, Yuan-yuan Ji
Summary: This study proposes an architecture that utilizes Ethereum to investigate the production-inventory-delivery problem in Physical Internet (PI), and develops an iterative heuristic algorithm that outperforms other algorithms. However, due to gas prices and consumption, blockchain technology may not always be the optimal solution.
COMPUTERS & INDUSTRIAL ENGINEERING
(2024)
Article
Computer Science, Interdisciplinary Applications
Paraskevi Th. Zacharia, Elias K. Xidias, Andreas C. Nearchou
Summary: This article discusses the assembly line balancing problem in production lines with collaborative robots. Collaborative robots have the potential to improve automation, productivity, accuracy, and flexibility in manufacturing. The article explores the use of a problem-specific metaheuristic to solve this complex problem under uncertainty.
COMPUTERS & INDUSTRIAL ENGINEERING
(2024)