Department of Mechanical Engineering, Pohang University of Science and Technology (POSTECH), 3 77 Cheonam-ro, Pohang, Gyeongbuk 37673, Republic of Korea
Auto-generated profile, based on publicly available information
Hyunsuk Huh
Department of Mechanical Engineering, Pohang University of Science and Technology (POSTECH), 3 77 Cheonam-ro, Pohang, Gyeongbuk 37673, Republic of Korea
In the era of the fourth industrial revolution (Industry 4.0) and the Internet of Things (IoT), real-time data is enormously collected and analyzed from mechanical equipment. By classifying and characterizing the measured signals, the fault condition of mechanical components could be identified. However, most current health monitoring techniques utilize time-consuming and labor-intensive feature engineering, i.e., feature extraction and selection, that are carried out by experts. This paper, on the contrary, deals with an automatic diagnosis method of machine monitoring using a convolutional neural network (CNN) with class activation maps (CAM). A class activation map enables us to discriminate the fault region in the images, thus allowing us to localize the fault precisely. The goal of the paper is to demonstrate how CNN and CAM could be employed to real-world vibration video to characterize the machine’s status, representing normal or fault conditions. The performance of the proposed model is validated with a base-excited cantilever beam dataset and a water pump dataset. This paper presents a novel industrial application by developing a promising method for automatic machine condition-based monitoring.
Detection and diagnosis of material degradation are of a complex and challenging task since it is presently hand-operated by a human. Therefore, it leads to misinterpretation and avoids correct classification and diagnosis. In this paper, we develop a computer-assisted detection method of material failure by utilizing a deep learning approach. A deep convolutional neural network (CNN) model, combined with an image processing technique, e.g., adaptive histogram equalization, is trained to classify a real-world turbine tube degradation image data set, which is retrieved from a power generation company. The experimental result demonstrates the effectiveness of the proposed approach with predictive classification accuracy is up to 99.99% in comparison with a shallow machine learning algorithm, e.g., linear SVM. Furthermore, performance evaluation of a deep CNN with and without an above-mentioned image processing technique is exhibited and benchmarked. We successfully demonstrate a novel application in constructing a deep-structure neural network model for material degradation diagnosis, which is not available in the current literature.
Cost-effective vertical etching of plug holes and word lines is crucial in enhancing 3D NAND device manufacturability. Even though multiscale technology computer-aided design (TCAD) methodology is suitable for effectively predicting etching processes and optimizing recipes, it is highly time-consuming. This article demonstrates that our deep learning platform called TCAD-augmented Generative Adversarial Network can reduce the computational load by 2 600 000 times. In addition, because well-calibrated TCAD data based on physical and chemical mutual reactions are used to train the platform, the etching profile can be predicted with the same accuracy as TCAD-only even when the actual experimental data are scarce. This platform opens up new applications, such as hot spot detection and mask layout optimization, in a chip-level area of 3D NAND fabrication.
2 July 2025 in Engineering Applications of Artificial Intelligence