http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
An Approach to Enhancement of Underwater Laser Cutting Process Monitoring Video
Sungmoon Joo,Dongjun Hyun,Ikjune Kim,Jaehyun Ha,Jonghwan Lee,Taeyoung Ko 한국방사성폐기물학회 2022 한국방사성폐기물학회 학술논문요약집 Vol.20 No.2
For highly contaminated elements such as reactor pressure vessels or reactor internals, it is a viable option to cool-down and dismantle these elements in submerged (e.g. underwater) state. Several tools and processes such as saw cutting, water jet cutting or plasma cutting are currently used for underwater cutting, with each of them having their own advantages and disadvantages. The main disadvantage of these existing methods, especially saw and water jet cutting, is the generation of secondary waste that then needs to be filtered out of the water. In addition, in the case of water jet cutting, a considerable amount of abrasive material is added, which must also be stored. To overcome this drawback, the feasibility of using laser cutting under water to minimize secondary waste production has been actively studied recently. One of the challenges with the underwater laser cutting is to visually monitor the cutting process. Flowing air bubbles generated by the cutting gas and the flashing light emitted from the laser and melting material prohibit visual monitoring of the cutting process. This study introduces a method to enhance the video from a monitoring camera. Air bubbles can be detected by computing optical flows and the video quality can be enhanced by selective removal of the detected bubbles. In addition, suppressing the frame image update from flashing light area can also effectively enhance the video quality. This paper describes the simple yet effective video quality enhancement method and reports preliminary results.
Development of Scan Simulator for Automation of Nuclear Facility 3D Modeling
Sungmoon Joo 한국방사성폐기물학회 2022 한국방사성폐기물학회 학술논문요약집 Vol.20 No.2
During the decommissioning of nuclear facilities, 3D digital model that precisely describes the work environment can expedite the accomplishment of the work. Thus, the workers’ exposure to radiation is minimized and the safety risk to the workers is reduced, while precluding inadvertent effects on the environment. However, it is common that the 3D model does not exist for legacy nuclear facilities as most of the initial design drawings are 2D drawings and even some of the 2D drawings are missing. Even in the case that all of the 2D drawings are intact, these initial design drawings need to be updated using asbuilt data because facilities get modified through years of operation. In those cases, 3D scanning can be a good option to quickly and accurately generate a structure’s actual 3D geometric information. 3D scanning is a technique used to capture the shape of an object in the form of point cloud. Point cloud is a collection of large number of points on the external surfaces of objects measured by 3D scanners. The conversion of point cloud to 3D digital model is a labor-intensive process as a human worker needs to recognize objects in the point cloud and convert the objects into 3D model, even though some of the conversion process can be automated by using commercial software packages. With the aim of full automation of scan-to-3D-model process, deep learning techniques that take point cloud as input and generate corresponding 3D model have been studies recently. This paper introduces an efficient scan simulation method. The simulator generates synthetic point cloud data used to train deep learning models for classifying reactor parts in robotic nuclear decommissioning system. The simulator is built by implementing a ray-casting mechanism using a python library called ‘Pycaster’. In order to improve the speed of simulation, multiprocessing is applied. This paper describes the ray casting simulation mechanism and compares the in-house scan simulator with an open source sensor simulation package called Blensor.
Sungmoon Joo,Ikjune Kim 한국방사성폐기물학회 2023 한국방사성폐기물학회 학술논문요약집 Vol.21 No.2
As nuclear decommissioning ventures become increasingly complex, the role of digitalization in facilitating and enhancing these operations is becoming indispensable. This transition to a more digitized approach presents a myriad of advantages, including: augmented avenues for data acquisition, analysis, and visualization to bolster dismantling strategies; simulations in virtual environments for operator training; precise forecasting of future waste emergence, culminating in refined cost estimations; and more immersive decommissioning visualizations for both operators and external stakeholders. Salient benefits conferred by the integration of digital technologies in decommissioning encompass improved collaboration, enriched knowledge transfer, clarity regarding present technological constraints, insights into key influencing factors, clearer criteria for technology selection, and a profound understanding of the potential challenges and merits of a broader incorporation of digital tools in decommissioning endeavors. Of paramount importance is the opportunity presented for superior workforce training and safety measures, exemplified by ALARAbased planning. Amidst the myriad facets of digital adoption, 3D modeling of nuclear facilities derived from laser-scanned point clouds stands out as a pivotal domain in the digitalization. The transformation of intricate point cloud data into a comprehensible 3D mesh remains the crux of this paper. The process of mesh generation, despite being simpler than its counterpart of converting to a 3D solid model, is crucial for multiple reasons. The resultant 3D mesh offers an enhanced visual representation compared to a sparse point cloud, paving the way for improved spatial perception. Furthermore, it serves as a rudimentary tool for approximating component volumes and the ensuing waste, thereby playing an instrumental role in waste manipulation strategies, notably in collision detection. This paper delves deep into the nuances of mesh generation, conducting an parametric study of mesh conversion algorithms, including down-sampling rates. Through this rigorous examination, we endeavor to shed light on optimal methodologies, hoping to catalyze advancements in the digitalization of nuclear decommissioning processes.