http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
플랜트 배관계 미세누출 지능형 감지를 위한 딥러닝 모델 구현
박재순(JaeSoon Park),여도엽(Doyeob Yeo),최유락(Yurak Choi),이종혁(JongHyunk Lee),배지훈(Ji-Hoon Bae) 한국정보기술학회 2021 Proceedings of KIIT Conference Vol.2021 No.11
본 논문에서는 플랜트 배관계의 저전력 센싱 모듈에서 수집한 미세누출에 대한 데이터를 이용하여 딥러닝 기반의 경량화된 누출진단 학습모델을 제안하고자 한다. 초기 건설 시에 설치되었던 플랜트 배관들의 노후화가 진행됨에 따라 배관계의 조기 누출탐지 요구가 증대되고 있지만, 플랜트에서 발생되는 기계잡음과 소음으로 인해 미세누출의 진위 여부를 판별하는 데에 어려움이 있다. 따라서 본 연구에서는 학습 데이터가 작고 기계잡음이 존재하는 상황에서 실제 누출 신호에 대한 이상감지를 수행하기 위해 전이학습 기반의 미세누출 판별 딥러닝 모델을 제안한다. 본 연구의 결과에 따르면 제안모델의 정확도 성능이 기존 신경망 기반의 모델들 보다 더 우수한 미세누출 판별 정확도를 제공할 수 있음을 실험적으로 관찰할 수 있었다. 또한 모델 경량화 작업을 수행한 후 라즈베리파이와 같은 저사양 하드웨어에 탑재하여 정상적인 기능 동작과 빠른 추론 성능도 검증하였다. In this paper, we propose a lightweight leak diagnosis learning model based on deep-learning using low-level leakage data collected from the low-power sensing module of the plant piping system. As the aging of the plant pipes installed during the initial construction progresses, the demand for early leak detection in the piping system is increasing. However, it is difficult to determine the authenticity of low-level leaks due to the mechanical noise and noise generated in the plant. Therefore, in this study, we propose a transer learning-based low-level leak detection deep-learning model to perform anomaly detection on actual leak signals in the presence of a small training dataset and machinery noise. According to the results of this study, it was experimentally observed that the accuracy performance of the proposed model could provide better low-level leak detection accuracy than the existing neural network-based models. In addition, after performing the lightweight model work, it was mounted on low-spec hardware such as Raspberry Pi to verify normal functional operation and fast inference performance.
황재순(Jaesoon Hwang),허형석(Hyungseok Heo),배석정(Sukjung Bae),이동혁(Donghyuk Lee),최해성(Haeseong Choi),김영삼(Youngsam Kim) 한국자동차공학회 2010 한국자동차공학회 학술대회 및 전시회 Vol.2010 No.11
Recently, works on compact heat exchangers have sparked interest amongst heat exchanger manufacturers to develop the compact heat exchangers with enhanced thermal performance in comparison to conventional ones. The purpose of this study is to design more efficient fin than commonly used louvered fin on the basis of same core size. Cooling capacity of louvered fin of conventional radiator is numerically analyzed and the performance is compared with that of the designed fin. Simulations were performed for different louver pitches with 0.9, 1.4 ㎜. In conclusion, cooling capacity of designed fin with 0.9 ㎜ louver pitch is increased more than 7% compared with the conventional fin even though the air side pressure drop increased.
Urea-SCR 시스템의 히팅장치 설계를 위한 빙결된 Urea 수용액의 해동 특성에 관한 연구
황재순(Jaesoon Hwang),허형석(Hyeongseok Heo),이천환(Chunhwan Lee) 한국자동차공학회 2010 한국자동차공학회 학술대회 및 전시회 Vol.2010 No.11
Among the after-treatment systems, urea-SCR system is one of the most efficient techniques to reduce nitrogen oxides emitted from diesel engines. However, urea-water solution is frozen at low ambient temperature level of below -11℃. The purpose of this study is to understand melting phenomena and to compare the melting characteristics between the two geometries - one is coil type and the other is zigzag type - of heater using engine coolant. Performance of the two kinds of heaters is analyzed by numerical study. It was found that melting time of zigzag type heater is shorter than coil type heater.