http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
동적 응답 기반 합성곱 신경망 모델의 구름 베어링 스폴 결함 분류 및 물리적 궤적 특성 시각화
황미경(Mikyung Hwang),최민석(Minseok Choi),오현석(Hyunseok Oh) 대한기계학회 2021 대한기계학회 춘추학술대회 Vol.2021 No.11
In this paper, the spall defect caused by fatigue, which is the main cause of failure, was applied as a direct feature for rolling element bearing condition monitoring. The various signal processing methods are used for the spall size estimation. However, in order to extract the physical trajectory of the spall defect, a filtering method for removing a background noise and interference is required, which required expert-based theory and domain knowledge. Therefore, to overcome this limitation, convolutional neural network structure was applied. A physics-based artificial intelligence model that reflects the text data of load and rotational speed is proposed in the deep learning model. Classification performance was verified for five conditions with different entry-exit angle of the defect bearing. In addition, we provided an explanation for the existing classical method by class activation map to visualize a specific frequency band and the time at which rolling element passes through the defect.
증기터빈 로터 중심공 묻힘 결함에 대한 해석적 수명 평가 기준 제시
황미경(Mikyung Hwang),조기현(Kihyun Cho) 대한기계학회 2021 대한기계학회 춘추학술대회 Vol.2021 No.11
The steam turbine rotor receives a radial temperature gradient when starting up and shutting down of the turbine, and fatigue and creep damage occur depending on the operation condition. Reliable steam turbine operation is necessary for stable power supply and as a basis for this, power generation companies are required to perform the life assessment for structural integrity in Article 32 of ELECTRIC UTILITY ACT. However, it is difficult to apply a standardized guideline because this criterion for the remaining life evaluation are based on the contract partners own standards. In this study, in order to provide the reliable life assessment method of steam turbines, heat transfer, and stress analysis and fracture mechanics based life assessment is performed on the high-pressure rotor bore. The results of the evaluation were verified by comparing SAFER-PC, Rotor Life Assessment Program, developed by Electric Power Research Institute (EPRI).
회전기계 상태감시 센서위치 변동에 강건한 특성인자 엔지니어링 기법 개발
최민석(Minseok Choi),황미경(Mikyung Hwang),오현석(Hyunseok Oh) 대한기계학회 2022 대한기계학회 춘추학술대회 Vol.2022 No.11
Recently, deep learning is widely employed for the fault diagnosis of rotating machines. Nonetheless, there are two challenges for implementing deep-learning-based fault diagnosis methods in real applications. First, in the field, it is challenging to collect sensory data from faulty rotating machines. Second, it is difficult to attach sensors on the identical location between different rotating machines. To overcome the challenges, this study proposes a new deep learning model that combines the conditional domain adversarial networks (CDAN) architecture with the Mixstyle module, namely, Mix CDAN. Multiple case studies are conducted. The performance of the proposed Mix CDAN is evaluated with existing deep learning models. It is confirmed that the proposed model outperforms the existing models.