http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
굴삭기 스윙 기어박스의 강건한 고장 감지를 위한 ARMED 필터의 오더 최적화
나규민(Kyumin Na),김건(Keon Kim) 대한기계학회 2017 대한기계학회 춘추학술대회 Vol.2017 No.11
This paper proposes the order parameter optimization technique of autoregressive minimum entropy deconvolution(ARMED). Traditionally, the best order is selected by Akaike information criterion(AIC) which is same as expected squared prediction error (ESPE). However, in practically, there are many cases that AIC value does not have minimum value and decreases monotonically. This is because residual signal could be disturbed by the other vibration source or noise even appropriate preprocessing. Therefore, AIC’s basic assumption could be violated and it does not work well. This paper proposes new order selection techniques based on health feature related with envelop frequency using Hilbert transform. The simulation and experimental result show that the selected order using these techniques make better performance in signal filtered by ARMED for differentiating normal and fault state than the order obtained by AIC.
음향신호 에너지 감소 메커니즘 기반 확률적 보일러 튜브 누설 위치 추정
나규민(Kyumin Na),김형민(Hyeongmin Kim),이현찬(Hyeonchan Lee),윤병동(Byeng D. Youn) 대한기계학회 2020 대한기계학회 춘추학술대회 Vol.2020 No.12
Estimation of leak location is important considering the labor cost of maintenance procedure and downtime cost in thermal power plant. The well-known approach, TDOA(Time difference of arrival) has lots of problem on practical situation such as limitation of leak signal extraction and numerically not solving issue. To solve these kinds of difficulties, we use probabilistic approach considering energy decaying effect of sound such as attenuation and geometric spreading. In addition, we use bayesian updating method to prevent bias error caused by unpredictable outsource energy variation such as soot-blowing. Finally, We validate our method with simulation data and acoustic emission sensor data of real power plant from installed BTLD(Boiler tube leak detection) system.
굴삭기 기어박스 고장 진단을 위한 고장 정보 강화 전처리 이미지 기반 트랜스퍼 러닝
나규민(Kyumin Na),박정호(Jungho Park),김윤한(Yunhan Kim),윤병동(Byeng D. Youn) 대한기계학회 2019 대한기계학회 춘추학술대회 Vol.2019 No.11
Excavator swing reduction gearbox is major component to rotate whole body. Therefore, they are vulnerable to breakdown by teeth in gearbox because they transmit huge power with high efficiency. For many years ago, there are lots of approach to detect these kinds of fault diagnosis of gearbox using feature extraction method based on fault information. However, most of them shows the result under the lab-level environment, so they have robustness problem when they apply to real industrial field. Also, deep learning based approach also have problems that they do not produce the result under different environmental situation, but they just truncate the signal in same situation. Therefore, they did not considering reassembling problem, so it has high possibility for over-fitting. Therefore, in our research, we conjoin preprocessing approach to eliminate the noise effect due to different environment with transfer learning to solve non-linear hyperplane classification problem. We show that deep learning approach have weakness to over-fitting by comparing how much accuracy under the preprocessing could increase. We also use gradCAM and t-SNE result to briefly explain the result of our algorithm. Further research to make algorithm robust under different operating condition are also discussed.
굴삭기 스윙 기어박스의 고장 진단을 위한 주파수 에너지 불확실성 기반 진동 신호 데이터 증폭기법
나규민(Kyumin Na),하종문(Jong Moon Ha),김건(Keon Kim),윤병동(Byeng D. Youn) 대한기계학회 2018 대한기계학회 춘추학술대회 Vol.2018 No.12
This study was proposed for detecting the fault of swing gearbox in the excavator, and especially fault exists in sun gear component. To prevent the accident, it is necessary to monitor the health condition of a gearbox in the excavator. Generally, the excavator used in the industrial field have lots of noise related with not only sensor system but other component consisting the excavator. Additionally, it is hard to obtain the vibration signal due to the difficulty of installing the sensor system in the excavator in the industrial field. Therefore, the general machine learning algorithm for classifying the state of the gearbox have difficulty in learning the pattern induced by the fault because of the shortage of data. To get over this difficulty, we introduce the data augmentation techniques on frequency domain by considering the uncertainty of physical quantity. The proposed method is validated with the experiment, which shows the accuracy of several machine learning algorithm is higher with using the augmented data.
윈도우 특성 이미지 기반 합성곱 신경망을 활용한 보일러 상태 진단
이현찬(Hyeonchan Lee),김형민(Hyeongmin Kim),나규민(Kyumin Na),윤병동(Byeng D. Youn) 대한기계학회 2020 대한기계학회 춘추학술대회 Vol.2020 No.12
Thermal power plant boiler is one of the main facility that produce steam. Long and thin tubes are inside boiler to generate steam efficiently. Because boiler operate under harsh condition, boiler tube is prone to leakage causing unexpected shutdown of the power plant. To detect leakage, various kind of signals are collected from boiler tube and acoustic signals are the most sensitive signal to tube leakage. However, some event unrelated to leakage cause increase in acoustic signal trend, making it harder to determine the status of boiler. Soot Blowing, cleaning procedure of soot deposited on the internal furnace tubes, is a representative event. In this paper, we propose a novel leakage detection method using Sliding Window Correlation(SWC) Matrix and Sliding Window Energy(SWE) Matrix. Two feature images are trained with Convolutional Neural Network(CNN). Test result from domestic power plant data shows that the proposed method can successfully classify normal, soot blowing and leakage.
위상 정보 활용 시간 영역 평균법을 이용한 산업용 로봇 고장 진단
김윤한(Yunhan Kim),박정호(Jungho Park),나규민(Kyumin Na),윤병동(Byeng D. Youn) 대한기계학회 2018 대한기계학회 춘추학술대회 Vol.2018 No.12
Recently, industrial robots play the crucial roles in the smart factory. However, the unexpected failure of the industrial robots causes a large amount of downtime and economic loss. Therefore, it is required to develop the fault detection method for the industrial robots. In this study, we focus on the mechanical fault from the gearbox, which works as the reducer in the industrial robots. Vibration signals are effectively used for detecting the fault on the gearbox. Vibration signals from the gearbox have strong deterministic signals due to the gear meshing. Therefore, the deterministic signals can be removed for improving the sensitivity of fault detection. Time domain average (TDA) is a method to obtain the deterministic signals from the gearbox. However, the signals from the gearbox in the industrial robots are not usually synchronized and the result of TDA is erroneous. To solve this problem, we propose TDA with phase. The phase is obtained from Fourier transform. The proposed method is demonstrated with the industrial robot.