http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
윤일채(I.C Yoon),박춘달(C.D Park),고태조(T.J Ko),윤인준(I.J Yoon) 한국생산제조학회 2011 한국생산제조시스템학회 학술발표대회 논문집 Vol.2011 No.4
Ultra precision rolling technology is developed to reduce machining cost and delivery in ball screw manufacturing. Process parameter of ultra precision rolling can be optimized by computer simulation(CAE). In this paper, the simulation parameters are determined to get accurate computer analysis. Simulation and prototype results are evaluated. The residual stress of calculation shows good agreement with measurement result.
Cutting Force of Ultrasonic Vibration-assisted Milling for Titanium Alloys
T. J. Ko(고태조),G. C. Park(박건철),R. Kurniawan(쿠르니아완 렌디),M. K. Choo(추민기),M. R. Xu(쉬모란),P. W. Han(한필완),I. J. Yoon(윤인준),Y. I. Kwak(곽예인) Korean Society for Precision Engineering 2021 한국정밀공학회 학술발표대회 논문집 Vol.2021 No.11월
Titanium alloys have great characteristics such as strong corrosion resistance, high strength-weight ratio and heat resistance. This material has been used in aviation and aerospace industries due to their superior properties. However, titanium alloys are a representative difficult-to-cut material. Ultrasonic vibration-assisted milling (UVAM) is a more effective machining method with noteworthy advantage such as prolongs tool life, low cutting force, good surface quality for difficult-to-cut material compared to the conventional milling (CM) method. In this paper, the cutting force on the titanium alloy was investigated by applying both the UVAM and CM method with the spindle rotational speed variation. As a result, the UVAM showed that the cutting force was lower than that of the CM, especially in the higher spindle rotational speed. The UVAM cutting force was decreased maximally for all cutting force components (Tangential : -11.2, Radial : -10.6, Axial : -9.1%) at 8,000 rpm spindle rotational speed.
T. J. Ko(고태조),R. Kurniawan(쿠르니아완 렌디),G. C. Park(박건철),M. R. Xu(쉬모란),P. W. Han(한필완),I. J. Yoon(윤인준) Korean Society for Precision Engineering 2021 한국정밀공학회 학술발표대회 논문집 Vol.2021 No.11월
This manuscript studies about preliminary numerical study in the 2D-UVAC (Two-dimensional Ultrasonic Vibration Assisted Cutting) mixed with the EDM (Electrical Discharged Machining) method. The commercial finite element analysis of ABAQUS 6.4 solver has been used in this research problem. The 2D-UVAC has been carried out with variation vibration frequency of 40-60 kHz, vibration amplitude for both directions about 10 μm, the constant cutting speed of 2 m/s, and constant cutting depth of 100 μm. Meanwhile, the EDM was assumed as a single Gaussian distribution heat flux with different pulsated frequency about 100 to 300 kHz with constant discharged voltage of 220 V, and discharged current of 1 A. According to numerical solution, the EDM maximum temperature can achieve about 1,200°C. By increasing pulsated frequency increases average EDM temperature thus it could decrease the average shear stress during cutting in 2D-UVAC. Therefore, the EDM brings benefit such as decreasing von-Mises stress during cutting in 2D-UVAC.
희소 데이터를 위한 강인 손실 함수를 이용한 준 지도 학습
안영준(Youngjun Ahn),심규석(Kyuseok Shim) 한국정보과학회 2021 정보과학회논문지 Vol.48 No.12
이 논문에서는 데이터의 레이블이 매우 부족한 상황에서 데이터 증강기법과 강인 손실 함수를 사용하여 준 지도 학습을 하는 방법을 제안한다. 기존 데이터 증강기법을 사용하는 준 지도 학습 방법은 레이블이 없는 데이터를 증강하고, 그 중 신뢰도가 높은 데이터에 대해서만 현재 모델이 예측한 레이블을 원 핫 벡터로 붙여 학습에 사용한다. 그래서 신뢰도가 낮은 데이터는 사용하지 않는 문제가 있었는데, 이를 해결하기 위해 강인 손실 함수를 이용하여 신뢰도가 낮은 데이터 또한 사용하는 연구도 진행되었다. 한편, 레이블이 매우 적은 상황에서는 모델이 예측한 레이블은 신뢰도가 높더라도 부정확하다는 문제가 있다. 이 논문에서는 레이블이 매우 적은 상황에서 원 핫 벡터가 아닌 모델이 예측한 확률을 레이블로 사용함으로써 분류 모델의 성능을 높일 수 있는 방법을 제시한다. 또한 이미지 분류 문제에 대한 실험을 통하여 제시된 방법이 분류 모델의 성능을 향상시킴을 보여준다. This paper proposes a semi-supervised learning method which uses data augmentation and robust loss function when labeled data are extremely sparse. Existing semi-supervised learning methods augment unlabeled data and use one-hot vector labels predicted by the current model if the confidence of the prediction is high. Since it does not use low-confidence data, a recent work has used low-confidence data in the training by utilizing robust loss function. Meanwhile, if labeled data are extremely sparse, the prediction can be incorrect even if the confidence is high. In this paper, we propose a method to improve the performance of a classification model when labeled data are extremely sparse by using predicted probability, instead of one hot vector as the label. Experiments show that the proposed method improves the performance of a classification model.