금형압축성형공정 해석용 CAE 프로그램 개발 및 적용

정석환,권영삼,Chung Suk-Hwan,Kwon Young-Sam 한국분말야금학회 2004 한국분말재료학회지 (KPMI) Vol.11 No.5

Crack generation during die compaction and distortion during sintering have been critical problems for the conventional pressing and sintering process. Until now, trial and error approach with engineers' industrial experiences has been only solution to protect the crack generation and distortion. However, with complexity in shape and process it is very difficult to design process conditions without CAE analysis. We developed the exclusive CAE software (PMsolver/Compaction) for die compaction process. The accuracy of PMsolver is verified by comparing the finite element simulation results with experimental results. The simplified procedures to find material properties are proposed and verified with iron based powder and tungsten carbide powder. Based on the accurate simulation result by PMsolver, the optimal process conditions are designed to get uniform density distribution in a powder compact after die compaction process by using a derivative based optimization scheme. In addition, the effect of non-uniform density distribution in a powder compact on distortion during sintering is shown in case of the fabrication of tungsten carbide insert.

전계 제어형 정전하 중화장치

鄭石煥(Suk-Hwan Chung),李大熙(Dae-Hee Lee),文在德(Jae-Duk Moon) 대한전기학회 1999 전기학회논문지C Vol.48 No.6

설계민감도를 이용한 비정상상태 소성가공공정 최적 설계

정석환(Suk-Hwan Chung),황상무(Sang-Moo Hwang) 한국소성가공학회 1997 한국소성가공학회 학술대회 논문집 Vol.1997 No.5

CRNN 기반의 소리 이벤트 검출 성능 분석

정석환(Suk-Hwan Chung),정용주(Yong-Joo Chung) 한국정보기술학회 2019 한국정보기술학회논문지 Vol.17 No.5

In this paper, we performed an intensive investigation on the CRNN for audio event detection. For this purpose, we tried to find out the optimal learning rate and input segment length for the training via performance experiments. In addition, we investigated the effects of variation in each iteration of the batch data on the performance and the effects of batch normalization and dropout were also observed. Finally, we figured out the performance variation depending on the optimization criterion for the convergence of training process. Through the aforementioned process, we tried to find out the optimal conditions for the CRNN. From the experimental results using the TUT Sound Events Synthetic 2016 database, we could obtain the best peformance with the learning rate 10<SUP>-4</SUP> and batch normalization was far more effective on performances and great performance improvement was obtained by varying the starting point of the batch data.

딥러닝 기반 자동작곡에서 구성을 갖춘 곡 생성방법

정석환(Suk-Hwan Chung),정성훈(Sung-Hoon Jung) 한국디지털콘텐츠학회 2021 한국디지털콘텐츠학회논문지 Vol.22 No.6

Although many deep learning based automatic music composition systems have been studied, it is hard to find a song generation system with the structure of songs. This paper addresses a method for generating songs with the structure of songs in deep learning-based automatic music generation by learning the structure together. We use a deep learning model that learns dynamic information which changes with the progress of the song and static information which is given according to the structure of the song together to learn the structure of the song. When a song is learned using the proposed method, it is possible to create a more natural song by using the composition information of the song when creating the song. We use METEOR and BLEU, which are used in natural language processing, to evaluate how similar the generated song is to a composers. The results showed that they were more similar to the song created by a composer when a song was generated with the proper structure of songs.

SVM과 GMM 기반의 비명인식 기법의 성능비교

정석환(Suk-Hwan Chung),정용주(Yong-Joo Chung) 한국정보기술학회 2017 한국정보기술학회논문지 Vol.15 No.3

In recent days, one of the issuing social problems is that violent crimes are frequent in places like public toilets and streets. Due to this problem, the importance of surveillance systems for the safety of pedestrians is increasing gradually. However, since the conventional visual surveillance systems have some limitations, many attempts have been tried to support the system by adding audio-based functionality. In this study, we try to detect scream sounds based on SVM and GMM, respectively, and compare the performance of the two methods through various experiments. From the experimental results, SVM could obtain 0.559% False Positive rate, which menas that there is a very low possibility of incorrectly deciding non-scream sound as a scream sound. In contrary, the GMM method could achieve 12.03% of False Negative rate, which implies that GMM has relatively good sensitivity to the scream sound compared with SVM. From these results, we could conclude that both GMM and SVM have, respectively, a distinctive merit from each other and the plausibility of further performance improvement by combining the two approaches is also observed.

증례 : 순환기 ; 경요골 동맥 관상동맥 조영술 시 생긴 이상성 우측 빗장밑 동맥의 합병증 1예

정석환 ( Suk Hwan Chung ),금동성 ( Dong Sung Kum ) 대한내과학회 2011 대한내과학회지 Vol.80 No.2S

우측 요골 동맥을 이용한 관상동맥 조영술은 환자의 편이성 등 여러 가지 장점 때문에 많이 시행되는 술식이다. 그러나 여러 장점에도 불구하고 우측 요골 동맥을 이용한 관상동맥 조영술에 따르는 여러 가지 합병증 역시 있을 수 있다. 이상성 우측 빗장밑 동맥을 가진 환자의 경우 이로 인한 시술상의 어려움, 시술의 실패 등은 잘 알려져 있으나, 시술 후에 생긴 우측 빗장밑 동맥 박리는 많이 알려져 있지 않다. 본 증례는 흔하지 않은 합병증을 경험했기에 이에 대한 증례 보고를 하는 바이다. Right transradial coronary catheterization is very popular and performed by many clinicians with few complications. An aberrant right subclavian artery (arteria lusoria) arising from the descending thoracic aorta is an uncommon congenital variant. Here, we present the case of a patient with an aberrant right subclavian artery dissection after right transradial coronary catheterization, which is a rare complication. We diagnosed the complication by chest CT, which showed the dissection of the aberrant right subclavian artery. Dissection of an aberrant right subclavian artery is a rare complication of right transradial coronary angiograph. (Korean J Med 2011;80:S157-S160)

SVM 기반의 오디오 이벤트 검출 성능 분석

정석환(Suk-Hwan Chung),정용주(Yong-Joo Chung) 한국정보기술학회 2018 한국정보기술학회논문지 Vol.16 No.6

In this paper, we performed deep investigation on the SVM method for audio event detection. We observed performance variation by varying the dimension and form of the input features of the SVM. In addition, the effects on the recognition performance of the parameters of the kernel function and the slack variables were observed, through which we tried to find the optimum conditions of the SVM classifiers for the audio event detection. From the classification results using the audio data in Task 3 of DCASE Challenge 2016, we found that the best performance is obtained when MFCC features are input to the SVM after averaging for 80ms, the MFCC feature dimension 13 and the parameters of the slack variables and kernel function are C=1, γ=0.00769, respectively. Through these optimization, we could find that SVM showed significant performance improvement in F-score compared with the conventional GMM.

Lyotard의 지식관과 수행적 지식관 비판

정석환(Chung Suk-Hwan) 한국교육철학회 2003 교육철학 Vol.24 No.-

The purpose of this thesis is to examine Lyotard's Postmodern theory of knowledge and its meanings in education. Specifically, the examination focused on The Critic of Performative Knowledge base on Lyotard's Postmodern theory of knowledge Conclusions are as follows. The result of the decline of meta narratives that have functioned as bases for the legitimation of knowledge, Lyotard argues, will step into the dominance of the performativity criterion. The power of technology dominating contemporary society has degraded reason into instrumental one and produced performativity standard as a meta narrative. Knowledge under performativity principle circulates not in its educational values and political importance, but in the same way as currency does, so the criteria judging the value of knowledge can be found in 'outgo knowledge' and 'investment knowledge, not in 'knowing' or 'not knowing'. For example, in the field of education the induction of training and technology according with the system's performance becomes more valuable. But the ascendancy of performativity underestimates a person's freedom and value independent of system. Lyotard criticizes every metanarratives, like system theory operating under this performativity principle in that it entails a kind of terror to make the incommensurable commensurable. To fight successfully against the terror of metanarratives Lyotard introduces a form of legitimation grounded on paralogy. Paralogy is a sphere that philosopher embodies the logic of dissent, not one that expert pursues consensus. In other words, it encourages new moves and possibilities of other purposes within one system. Also, not excluding other voices, it accepts the possibility of various utterances and strengthens patience to tolerate difference. The delegitimation by the logic of paralogy is necessarily tied to the possibility of knowledge based on difference. Lyotard's theory of knowledge is grounded upon a tolerance to difference. Here difference is justified by different modes of utterences existing among rules inherent modes of discourses. Thus, Lyotard's postmodern theory of knowledge presents knowledge of difference and aims at showing the differences, and suggests a man accepting them as they are with tolerance.

CNN과 CRNN을 이용한 오디오 이벤트 검출 성능 비교

정석환(Suk-Hwan Chung ),정용주(Yong-Joo Chung) 한국정보기술학회 2018 Proceedings of KIIT Conference Vol.2018 No.6