전방향 신경망에 의한 비선형 동적 시스템의 동정

심귀보,전효병,이상환 중앙대학교 생산공학연구소 1998 생산공학연구소 논문집 Vol.7 No.1

최근에 신경망을 이용한 비선형 동적 시스템의 동정에 관한 연구가 활발히 진행되고 있다. 그러나 일반적인 비선형 동적 시스템의 동정에 있어서는 신경망의 동적 특성을 높이기 위해 과거 데이터를 저장하는 과도한 메모리의 사용이 요구되며, 역전파 알고리즘의 수행에 많은 복잡한 계산이 수행된다. 따라서, 본 논문에서는 단순한 전방향 신경망을 이용한 새로운 비선형 동적 시스템의 동정 방법을 제시한다. 제안된 방법에서는 전방향 신경망의 출력을 현재의 상태변수값에 얼마만큼 증감할 것인가 하는 증감량으로 사용한다. 연결강도의 조정을 위해서 많은 국소해가 존재하는 문제에서 보다 쉽고 간단하게 최적해를 구할 수 있는 진화전략을 사용한다. 제안된 방법의 유효성은 2축 로봇 메니퓰레이터의 시뮬레이션을 통해 검증된다. Recently, the identification of nonlinear dynamic systems using neural networks has been researched extensively. For the identification of nonlinear dynamic systems, generally, excessive memories storing past data set are required to give dynamic property to neural networks and lots of complex calculations are executed for the backpropagation algorithm. So, we propose a new approach to identify nonlinear dynamic systems using simple feedforward neural networks. We use the outputs of feedforward neural networks as the value of what is added to or subtracted from the current states. In order to adjust the weights of neurons, we use evolution strategies, which makes it easier and simpler to solve an optimal problem having many local minima. Simulation results using two-link robot manipulator data reveal the performance and efficiency of this proposed approach.

韓國産 Trichoderma 에 關한 分類學的 硏究

沈聖輔 건국대학교 1979 論文集 Vol.10 No.1

31 strains of Trichoderma sap. have been obtained either by bringing into culture colonies growing naturally on wood, mushroom and compost or by isolating them from soil samples taken from various localities. This study is based largely on morphological and cultural characters of Trichoderma grown on 2% malt extract agar, which throughout this work has been used as standard medium. Repeated examination of each isolates was made both macroscopically on the plate culture and microscopically on the slide culture. All isolates of this studies identified using the group key of Rifai. The results of these studies are as follows; 1.One strain isolated from mushroom was identified as T. viride. 2.One strain isolated from compost was identified as T. harzianum. 3.One strain isolated from wood was identified as T. viride. 4.27 strains isolated from soil samples were identified as T. piluliferum (4 strains), T. hamatum(3 strains), T. polysporum(3 strains), T. longibrachiatum (2 stranis), T. viride (2 strains), T. aureoviride (4 strains), T. koningii (4 strains), T. puseudokoningii (1 strain), T. harzianum (4 strains).

정밀금형의 엔드밀 가공시 최적가공조건의 결정에 관한 연구

심성보,윤문철,이종두 釜山工業大學 1991 論文集 Vol.33 No.-

In this study, We determine the optimum cutting condition of end milling operation for precision die in machining center. To obtain more complete solution of optimum design vector in end milling operation, several constraints of the actual endmill-operation are considered. As a constraints considered, surface roughness, cutting force in feeding direction, machining power and tool life are considered. As a searching method for absolute optimum, nonlinear mathematical programming(NLP:hypersphere inscribed method) is used. also to obtain required model of constraints and to simplify experiments, central composite design is used.

동적 귀환 신경망에 의한 비선형 시스템의 동정

심귀보,이상환,전효병 중앙대학교 생산공학연구소 1997 생산공학연구소 논문집 Vol.6 No.2

최근들어 비선형 동적 시스템의 동정을 위한 동적 귀환 신경망에 대한 연구가 활발히 이루어 지고 있다. 일반적으로, 신경망의 연결강도를 학습시키기 위해서는 역전파 학습규칙이 사용되고 있다. 그러나, 이 방법은 많은 복잡한 계산과정을 필요로 하게 되고, 지역 최소값에 빠질 가능성이 매우 높다. 따라서 본 논문에서는 동적 귀환 신경망을 이용하여 비선형 동적 시스템을 동정하기 위한 새로운 방법을 제안한다. 뉴런의 연결강도를 학습시키기 위해서 본 논문에서는 진화전략을 사용한다. 주 연산자로서 돌연변이를 갖는 진화전략에 의해 학습된 동적 귀환 신경망은 플랜트 에뮬레이터로서 역할할 수 있게 된다. 또한 진화전략의 적합도 평가 함수는 플랜트의 출력과 신경망의 출력간의 오차에 기반을 둔다. 비선형 동적 시스템을 동정하는 새로운 방법은 2축 로봇 메니퓰레이터의 시뮬레이션에 적용하여 그 가능성과 유효성을 검토한다. Recently, dynamic recurrent neural networks(DRNN) for identification of nonlinear dynamic systems have been researched extensively. In general, dynamic backpropagation was used to adjust the weights of neural networks. But, this method requires many complex calculations and has the possibility of falling into a local minimum. So, we propose a new approach to identify nonlinear dynamic systems using DRNN. In order to adjust the weights of nonlinear dynamic systems using DRNN. In order to adjust the weights of neurons, we use evolution strategies, which is a method used to solve and optimal problem having many local minimums. DRNN trained by evolution strategies with mutation as the main operator can act as a plant emulator. And the fitness function of evolution strategies is based on the difference of the plant's output and DRNN's outputs. Thus, this new approach at identifying nonlinear dynamic system, when applied to the simulation of a tow-link robot manipulator, demonstrates the performance and efficiency of this proposed approach.

摩擦壓接된 SUH 3 - SUS 303鋼의 크리이프 强度 및 Fractography에 대한 硏究

沈成輔 釜山工業大學校 1985 論文集 Vol.27 No.-

The effects of aging treatment on the creep strength and fractography of domestic SUH3-SUS303 friction welded steel were studied on and the relation ship between strength and structure in the SUH3-SUS303 friction welded steel was examined. The creep test specimens were 0hr, 10hr, 100hr aging treated ones(at 700℃) after the solution treatment of Keeping 1060℃ for 1hr in siliconit muffle furnace and then quenching in water of constant temperature 14℃ Main result obtained were as follows: 1) In the case of 10hr aging specimens, the creep strength was highest a mong all. 2) According to each aging time, the equations between stresses and creep rate were determined as follows. 0=1.4163×10??(σ)?? 10=4.2304×10??(σ)?? 100=2.2033×10??(σ)?? 3) the creep fracture appearance consists of equi-axed dimple and elongatad dimple type rupture. The types of fracture observed by scanning electronic microscope were grain boundary ductile fracture.

STS304鋼에 對한 STD11鋼의 摩擦 磨滅 特性에 關한 硏究

沈成輔 釜山工業大學校 1990 論文集 Vol.32 No.-

In Order to increase to tool lifetime, this experiment was performed on lubricated frition between the test specimen of STD11 hardened tool steel and revolving disc of STS304 stainless steel. The following conclusions were obtained. 1) Situation of highest wear recognized in the final load 18.9kg and sliding friction speed 1,989m/sec. 2) Situation of lowest wear recognized in final load 3.2kg and sliding friction speed 0.065m/sec.

金型用鋼의 摩擦 磨耗에 關한 硏究(Ⅱ)

沈成輔 釜山工業大學校 1990 論文集 Vol.32 No.-

In order to increase to tool life time, this experiment was Performed on lubricated friction between the test specimen of STDII hardened tool steel and revolving disc of SM45C steel. The following conclusions were obtained 1) The situation of highest wear was recognized in the final load 3.2kg and sliding friction speed 2.880m/sec. 2) The situation of lowest wear was recognized in the final load 3.2kg and sliding friction speed 0.25m/sec.

두경부 유리피판 26례 분석

정복성,윤근철,심경보,고경석,한상훈 大韓成形外科學會 1993 Archives of Plastic Surgery Vol.20 No.3

Twenty four patients(21 males and 3 females) aged 14 to 70 years underwent 26 microvascular free tissue transfers in head and neck reconstructions from April 1990 to July 1992 (27 months) . Nineteen cases were tumors in head and neck area, and four were acquired deformities such as burn alopecia, burn scar contracture of the neck, Romberg disease and frontal sinus infection, and one was hemifaclal microsomia. The free flaps were executed successfully in 15 upper aerodigestive tracts, in 2 scalps and 3 skull bases and in contouring and resurfacing six facial and cervical lesions. The selection of donor site was bases upon the location of defect and the volume of tissue required : the forearm free flap was used in 11 cases, the rectus abdominis in 5, the groin in 2, the jejunum in 2, the occipital scalp in one, the first toe web space in one, the latissimus dorsi in one, the serratus anterior in one, the parietotemporalis fascia in one and the omentum in one. The free tissue transfers were primarily indicated in all patients in our series and there was no flap failure. The complications were minimal and the morbidities low. A free flap in the head and neck area is not only the most expedient method of reconstruction but also either the most reliable or the only available one in some clinical situations.

인터넷을 이용한 멀티미디어 원격 교육 시스템 개발

선상준,심귀보 중앙대학교정보통신연구소 2000 정보통신연구소논문집 Vol.2 No.1

인터넷은 전세계를 동일한 프로토콜일 TCP/IP을 이용하여 연결하고 있다. 따라서 인터넷을 이용하면 어느 곳에서나 고품질의 교육을 받을 수 있다. 본 논문에서는 이러한 인터넷의 광범위한 연결성을 이용해 고품질의 교육이 가능한 멀티미디어 원격 교육 시스템을 구현하였다. 구현한 시스템은 크게 두 부분으로 구성되었다. 평면에서 자유로운 이동이 가능한 로봇을 이용한 카메라를 통해 멀티미디어 정보를 얻으며 이 정보를 TCP/IP를 이용하여 전송하는 서버 시스템과 멀티미디어 정보를 사용자에게 제공하며 이동 가능한 카메라 시스템의 조작 인터페이스를 제공하는 클라이언트 시스템으로 나누어 구현하였다. 이동 가능한 카메라 시스템으로 구현하고 교육 공간의 활용을 높이며 능동적인 교육효과를 얻을 수 있다. 또한 연결방법에 있어서 소켓을 이용하여 서버 시스템에 직접 연결하는 방법과 웹 서버를 통해 웹 브라우저를 서버 시스템에 연결하는 방법의 두 가지로 구현하였다.

컴퓨터 면역시스템 개발을 위한 생체 면역시스템 모델링

선상준,심재윤,심귀보 중앙대학교 정보통신연구소 2000 정보통신연구소논문집 Vol.2 No.2

본 논문은 생체 면역시스템의 자기와 비자기를 구분하여 인식하는 세포독성 T세포의 특징을 이용하여 컴퓨터에서 자기 파일과 비자기 파일을 구분할 수 있는 T세포를 모델링하였다. 모델링한 T세포는 MHC 인식부와 향원 인식부로 구성되며 각각의 인식부를 이용하여 비자기 파일에 적용시켜 자기-비자기 구분이 가능함을 컴퓨터 시뮬레이션을 통하여 입증하였다. Biological immune system distinguishes nonself from self-body and protects self intrusion such as virus, pathogenic bacteria. In this paper, we make a model of the T-cell of immune system to apply to computer immune system. The modeled T-cell possesses the ability to be able to distinguishes self files from random files in computer. The T-cell consists of MHC perception and antigen perception, and distinguishes self from non-self through matching file which we must test. We verify the effectiveness of the proposed method by applying self-nonself discrimination in a computer virus.