• 대한전기학회:학술대회논문집
• /
• 대한전기학회 2006년도 심포지엄 논문집 정보 및 제어부문
• /
• pp.102-104
• /
• 2006

Auto-Correlated time series (ATS) model was constructed by using the backpropagation neural network. The performance of ATS model was evaluated with sensor information collected from a large volume, industrial plasma-enhanced chemical vapor deposition system. A total of 18 sensor information were collected. The effect of inclusion of past and future information were examined. For all but three sensor information with a large data variance demonstrated a prediction error less than 4%. By integrating ATS model into equipment software, process quality can be more stringently monitored while improving device throughput.

• 한국전산구조공학회:학술대회논문집
• /
• 한국전산구조공학회 1998년도 봄 학술발표회 논문집
• /
• pp.216-225
• /
• 1998

Although artificial neural network based on backpropagation algorithm is an excellent system simulator, it has still unsolved problems of its structure-decision and learning method. That is, we cannot find a general approach to decide the structure of the neural network and cannot train it satisfactorily because of the local optimum point which it frequently falls into. In addition, although there are many successful applications using backpropagation learning algorithm, there are few efforts to improve the learning algorithm itself. In this study, we suggest a general way to construct the hidden layer of the neural network using binary genetic algorithm and also propose the various learning methods by which the global minimum value of the teaming error can be obtained. A XOR problem and line heating problems are investigated as examples.

• Journal of Advanced Marine Engineering and Technology
• /
• 제20권4호
• /
• pp.11-19
• /
• 1996

Recently, fuzzy and neural network techniques have been successfully applied to control of complex and ill-defined system in a wide variety of areas, such as robot, water purification, automatic train operation system and automatic container crane operation system, etc. In this paper, we present a neuro-fuzzy controller which unifies both fuzzy logic and multi-layered feedforward neural networks. Fuzzy logic provides a means for converting linguistic control knowledge into control actions. On the other hand, feedforward neural networks provide salient features, such as learning and parallelism. In the proposed neuro-fuzzy controller, the parameters of membership functions in the antecedent part of fuzzy inference rules are identified by using the error backpropagation algorithm as a learning rule, while the coefficients of the linear combination of input variables in the consequent part are determined by using the least square estimation method. Finally, the effectiveness of the proposed controller is verified through computer simulation of an inverted pendulum system.

• 한국정보과학회논문지:컴퓨팅의 실제 및 레터
• /
• 제8권3호
• /
• pp.344-355
• /
• 2002

MLP(multilayer perceptron)는 다른 패턴인식 방법에 비해 여러 가지 훌륭한 특성을 가지고 있어 패턴인식에서 폭넓게 사용되고 있다. 그러나 MLP의 학습에 일반적으로 사용되는 EBP(error backpropagation) 알고리즘은 학습시간이 비교적 오래 걸린다는 단점이 있다. 패턴인식에 사용되는 학습 데이타는 풍부한 중복특성을 내포하고 있으므로 패턴마다 MLP의 내부변수를 갱신하는 온라인 계열의 학습방식이 속도의 향상에 상당한 효과가 있다. 일반적인 온라인 EBP 알고리즘에서는 내부변수 갱신시 고정된 학습률을 적용한다. 고정 학습률을 적절히 선택함으로써 패턴인식 웅용에서 상당한 속도개선을 얻을 수 있지만, 학습률이 고정되고 학습이 진행됨에 따라 학습패턴 영역이 달라지는 학습과정의 각 단계에 효과적으로 대웅하지 못하는 문제가 있다. 이 문제에 대해 본 논문에서는 학습과정을 세 단계로 정의하고, 각 단계별로 필요한 패턴만을 학습에 반영하는 패턴별 가변학습속도 및 학습생략(ILVRS) 방법을 제안한다. ILVRS의 기본개념은 다음과 같다. 학습단계마다 학습에 필요한 패턴의 부분이 달라지므로 이를 구별 하여 학습에 적용할 수 있도록 (1)패턴마다 발생하는 오류치를 적절한 범위 이내로 제한하여 가변 학습률로 사용하고, (2)학습이 진행됨에 따라 불필요한 부분의 패턴을 학습에서 생략한다. 제안한 ILVRS의 성능을 입증하기 위해 본 논문에서는 패턴인식 응용의 한 갈래인 화자증명을 실험하고 그 결과를 제시한다.

• ETRI Journal
• /
• 제22권3호
• /
• pp.10-18
• /
• 2000

Although an n-th order cross-entropy (nCE) error function resolves the incorrect saturation problem of conventional error backpropagation (EBP) algorithm, performance of multilayer perceptrons (MLPs) trained using the nCE function depends heavily on the order of nCE. In this paper, we propose an adaptive learning rate to markedly reduce the sensitivity of MLP performance to the order of nCE. Additionally, we propose to limit error signal values at out-put nodes for stable learning with the adaptive learning rate. Through simulations of handwritten digit recognition and isolated-word recognition tasks, it was verified that the proposed method successfully reduced the performance dependency of MLPs on the nCE order while maintaining advantages of the nCE function.

• 대한기계학회논문집A
• /
• 제26권12호
• /
• pp.2475-2482
• /
• 2002

The measurement of residual stresses by the hole-drilling method has been commonly used to evaluate residual stresses in structural members. In this method, eccentricity can usually occur between the hole center and rosette gage center. In this study, the error due to the hole eccentricity is compensated using the neural network. The neural network has trained training examples of normalized eccentricity, eccentric direction and direction of maximum stress at eccentric case using backpropagation learning process. The trained neural network could compensated the error of measured residual stress in experiments with hole eccentricity. The proposed neural network is very useful for compensation of the error due to hole eccentricity in hole-drilling method.

• 대한기계학회:학술대회논문집
• /
• 대한기계학회 2001년도 추계학술대회논문집A
• /
• pp.412-418
• /
• 2001

The measurement of residual stresses by the hole-drilling method has been commonly used to evaluate residual stresses in structural members. In this method, eccentricity can usually occur between the hole center and rosette gage center. In this study, the error due to the hole eccentricity is corrected using the neural network. The neural network has trained training examples of normalized eccentricity, eccentric direction and direction of maximum stress at eccentric case using backpropagation learning process. The trained neural network could corrected the error of measured residual stress in experiments with hole eccentricity. The proposed neural network is very useful for correction of the error due to hole eccentricity in hole-drilling method.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 1990년도 한국자동제어학술회의논문집(국내학술편); KOEX, Seoul; 26-27 Oct. 1990
• /
• pp.30-35
• /
• 1990

Learning control of a robot manipulator is proposed using the backpropagation neural network. The learning controller is composed of both a linear feedback controller and a neural network-based feedforward controller. The stability analysis of the learning controller is presented. Three energy functions are selected in teaching the neural network controller : 1/2.SIGMA.vertical bar torque error vertical bar $^{2}$, 1/2.SIGMA..alpha. vertical bar position error vertical bar $^{2}$ + .betha. vertical bar velocity error vertical bar $^{2}$ + .gamma. vertical bar acceleration error vertical bar $^{2}$ and learning methods are presented. Simulation results show that the learning controller which is learned to minimize the third energy function performs better than the others in tracking problems. Some properties of the learning controller are discussed with simulation results.

• 대한기계학회:학술대회논문집
• /
• 대한기계학회 2001년도 춘계학술대회논문집A
• /
• pp.956-963
• /
• 2001

The measurement of residual stresses by the hole-drilling method has been commonly used to evaluate residual stresses in structural members. In this method, eccentricity can usually occur between the hole center and rosette gage center. In this study, the error due to the hole eccentricity is predicted using the artificial neural network. The neural network has trained training examples of stress ratio, normalized eccentricity, off-centered direction and stress error using backpropagation loaming process. The prediction results of the error using the trained neural network are good agreement with FE analyzed ones.

• 한국음향학회지
• /
• 제20권4호
• /
• pp.48-53
• /
• 2001

본 논문에서는 에러 역전파 알고리듬에 기반한 다층 퍼셉트론의 학습 속도를 개선하기 위해 선택적 주의 학습방식을 제안한다. 제안된 방식은 학습 과정에서 세 가지 선택적 주의 기준을 적용하여 학습 데이터베이스 내의 일부 데이터만을 입력 패턴으로 사용하거나 주어진 입력 패턴에 대해 신경회로망내의 특정 영역만 선택적으로 학습이 이루어지도록 한다. 이러한 선택적 주의 기준은 다층 퍼셉트론의 출력층에서 계산된 평균 자승 에러와 은닉층의 각 노드에서 획득된 클래스 의존적인 적합도(relevance)를 이용하여 설정된다. 학습 속도의 개선은 학습 반복 횟수 당 계산량을 줄임으로써 이루어진다. 본 논문에서는 고립 단어 인식시스템에서의 화자 적응 문제에 대해 제안한 선택적 주의 학습방법을 적용하여 그 유효성을 알아보았다. 실험 결과로부터 제안한 선택적 주의 기법이 학습 속도를 평균 60%이상 개선시킬 수 있음을 확인하였다