• 한국산학기술학회논문지
• /
• 제21권2호
• /
• pp.189-194
• /
• 2020

본 논문에서는 신경망을 학습하는 데 가장 많이 사용되고 있는 경사하강법에 대해 분석하였다. 학습이란 손실함수가 최소값이 되도록 매개변수를 갱신하는 것이다. 손실함수는 실제값과 예측값의 차이를 수치화 해주는 함수이다. 경사하강법은 오차가 최소화되도록 매개변수를 갱신하는데 손실함수의 기울기를 사용하는 것으로 현재 최고의 딥러닝 학습알고리즘을 제공하는 라이브러리에서 사용되고 있다. 그러나 이 알고리즘들은 블랙박스형태로 제공되고 있어서 다양한 경사하강법들의 장단점을 파악하는 것이 쉽지 않다. 경사하강법에서 현재 대표적으로 사용되고 있는 확률적 경사하강법(Stochastic Gradient Descent method), 모멘텀법(Momentum method), AdaGrad법 그리고 Adadelta법의 특성에 대하여 분석하였다. 실험 데이터는 신경망을 검증하는 데 널리 사용되는 MNIST 데이터 셋을 사용하였다. 은닉층은 2개의 층으로 첫 번째 층은 500개 그리고 두 번째 층은 300개의 뉴런으로 구성하였다. 출력 층의 활성화함수는 소프트 맥스함수이고 나머지 입력 층과 은닉 층의 활성화함수는 ReLu함수를 사용하였다. 그리고 손실함수는 교차 엔트로피 오차를 사용하였다.

• 한국신뢰성학회지:신뢰성응용연구
• /
• 제18권3호
• /
• pp.271-279
• /
• 2018

Purpose: Probabilistic safety analysis was performed to enhance the safety and reliability of nuclear power plants because traditional deterministic approach has limitations in predicting the risk of failure by crack growth. The study introduces a probabilistic approach to establish a basis for probabilistic safety assessment of passive components. Methods: For probabilistic modeling of fatigue crack growth rate (FCGR), various FCGR tests were performed either under constant load amplitude or constant ${\Delta}K$ conditions by using heat treated X-750 at low temperature with adequate cathodic polarization. Bayesian inference was employed to update uncertainties of the FCGR model using additional information obtained from constant ${\Delta}K$ tests. Results: Four steps of Bayesian parameter updating were performed using constant ${\Delta}K$ test results. The standard deviation of the final posterior distribution was decreased by a factor of 10 comparing with that of the prior distribution. Conclusion: The method for developing a probabilistic crack growth model has been designed and demonstrated, in the paper. Alloy X-750 has been used for corrosion fatigue crack growth experiments and modeling. The uncertainties of parameters in the FCGR model were successfully reduced using the Bayesian inference whenever the updating was performed.

• 한국정보통신학회:학술대회논문집
• /
• 한국정보통신학회 2022년도 춘계학술대회
• /
• pp.587-588
• /
• 2022

본 논문에서는 다수의 모션 캡처 카메라의 2D 패시브마커 영상을 이용한 3차원 공간에서의 리지드 바디(Rigid Body) 추적 방법을 제안한다. 1차로 개별 카메라의 내부 변수를 구하기 위해 체스보드를 이용한 칼리브레이션 과정을 수행하고, 2차 보정 과정에서 3개의 마커가 있는 삼각형 구조물을 모든 카메라가 관찰 가능하도록 움직인 후 프레임별 누적된 데이터를 계산하여 카메라간의 상대적인 위치정보의 보정 및 업데이트를 진행한다. 이 후 각 카메라의 좌표계를 3D월드 좌표계로 변환하는 과정을 통해 3개 마커의 3차원 좌표를 복원하여 각 마커간 거리를 계산하여 실제 거리와의 차이를 비교한 결과 2mm 이내의 오차를 측정하였다.

• Structural Engineering and Mechanics
• /
• 제32권5호
• /
• pp.667-683
• /
• 2009

FE model-based dynamic analysis has been widely used to predict the dynamic characteristics of civil structures. In a physical point of view, an FE model is unavoidably different from the actual structure as being formulated based on extremely idealized engineering drawings and design data. The conventional model updating methods such as direct method and sensitivity-based parameter estimation are not flexible for model updating of complex and large structures. Thus, it is needed to develop a model updating method applicable to complex structures without restriction. The main objective of this paper is to present the model updating method based on the hybrid genetic algorithm (HGA) by combining the genetic algorithm as global optimization method and modified Nelder-Mead's Simplex method as local optimization method. This FE model updating method using HGA does not need the derivation of derivative function related to parameters and without application of complicated inverse analysis methods. In order to allow its application on diversified and complex structures, a commercial FEA tool is adopted to exploit previously developed element library and analysis algorithms. Moreover, an output-level objective function making use of measurement and analytical results is also presented to update simultaneously the stiffness and mass of the analysis model. The numerical examples demonstrated that the proposed method based on HGA is effective for the updating of the FE model of bridge structures.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2003년도 ICCAS
• /
• pp.1706-1709
• /
• 2003

This paper demonstrates a unique method for measuring vehicle states such as body sideslip angle and tire sideslip angle using Global Positioning System(GPS) velocity information in conjunction with other sensors. A method for integrating Inertial Navigation System (INS) sensors with GPS measurements to provide higher update rate estimates of the vehicle states is presented, and the method can be used to estimate the tire cornering stiffness. The experimental results for the GPS velocity-based sideslip angle measurement. From the experimental results, it can be concluded that the proposed method has an advantage for future implementation in a vehicle safety system.

• IEIE Transactions on Smart Processing and Computing
• /
• 제1권2호
• /
• pp.78-87
• /
• 2012

In this paper, we address the tracking problem caused by camera motion and rolling shutter effects associated with CMOS sensors in consumer handheld cameras, such as mobile cameras, digital cameras, and digital camcorders. A modified particle filtering method is proposed for simultaneously tracking objects and compensating for the effects of camera motion. The proposed method uses an elastic registration algorithm (ER) that considers the global affine motion as well as the brightness and contrast between images, assuming that camera motion results in an affine transform of the image between two successive frames. By assuming that the camera motion is modeled globally by an affine transform, only the global affine model instead of the local model was considered. Only the brightness parameter was used in intensity variation. The contrast parameters used in the original ER algorithm were ignored because the change in illumination is small enough between temporally adjacent frames. The proposed particle filtering consists of the following four steps: (i) prediction step, (ii) compensating prediction state error based on camera motion estimation, (iii) update step and (iv) re-sampling step. A larger number of particles are needed when camera motion generates a prediction state error of an object at the prediction step. The proposed method robustly tracks the object of interest by compensating for the prediction state error using the affine motion model estimated from ER. Experimental results show that the proposed method outperforms the conventional particle filter, and can track moving objects robustly in consumer handheld imaging devices.

• 정보학연구
• /
• 제4권1호
• /
• pp.71-79
• /
• 2001

본 논문에서는 역 전파 신경회로망의 학습파라미터를 발생한 오차에 따라서 유동성 있게 갱신할 수 있고 이 학습알고리즘의 효율을 향상시킬 수 있는 초기연결강도 설정 방법을 제안하였다. 제안한 알고리즘은 국소 점을 벗어날 수 있는 것으로 기대되고, 수렴환경에 알맞은 초기 연결강도 발생을 설정할 수 있다. 모의실험에서는 세 가지의 학습패턴을 가지고 실험하였다. 첫 번째 3-패리티 문제에 대한 학습을 수행하였고, 두 번째는 $7{\times}5$ 알파벳 영문자 폰트에 대한 학습이고 세 번째는 필기체 숫자 및 한글의 기본 획에 적용하였다. 모의실험결과, 제안된 방법은 기존의 표준 역 전파 알고리즘에 비하여 약 27%~57.2%정도 학습효율이 향상됨을 고찰하였다

• 한국소음진동공학회:학술대회논문집
• /
• 한국소음진동공학회 1998년도 춘계학술대회논문집; 용평리조트 타워콘도, 21-22 May 1998
• /
• pp.340-346
• /
• 1998

In order to improve the prediction of dynamic behavior in structures, several lower vibration modes from FFT analysis through experiments are used to update the mechanical properties followed by the updated frequencies from numerical analysis. Performance index consists of the sum of error norms between the chosen frequencies and corresponding frequencies from numerical analysis. As an updating process of the natural frequencies, the optimization algorithm based on conjugate gradient method is adopted. The gradient of performance index is calculated using the sensitivity of selected eigenvalues with respect to each design parameter. The mechanical properties of lamina, E$\_$l/, E$\_$2/, .nu.$\_$12/ and G$\_$12/, are design parameters for the updating process. The proposed method is applied to predict the dynamic behavior of composite laminated plates of [0]$\_$8T/ and [.+-.45]$\_$2S/ separately or interchangeably. Also, the mixed case for [0]$\_$8T/ and [.+-.45]$\_$2S/ is exarm'ned to check the possibility for the improved prediction generally. The good agreement is obtained between the measured frequencies and the numerical ones. Based on the results for all the cases studied, the proposed approach has a clear potential in characterizing the mechanical properties of composite lamina.

• 한국산학기술학회논문지
• /
• 제13권6호
• /
• pp.2421-2427
• /
• 2012

본 논문은 쿼드로터 무인기의 비선형 적응 자세 제어기 설계에 대한 것이다. 쿼드로터 무인기는 4 개의 추력기를 갖는 불완전한 비선형시스템으로써 모델 파라미터가 변화하기 때문에 특별한 제어기가 필요하다. 본 연구에서는 산업 응용에 따른 페이로드의 변화에 따라서 무인기의 모델파라미터가 변화하더라도 페이로드에 따른 튜닝을 자동으로 할 수 있도록 적응제어기를 설계하였다. 특히, 무인기의 수학적 모델을 이용하여 비선형 적응제어기를 설계하였고 리야프노프 함수를 이용하여 이론적으로 안정성을 분석하였고 페이로드 변화에 따른 시뮬레이션을 통해서 제어기의 성능과 파라미터 추정법칙의 유효성을 입증하였다.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 1996년도 한국자동제어학술회의논문집(국내학술편); 포항공과대학교, 포항; 24-26 Oct. 1996
• /
• pp.1352-1355
• /
• 1996

The control of diamond turning is usually achieved through a laser-interferometer feedback of slide position. If the tool post is rigid and the material removal process is relatively static, then such a non-collocated position feedback control scheme may surface. However, as the accuracy requirement gets tighter and desired surface contours become more complex, the need for a direct tool-tip sensing becomes inevitable. The physical constraints of the machining process prohibit any reasonable implementation of a tool-tip motion measurement. It is proposed that the measured force normal to the face of the workpiece can be filtered through an appropriate admittance transfer function to result in the estimated depth of cut. This can be compared to the desired depth of cut to generate the adjustment control action in addition to position feedback control. In this work, the design methodology on the admittance model-based control with a conventional controller is presented. The recursive least-squares algorithm with forgetting factor is proposed to identify the parameters and update the cutting process in real time. The normal cutting forces are measured to identify the cutting dynamics in the real diamond turning process using the precision dynamometer. Based on the parameter estimation of cutting dynamics and the admittance model-based nanodynamic control scheme, simulation results are shown.