• Proceedings of the KSME Conference
• /
• 2008.11a
• /
• pp.695-700
• /
• 2008

Artificial neural network (ANN) has been extensively used in areas of nonlinear system modeling, analysis and design applications. Basically, ANN has its distinct capabilities of implementing system identification and/or function approximation using a number of input/output patterns that can be obtained via numerical and/or experimental manners. The paper describes a role of ANN, especially a back-propagation neural network (BPN) in the context of engineering analysis, design and optimization. Fundamental mechanism of BPN is briefly summarized in terms of training procedure and function approximation. The BPN based causality analysis (CA) is further discussed to realize the problem decomposition in the context of multidisciplinary design optimization. Such CA is also applied to quantitatively evaluate the uncoupled or decoupled design matrix in the context of axiomatic design with the independence axiom.

• Phonetics and Speech Sciences
• /
• v.8 no.2
• /
• pp.23-30
• /
• 2016

This paper describes some implementation schemes of CNN in view of mini-batch DNN training for efficient second order optimization. This uses same procedure updating parameters of DNN to train parameters of CNN by simply arranging an input image as a sequence of local patches, which is actually equivalent with mini-batch DNN training. Through this conversion, second order optimization providing higher performance can be simply conducted to train the parameters of CNN. In both results of image recognition on MNIST DB and syllable automatic speech recognition, our proposed scheme for CNN implementation shows better performance than one based on DNN.

• KSBB Journal
• /
• v.14 no.4
• /
• pp.495-502
• /
• 1999

The optimization of fed-batch yeast fermentation process has been performed using genetic algorithm(GA). Three strategies were designed and applied to obtain the optimal feed rate profiles. Genes in the chromosome (input variables for optimization) included feed rates on fixed time intervals (strategy I), or swiching times $t_s1\;and\;t_s2$, and feed rates on singular arc (strategy II), or feed rates and the length of time interval (strategy III). Strategy III showed the best results for all initial conditions due to efficient utilization of genetic information. Simulation results using GA showed similar or better performance compared with previous results by variational caculus and singular control approach.

• The Transactions of the Korean Institute of Electrical Engineers B
• /
• v.52 no.7
• /
• pp.307-314
• /
• 2003

This paper presents a 3D shape optimization algorithm for electromagnetic devices using the design sensitivity analysis with finite element method. The structural deformation analysis based on the deformation theory of the elastic body under stress is used for mesh renewing. The design sensitivity and adjoint variable formulae are derived for the 3D finite element method with edge element. The results of sensitivity analysis are used as the input data of the structural analysis to calculate the relocation of the nodal points. This method makes it possible that the new mesh of analysis region can be obtained from the initial mesh without regeneration. The proposed algorithm is applied to the shape optimization of 3D electromagnet pole to net a uniform flux density at the target region.

• Structural Engineering and Mechanics
• /
• v.53 no.6
• /
• pp.1127-1141
• /
• 2015

The optimum design of liquid column dampers in seismic vibration control considering system parameter uncertainty is usually performed by minimizing the unconditional response of a structure without any consideration to the variation of damper performance due to uncertainty. However, the system so designed may be sensitive to the variations of input system parameters due to uncertainty. The present study is concerned with robust design optimization (RDO) of liquid column vibration absorber (LCVA) considering random system parameters characterizing the primary structure and ground motion model. The RDO is obtained by minimizing the weighted sum of the mean value of the root mean square displacement of the primary structure as well as its standard deviation. A numerical study elucidates the importance of the RDO procedure for design of LCVA system by comparing the RDO results with the results obtained by the conventional stochastic structural optimization procedure and the unconditional response based optimization.

• Journal of Welding and Joining
• /
• v.10 no.2
• /
• pp.32-42
• /
• 1992

A method of optimization of process parameters in Arc Welding has been discussed in this paper. The method of investigation is based on the numerical calculation of weld bead by a finite element method and non-linear optimization technique is applied to estimated the optimization process parameters from the numerical calculation. The common package program(ANSYS 4.4A) was used to obtain the process parameters for a thin plate arc welding (TIG, CO$_{2}$). The results on some test are satisfactory and the used method of this paper is a useful guide to the optimum welding condition.

• Transactions of the Korean Society of Mechanical Engineers
• /
• v.18 no.1
• /
• pp.12-23
• /
• 1994

This paper proposes an optimal design software for the open-chain dynamic systems whose governing equations are expressed as differential equation. In this software, an input module and an automatic creation module of the equation of motion are developed to contrive the user's convenience. To analyze the equation of motion of the dynamic systems, variable-order and variable-stepsize Adams-Bashforth-Moulton predictor-corrector method is used to improve the efficiency. For the optimization and the design sensitivity analysis, ALM(augmented lagrange multiplier)method and adjoint variable method are adopted respectively. An output module with which the user can compare and investigate the analysis and the optimization results through tables and graphs is also provided. The developed software is applied to three typical dynamic response optimization problems, and the results compare very well with those available in the literature, demonstrating its effectiveness.

• Speech Sciences
• /
• v.14 no.2
• /
• pp.105-114
• /
• 2007

The performance of speaker identifier is severely degraded in noisy environments. A study suggested the concept of observation membership for enhancing performances of speaker identifier with noisy speech [1]. The method scaled observation probabilities of input speech by observation identification values decided by SNR. In the paper [1], the authors suggested heuristic parameter values for membership function. In this paper we attempt to apply particle swarm optimization (PSO) for obtaining the optimal parameters for speaker identification in noisy environments. With the speaker identification experiments using the ETRI database we prove that the optimization approach can yield better performance than using only the original membership function.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 1993.06a
• /
• pp.1230-1233
• /
• 1993

Numerical solution of inverse problem of Takagi-Sugeno fuzzy model is proposed. The method is located on the application of numerical optimization to the fuzzy model. Steepest descent method is used for the numerical optimization. We use the linear approximation of fuzzy model, called pseudo first order approximation, by fixing the membership value on the neighborhood of the corresponding input. It is introduced in order to reduce the difficulty of optimization process. The efficiency of this method is shown by a numerical experiment.

• Knowledge Management Research
• /
• v.21 no.1
• /
• pp.27-40
• /
• 2020

Response surface methodology (RSM) empirically studies the relationship between a response variable and input variables in the product or process development phase. The ultimate goal of RSM is to find an optimal condition of the input variables that optimizes (maximizes or minimizes) the response variable. RSM can be seen as a knowledge management tool in terms of creating and utilizing data, information, and knowledge about a product production and service operations. In the field of product or process development, most real-world problems often involve a simultaneous consideration of multiple response variables. This is called a multiple response surface (MRS) problem. Various approaches have been proposed for MRS optimization, which can be classified into loss function approach, priority-based approach, desirability function approach, process capability approach, and probability-based approach. In particular, the loss function approach is divided into univariate and multivariate approaches at large. This paper focuses on the univariate approach. The univariate approach first obtains the mean square error (MSE) for individual response variables. Then, it aggregates the MSE's into a single objective function. It is common to employ the weighted sum or the Tchebycheff metric for aggregation. Finally, it finds an optimal condition of the input variables that minimizes the objective function. When aggregating, the relative weights on the MSE's should be taken into account. However, there are few studies on how to determine the weights systematically. In this study, we propose an interactive procedure to determine the weights through considering a decision maker's preference. The proposed method is illustrated by the 'colloidal gas aphrons' problem, which is a typical MRS problem. We also discuss the extension of the proposed method to the weighted MSE (WMSE).