• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.30 no.8 s.251
• /
• pp.1001-1007
• /
• 2006

In this paper we present robust optimal supporting positions for large glass panels used for TFT-LCD monitors when they are stored in a cassette during manufacturing process. The criterion taken is to minimize their maximum deflection. Since they are supported by some supports and have large deformations, contact analysis with a geometrically nonlinear effect is necessary. In addition, the center of a panel can not be positioned exactly as intended and should be considered as uncertainties. To take into account of these effects, the mean and the standard deviation of system response functions, particularly the deflection of the panels, need be calculated. A function approximation moment method (FAMM) is utilized to estimate them. It is a special type of response surface methodology for structural reliability analysis and can be efficiently used to estimate the two stochastic properties, that is, the system performance and the perturbations caused by uncertainties. For a design purpose, they are to be minimized simultaneously by some optimization algorithm to obtain robust optimal supporting positions.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.8
• /
• pp.3519-3533
• /
• 2020

Tuberculosis is a chronic and delayed infection which is easily experienced by young people. According to the statistics of the World Health Organization (WHO), there are nearly ten million fell ill with tuberculosis and a total of 1.5 million people died from tuberculosis in 2018 (including 251000 people with HIV). Tuberculosis is the largest single infectious pathogen that leads to death. In order to help doctors with tuberculosis diagnosis, we compare the tuberculosis classification abilities of six popular convolutional neural network (CNN) models in the same data set to find the best model. Before training, we optimize three parts of CNN to achieve better results. We employ sigmoid function to replace the step function as the activation function. What's more, we use binary cross entropy function as the cost function to replace traditional quadratic cost function. Finally, we choose stochastic gradient descent (SGD) as gradient descent algorithm. From the results of our experiments, we find that Densenet121 is most suitable for tuberculosis diagnosis and achieve a highest accuracy of 0.835. The optimization and expansion depend on the increase of data set and the improvements of Densenet121.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2018.05a
• /
• pp.327-327
• /
• 2018

돌발홍수, 집중호우 등 강우가 발생 원인되는 자연재해에 효과적으로 대응하기 위한 연구가 활발히 이루어지고 있으나 강우의 시공간 변동성과 발생과정의 복잡한 물리과정으로 인해 강우 추정에 한계를 가진다. 일반적으로 강우 추정은 물리적, 추계학적 모형을 이용하며 추계학적 모형의 점과정(point process)을 이용하여 강우를 생산한다. 추계학적 강우 모형은 관측 강우의 시간 스케일, 강우발생 빈도, 강우 강도 등 강우 구조의 특성을 반영 할 수 있다는 장점을 가지고 있으나 생산되는 강우의 구조가 추정되는 매개변수에 크게 의존한다는 점에서 실제 강우에 적합한 매개변수 추정이 중요하다. 본 연구에서는 낙동강 유역내에 있는 20개의 강우관측 지점을 대상으로 1973년-2017년까지의 강우 관측자료를 수집하였으며 추계학적 강우생성 모형으로 점과정을 이용하는 추계학적 강우생성 모형인 NSRPM(Neymann-Scott rectangular pulse model)을 선정하였다. NSRPM모형의 매개변수를 추정하기위한 최적기법으로 DFP(Davidon-Fletcher-Powell), GA(genetic algorithm), Nelder-Mead, DE(differential evolution)를 이용하여 추정된 매개변수의 적합성을 분석하고 지역특성을 고려한 매개변수 추정 기법을 제시하였다. 추정된 모형의 매개변수를 분석한 결과 DE와 Nelder-Mead 기법이 높은 적합성을 보였으며 DFP, GA기법이 상대적으로 낮은 적합도를 보였다.

• International Journal of Computer Science & Network Security
• /
• v.23 no.10
• /
• pp.115-128
• /
• 2023

Deep learning has been incorporating various optimization techniques motivated by new pragmatic optimizing algorithm advancements and their usage has a central role in Machine learning. In recent past, new avatars of various optimizers are being put into practice and their suitability and applicability has been reported on various domains. The resurgence of novelty starts from Stochastic Gradient Descent to convex and non-convex and derivative-free approaches. In the contemporary of these horizons of optimizers, choosing a best-fit or appropriate optimizer is an important consideration in deep learning theme as these working-horse engines determines the final performance predicted by the model. Moreover with increasing number of deep layers tantamount higher complexity with hyper-parameter tuning and consequently need to delve for a befitting optimizer. We empirically examine most popular and widely used optimizers on various data sets and networks-like MNIST and GAN plus others. The pragmatic comparison focuses on their similarities, differences and possibilities of their suitability for a given application. Additionally, the recent optimizer variants are highlighted with their subtlety. The article emphasizes on their critical role and pinpoints buttress options while choosing among them.

• Journal of Auto-vehicle Safety Association
• /
• v.16 no.1
• /
• pp.6-11
• /
• 2024

This paper presents a motion planning algorithm of autonomous racing vehicles for mimicking the characteristics of a human driver. Time optimal maneuver of a race car has been actively studied as a major research area over the past decades. Although the time optimization problem yields a single time series solution of minimum time maneuver inputs for the vehicle, human drivers achieve similar lap times while taking various racing lines and velocity profiles. In order to model the characteristics of a specific driver and reproduce the motion, a stochastic motion planning framework based on kernelized motion primitive is introduced. The proposed framework imitates the behavior of the generated reference motion, which is based on a small number of human demonstration laps along the racetrack using Gaussian mixture model and Gaussian mixture regression. The mean and covariance of the racing line and velocity profile mimicking the driver are obtained by accumulating the outputs tested at equidistantly sampled input points. The results confirmed that the obtained lateral and longitudinal motion simulates the driver's driving characteristics, which are feasible for actual vehicle test environments.

• Journal of IKEEE
• /
• v.5 no.1 s.8
• /
• pp.1-7
• /
• 2001

In this paper, we have investigated the CDMA(Code Division Multiple Access) Cellular System with non-linear equalizer in reverse link channel. In general, due to unknown characteristics of channel in the wireless communication, the distribution of the observables cannot be specified by a finite set of parameters; instead, we partitioned the m-dimensional sample space Into a finite number of disjointed regions by using quantiles and a vector quantizer based on training samples. The algorithm proposed is based on a piecewise approximation to regression function based on quantiles and conditional partition moments which are estimated by Robbins Monro Stochastic Approximation (RMSA) algorithm. The resulting equalizers and detectors are robust in the sense that they are insensitive to variations in noise distributions. The main idea is that the robust equalizers and robust partition detectors yield better performance in equiprobably partitioned subspace of observations than the conventional equalizer in unpartitioned observation space under any condition. And also, we apply this idea to the CDMA system and analyze the BER performance.

• Journal of Korea Water Resources Association
• /
• v.33 no.3
• /
• pp.365-374
• /
• 2000

This paper is to present the determination of the optimal Joss rate parameters and urnt bydrographs from the observed single rainfall-runoff event using optimization models coupled with a stochastic technique for the global solution. Two kinds of the linear program models are formulated to derive the optimal unit hydrographs and loss rate parameters for gaged basins; one mimmizes the summation of the absolute residual between predlCted and observed runoff ordinates and the other, the maximum absolute residuaL Multistart algorithm which is one or stochastic techniques for the global optimum is adopted to perturb the parameters of the loss rate equations. Multistart efficiently searches the feasIble region to identify the global optimlUll for loss rate parameters, which yields the optimal loss rate parameters and unit hydrograph for Kostiakov's, Plulip's, and Horton's equation. The unique unit hydrograph ordinates for a gIven rainfall-runoff event iS exclusrvely obtained WIth $\Phi$ index, but unit hydrograph ordinates depend upon the parameters [or each loss rate equations. The parameters of Green-Ampt's are determined through a trial and error method. In this paper the single rainfall-nmoff event observed from a watershed is considered to test the proposed method. The optimal unit hydrograph herein found has smaller deviations than the ones reported previously by other researchers.

• Clean Technology
• /
• v.24 no.3
• /
• pp.239-248
• /
• 2018

In this study, pervaporation experiments of water, ethanol and IPA (Isopropyl alcohol) single components and water/ethanol, water/IPA mixtures were carried out using zeolite 4A membranes developed by Fine Tech Co. Ltd. Those membranes were fabricated by hydrothermal synthesis (growth in hydrothermal condition) after uniformly dispersing the zeolite seeds on the tubular alumina supports. They have a pore size of about $4{\AA}$ by ion exchange of $Na^+$ to the LTA structure with Si/Al ratio of 1.0, and shows strong hydrophilic property. Physical characteristics of prepared membranes were evaluated by using SEM (surface morphology), porosimetry (macro- or meso- pore analysis), BET (micropore analysis), and load tester (compressive strength). Pervaporation experiments with various temperature and concentration conditions confirmed that the zeolite 4A membrane can selectively separate water from ethanol and IPA. Water/ethanol separation factor was over 3,000 and water/IPA separation factor was over 1,500 (50 : 50 wt%, initial feed concentration). Pervaporation behaviors of single components and binary mixtures were predicted using ACM (activity coefficient model), GMS (generalized Maxwell Stefan) model and DGM (Dusty Gas Model). The adsorption and diffusion coefficients of the zeolite top layer were obtained by parameter estimation using GA (Genetic Algorithm, stochastic optimization method). All the calculations were carried out using MATLAB 2018a version.

• Korean Chemical Engineering Research
• /
• v.52 no.4
• /
• pp.467-474
• /
• 2014

Preliminary design in chemical process furnishes economic feasibility through calculation of both mass balance and energy balance and makes it possible to produce a desired product under the given conditions. Through this design stage, the process possesses unchangeable characteristics, since the materials, reactions, unit configuration, and operating conditions were determined. Unique characteristics could be very economic, but it also implies various potential risk factors as well. Therefore, it becomes extremely important to design process considering both economics and safety by integrating process simulation and quantitative risk analysis during preliminary design stage. The target of this study is LNG liquefaction process. By the simulation using Aspen HYSYS and quantitative risk analysis, the design variables of the process were determined in the way to minimize the inherent explosion risks and operating cost. Instead of the optimization tool of Aspen HYSYS, the optimization was performed by using stochastic optimization algorithm (Covariance Matrix Adaptation-Evolution Strategy, CMA-ES) which was implemented through automation between Aspen HYSYS and Matlab. The research obtained that the important variable to enhance inherent safety was the operation pressure of mixed refrigerant. The inherent risk was able to be reduced about 4~18% by increasing the operating cost about 0.5~10%. As the operating cost increases, the absolute value of risk was decreased as expected, but cost-effectiveness of risk reduction had decreased. Integration of process simulation and quantitative risk analysis made it possible to design inherently safe process, and it is expected to be useful in designing the less risky process since risk factors in the process can be numerically monitored during preliminary process design stage.