• The Transactions of the Korean Institute of Electrical Engineers D
• /
• v.53 no.2
• /
• pp.86-93
• /
• 2004

The DCT algorithm needs an efficient hardware architecture to compute inner product. The conventional design method, like ROM-based DA(Distributed Arithmetic), has large hardware complexity. Because of this reason, a CSHM(Computation Sharing Multiplication) was proposed for implementing inner product by Park. However, the Park's CSHM has inefficient hardware architecture in the precomputer and select units. Therefore it degrades the performance of the multiplier. In this paper, we presents the optimization design method for inner product using CSHM algorithm and applied it to implementation of 1-D DCT processor. The experimental results show that the proposed multiplier is more efficient than Park's when hardware architectures and logic synthesis results were compared. The designed 1-D DCT processor by using proposed design method is more high performance than typical methods.

• Proceedings of the SAREK Conference
• /
• 2006.06a
• /
• pp.731-736
• /
• 2006

Groundwater heat pump (GWHP) system has been expected to achieve the higher coefficient of performance (COP) and more energy-saving than the conventional air-source heat pump (ASHP) system. Its performance significantly depends on the characteristics of groundwater and the underground thermal properties. Furthermore, there is a large difference of COP in utilizing groundwater between as a heat resource and as a thermal storage medium. For properties of groundwater there is suitable utilizing system. However, many of GWHP systems have not been considered sufficiently such properties. This research describes optimization of GWHP system according to the properties of groundwater based on 3D numerical heat and water transport simulation.

• Journal of the Semiconductor & Display Technology
• /
• v.18 no.2
• /
• pp.17-22
• /
• 2019

The method of electrical discharge machining (EDM), one of the processing methods based on non-traditional manufacturing procedures, is gaining increased popularity, since it does not require cutting tools and allows machining involving hard, brittle, thin and complex geometry. This present investigation details the determination of optimum process parameter to attain the better machining performance in EDM of AISI 4340 steel with graphite as a tool electrode. The experimental combinations are planned and analyzed by Taguchi's design of experiments approach. To predict the optimal condition, the experiments are conducted by using Taguchi's L27 orthogonal array. The influence of process variables such as discharge current, pulse on and pulse off time, voltage and spark speed were investigated to control the various desired performance measures such as surface roughness. Analysis of Variance (ANOVA) has to be performed to know the magnitude of each factor. Investigations indicate that the surface roughness is strongly depend on pulsed current.

• Electronics and Telecommunications Trends
• /
• v.34 no.4
• /
• pp.32-42
• /
• 2019

The performance of machine learning algorithms significantly depends on how a configuration of hyperparameters is identified and how a neural network architecture is designed. However, this requires expert knowledge of relevant task domains and a prohibitive computation time. To optimize these two processes using minimal effort, many studies have investigated automated machine learning in recent years. This paper reviews the conventional random, grid, and Bayesian methods for hyperparameter optimization (HPO) and addresses its recent approaches, which speeds up the identification of the best set of hyperparameters. We further investigate existing neural architecture search (NAS) techniques based on evolutionary algorithms, reinforcement learning, and gradient derivatives and analyze their theoretical characteristics and performance results. Moreover, future research directions and challenges in HPO and NAS are described.

• International Journal of Internet, Broadcasting and Communication
• /
• v.16 no.3
• /
• pp.192-198
• /
• 2024

This paper investigates the application of multi-agent deep reinforcement learning in the fighting game Samurai Shodown using Proximal Policy Optimization (PPO) and Advantage Actor-Critic (A2C) algorithms. Initially, agents are trained separately for 200,000 timesteps using Convolutional Neural Network (CNN) and Multi-Layer Perceptron (MLP) with LSTM networks. PPO demonstrates superior performance early on with stable policy updates, while A2C shows better adaptation and higher rewards over extended training periods, culminating in A2C outperforming PPO after 1,000,000 timesteps. These findings highlight PPO's effectiveness for short-term training and A2C's advantages in long-term learning scenarios, emphasizing the importance of algorithm selection based on training duration and task complexity. The code can be found in this link https://github.com/Lexer04/Samurai-Shodown-with-Reinforcement-Learning-PPO.

• Journal of the Computational Structural Engineering Institute of Korea
• /
• v.37 no.3
• /
• pp.217-224
• /
• 2024

This paper proposes an efficient dimension reduction method (DRM) that considers the nonlinearity of the performance functions in reliability-based design optimization (RBDO). The dimension reduction method evaluates the reliability more accurately than the first-order reliability method (FORM) using integration quadrature points and weights. However, its efficiency is hindered as the number of quadrature points increases owing to the need for an additional evaluation of the performance function. In this study, we assessed the nonlinearity of the performance function in RBDO and proposed criteria for determining the number of quadrature points based on the degree of nonlinearity. This approach suggests adjusting the number of quadrature points during each iteration of the RBDO process while maintaining the accuracy of theDRM while improving the computational efficiency. The nonlinearity of the performance function was evaluated using the angle between the vectors used in the maximum probable target point (MPTP) search. Numerical tests were conducted to determine the appropriate number of quadrature points according to the degree of nonlinearity. Through a 2D numerical example, it is confirmed that the proposed method improves the efficiency while maintaining the accuracy of the dimension reduction method or Monte Carlo Simulation (MCS).

• The KSFM Journal of Fluid Machinery
• /
• v.18 no.4
• /
• pp.43-48
• /
• 2015

In this study, a method for optimal design of impeller shroud for centrifugal compressor using response surface method (RSM) and multi-objective genetic algorithm (MOGA) was studied. Numerical simulation was conducted using ANSYS CFX with various configurations of shroud. Each of the design parameters was divided into 3 levels. Total 15 design points were planned by central composite design (CCD) method, which is one of the design of experiment (DOE) techniques. Response surfaces based on the results of DOE were used to find the optimal shape of impeller shroud for high aerodynamic performance. The whole process of optimization was conducted using ANSYS Design Xplorer (DX). Results showed that the isentropic efficiency, which is the main performance parameter of the centrifugal compressor, was increased 0.4% through the optimization.

• Journal of the Korean Society of Safety
• /
• v.33 no.2
• /
• pp.61-67
• /
• 2018

This study proposes a new vibration control approach for adjacent twin structures, which is termed as viscous damper asymmetric coupling system in this paper. The proposed system takes a concept that the diagonal bracing viscous dampers are asymmetrically distributed in two buildings to break the behavior symmetry of the twin buildings and then the coupling viscous damper is additionally installed at the top floor of the two buildings to couple both buildings and interactively transfer the asymmetric behavior-caused damping forces into both buildings. These asymmetric damping distributions and interacting damping forces of the connection damper efficiently suppress the overall vibration of the damper-coupled adjacent twin buildings efficiently. Genetic algorithm (GA) based multi-objective optimization technique is adopted for optimal design of the proposed system. In the numerical example of adjacent twin 10-story building structures, the conventional control approach, that is, uniform damping distribution system (UDS) is also taken into account for comparison purpose. The optimization results verify that the proposed system either can improve the control performance over the UDS with the same damping capacity, or can save the damping capacity significantly while maintaining the similar level of control performance to the UDS.

• Proceedings of the Korean Institute of Navigation and Port Research Conference
• /
• 2015.07a
• /
• pp.171-172
• /
• 2015

Vinalines is actually known as not only one of Vietnam's state-sponsored economic giants but also the largest shipowner by tonnage in Vietnamese shipping industry. Therefore, a question of how to improve business performance of the corporation is always received deep attention by Vietnamese government, specially after the seriously economic scandal of Vinalines in a last few years. Among development strategies, the study focuses on short-term one in which Vinalines is recommended to restructure its own fleet in order to optimize performance of fleet operation and minimize costs while meeting the customer's shipping demand in near future. The first section is of introduction. Via method of statistical data analysis, section 2 brings to readers a panorama about the development profile and the current situation of development of Vinalines. In section 3, the authors use linear programming for setting a cost-minimization model optimizing Vinalines fleet structure based on available statistics and forecast information by Vinalines. The optimization problem is solved by applying AIMMS software in section 4. Finally, some conclusions and proposals by authors for the development of Vinalines are given.

• Journal of Mechanical Science and Technology
• /
• v.19 no.3
• /
• pp.768-777
• /
• 2005

This paper shows that the performance of a nonlinear fluid engine mount can be improved by an optimal design process. The property of a hydraulic mount with inertia track and decoupler differs according to the disturbance frequency range. Since the excitation amplitude is large at low excitation frequency range and is small at high excitation frequency range, mathematical model of the mount can be divided into two linear models. One is a low frequency model and the other is a high frequency model. The combination of the two models is very useful in the analysis of the mount and is used for the first time in the optimization of an engine mount in this paper. Normally, the design of a fluid mount is based on a trial and error approach in industry because there are many design parameters. In this study, a nonlinear mount was optimized to minimize the transmissibilities of the mount at the notch and the resonance frequencies for low and high-frequency models by a popular optimization technique of sequential quadratic programming (SQP) supported by $MATLAB^{(R)}$subroutine. The results show that the performance of the mount can be greatly improved for the low and high frequencies ranges by the optimization method.