• Journal of Electrical Engineering and Technology
• /
• v.7 no.4
• /
• pp.636-645
• /
• 2012

This study introduces an information granular-based fuzzy radial basis function neural networks (FRBFNN) based on multiobjective optimization and weighted least square (WLS). An improved multiobjective space search algorithm (IMSSA) is proposed to optimize the FRBFNN. In the design of FRBFNN, the premise part of the rules is constructed with the aid of Fuzzy C-Means (FCM) clustering while the consequent part of the fuzzy rules is developed by using four types of polynomials, namely constant, linear, quadratic, and modified quadratic. Information granulation realized with C-Means clustering helps determine the initial values of the apex parameters of the membership function of the fuzzy neural network. To enhance the flexibility of neural network, we use the WLS learning to estimate the coefficients of the polynomials. In comparison with ordinary least square commonly used in the design of fuzzy radial basis function neural networks, WLS could come with a different type of the local model in each rule when dealing with the FRBFNN. Since the performance of the FRBFNN model is directly affected by some parameters such as e.g., the fuzzification coefficient used in the FCM, the number of rules and the orders of the polynomials present in the consequent parts of the rules, we carry out both structural as well as parametric optimization of the network. The proposed IMSSA that aims at the simultaneous minimization of complexity and the maximization of accuracy is exploited here to optimize the parameters of the model. Experimental results illustrate that the proposed neural network leads to better performance in comparison with some existing neurofuzzy models encountered in the literature.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.8 no.4
• /
• pp.561-567
• /
• 2013

Some of the feature vectors for speech recognition are rare and unusual. These patterns lead to overfitting for the parameters of the speech recognition system and, as a result, cause structural risks in the system that hinder the good performance in recognition. In this paper, as a method of removing these unusual patterns, we try to exclude vectors whose norms are larger than a specified cutoff value and then train the speech recognition system. The objective of this study is to exclude as many unusual feature vectors under the condition of no significant degradation in the speech recognition error rate. For this purpose, we introduce a cutoff parameter and investigate the resultant effect on the speaker-independent speech recognition of isolated words by using FVQ(Fuzzy Vector Quantization)/HMM(Hidden Markov Model). Experimental results showed that roughly 3%~6% of the feature vectors might be considered as unusual, and therefore be excluded without deteriorating the speech recognition accuracy.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.9 no.4
• /
• pp.929-935
• /
• 2008

As internet telephony services based on Session initiation Protocol (SIP) enter the spotlight as marketable technology, many products based on SIPs have been developed and utilized for home and office telephony services. The call connection of an internet phone is classified into specific call connections and group call connections. Group call connections have a forking function which delivers the message to all of the group members. This function requires excessive message control for a call connection and creates heavy traffic in the network. In the internet cail system model. most of the call-setup messages are directed to the proxy server during a short time period. This heavy message load brings an unwanted delay in message processing and. as a result, call setup can not be made. To solve the delay problem, we simplified the analysis of the call-setup message in the proxy server, and processed the forking function distributed for the group call-setup message. In this thesis, a new system model to minimize the load is proposed and the subsequent implementation of this model demonstrates the performance improvement.

• Journal of the Society of Naval Architects of Korea
• /
• v.34 no.2
• /
• pp.104-110
• /
• 1997

Increase of marine transpotation in coastal area frequently yields oil spill accidents due to collision or grounding of oil tankers, which affects great deal of damages on ocean environments. Exact prediction of oil pollution area in time domain, which is called oil map, is very important for effective and efficient oil recovery and minimization of environmental damage. The prediction is carried out by considering the two distinct processes which are initial diffusion on the still water surface and advection due to tide, wind wave induced surface currents. In the present paper, only the initial diffusion is dealt with. Somewhat new simulation model and its numerical scheme are proposed to predict it. Simple diffusion experiment is also carried out to check the validity of the present method. Furthermore, some example simulations are performed for virtual oil spill accident. Quite realistic oil map including oil thickness distributions can be obtained by the present model.

• Transactions of the Korean Society of Automotive Engineers
• /
• v.9 no.2
• /
• pp.107-115
• /
• 2001

It is required more precise analysis for practical load because of complexities and varieties of vehicle structure. To establish the numerical model, many researchers have been developed designing tools for linking F.E. Analysis results and experimental results. There studies have generally focused on each experimental method or analytical method separately. There are few studies based on both methods. This paper conceives new procedure for the determination of the load direction and magnitude applied on mechanical structures. New procedure is the combination of the analytical and empirical method with analyzed strain by F.E. Analysis under unit load and with measured principal stress by strain gages under driving load, respectively. In this paper, we theorize the procedure of practical load determination and make the validity and the practicality of the procedure with the application to T-shape jointed structure. F.E. Analysis is conducted to get the principal stress on arbitrary points in the F.E. model of T-shape joint under unit load. Then experiment is carried out to get the principal stress on the same points of F.E. model. To demonstrate the actual driving condition, the load conditions are bending and torsion. From these two data sets, the magnitude, the direction and the position of load can be obtained. Theory and practice do not always coincide; since there are some errors such as ill-poseness, measuring error and modeling error in experimental data, we examine the proper method of error minimization.

• Structural Engineering and Mechanics
• /
• v.56 no.6
• /
• pp.959-982
• /
• 2015

In this study, a non-stationary random earthquake Clough-Penzien model is used to describe earthquake ground motion. Using stochastic direct integration in combination with an equivalent linear method, a solution is established to describe the non-stationary response of lead-rubber bearing (LRB) system to a stochastic earthquake. Two parameters are used to develop an optimization method for bearing design: the post-yielding stiffness and the normalized yield strength of the isolation bearing. Using the minimization of the maximum energy response level of the upper structure subjected to an earthquake as an objective function, and with the constraints that the bearing failure probability is no more than 5% and the second shape factor of the bearing is less than 5, a calculation method for the two optimal design parameters is presented. In this optimization process, the radial basis function (RBF) response surface was applied, instead of the implicit objective function and constraints, and a sequential quadratic programming (SQP) algorithm was used to solve the optimization problems. By considering the uncertainties of the structural parameters and seismic ground motion input parameters for the optimization of the bearing design, convex set models (such as the interval model and ellipsoidal model) are used to describe the uncertainty parameters. Subsequently, the optimal bearing design parameters were expanded at their median values into first-order Taylor series expansions, and then, the Lagrange multipliers method was used to determine the upper and lower boundaries of the parameters. Moreover, using a calculation example, the impacts of site soil parameters, such as input peak ground acceleration, bearing diameter and rubber shore hardness on the optimization parameters, are investigated.

• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.38 no.3
• /
• pp.117-126
• /
• 2015

This paper is to analyze an integrated production and inventory model in a single-vendor multi-buyer supply chain. The vendor is defined as the manufacturer and the buyers as the retailers. The product that the manufacturer produces is supplied to the retailers with constant periodic time interval. The production rate of the manufacturer is constant for the time. The demand of the retailers is constant for the time. The cycle time of the vendor is defined as the elapsed time from the start of the production to the start of the next production, while the cycle times of the buyer as the elapsed time between the adjacent supply times from the vendor to the buyer. The cycle times of the vendor and the buyers that minimizes the total cost in a supply chain are analyzed. The cost factors are the production setup cost and the inventory holding cost of the manufacturer, the ordering cost and the inventory holding cost of the retailers. The cycle time of the vendor is investigated through the cycle time that satisfies economic production quantity with the production setup cost and the inventory holding cost of the manufacturer. An integrated production and inventory model is formulated, and an algorithm is developed. An numerical example is presented to explain the algorithm. The solution of the algorithm for the numerical examples is compared with that of genetic algorithm. Numerical example shows that the vendor and the buyers can save cost by integrated decision making.

• Fire Science and Engineering
• /
• v.20 no.4 s.64
• /
• pp.13-18
• /
• 2006

The fire protection regulation for the nuclear power plants is based on the qualitative fire hazard assessment and the quantitative fire risk analysis, and the fire risk is managed by the fire protection plan with the appropriate balance among the fire prevention, fire suppression and the minimization of the fire effect. In these days, the zone model or the field model is generally used for the detail evaluation for the fire risk. At this paper, with consideration of the present trend, we evaluate whether the quantitative fire risk analysis and the assessment of fire result for fire areas at nuclear power plants can be possible by use of Fire Dynamics Simulator (FDS) that is the state-of-the-art fire modeling tool. Consequently, it is expected that the quantitative fire risk evaluation propelled by the fire modeling can be available as an applicable tool to improve the core damage frequency as well as the quantitative fire risk analysis.

• Journal of the Korea Institute of Military Science and Technology
• /
• v.24 no.4
• /
• pp.382-391
• /
• 2021

Unsupervised domain adaptation often gives impressive solutions to handle domain shift of data. Most of current approaches assume that unlabeled target data to train is abundant. This assumption is not always true in practices. To tackle this issue, we propose a general solution to solve the domain gap minimization problem without any target data. Our method consists of two regularization steps. The first step is a pixel regularization by arbitrary style transfer. Recently, some methods bring style transfer algorithms to domain adaptation and domain generalization process. They use style transfer algorithms to remove texture bias in source domain data. We also use style transfer algorithms for removing texture bias, but our method depends on neither domain adaptation nor domain generalization paradigm. The second regularization step is a feature regularization by feature alignment. Adding a feature alignment loss term to the model loss, the model learns domain invariant representation more efficiently. We evaluate our regularization methods from several experiments both on small dataset and large dataset. From the experiments, we show that our model can learn domain invariant representation as much as unsupervised domain adaptation methods.

• Steel and Composite Structures
• /
• v.52 no.1
• /
• pp.31-43
• /
• 2024

A very widely used analytical method (mathematical model), mentioned in Eurocode 3, to examine the connections' bending behavior is the component-based method that has certain weak points shown in the plastic behavior part of the moment-rotation curves. In the component method available in Eurocode 3, for simplicity, the effect of strain hardening is omitted, and the bending behavior of the connection is modeled with the help of a two-line diagram. To make the component method more efficient and reliable, this research proposed its advanced version, wherein the plastic part of the diagram was developed beyond the guidelines of the mentioned Regulation, implemented to connect the end plate, and verified with the moment-rotation curves found from the laboratory model and the finite element method in ABAQUS. The findings indicated that the advanced component method (the method developed in this research) could predict the plastic part of the moment-rotation curve as well as the conventional component-based method in Eurocode 3. The comparison between the laboratory model and the outputs of the conventional and advanced component methods, as well as the outputs of the finite elements approach using ABAQUS, revealed a different percentage in the ultimate moment for bolt-extended end-plate connections. Specifically, the difference percentages were -31.56%, 2.46%, and 9.84%, respectively. Another aim of this research was to determine the optimal dimensions of the end plate joint to reduce costs without letting the mechanical constraints related to the bending moment and the resulting initial stiffness, are not compromised as well as the safety and integrity of the connection. In this research, the thickness and dimensions of the end plate and the location and diameter of the bolts were the design variables, which were optimized using Particle Swarm Optimization (PSO), Snake Optimization (SO), and Teaching Learning-Based Optimization (TLBO) to minimization the connection cost of the end plate connection. According to the results, the TLBO method yielded better solutions than others, reducing the connection costs from 43.97 to 17.45€ (60.3%), which shows the method's proper efficiency.