• The KIPS Transactions:PartB
• /
• v.11B no.4
• /
• pp.449-456
• /
• 2004

This paper proposes a novel recognition model, a recurrent fuzzy associative memory(RFAM), for recognizing time-series patterns contained an ambiguity. RFAM is basically extended from FAM(Fuzzy Associative memory) by adding a recurrent layer which can be used to deal with sequential input patterns and to characterize their temporal relations. RFAM provides a Hebbian-style learning method which establishes the degree of association between input and output. The error back-propagation algorithm is also adopted to train the weights of the recurrent layer of RFAM. To evaluate the performance of the proposed model, we applied it to a word boundary detection problem of speech signal.

• Transactions of the Korean Society of Automotive Engineers
• /
• v.21 no.3
• /
• pp.1-14
• /
• 2013

In this paper, a finite element analysis technique of rolling stock models for collision-induced derailments was suggested using rolling contacts for wheel-rail interaction. The collision-induced derailments of rolling stock can be categorized into two patterns of wheel-climb and wheel-lift according to the friction direction between wheel flange and rail. The wheel-climb derailment types are classified as Climb-up, Climb/roll-over and Roll-over-C types, and the wheel-lift derailment types as Slip-up, Slip/roll-over and Roll-over-L types. To verify the rolling contact simulations for wheel-rail interaction, dynamic simulations of a single wheelset using Recurdyn of Functionbay and Ls-Dyna of LSTC were performed and compared for the 6-typical derailments. The collision-induced derailment simulation of the finite element model of KHST (Korean High Speed Train) was conducted and verified using the theoretical predictions of a simplified wheel-set model proposed for each derailment type.

• Journal of KIISE:Software and Applications
• /
• v.30 no.7_8
• /
• pp.752-763
• /
• 2003

In the large scale B2B transaction like buying ´Express-Train´ or selling ´Daewoo Motor`, a tremendous amount of variables and factors of chaos functionate in it directly or indirectly. To get the effective information processing on the so called complex system like this, it should be possible to unite the human´s ability on the implicit information processing and the agent´s ability on the explicit information processing. In this paper, we suggested a union model for uniting these two heterogeneous abilities and showed how the suggested model can be used for processing the information of such a complex system as B2B negotiation.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.15 no.3
• /
• pp.119-127
• /
• 2020

The detection and recognition of abnormal driving becomes crucial for achieving safety in Intelligent Transportation Systems (ITS). This paper presents a feature extraction method based on spectral data to train a neural network model for driving behavior recognition. The proposed method uses a two stage signal processing approach to derive time-saving and efficient feature vectors. For the first stage, the feature vector set is obtained by calculating variances from each frequency bin containing the power spectrum data. The feature set is further reduced in the second stage where an intersection method is used to select more significant features that are finally applied for training a neural network model. A stream of live signals are fed to the trained model which recognizes the abnormal driving behaviors. The driving behaviors considered in this study are weaving, sudden braking and normal driving. The effectiveness of the proposed method is demonstrated by comparing with existing methods, which are Particle Swarm Optimization (PSO) and Convolution Neural Network (CNN). The experiments show that the proposed approach achieves satisfactory results with less computational complexity.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.18 no.5
• /
• pp.171-177
• /
• 2018

Recently there are many image datasets which has variety of data class and point to extract general features. But in order to this variety data class and point, deep learning model trained this dataset has not good performance in heterogeneous data feature local area. In this paper, we propose the structure which use sub-category and openset object detection methods to train more robust model, named multi-branch tree using ASSL. By using this structure, we can have more robust object detection deep learning model in heterogeneous data feature environment.

• Journal of Institute of Control, Robotics and Systems
• /
• v.3 no.5
• /
• pp.526-531
• /
• 1997

To control nonlinear chemical processes, a new neural linearizing control scheme is proposed. This is a hybrid of a radial basis function(RBF) network and a linear controller, thus the control action applied to the process is the sum of both control actions. Firstly, to train the RBF newtork a linear reference model is determined by analyzing the past operating data of the process. Then, the training of the RBF newtork is iteratively performed to minimize the difference between outputs of the process and the linear reference model. As a result, the apparent dynamics of the process added by the RBF newtork becomes similar to that of the linear reference model. After training, the original nonlinear control problem changes to a linear one, and the closed-loop control performance is improved by using the optimum tuning parameters of the linear controller for the linear dynamics. The proposed control scheme performs control and training simultaneously, and shows a good control performance for nonlinear chemical processes.

• Proceedings of the Korean Society for Noise and Vibration Engineering Conference
• /
• 2012.04a
• /
• pp.606-611
• /
• 2012

An analysis and design method for reducing aerodynamic noise in high-speed trains based on biomimetics of noiseless flight of owl is proposed. Wind tunnel testing and numerical CFD (Computational Fluid Dynamics) simulation for the basic inter-coach spacing model are carried out, and their results compared. To determine the effect of scaling of the owl's flight feather on the noise reduction, two-fold and a four-fold scaled up model of the feather are constructed, and the numerical simulations are carried out to obtain the aerodynamic noise levels for each scale. Original model is found to reduce the noise level by 10 dB, while two-fold increase in length dimensions reduces the noise by 12 dB. Validation of numerical solution using wind tunnel experimental measurements are presented as well.

• 한국신재생에너지학회:학술대회논문집
• /
• 2008.10a
• /
• pp.353-356
• /
• 2008

HEV(Hybrid Electrical Vehicle) is considered as one of the next generation vehicles. To develop the HEV, there must be a reliable simulator, by which the capacities of the power resources are tested, and the parameters of the HEV are optimized before developing the real model of the HEVs. This process can save the money for designing the HEV system and improve the system without experiments. Matlab Simulink is familiar to mechanical engineers and the program can simultaneously provide a system model and a controller in one program. Nowadays, the Simdriveline toolbox which is used for analysis a power-train system is applied to build a dynamic model for a HEV system. In this study, we make a HEV simulator with the Simdriveline toolbox and develop a controller. There are two simple strategies, applied to the controller. One strategy includes a power split ratio and a shift map which are created by user. Other strategy calculated an appropriate amount of resource's torque along specific results, and this is useful when users can't develop a fitting controller. The methodologies for configuring the simulator and its control system are presented in this paper.

• Journal of Electrical Engineering and Technology
• /
• v.1 no.4
• /
• pp.513-519
• /
• 2006

As a very important part in development of the protocol, verifications for developed protocol specification are complementary techniques that are used to increase the level of confidence in the system functions by their specifications. Using the informal method for specifying the protocol, some ambiguity may be contained therein. This indwelling ambiguity in control systems can cause the occurrence of accidents, especially in the case of safety-critical systems. To clear the vagueness contained in the designed protocol, we use the LTS (Labeled Transition System) model to design the protocol for railway signaling. And then, we verify the safety and the liveness properties formally through the model checking method. The modal ${\mu}$-calculus, which is an expressive method of temporal logic, has been applied to the model checking method. We verify the safety and liveness properties of Korean standard protocol for railway signaling systems. To perform automatic verification of the safety and liveness properties of the designed protocol, a communication verification tool is implemented. The developed tools are implemented by C++ language under Windows XP. It is expected to increase the safety and reliability of communication protocol for signaling systems by using the developed communication verification tool.

• Journal of Korean Society on Water Environment
• /
• v.24 no.1
• /
• pp.1-6
• /
• 2008

Biological and chemical sensors are the two most frequently used sensors to monitor the water resource. Chemical sensor is very accurate to pick up the types and to measure the concentration of the chemical substance. Drawback is that it works for just one type of chemical substance. Therefore a lot of expensive monitoring system needs to be installed to determine the safeness of the water, which costs too much expense. Biological sensor, on the contrary, can judge the degree of pollution of the water with just one monitoring system. However, it is not easy to figure out the type of contaminant with a biological sensor. In this study, an endeavor is made to identify the toxicant in the water using the shape of the chlorophyll fluorescence induction curve (FIC) from a biological monitoring system. Wem-tox values are calculated from the amount of flourescence of contaminated and reference water. Curve fitting is executed to find the representative curve of the raw data of Wem-tox values. Then the curves are digitalized at the same interval to train the neural network model. Taguchi method is used to optimize the neural network model parameters. The optimized model shows a good capacity to figure out the toxicant from FIC.