• Geomechanics and Engineering
• /
• v.8 no.5
• /
• pp.663-674
• /
• 2015

Many seismically vulnerable regions in India and worldwide are located on deep soil deposits which extend to several hundred meters of depth. It has been well recognized that the earthquake shaking is altered by geological conditions at the location of building. As seismic waves propagates through uppermost layers of soil and rock, these layers serve as filter and they can increase the duration and amplitude of earthquake motion within narrow frequency bands. The amplification of these waves is largely controlled by mechanical properties of these layers, which are function of their stiffness and damping. Stiffness and damping are further influenced by soil type and thickness. In the current study, an attempt has been made to study the seismic site response of deep soils. Three hypothetical homogeneous soil models (e.g., soft soil, medium soil and hard soil) lying on bedrock are considered. Depth of half space is varied from 30 m to 2,000 m in this study. Controlled synthetic motions are used as input base motion. One dimensional equivalent linear ground response analyses are carried out using a computer package DEEPSOIL. Conventional approach of analysing up to 30 m depth has been found to be inadequate for deep soil sites. PGA values are observed to be higher for deeper soil profiles as compared to shallow soil profiles indicating that deeper soil profiles are more prone to liquefaction and other related seismic hazards under earthquake ground shaking. The study recommends to deal the deeper soil sections more carefully for estimating the amplification factors for seismic hazard assessment at the surface.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.14 no.2
• /
• pp.121-136
• /
• 1989

In this paper, an attempt has been made to analyze the theoretically and verify experimentally the effect of curvature on the radiation characteristics of microstrip array antennas mounted conformally on the concave surface and the convex surface of the cylindrical body. The analysis of single element microstrip antenna is made by using the analysis method of Transmission Line Model. The theory of array antennas is established by application of the method of transformed coordinates, in which the translation and the ratation about each single element arrayed two-demensionally on the nonplanar surface are under consideration, and it is investigated by computation of the synthetic electric field strength in the far zone. In addition, various radiation characteristics, such as return loss, resonant frequency, radiation pattern, half-power, beamwidth, gain, are measrued and compared with the theroetical values according to the variation of curvature, by designing and building 4-element array microstrip antenna operating at 10 GHz, and microstrip feed lines. As predicted in theroy, it is verified that radiation pattern of antennas mounted on the concave and the convex surfaces alike broadens as the radius of curvature decreases. And for the curved surfaces, aggrement between computed values of the total synthetic radiation power pattern by the method of transformed coordinates and measured valuse is good. Besides, it is found that resonant frequency, input impedance and gain are hardly affected by the radius of curvature.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.21 no.3
• /
• pp.307-316
• /
• 2011

The majority of maritime accidents happened by human factor. For that reason, navigation experts want to an intelligent support system for navigation safety, without officer involvement. The expert system which is one of artificial intelligence skills for navigation support is an important tool that a machine can substitute for an expert through the design of a knowledge base and inference engine using the experience or knowledge of an expert. Further, in the real world, a complex situation requires synthetic estimation with the input of experts in various fields for the correct estimation of the situation, not any one expert. In particular, synthetic estimation is more important for navigation situations than in other cases, because of diverse potential threats. This paper presents the method of knowledge fusion pertaining to navigation safety knowledge from various expert systems, using a blackboard system. Then we will show the validity of the method via a design and implementation of test system effort.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.25 no.1
• /
• pp.75-80
• /
• 2021

Positioning technology is performing important functions in augmented reality, smart factory, and autonomous driving. Among the positioning techniques, the positioning method using beacons has been considered a challenging task due to the deviation of the RSSI value. In this study, the position of a moving object is predicted by training a neural network that takes the RSSI value of the receiver as an input and the distance as the target value. To do this, the measured distance versus RSSI was collected. A neural network was introduced to create synthetic data from the collected actual data. Based on this neural network, the RSSI value versus distance was predicted. The real value of RSSI was obtained as a neural network for generating synthetic data, and based on this value, the coordinates of the object were estimated by learning a neural network that tracks the location of a terminal in a virtual environment.

• The Journal of the Convergence on Culture Technology
• /
• v.10 no.2
• /
• pp.465-470
• /
• 2024

The advancement of AI-based image generation technology has resulted in the creation of various images, emphasizing the need for technology capable of accurately discerning them. The amount of generated image data is limited, and to achieve high performance with a limited dataset, this study proposes a model for discriminating generated images using transfer learning. Applying pre-trained models from the ImageNet dataset directly to the CIFAKE input dataset, we reduce training time cost followed by adding three hidden layers and one output layer to fine-tune the model. The modeling results revealed an improvement in the performance of the model when adjusting the final layer. Using transfer learning and then adjusting layers close to the output layer, small image data-related accuracy issues can be reduced and generated images can be classified.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.41 no.2
• /
• pp.23-32
• /
• 2004

The query optimizer's important task while a query is invoked is to estimate the fraction of records in the databases that satisfy the given query condition. The query result size estimation in spatial databases, like relational databases, proceeds to partition the whole input into a small number of subsets called “buckets” and then estimate the fraction of the input in the buckets. The accuracy of estimation is determined by the difference between the real data counts and approximations in the buckets, and is dependent on how to partition the buckets. Existing techniques for spatial databases are equi-area and equi-count techniques, which are respectively analogous in relation databases to equi-height histogram that divides the input value range into buckets of equal size and equi-depth histogram that is equal to the number of records within each bucket. In this paper we propose a new partitioning technique that determines buckets according to the maximal difference of area which is defined as the product of data ranges End frequencies of input. In this new technique we consider both data values and frequencies of input data simultaneously, and thus achieve substantial improvements in accuracy over existing approaches. We present a detailed experimental study of the accuracy of query result size estimation comparing the proposed technique and the existing techniques using synthetic as well as real-life datasets. Experiments confirm that our proposed techniques offer better accuracy in query result size estimation than the existing techniques for space query size, bucket number, data number and data size.

• Ecology and Resilient Infrastructure
• /
• v.10 no.4
• /
• pp.107-115
• /
• 2023

High turbidity in source water can have adverse effects on water treatment plant operations and aquatic ecosystems, necessitating turbidity management. Consequently, research aimed at predicting river turbidity continues. This study developed a multi-class classification model for prediction of turbidity using LightGBM (Light Gradient Boosting Machine), a representative ensemble machine learning algorithm. The model utilized data that was classified into four classes ranging from 1 to 4 based on turbidity, from low to high. The number of input data points used for analysis varied among classes, with 945, 763, 95, and 25 data points for classes 1 to 4, respectively. The developed model exhibited precisions of 0.85, 0.71, 0.26, and 0.30, as well as recalls of 0.82, 0.76, 0.19, and 0.60 for classes 1 to 4, respectively. The model tended to perform less effectively in the minority classes due to the limited data available for these classes. To address data imbalance, the SMOTE (Synthetic Minority Over-sampling Technique) algorithm was applied, resulting in improved model performance. For classes 1 to 4, the Precision and Recall of the improved model were 0.88, 0.71, 0.26, 0.25 and 0.79, 0.76, 0.38, 0.60, respectively. This demonstrated that alleviating data imbalance led to a significant enhancement in Recall of the model. Furthermore, to analyze the impact of differences in input data composition addressing the input data imbalance, input data was constructed with various ratios for each class, and the model performances were compared. The results indicate that an appropriate composition ratio for model input data improves the performance of the machine learning model.

• Journal of the Korean Society for Aeronautical & Space Sciences
• /
• v.41 no.8
• /
• pp.639-648
• /
• 2013

This paper describes on the NOx/CO emission characteristics, temperature characteristics and flame structures when firing coal derived synthetic gas especially for gases of Buggenum and Taean IGCC. These combustion characteristics were observed by conducting ambient-pressure elevated-temperature combustion tests in GE7EA model combustor when varying heat input and nitrogen dilution ratio. Nitrogen addition caused decrement in adiabatic flame temperature, thus resulting in the NOx reduction. At low heat input condition, nitrogen dilution raised the CO emission dramatically due to incomplete combustion. These NOx reduction and CO arising phenomena were observed at certain flame temperature of $1500^{\circ}C$ and $1250^{\circ}C$, respectively. As increasing nitrogen dilution, adiabatic flame temperature and combustor liner temperature were decreased and singular points were detected due to change in flame structure such as flame lifting. From the results, the effect of nitrogen dilution on the NOx/CO and flame structure was examined, and the test data will be utilized as a reference to achieve optimal operating condition of the Taean IGCC demonstration plant.

• Korean Journal of Cognitive Science
• /
• v.12 no.1_2
• /
• pp.77-89
• /
• 2001

Applying perceptual hierarchy of facial feature points, a neural network model for recognizing facial expressions was designed. Input data were convolution values of 150 facial expression pictures by Gabor-filters of 5 different sizes and 8 different orientations for each of 39 mesh points defined by MPEG-4 SNHC (Synthetic/Natural Hybrid Coding). A set of multiple regression analyses was performed with the rating value of the affective states for each facial expression and the Gabor-filtered values of 39 feature points. The results show that the pleasure-displeasure dimension of affective states is mainly related to the feature points around the mouth and the eyebrows, while a arousal-sleep dimension is closely related to the feature points around eyes. For the filter sizes. the affective states were found to be mostly related to the low spatial frequency. and for the filter orientations. the oblique orientations. An optimized neural network model was designed on the basis of these results by reducing original 1560(39x5x8) input elements to 400(25x2x8) The optimized model could predict human affective rating values. up to the correlation value of 0.886 for the pleasure-displeasure, and 0.631 for the arousal-sleep. Mapping the results of the optimized model to the six basic emotional categories (happy, sad, fear, angry, surprised, disgusted) fit 74% of human responses. Results of this study imply that, using human principles of recognizing facial expressions, a system for recognizing facial expressions can be optimized even with a a relatively little amount of information.

• Journal of Broadcast Engineering
• /
• v.26 no.4
• /
• pp.453-462
• /
• 2021

A lot of license plate data is required for car number recognition. License plate data needs to be balanced from past license plates to the latest license plates. However, it is difficult to obtain data from the actual past license plate to the latest ones. In order to solve this problem, a license plate recognition study through deep learning is being conducted by creating a synthetic license plates. Since the synthetic data have differences from real data, and various data augmentation techniques are used to solve these problems. Existing data augmentation simply used methods such as brightness, rotation, affine transformation, blur, and noise. In this paper, we apply a style transformation method that transforms synthetic data into real-world data styles with data augmentation methods. In addition, real license plate data are noisy when it is captured from a distance and under the dark environment. If we simply recognize characters with input data, chances of misrecognition are high. To improve character recognition, in this paper, we applied the DeblurGANv2 method as a quality improvement method for character recognition, increasing the accuracy of license plate recognition. The method of deep learning for license plate detection and license plate number recognition used YOLO-V5. To determine the performance of the synthetic license plate data, we construct a test set by collecting our own secured license plates. License plate detection without style conversion recorded 0.614 mAP. As a result of applying the style transformation, we confirm that the license plate detection performance was improved by recording 0.679mAP. In addition, the successul detection rate without image enhancement was 0.872, and the detection rate was 0.915 after image enhancement, confirming that the performance improved.