• The Transactions of the Korean Institute of Electrical Engineers P
• /
• v.58 no.4
• /
• pp.385-390
• /
• 2009

This paper presents a novel active noise control (ANC) system using least mean square (LMS) algorithm and neural network approach for decreasing acoustic noise signals inside high-speed trains. We construct a LMS framework as a nominal ANC system and additionally design an artificial single-layered perceptron model as an auxiliary ANC which is aimed to reduce real-time residuary noise due to its nonstationary and uncertain nature. Parameter vector of the hybrid ANC is determined through online estimation to realize an adaptive ANC configuration by means of the steepest descent algorithm. We achieve simulation experiment to demonstrate the proposed ANC system employing realistic acoustic noise signals measured in Korea Train eXpress (KTX).

• Proceedings of the IEEK Conference
• /
• summer
• /
• pp.371-375
• /
• 2004

Fault diagnosis is one of the most studied problems in process engineering. Recently, great research interest has been devoted to approaches that use classification methods to detect faults. This paper presents an application of a newly developed classification method - support vector machines - for fault diagnosis in an industrial case. A real set of operation data of a motor pump was used to train and test the support vector machines. The experiment results show that the support vector machines give higher correct detection rate of faults in comparison to rule-based diagnostics. In addition, the studied method can work with fewer training instances, what is important for online diagnostics.

• Proceedings of the Korea Society for Simulation Conference
• /
• 2002.11a
• /
• pp.101-105
• /
• 2002

In a traditional hierarchical inventory system, direct orders are the only information for inventory management that is exchanged between the firms involved. But due to the rapid development of modern information technology, it becomes possible for the firms to share more information in real time, e.g. demand and inventory status data. And so the term Supply Chain has emerged because it is seen as an important source of competitive advantage. Now it is possible to challenge traditional approaches to inventory management. In the past, one of the de-facto assumptions for inventory management was that the demand pattern follows a specific distribution function. However, it is undesirable to apply this assumption in real situations because the demand information in the supply chain tends to be distorted due to the bullwhip effect in a supply chain. To overcome this weakness, we propose a new solution method using NN (Neural Network). Our method proceeds in three steps. First, we find the patterns of optimal reorder points by analyzing past data. Second. train the NN using these pattern data and finally decide the reorder point. Using simulation experiment, we show that the proposed solution method gives better result than that of traditional research.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2009.01a
• /
• pp.803-807
• /
• 2009

The bobsleigh is a winter sport which use a sled to slide down an ice-covered course. There is a big expectation for having a training environment and being able to train year round. At present, training is very limited due to the season or course facilities. A variety of VR (Virtual Reality) equipment has been developed in recent years, and it is beginning to spread. We have also made our contribution in bobsleigh simulation. The reactive force applied in our bobsleigh simulation is much smaller than that of a real bobsleigh. This paper proposes a method to enhance reactive force of bobsleigh simulation in real time. The reactive force is magnified instantly in the physically-based simulation. The Laplacian filter is applied to the sequence of reactive force, this technique is often used in the field of image processing. The simulation is comprised of four large scale surround screens and a 6-D.O.F. (Degree Of Freedom) motion system. We also conducted an experiment with several motion patterns to evaluate the effectiveness of enhancement. The experimental results proved useful in some cases.

• Current Optics and Photonics
• /
• v.1 no.3
• /
• pp.196-202
• /
• 2017

Corona discharge is always a sign of failure processes of high-voltage electrical apparatus, including those utilized in electric railway systems. Solar-blind ultraviolet (UV) cameras are effective tools for corona inspection. In this work, we present an automatic railway corona-discharge detection system based on solar-blind ultraviolet detection. The UV camera, mounted on top of a train, inspects the electrical apparatus, including transmission lines and insulators, along the railway during fast cruising of the train. An algorithm based on the Hough transform is proposed for distinguishing the emitting objects (corona discharge) from the noise. The detection system can report the suspected corona discharge in real time during fast cruises. An experiment was carried out during a routine inspection of railway apparatus in Xinjiang Province, China. Several corona-discharge points were found along the railway. The false-alarm rate was controlled to less than one time per hour during this inspection.

• Journal of Korea Multimedia Society
• /
• v.22 no.12
• /
• pp.1447-1456
• /
• 2019

As artificial intelligence technology is actively used in image processing, it is possible to generate high-quality fake images based on deep learning. Fake images generated using GAN(Generative Adversarial Network), one of unsupervised learning algorithms, have reached levels that are hard to discriminate from the naked eye. Detecting these fake images is required as they can be abused for crimes such as illegal content production, identity fraud and defamation. In this paper, we develop a deep-learning model based on CNN(Convolutional Neural Network) for the detection of StyleGAN fake images. StyleGAN is one of GAN algorithms and has an excellent performance in generating face images. We experiment with 48 number of experimental scenarios developed by combining parameters of the proposed model. We train and test each scenario with 300,000 number of real and fake face images in order to present a model parameter that improves performance in the detection of fake faces.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.41 no.6
• /
• pp.567-573
• /
• 2017

The drive shaft of passenger vehicle has an important role in transmitting the torque between the power train system and the wheels. Torsional fatigue failures occur generally in the connection parts of the spline edge of the drive shaft, when there is significant fatigue damage under repeated twisting loads. A heat treatment, an induction hardening process, has been adopted to increase the torsional strength as well as the fatigue life of the drive shaft. However, it is still unclear how the extension of the induction hardening process in a used material relates to its shear-strain fatigue life range. In this study, a shear-strain controlled torsional-fatigue test with a specially designed specimen was conducted by an electro-dynamic torsional fatigue test machine. A finite element analysis of the drive shaft was carried out using the results obtained by the fatigue experiment. The estimated fatigue life was verified through a twisting load test of the real drive shaft in a test rig.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.5
• /
• pp.55-67
• /
• 2022

Sensor data can provide fault diagnosis for equipment. However, the cause analysis for fault results of equipment is not often provided. In this study, we propose an explainable convolutional neural network framework for the sensor-based time series classification model. We used sensor-based time series dataset, acquired from vehicles equipped with sensors, and the Wafer dataset, acquired from manufacturing process. Moreover, we used Cycle Signal dataset, acquired from real world mechanical equipment, and for Data augmentation methods, scaling and jittering were used to train our deep learning models. In addition, our proposed classification models are convolutional neural network based models, FCN, 1D-CNN, and ResNet, to compare evaluations for each model. Our experimental results show that the ResNet provides promising results in the context of time series classification with accuracy and F1 Score reaching 95%, improved by 3% compared to the previous study. Furthermore, we propose XAI methods, Class Activation Map and Layer Visualization, to interpret the experiment result. XAI methods can visualize the time series interval that shows important factors for sensor data classification.

• Journal of Broadcast Engineering
• /
• v.28 no.3
• /
• pp.302-313
• /
• 2023

This paper aims to develop a framework that can fully automate the quality management of training data used in large-scale Artificial Intelligence (AI) models built by the Ministry of Science and ICT (MSIT) in the 'AI Hub Data Dam' project, which has invested more than 1 trillion won since 2017. Autonomous driving technology using AI has achieved excellent performance through many studies, but it requires a large amount of high-quality data to train the model. Moreover, it is still difficult for humans to directly inspect the processed data and prove it is valid, and a model trained with erroneous data can cause fatal problems in real life. This paper presents a dataset reconstruction framework that removes abnormal data from the constructed dataset and introduces strategies to improve the performance of AI models by reconstructing them into a reliable dataset to increase the efficiency of model training. The framework's validity was verified through an experiment on the autonomous driving dataset published through the AI Hub of the National Information Society Agency (NIA). As a result, it was confirmed that it could be rebuilt as a reliable dataset from which abnormal data has been removed.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.18 no.2
• /
• pp.231-240
• /
• 2023

It remains a challenge for robots to learn avoiding obstacles automatically in path planning using deep reinforcement learning (DRL). More and more researchers use DRL to train a robot in a simulated environment and verify the possibility of DRL to achieve automatic obstacle avoidance. Due to the influence factors of different environments robots and sensors, it is rare to realize automatic obstacle avoidance of robots in real scenarios. In order to learn automatic path planning by avoiding obstacles in the actual scene we designed a simple Testbed with the wall and the obstacle and had a camera on the robot. The robot's goal is to get from the start point to the end point without hitting the wall as soon as possible. For the robot to learn to avoid the wall and obstacle we propose to use the double deep Q networks (DDQN) to verify the possibility of DRL in automatic obstacle avoidance. In the experiment the robot used is Jetbot, and it can be applied to some robot task scenarios that require obstacle avoidance in automated path planning.