• International Journal of Korean Welding Society
• /
• v.1 no.1
• /
• pp.31-38
• /
• 2001

For the horizontal fillet welding with one plate in a vertical position, there will be a higher tendency of weld metal falling down rather than for the butt-welding in flat position. Such phenomenon could bring about the overlap or deflection of weld pool, and consequently induce the poor mechanical strength of weldments. Therefore, a precise position control of welding torch in conjunction with the weld qualify plays an important role in welding robot applications. In the present study, an experimental method was proposed for deriving a mathematical model between the leg length and the welding conditions. Finally, an algorithm was proposed for weld seam tracking and improvement of the weld quality. The reliability of the proposed algorithm was evaluated through various experiments, which showed that the proposed algorithm can be very effective for tracking the weld line and simultaneously achieving the sound weld bead.

• Journal of the Korea Society of Computer and Information
• /
• v.23 no.10
• /
• pp.23-31
• /
• 2018

Recently, there have been many studies on machine learning. Among them, studies on reinforcement learning are actively worked. In this study, we propose a controller to control bicycle using DDPG (Deep Deterministic Policy Gradient) algorithm which is the latest deep reinforcement learning method. In this paper, we redefine the compensation function of bicycle dynamics and neural network to learn agents. When using the proposed method for data learning and control, it is possible to perform the function of not allowing the bicycle to fall over and reach the further given destination unlike the existing method. For the performance evaluation, we have experimented that the proposed algorithm works in various environments such as fixed speed, random, target point, and not determined. Finally, as a result, it is confirmed that the proposed algorithm shows better performance than the conventional neural network algorithms NAF and PPO.

• Proceedings of the KIEE Conference
• /
• 2005.07b
• /
• pp.1086-1088
• /
• 2005

It is necessary to modify the state-of-the-art of vector control theory, because of the phase asymmetry(static end effect) in the Linear Induction Motor(LIM). The dynamic characteristic analysis method of the vector controlled LIM using coupled FEM and control algorithm taking into account the movement is proposed. The focus of this paper is the analysis relative to phenomena of static end effect due to the motor structure. To prove the propriety of the proposed method, the Digital Signal Processor(DSP) installed experimental devices are equipped and the experiment is performed.

• Proceedings of the Korean Society for Noise and Vibration Engineering Conference
• /
• 2013.10a
• /
• pp.346-353
• /
• 2013

Industrial processes need to be monitored in real-time based on the input-output data observed during their operation. Abnormalities in an induction motor should be detected early in order to avoid costly breakdowns. To early identify induction motor faults, this paper effectively estimates spectral envelopes of each induction motor fault by utilizing a linear prediction coding (LPC) analysis technique and an expectation maximization (EM) algorithm. Moreover, this paper classifies induction motor faults into their corresponding categories by calculating Mahalanobis distance using the estimated spectral envelopes and finding the minimum distance. Experimental results shows that the proposed approach yields higher classification accuracies than the state-of-the-art approach for both noiseless and noisy environments for identifying the induction motor faults.

• Journal of Information Processing Systems
• /
• v.10 no.2
• /
• pp.256-270
• /
• 2014

A new morphed steganographic algorithm is proposed in this paper. Image security is a challenging problem these days. Steganography is a method of hiding secret data in cover media. The Least Significant Bit is a standard Steganographic method that has some limitations. The limitations are less capacity to hide data, poor stego image quality, and imperceptibility. The proposed algorithm focuses on these limitations. The morphing concept is being used for image steganography to overcome these limitations. The PSNR and standard deviation are considered as a measure to improve stego image quality and morphed image selection, respectively. The stego keys are generated during the morphed steganographic embedding and extracting process. Stego keys are used to embed and extract the secret image. The experimental results, which are based on hiding capacity and PSNR, are presented in this paper. Our research contributes towards creating an improved steganographic method using image morphing. The experimental result indicates that the proposed algorithm achieves an increase in hiding capacity, stego image quality, and imperceptibility. The experimental results were compared with state of the art steganographic methods.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.14 no.4
• /
• pp.197-205
• /
• 2019

In this paper, an efficient single dehazing algorithm is proposed based on linear transformation by assuming that a linear relationship exists in saturation component between the haze image and haze-free image. First, we analyze the linearity of saturation channel, estimate the medium transmission map in terms of the saturation component. Then, the intensity of haze-free image is assumed by using CLAHE to enhance contrast of haze image. Experimental results demonstrate that proposed algorithm can naturally recover the image, especially can remove color distortion caused by conventional methods. Therefore, our approach is competitive with other state-of-the art single dehazing methods.

• Electronics and Telecommunications Trends
• /
• v.37 no.1
• /
• pp.53-62
• /
• 2022

With increasing size of transformer-based neural networks, a light-weight algorithm and efficient AI accelerator has been developed to train these huge networks in practical design time. In this article, we present a survey of state-of-the-art research on the low-precision computational algorithms especially for floating-point formats and their hardware accelerator. We describe the trends by focusing on the work of two leading research groups-IBM and Seoul National University-which have deep knowledge in both AI algorithm and hardware architecture. For the low-precision algorithm, we summarize two efficient floating-point formats (hybrid FP8 and radix-4 FP4) with accuracy-preserving algorithms for training on the main research stream. Moreover, we describe the AI processor architecture supporting the low-bit mixed precision computing unit including the integer engine.

• Nuclear Engineering and Technology
• /
• v.54 no.5
• /
• pp.1747-1753
• /
• 2022

Compton imaging is the main method for locating radioactive hot spots emitting high-energy gamma-ray photons. In particular, this imaging method is crucial when the photon energy is too high for coded-mask aperture imaging methods to be effective or when a large field of view is required. Reconstruction of the photon source requires advanced Compton event processing algorithms to determine the exact position of the source. In this study, we introduce a novel method based on a Deep Learning algorithm with a Convolutional Neural Network (CNN) to perform Compton imaging. This algorithm is trained on simulated data and tested on real data acquired with Caliste, a single planar CdTe pixelated detector. We show that performance in terms of source location accuracy is equivalent to state-of-the-art algorithms, while computation time is significantly reduced and sensitivity is improved by a factor of ~5 in the Caliste configuration.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.17 no.4
• /
• pp.209-215
• /
• 2022

This paper proposes a novel approach for training an unsupervised depth estimation algorithm. The objective of unsupervised depth estimation is to estimate pixel-wise distances from camera without external supervision. While most previous works focus on model architectures, loss functions, and masking methods for considering dynamic objects, this paper focuses on the training framework to effectively use depth cue. The main loss function of unsupervised depth estimation algorithms is known as the photometric error. In this paper, we claim that direct depth cue is more effective than the photometric error. To obtain the direct depth cue, we adopt the technique of knowledge distillation which is a teacher-student learning framework. We train a teacher network based on a previous unsupervised method, and its depth predictions are utilized as pseudo labels. The pseudo labels are employed to train a student network. In experiments, our proposed algorithm shows a comparable performance with the state-of-the-art algorithm, and we demonstrate that our teacher-student framework is effective in the problem of unsupervised depth estimation.

• Advances in Computational Design
• /
• v.8 no.1
• /
• pp.37-59
• /
• 2023

Squat reinforced concrete (SRC) shear walls are a critical part of the structure for both office/residential buildings and nuclear structures due to their significant role in withstanding seismic loads. Despite this, empirical formulae in current design standards and published studies demonstrate a considerable disparity in predicting SRC wall shear strength. The goal of this research is to develop and evaluate hybrid and ensemble artificial neural network (ANN) models. State-of-the-art population-based algorithms are used in this research for hybrid intelligence algorithms. Six models are developed, including Honey Badger Algorithm (HBA) with ANN (HBA-ANN), Hunger Games Search with ANN (HGS-ANN), fitness-distance balance coyote optimization algorithm (FDB-COA) with ANN (FDB-COA-ANN), Averaging Ensemble (AE) neural network, Snapshot Ensemble (SE) neural network, and Stacked Generalization (SG) ensemble neural network. A total of 434 test results of SRC walls is utilized to train and assess the models. The results reveal that the SG model not only minimizes prediction variance but also produces predictions (with R²= 0.99) that are superior to other models.