Jo May 22, 2026
Bio-identification by deep neural networks in such devices with limited computational power and memory capacity as mobile phones has become an essential but challenging task today. As faces have rather invariable features among human biometric features, face recognition is considered as the most important biometric identification task. Face recognition has been widely used for user authentication in security systems such as electronic payment systems as it can identify faces from facial images from photographs or videos, and it has been studied for years.
The face recognition system using convolutional neural networks is considered as the best method among the existing ones. Face recognition network models that have been developed recently and have proved to be superior in performance cannot be used for real-time face recognition in devices with limited computational resources such as low base computers or mobile phones because their structure is very complex and they need a large amount of computation. What is more, reducing the number of layers continuously to reduce computational burden affects recognition performance.
In previous studies, several methods to improve the trade-off between speed and recognition performance were proposed. One of them is GhostFaceNets which uses Ghost module to reduce the feature map redundancy, where the trade-off between speed and recognition performance is improved by extracting less repetitive feature maps with small amount of computation. In GhostFaceNets, they improved the trade-off between speed and accuracy by performing the attention operation using a DFC (decoupled fully-connected) attention. However, the DFC attention has limitations in capturing wide spatial information, which may lead to the degradation of recognition performance.
Jo Kwang Chol, a researcher at the Institute of Information Technology, has designed a network structure with low computational cost and improved performance by combining the self-attention module with the extended Ghost module based on the backbone of GhostFaceNets, and verified its accuracy using international standard databases.
The results showed that the proposed network model brings significant improvement in face recognition performance with 99.74% in LFW and 97.7% in AgeDB-30 and that with 42 MFLOP, it can support stable real-time face recognition in embedded devices.
For more details, you can refer to his paper “GhostFormerNet: A Lightweight Face Recognition Method based on Extended Ghost Module and Self-Attention” in “2025 International Conference on Graphics and Signal Processing” (EI).
...
Jo May 21, 2026
Rare earth minerals including pyrochlore contain uranium, thorium and rare earths, and the processing of these rare earth ores must result in a mixed solution containing uranium, thorium and rare earths.
Many studies have been carried out to determine the concentration of individual components in the mixtures of uranium, thorium and rare earth, and thus, various instrumental methods have been widely used. However, these methods have disadvantages such as the need for special analytical tools, the need for sample solidification and the high cost of analysers concerned.
Due to the similar spectroscopic properties of uranium, thorium and rare earths, some analytical methods have been developed to separate uranium, thorium and rare earths and determine the concentration of individual components using various separation methods including extraction and ion exchange. However, these methods have other disadvantages such as long analysis time and complicated operation.
Kwon Myong Gang, a researcher at the Institute of Analysis, has investigated analytical methods for simultaneously determining individual components in a mixture of uranium, thorium and cerium obtained during the pyrochlore hydrometallurgical process, using spectrophotometric methods.
First, he proved that the maximum absorption wavelengths of uranium, thorium and cerium in three different solutions (3, 0.1mol/L HCl, 1mol/L CH3COOH) were 660, 652 and 660nm, respectively. Then, he used the relationship between the concentration of individual components and the absorbance to determine the absorption coefficients of individual components in different solutions. After that, based on the principle of absorbance additivity, he established a simulation equation between absorbance and concentration and used it to determine the concentration of individual components in the mixture solution. The concentration limit within which the additivity of absorbance in the mixed solution is established was less than 2mg/L of uranium and cerium, and 1.5mg/L of thorium. The error of analysis was less than 3%.
For further details, you can refer to his paper “Simultaneous Spectrophotometric Analysis of Uranium, Thorium, Cerium During Pyrochlore Hydrometallurgical Process” in “Proceedings of KUTIC-2025”.
...
Jo May 20, 2026
At present, in rare earth production, double salt precipitation is widely used for separation of rare and non-rare earth elements, and separation of quaternary cerium and trivalent rare earth elements. Rare earth double sulphates are commonly treated by hydroxide, but the filtration is difficult because the rare earth hydroxide obtained during the hydroxide conversion is an amorphous precipitate. In contrast, if RE double sulphates are converted into carbonates, filtration and dissolution by acids become easier because RE carbonates are crystalline precipitates.
Ri Sun Chol, a researcher at the Faculty of Chemistry, conducted a study to convert rare earth double sulphates obtained from monazite sulphate leachates to rare earth carbonates using sodium carbonate.
Aiming to improve the conversion of rare earths and to effectively separate U and Th from RE carbonates, he investigated the influence of reaction parameters using Taguchi-Grey Relation Analysis (Taguchi-GRA), determined the optimum conditions and studied the separation of U and Th.
When reaction temperature is 80℃, reaction time 2h, additive amount of sodium carbonate 1.5 times the theoretical amount, the ratio of solid to liquid 2:1 and stirring speed 200r/min, the RE carbonate conversion was 97.1% and the U and thorium removal 80.8% and 0.3%, respectively.
For more information, please refer to his paper “Study on the Carbonate Conversion of Rare Earth Double Sulfates” in “Proceedings of KUTIC-2025”.
...
Jo May 19, 2026
Compared to monochannel membranes, multichannel membranes have the advantage of being able to generate higher permeate flux per unit volume of membrane elements by providing higher mechanical strength and larger filtration area within a given volume. However, the permeability of multichannel membranes is not directly proportional to the filtration area because the contributions of the middle and central channels and wall channels to the permeate flux are not the same. Since the permeated flow through the inner channels is not drained well towards the outer surface, proper placement of exit channels can increase permeate flux.
Through previous studies, it is found that permeate flux is mainly controlled by the permeability ratio of skin layer to porous support, and the geometry, and that proper arrangement of exit channels can increase permeate flux.
Kim Un Ok, a researcher at the Faculty of Applied Mathematics, has investigated the effect of exit channel placement on permeate flow in square 64-channel ceramic membranes which are now in wide use, and determined proper placement for various cases.
She assumed permeate flow to be two-dimensional potential flow and solved the mathematical model by finite element method.
The numerical simulation results show that the placement of exit channels affects permeate flux and it is possible to determine reasonable placement of exit channels according to the permeability ratio of skin layer to porous support.
For more information, you can refer to her paper “Effect of Exit Channel Placement on Permeate Flow in Square 64-Channel Ceramic Membrane” in “Proceedings of KUTIC-2025”.
...
Jo May 18, 2026
Several computational methods have been applied to the research to develop new materials. For example, according to the considered scale, they are divided into finite element methods, Monte Carlo methods, molecular dynamics methods, etc. In addition, the experimental data analysis methods include experimental design, neural network, genetic algorithm, etc., which require high-performance computing devices, long computational time and a considerable amount of experimental analysis.
In contrast, grey models are widely used in data analysis as they can improve prediction accuracy from a small amount of poor data. Grey models are mainly applied in various fields such as energy consumption and prediction and CO2 gas emission, attracting a great deal of interest of many researchers. This has led to active research into grey models but few of the results have been applied to the study of material properties.
Pang Chol Ho, a researcher at the Faculty of Material Science and Technology, proposed a hybrid exponential smoothing method, developed a grey model combined with a structural adaptive discrete grey Bernoulli model, and predicted some properties of material.
The comparison analysis with other grey models showed that the proposed model has the highest predictive accuracy.
He used the proposed model to predict various properties of material. The results showed that the predictive accuracy (MAPE) was 0.007 65 for tensile strength, 0.016 52 for Branell hardness and 0.025 15 for the thermoelectric performance parameters of high-entropy alloy AgSnSbSe1.5. This means that the proposed model is effective for predicting material properties.
You can find more information in his paper “Application of Hybrid Grey Bernoulli Model in Material Research” in “Proceedings of KUTIC-2025”.
...
Jo May 17, 2026
Selective laser melting (SLM) is the most popular AM technique that enables production of complicated metal products with high precision, flexibility, acceptable surface finish and material efficiency by melting metal powders using laser energy.
Since SLM process parameters affect the multiple quality attributes of SLM parts, it is very important to perform optimization and effect assessment of SLM process parameters. AlSi10Mg alloy is widely used in the industrial fields such as automotive and aerospace industries, and it is suitable for SLM processes due to its good castability, good strength, lower thermal expansion coefficient, and excellent wear and corrosion resistance.
Yang Ji Yon, a post-graduate student at the Faculty of Material Science and Technology, performed an analysis of the relationship between the properties (tensile strength, hardness and relative density) and process parameters (laser power LP, scan speed SS, and overlap rate OR) of SLM-manufactured AlSi10Mg alloy using Taguchi and TOPSIS methods.
For the individual properties, the optimal process parameters were LP 320-360W, SS 600mm/s and OR 35%, and the effect rankings varied according to the properties. For the TOPSIS-based overall quality index, the optimal process parameters were LP 320W, SS 600mm/s and OR 35%, and the effect ranking was SS, OR and LP.
For further details, you can refer to her paper “Relationship between Properties and Process Parameters for AlSi10Mg Alloy Manufactured by Selective Laser Melting” in “Proceedings of KUTIC-2025”.
...