• Journal of the Korean Ceramic Society
• /
• v.31 no.8
• /
• pp.879-885
• /
• 1994

Coprecipitation technique was used to synthesize ZrO2+12 mol% CeO2 powders with ZrOCl2.8H2O and Ce(NO3)3.6H2O as starting materials. The powders were dried on different conditions such as distilled water, ethanol, and azeotropic distillation. The powders prepared by azeotropic distillation showed weak aggregation of particles and the average particle size of powders calcined at 85$0^{\circ}C$ for 1 hour was 0.19 ${\mu}{\textrm}{m}$. The optimum sintering temperature and holding time are 130$0^{\circ}C$ and 2.5~10 hours, respectively. Beyond the optimum conditions, a phase transition from tetragonal to monoclinic causes to produce cracks in the sintered bodies and to decrease the density.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.16 no.4
• /
• pp.145-151
• /
• 2021

In this paper, we propose a novel visualization technique to explain the predictions of deep neural networks. We use knowledge distillation (KD) to identify the interior of a black-box model for which we know only inputs and outputs. The information of the black box model will be transferred to a white box model that we aim to create through the KD. The white box model will learn the representation of the black-box model. Second, the white-box model generates attention maps for each of its layers using Grad-CAM. Then we combine the attention maps of different layers using the pixel-wise summation to generate a final saliency map that contains information from all layers of the model. The experiments show that the proposed technique found important layers and explained which part of the input is important. Saliency maps generated by the proposed technique performed better than those of Grad-CAM in deletion game.

• Journal of Radiation Protection and Research
• /
• v.33 no.1
• /
• pp.35-40
• /
• 2008

A distillation tower is one of the important facilities which separates and refines a crude oil stream according to certain boiling points. Its operation efficiency can affect the productivity of a refinery substantially. The objective of this study is to elucidate some operational information on the internal conditions of a distillation tower from a measurement of density profile by using a sealed gamma-ray source and a radiation detector. Gamma radiation counts were measured by a BGO detector positioned diametrically outside the tower-wall, opposite to the gamma source(Co-60) as the detector and the source were lowered concurrently. From the results, structural abnormality of the trays was not found inside the tower. Considering the flow distribution patterns, however, a vapor phase was dominantly formed at the upper part of the tower and a liquid phase at the lower part. From the gamma scanning of the distillation tower, it is anticipated that the gamma absorption technique can be used as an important tool for confirming the structural soundness of trays and investigating flow distribution in refinery facilities.

• Membrane and Water Treatment
• /
• v.13 no.4
• /
• pp.173-181
• /
• 2022

Polyvinylidene fluoride (PVDF) flat sheet hydrophobic membranes were prepared using 16 wt% PVDF in Dimethyl acetamide (DMAc) by phase inversion technique for desalination application using Membrane Distillation (MD). In this work, the effect of coagulation mediums such as ethanol and water as well their synergistic behavior on the fabricated PVDF membrane morphology was studied using SEM. Moreover, other characteristics required for the membrane distillation applications namely porosity, hydrophobicity and tensile strength were measured using the gravimetric method, sessile drop method and universal testing machine respectively. It was observed that the membrane morphology paradigm shifted from the finger-like structure to the sponge-like structure on increasing the ethanol concentration in coagulant. The porosity of the fabricated membrane was under the required MD range and found to be 57.3% at 16 weight % of PVDF in DMAc solvent under a pure ethanol coagulant bath. Moreover, the top surface contact angle ranges from 85° to 115° on increasing the bath concentration from CBC 0 to CBC 100 at 16 weight % of PVDF in DMAc solvent.

• Journal of Radiation Industry
• /
• v.7 no.2_3
• /
• pp.115-119
• /
• 2013

The density profile measurement technology by gamma transmission has been widely used to diagnose processes in the field of refinery and petrochemical industry. This technology can reveal a clue and position of abnormal phenomenon of industrial processes during their operation. In this paper, the feasibility of the gamma transmission technology for detecting changes in the amount of fluid in a distillation column was evaluated by using Monte Carlo simulations. The simulations assumed that $^{60}Co$ (1.17, 1.33 MeV) sources and NaI (Tl) detectors (${\Phi}5{\times}5cm$) are located in opposite sides of a column and it concurrently moves in vertical direction. To determine the dependency of a spatial resolution on aperture size of a collimator, the simulation model for a tray in a column were simulated with the aperture sizes of 1 and 2 cm. The thickness of the high density area including a tray and fluid was 7.6 cm in the simulation. The spatial resolution of the tray was 8.2 and 8.5 cm, respectively. As a result, it was revealed that the conventional density profile measurement technique is not able to show the deviation of liquid level on a tray in a column.

• Journal of Broadcast Engineering
• /
• v.28 no.3
• /
• pp.333-336
• /
• 2023

Recently, the transformer model used in natural language processing is also applied to the image super resolution field, showing good performance. However, these transformer based models have a disadvantage that they are difficult to use in small mobile devices because they are complex and have many learning parameters and require high hardware resources. Therefore, in this paper, we propose a knowledge distillation technique that can effectively reduce the size of a transformer based super resolution model. As a result of the experiment, it was confirmed that by applying the proposed technique to the student model with reduced number of transformer blocks, performance similar to or higher than that of the teacher model could be obtained.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.05a
• /
• pp.479-481
• /
• 2023

Generative Adversarial Networks (GANs) have facilitated substantial improvement in single-image super-resolution (SR) by enabling the generation of photo-realistic images. However, the high memory requirements of GAN-based SRs (mainly generators) lead to reduced performance and increased energy consumption, making it difficult to implement them onto resource-constricted devices. In this study, we propose an efficient and compressed architecture for the SR-GAN (generator) model using the model compression technique Knowledge Distillation. Our approach involves the transmission of knowledge from a heavy network to a lightweight one, which reduces the storage requirement of the model by 58% with also an increase in their performance. Experimental results on various benchmarks indicate that our proposed compressed model enhances performance with an increase in PSNR, SSIM, and image quality respectively for x4 super-resolution tasks.

• Membrane and Water Treatment
• /
• v.1 no.3
• /
• pp.215-230
• /
• 2010

Poly(vinylidene fluoride-co-hexafluoropropylene), PVDF-HFP, hollow fiber membranes were prepared by the dry/wet spinning technique using different polyethylene glycol (PEG) concentrations as non-solvent additive in the dope solution. Two different PEG concentrations (3 and 5 wt.%). The morphology and structural characteristics of the hollow fiber membranes were studied by means of optical microscopy, scanning electron microscopy, atomic force microscopy (AFM) and void volume fraction. The experimental permeate flux and the salt (NaCl) rejection factor were determined using direct contact membrane distillation (DCMD) process. An increase of the PEG content in the spinning solution resulted in a faster coagulation of the PVDF-HFP copolymer and a transition of the cross-section internal layer structure from a sponge-type structure to a finger-type structure. Pore size, nodule size and roughness parameters of both the internal and external hollow fiber surfaces were determined by AFM. It was observed that both the pore size and roughness of the internal surface of the hollow fibers enhanced with increasing the PEG concentration, whereas no change was observed at the outer surface. The void volume fraction increased with the increase of the PEG content in the spinning solution resulting in a higher DCMD flux and a smaller salt rejection factor.

• Journal of Korean Society on Water Environment
• /
• v.38 no.3
• /
• pp.143-149
• /
• 2022

Membrane distillation (MD) has emerged as a sustainable desalination technology to solve the water and energy problems faced by the modern society. In particular, the surface wetting properties of the membrane have been recognized as a key parameter to determine the performance of the MD system. In this study, a novel surface modification technique was developed to induce a Janus-type hydrophilic/hydrophobic layer on the membrane surface. The hydrophilic layer was created on a porous PVDF membrane by vapor phase polymerization of the pyrrole monomer, forming a thin coating of polypyrrole on the membrane walls. A rigid polymeric coating layer was created without compromising the membrane porosity. The hydrophilic coating was then followed by the in-situ growth of siloxane nanoparticles, where the condensation of organosilane provided quick loading of hydrophobic layers on the membrane surface. The composite layers of dual coatings allowed systematic control of the surface wettability of porous membranes. By the virtue of the photothermal property of the hydrophilic polypyrrole layer, the desalination performance of the coated membrane was tested in a solar MD system. The wetting properties of the dual-layer were further evaluated in a direct-contact MD module, exploring the potential of the Janus membrane structure for effective and low-energy desalination.

• Smart Media Journal
• /
• v.12 no.9
• /
• pp.89-94
• /
• 2023

Recommender systems provide users with the most favorable items by analyzing explicit or implicit feedback of users on items. Recently, as the size of deep-learning-based models employed in recommender systems has increased, many studies have focused on reducing inference time while maintaining high recommendation accuracy. As one of them, a study on recommender systems with a knowledge distillation (KD) technique is actively conducted. By KD, a small-sized model (i.e., student) is trained through knowledge extracted from a large-sized model (i.e., teacher), and then the trained student is used as a recommendation model. Existing studies on KD for recommender systems have been mainly performed only for implicit feedback settings. Thus, in this paper, we try to investigate the performance and accuracy when applied to explicit feedback settings. To this end, we leveraged a total of five state-of-the-art KD methods and three real-world datasets for recommender systems.