• The Journal of the Convergence on Culture Technology
• /
• v.10 no.3
• /
• pp.909-914
• /
• 2024

This study suggests a quantitative evaluation of transfer learning, which is widely used in various AI fields, including image recognition for robot vision. Quantitative and qualitative analyses of results applying transfer learning are presented, but transfer learning itself is not discussed. Therefore, this study proposes a quantitative evaluation of transfer learning itself based on MNIST, a handwritten digit database. For the reference network, the change in recognition accuracy according to the depth of the transfer learning frozen layer and the ratio of transfer learning data and pre-training data is tracked. It is observed that when freezing up to the first layer and the ratio of transfer learning data is more than 3%, the recognition accuracy of more than 90% can be stably maintained. The transfer learning quantitative evaluation method of this study can be used to implement transfer learning optimized according to the network structure and type of data in the future, and will expand the scope of the use of robot vision and image analysis AI in various environments.

• Proceedings of the KSME Conference
• /
• 2003.04a
• /
• pp.1461-1466
• /
• 2003

This paper presents the effects of the fin array and pitch on the frost layer growth of a heat exchanger. The numerical results are compared with experimental data of a cold plate to validate the present model, and agree well with experimental data within a maximum error of 8%. The characteristics of the frost formation on staggered fin array are somewhat different from those of in-line array. The frost layer at the first fin of the in-line array grows rapidly, compared to second fin, whereas the difference of the frost layer growth between the fins of the staggered array is small. For fin pitch below 10 mm, the frost layer growth of second fin in the staggered array is affected by that of first fin. The frost layer growth and heat transfer of single fin deteriorate with decreasing fin pitch regardless of fin array, however, the thermal performance of a heat exchanger, considering increase of heat surface area, becomes better.

• International Journal of Aerospace System Engineering
• /
• v.2 no.2
• /
• pp.6-9
• /
• 2015

A new cross disciplinary conception of transitional aero-optics is built up during analyzing and measuring the linkage between the hypersonic boundary layer transition on a flat plate and the jittering characteristics of the small-aperture beam through that boundary layer. Based on that conception, the Small-Aperture Beam Technique (SABT) and high-speed Imaging Camera System (ICS) used in aero-optical studies are considered as new techniques for the assessment of the hypersonic transition in the boundary layer on a flat plate. In the FD-20 gun tunnel, for the free stream parameters with Mach number of 8 and unit Reynolds number of $1{\times}10^7$ (1/m), those two optical techniques are used to measure the jitter of the small-aperture beam. At the same free stream parameters, the distribution of the heat transfer along the centerline of the flat plate is also measured by the thin film resistance gauge technique. The results show the similarity of the increase trend between the heat transfer and the jitter of the small-aperture beam in the transitional region. It helps us to surmise that it may be feasible to diagnose the transition in a hypersonic boundary layer on a flat plate by means of those above optical techniques.

• Structural Engineering and Mechanics
• /
• v.13 no.2
• /
• pp.187-209
• /
• 2002

The infrastructure system in the United States has been aging faster than the resource available to restore them. Therefore decision for allocating the resources is based in part on the condition of the structural system. This paper proposes to use neural network to predict the overall rating of the structural system because of the successful applications of neural network to other fields which require a "symptom-diagnostic" type relationship. The goal of this paper is to illustrate the potential of using neural network in civil engineering applications and, particularly, in bridge evaluations. Data collected by the Tennessee Department of Transportation were used as "test bed" for the study. Multi-layer feed forward networks were developed using the Levenberg-Marquardt training algorithm. All the neural networks consisted of at least one hidden layer of neurons. Hyperbolic tangent transfer functions were used in the first hidden layer and log-sigmoid transfer functions were used in the subsequent hidden and output layers. The best performing neural network consisted of three hidden layers. This network contained three neurons in the first hidden layer, two neurons in the second hidden layer and one neuron in the third hidden layer. The neural network performed well based on a target error of 10%. The results of this study indicate that the potential for using neural networks for the evaluation of infrastructure systems is very good.

• 제어로봇시스템학회:학술대회논문집
• /
• 2001.10a
• /
• pp.164.2-164
• /
• 2001

Sliding mode control (SMC) with variable nonlinear boundary layer is proposed. Two Fuzzy logic controllers (FLCs) are used to decide both boundary layer thickness and nonlinear interpolation using sigmoid function in the boundary layer. The nonlinear interpolation in the boundary layer suing FLC reduces stead state error and chattering. Sigmoid function is used to nonlinear interpolation in the boundary layer sigmoid function parameter with FLC. To demonstrate its performance, the Proposed control algorithm is applied to a simple nonlinear system.

• Journal of the Korean Society of Clothing and Textiles
• /
• v.13 no.4
• /
• pp.400-411
• /
• 1989

Moisture related properties of fabrics in dynamic modes are considered to be important in the judgement of the subjective comfort characteristics of clothing fabrics. In the current study, an attempt to improve the cobaltous chloride test method was made which has been known as a convenient screening test for dynamic surface wetness. The color changes of cobaltous chloride treated fabrics on the simulated sweating skin were calibrated against standard color strips. The standard color strips were made of all typs of test fabrics and installed inside of the test tubes containing a series of saturated salt solutions, which gives more quantitative informations on dynamic moisture transfer Influences of fiber types and finishes on dynamic moisture transfer in textiles were studied using a single layer of fabric samples. Fiber types included $100\%$ cotton, C/P 50/50, C/P 35/65. Durable press and soil release finished cotton and C/P 50/50 fabrics were also included. There were significant fiber effects on the dynamic moisture transfer. The order of time taken to reach to the specified $\%$ RH was C/P 35/65$100\%$ cotton fabrics. It was possible to detect significant finish effects by increasing the concentrations of cobaltous chloride solutions. The order of time taken to reach to the specified $\%$ R.H was durable press$100\%$ cotton, C/P 50/50, C/P 35/65 and $100\%$ PET were placed at the inner side of the outer layer and tested. It was shown that cobaltous chloride treated $100\%$ cotton fabirc was easier to detect color changes than C/P blend fabic in the double layer experiments. By placing test sample under the cobaltous chloride treated cotton fabrics, it was able to detect the differences among the test samples, some of which were known to be difficult in padding with cobaltous chloride solutions. Besides, the double layer method would provide with the broader application of the cobaltous chloride method in !uture, since it is possible to test the dynamic moisture transfer of clothing as worn.

• International Journal of Automotive Technology
• /
• v.8 no.6
• /
• pp.761-769
• /
• 2007

The configuration of the flow channel on a bipolar plate of a proton exchange membrane fuel cell(PEMFC) for efficient reactant supply has great influence on the performance of the fuel cell. Recent demand for higher energy density fuel cells requires an increase in current density at mid voltage range and a decrease in concentration overvoltage at high current density. Therefore, an interdigitated flow channel where mass transfer rate by convection through a gas diffusion layer is greater than the mass transfer by a diffusion mechanism through a gas diffusion layer was recently proposed. This study attempts to analyze the i-V performance, mass transfer and pressure drop in interdigitated flow channels by developing a fully three dimensional simulation model for PEMFC that can deal with anode and cathode flow together. The results indicate that the trade off between performance and pressure loss should be considered for efficient design of flow channels. Although the performance of the fuel cell with interdigitated flow is better than that with conventional flow channels due to a strong mass transfer rate by convection across a gas diffusion layer, there is also an increase in friction due to the strong convection through the porous diffusion layer accompanied by a larger pressure drop along the flow channel. It was evident that the proper selection of the ratio of channel and rib width under counter flow conditions in the fuel cell with interdigitated flow are necessary to optimize the interdigitated flow field design.

• Journal of the Korean Solar Energy Society
• /
• v.29 no.5
• /
• pp.35-44
• /
• 2009

In this study, boiling heat transfer coefficients(HTCs) and critical heat flux(CHF) are measured on a smooth square flat copper heater in a pool of pure water with and without carbon nano tubes(CNTs) dispersed at $60^{\circ}C$. Tested aqueous nanofluids are prepared using multi-walled CNTs whose volume concentrations are 0.0001, 0.001, 0.01, and 0.05%. For dispersion of CNTs, polyvinyl pyrrolidone(PVP) is used in distilled water. Pool boiling HTCs are taken from $10kW/m^2$ to critical heat flux for all nanofluids. Test results show that the pool boiling HTCs of the nanofluids are lower than those of pure water in entire nucleate boiling regime. On the other hand, critical heat flux is enhanced greatly showing up to 200% increase at volume concentration of 0.001% CNTs as compared to that of pure water. This is related to the change of surface characteristics by the deposition of CNTs. This deposition makes a thin CNT layer on the surface and the active nucleation sites of heat transfer surface are decreased due to this layer. The thin layer acts as the thermal resistance and also decreases the bubble generation rate resulting in a decrease in pool boiling HTCs. The same layer, however, maintains the nucleate boiling even at very high heat fluxes and reduces the formation of large vapor canopy at near CHF resulting in a significant increase in CHF.

• The KSFM Journal of Fluid Machinery
• /
• v.18 no.3
• /
• pp.38-45
• /
• 2015

In this study, two computational methodologies were compared to consider an effective conjugate heat transfer analysis technique for the cooled vane with thermal barrier coating. The first one is the physical modeling method of the TBC layer on the vane surface, which means solid volume of the TBC on the vane surface. The second one is the numerical modeling method of the TBC layer by putting the heat resistance interface condition on the surface between the fluid and solid domains, which means no physical layer on the vane surface. For those two methodologies, conjugate heat transfer analyses were conducted for the cooled vane with TBC layer having various thickness from 0.1 mm to 0.3 mm. Static pressure distributions for two cases show quite similar patterns in the overall region while the physical modeling shows quite a little difference around the throat area. Thermal analyses indicated that the metal temperature distributions are quite similar for both methods. The results show that the numerical modeling method can reduce the computational resources significantly and is quite suitable method to evaluate the overall performance of TBC even though it does not reflect the exact geometry and flow field characteristics on the vane surface.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.22 no.5
• /
• pp.554-560
• /
• 2021

CRUD (chalk river unidentified deposits) is a porous material deposited on the surface of nuclear fuel during nuclear power plant operation. The CRUD is composed of metal oxides, such as iron, nickel, and chromium. It is essential to investigate the effects of the CRUD layer on the wall heat transfer between the nuclear fuel surface and the coolant in the event of a nuclear accident. CRUD only negatively affects the temperature of the nuclear fuel due to heat resistance because the effects of the CRUD layer on two-phase boiling heat transfer are not considered. In this study, the physical property models for the porous CRUD layer were developed and implemented into the SPACE code. The effects of boiling heat transfer models on the peak cladding temperature and quenching were investigated by simulating a reflood experiment. The calculation results showed some positive effects of the CRUD layer.