• Nuclear Engineering and Technology
• /
• v.52 no.2
• /
• pp.287-295
• /
• 2020

Group method of data handling (GMDH) is considered one of the earliest deep learning methods. Deep learning gained additional interest in today's applications due to its capability to handle complex and high dimensional problems. In this study, multi-layer GMDH networks are used to perform uncertainty quantification (UQ) and sensitivity analysis (SA) of nuclear reactor simulations. GMDH is utilized as a surrogate/metamodel to replace high fidelity computer models with cheap-to-evaluate surrogate models, which facilitate UQ and SA tasks (e.g. variance decomposition, uncertainty propagation, etc.). GMDH performance is validated through two UQ applications in reactor simulations: (1) low dimensional input space (two-phase flow in a reactor channel), and (2) high dimensional space (8-group homogenized cross-sections). In both applications, GMDH networks show very good performance with small mean absolute and squared errors as well as high accuracy in capturing the target variance. GMDH is utilized afterward to perform UQ tasks such as variance decomposition through Sobol indices, and GMDH-based uncertainty propagation with large number of samples. GMDH performance is also compared to other surrogates including Gaussian processes and polynomial chaos expansions. The comparison shows that GMDH has competitive performance with the other methods for the low dimensional problem, and reliable performance for the high dimensional problem.

• Journal of Ocean Engineering and Technology
• /
• v.30 no.1
• /
• pp.44-50
• /
• 2016

An experimental study was performed to investigate the behavior of green water on a structure with a rectangular cross section under wave conditions, along with the flow characteristics in bubbly water flow. An experiment was conducted in a two-dimensional wave flume using an acrylic model (1/125) of FPSO BW Pioneer operating in the Gulf of Mexico under its design wave condition. The occurrence of green water, including its development, in front of the model was captured using a high-speed Charge Coupled Device (CCD) camera with the shadowgraph technique. Using consecutive images, the generation procedure for green water on the model was divided into five phases: flip through, air entrapment, wave run-up, wave overturning, and water shipping. In addition, the distinct water elevations of the green water were defined as the height of flip through, height of splashing jet, and height of freeboard exceedance, and showed a linear relationship with the incoming wave height.

• Journal of Energy Engineering
• /
• v.12 no.1
• /
• pp.35-41
• /
• 2003

Heat regenerator occupied by regenerative materials improves thermal efficiency of regenerative combustion system through the recovery of heat of exhaust gaset. By using one-dimensional two-phase fluid dynamics model, the unsteady thermal flow of heat regenerator with spherical particles, was numerically simulated to evaluate the heat transfer and pressure drop and thereby to suggest the parameter for designing heat regenerator. It takes about 7 hours for the steady state of the flow field in regenerator, in which heat absorption of regenerative particle is concurrent with the same magnitude of heat desorption. The regenerative particle experiences small temperature fluctuation below 10 K during the reversing process. The performance of thermal flow in heat regenerator varies with inlet velocity of exhaust gas and air, configuration of regenerator (cross-sectional area and length) and diameter of regenerative particle. As the gas velocity increases, the heat transfer between gas and particle enhances and with the increase the pressure losses. As particle diameter decreases, the air is preheated higher and the exhaust gases are cooled more with the increase of pressure losses.

• Journal of IKEEE
• /
• v.28 no.2
• /
• pp.158-167
• /
• 2024

In this study, U-net model based on electrical capacitance is applied to monitor the condition inside the pipeline of semiconductor rear end process implemented in the numerical simulation. Capacitance values measured from electrodes attached to the pipeline is used as input data for the U-net network model and estimated permittivity distribution by the U-net model is used to reconstructed cross-sectional image at the pipeline. In the numerical simulation, images reconstructed by U-net model, Fully-connected neural network (FCNN) model and Newton-Raphson method are compared for evaluation. U-net model shows good results as compared to other models.

• Wind and Structures
• /
• v.38 no.3
• /
• pp.161-170
• /
• 2024

In recent years, there have been an increasing number of experimental studies showing the need to include robustness criteria in the design process to develop complex active control designs for practical implementation. The paper investigates the crosswind aerodynamic parameters after the blocking phase of a two-dimensional square cross-section structure by measuring the response in wind tunnel tests under light wind flow conditions. To improve the accuracy of the results, the interpolation of the experimental curves in the time domain and the analytical responses were numerically optimized to finalize the results. Due to this combined effect, the three aerodynamic parameters decrease with increasing wind speed and asymptotically affect the upper branch constants. This means that the aerodynamic parameters along the density distribution are minimal. Taylor series are utilized to describe the fuzzy nonlinear plant and derive the stability analysis using polynomial function for analyzing the aerodynamic parameters and numerical simulations. Due to it will yield intricate terms to ensure stability criterion, therefore we aim to avoid kinds issues by proposing a polynomial homogeneous framework and utilizing Euler's functions for homogeneous systems. Finally, we solve the problem of stabilization under the consideration by SOS (sum of squares) and assign its fuzzy controller based on the feasibility of demonstration of a nonlinear system as an example.

• Proceedings of the KSME Conference
• /
• 2000.04b
• /
• pp.83-89
• /
• 2000

This paper presents a pressure drop correlation for evaporation and condensation of alternative refrigerant with oil in micro-fin tubes. The correlation was developed from a data base consisting of oil-free pure and mixed refrigerants in micro-fin tube; Rl25 R134a. R32 R410a(R32/R125 50/50% mass), R22, R407c(R32/R125/R134a, 23/25/52% mass) and R32/R134a(25/75% mass). The micro-fin tube used in this paper had 60 0.2mm high fins with a 18 helix angle. The cross sectional flow area $(A_c)$ was $60.8 mm^2$ giving an equivalent smooth diameter$(D_e)$ of 8.8mm. The hydraulic diameter $(D_h)$ was estimated to the 5.45mm. The new correlation was obtained by replacing the friction factor and the tube-diameter in Bo Pierre correlation by a friction factor derived from pressure drop data for a micro-fin tube and the hydraulic diameter, respectively. This correlation was also used to predict some pressure data with a lubricant after using a mixing viscosity rule of lubricants and refrigerants. As a result, the new correlation was also well predicted to the measured data within a mean deviation of 19.0%.

• Journal of Korean Society of Environmental Engineers
• /
• v.30 no.9
• /
• pp.948-954
• /
• 2008

Membrane bioreactor(MBR) processes have been widely applied to wastewater treatment for last decades due to its excellent capability of solid-liquid separation. However, membrane fouling was considered as a limiting factor in wide application of the MBR process. Excess aeration into membrane surface is a common way to control membrane fouling in most MBR. However, the excessively supplied air is easily dissipated in the reactor, which results in consuming energy and thus, it should be modified for effective control of membrane fouling. In this study, cylindrical tube was introduced to MBR in order to use the supplied air effectively. Membrane fibers were immersed into the cylindrical tube. This makes the supplied air non-dissipated in the reactor so that membrane fouling could be controlled economically. Two different air supplying method was employed and compared each other; nozzle and porous diffuser which were located just beneath the membrane module. Transmembrane pressure(TMP) was monitored as a function of airflow rate, flux, and ratio of the tube area and cross-sectioned area of membrane fibers(A$_m$/A$_t$). Flow rate of air and liquid was regulated to obtain slug flow in the cylindrical tube. With the same flow of air supply, nozzle was more effective for controlling membrane fouling than porous diffuser. Accumulation of sludge was observed in the tube with the nozzle, if the air was not suppled sufficiently. Reduction of membrane fouling was dependent upon the ratio, A$_m$/A$_t$. For diffuser, membrane fouling was minimized when A$_m$/A$_t$ was 0.27, but 0.55 for nozzle.

• Transactions of the Korean Society of Mechanical Engineers
• /
• v.10 no.6
• /
• pp.823-829
• /
• 1986

The Large-scale structure of the circular jet in the transition region, which influences the subsequent flow in the turbulent region, was studied experimentally. Measuring equipments are composed of the two channel hot-wire anemometer, the computer controlled two-directional traverse mechanism, the data acquisition system, and FFT-analyzer. The circular jet has 50mm diameter. The mean velocity distribution, the velocity fluctuation, the auto 'cross correlations and the power spectra were acquired at moderate Reynolds number of 10$^{4}$. And the VITA method was used to measure the convection velocity of Large-scale eddy. The phase of u'is in advance of that of v'in all regions. .over bar. $R_{u}$(.tau.=0) is approximately zero in the potential core region, but a small regular deviation is observed. At a position in the mixing layer region the convection velocity is different along the part of the eddy, and in this experiment the convection velocity of the inner region is larger than the outer region. The averge convection velocity of the eddy along y/D=0 was approximately constant in the transition region.D=0 was approximately constant in the transition region.

• Journal of Acupuncture Research
• /
• v.17 no.2
• /
• pp.85-93
• /
• 2000

We came to the conclusion after considering all the information from many kinds of books on the circulation courses, cross-link points, functions and the symptoms of disease of Chongmai. The results were as follows : 1. The Chongmai that starts from a Uterus flows to Dazhu through the inside of vertebra after joining Renmai and Duimai at Huiyin. The Chongmais up-line that comes out from Qichong into a surface of body arrives and is scattered at a chest through an abdomen. One strand of them goes upward again and is connected to a throat and lips area. After coming out from Qichong, separated down-line is divided into two parts when it arrives a inner part of a heel through the inner part of a leg. One is for an instep, the other is for the sole of a foot. 2. We call it "Sea of Twelve Meridians" or "Meridian's Sea". Because Chongmai controls all of Meridian by acquired "Basic energy" as getting Stomach's energy, Kidney's energy and air-energy, and there are responsible of physiological phenomenon control. And also we name it "Sea of Blood", because it starts from and provides a nutrition to Uterus. 3. All of these four Meridian such as Renmai, Kidney Meridian, Stomach Meridian and Spleen Meridian are ones that flow around the part of a chest and an abdomen. Chongmai makes energy and blood circulation of a chest and an abdomen be stronger and be controlled. Therefore it makes viscera, bowels and body surface be warm and given a nutrition. So Chongmai becomes "Sea of Viscera and Bowels". 4. Chongmai provides a nutrition for ligament and muscle and makes legs get warm as making energy and blood circulate from head to foot. If Chongmai is energetic, hair is completely easy to grow. 5. To see in pathological phase, Chongmais failure or weakness causes like a chest pain, stomachache, heart attack, a menstrual irregularity and sterility and so on. And also if Chongmai is damaged, it happens that giving a nutrition for lips area is stopped, and then mustache doesn't grow any more.

• Progress in Superconductivity and Cryogenics
• /
• v.15 no.2
• /
• pp.34-38
• /
• 2013

AC Losses for face to face stacks of four identical coated conductors (CCs) were numerically calculated using the H-formulation combined with the E-J power law and the Kim model. The motive sample was the face to face stack of four 2 mm-wide CC tapes with 2 ${\mu}m$ thick superconducting layer of which the critical current density, $J_c$, was $2.16{\times}10^6A/cm^2$ on IBAD-MgO template, which was suggested for the mitigation of ac loss as a round shaped wire by Korea Electrotechnology Research Institute. For the calculation the cross section of the stack was simply modeled as vertically aligned 4 rectangles of superconducting (SC) layers with $E=E_o(J(x,y,t)/J_c(B))^n$ in x-y plane where $E_o$ was $10^{-6}$ V/cm, $J_c$(B) was the field dependence of current density and n was 21. The field dependence of the critical current of the sample measured in four-probe method was employed for $J_c$(B) in the equation. The model was implemented in the finite element method program by commercial software. The ac loss properties for the stacks were compared with those of single 4 cm-wide SC layers with the same critical current density or the same critical current. The constraint for the simulation was imposed in two different ways that the total current of the stack obtained by integrating J(x,y,t) over the cross sections was the same as that of the applied transport current: one is that one fourth of the external current was enforced to flow through each SC. In this case, the ac loss values for the stacks were lower than those of single wide SC layer. This mitigation of the loss is attributed to the reduction of the normal component of the magnetic field near the SC layers due to the strong expulsion of the magnetic field by the enforced transport current. On the contrary, for the other case of no such enforcement, the ac loss values were greater than those of single 4cm-wide SC layer and. In this case, the phase difference of the current flowing through the inner and the outer SC layers of the stack was observed as the transport current was increased, which was a cause of the abrupt increase of ac loss for higher transport current.