• Proceedings of the Korean Society for Technology of Plasticity Conference
• /
• 2002.11a
• /
• pp.63-70
• /
• 2002

The purpose of the present study is to propose a new way of manufacturing curved metal tubes with arbitrary sections and way of eliminating the conventional bending defects such as thinning and thickening, in the wall of tube, distortion of the section, and wrinkling and folding on the surface by the extrusion bending process that can extrude and weld together one or more billets inside dies cavity, and can bend them during extrusion due to the gradient of extrusion velocities controlled by the eccentricity of the cavity sections between the entrance and the exit of the eccentric conical extrusion bending dies and conical plug, or by the relative size of the holes of multi-hole container, or by the relative moving velocity of multi-punches.

• Structural Engineering and Mechanics
• /
• v.60 no.4
• /
• pp.707-727
• /
• 2016

In this paper, thermal vibration of a nonlocal functionally graded (FG) plates with arbitrary boundary conditions under linear and non-linear temperature fields is explored by developing a refined shear deformation plate theory with an inverse cotangential function in which shear deformation effect was involved without the need for shear correction factors. The material properties of FG nanoplate are considered to be temperature-dependent and graded in the thickness direction according to the Mori-Tanaka model. On the basis of non-classical higher order plate model and Eringen's nonlocal elasticity theory, the small size influence was captured. Numerical examples show the importance of non-uniform thermal loadings, boundary conditions, gradient index, nonlocal parameter and aspect and side-to-thickness ratio on vibrational responses of size-dependent FG nanoplates.

• Proceedings of the Korean Society of Propulsion Engineers Conference
• /
• 2001.11a
• /
• pp.50-53
• /
• 2001

Flame propagation during the initial stages of ignition in a non-premixed swirl, having some of characteristics of the primary zone of an aero gas turbine combustor, has been investigated. Nd:YAG laser was adopted as the principal ignition source to allow arbitrary placing of the ignition site i subsequent flame development was monitored using a natural light high speed filming technique for many ignition site at two different swirl ratios and an overall equivalence ratio of 0.9. For ignition offset from the burner centreline, buoyancy force associated with radial pressure gradient produced a strong inward movement of the flame kernel. At the burner exit. flame kernels invariably developed into cylindrical form and a 'radial confinement /axia expansion' (RCAE) process was observed.

• Korean Journal of Optics and Photonics
• /
• v.17 no.6
• /
• pp.493-499
• /
• 2006

Arbitrary graded-index HOF(Hollow Optical Fibers) are analyzed using the modified Airy function, and the corresponding eigenvalue equation that renders precise results is derived. For graded index HOF, the gradient of an evanescent field in hollow region could be adjusted more sharply than the conventional step-index HOF and the feasibility of more effective atom-guiding is confirmed.

• Structural Engineering and Mechanics
• /
• v.83 no.6
• /
• pp.825-834
• /
• 2022

The effectiveness and accuracy of the strain-based approach applied for analysis of two kinds of heterogeneous hollow cylinders subjected to thermal and mechanical loads are examined in this study. One is a multilayer cylinder in which the material in each layer is assumed to be linearly elastic, homogeneous and isotropic. Another is a hollow cylinder made of functionally graded materials with arbitrary gradient. The steady state condition without heat generation is considered. A sector in-plane finite element in the polar coordinate system based on strain approach is used. This element has only three degrees of freedom at each corner node. Analytical solutions available in the literature are presented to illustrate the accuracy of the sector element used. The obtained results for displacements and stresses are shown to be in good agreement with the analytical solutions.

• Investigative Magnetic Resonance Imaging
• /
• v.14 no.2
• /
• pp.87-94
• /
• 2010

Purpose : To calculate the appearance of the image distortion from metallic artifacts and to determine the location of a metallic needle from a distorted MR image. Materials and Methods : To examine metal artifacts, an infinite metal cylinder in a strong magnetic field are assumed. The cylinder’s axis leaned toward the magnetic field along some arbitrary angle. The Laplace equation for this situation was solved to investigate the magnetic field distortion, and the simulation was performed to evaluation the image artifact caused by both readout and slice-selection gradient field. Using the result of the calculation, the exact locations of the metal cylinder were calculated from acquired images. Results : The distances between the center and the folded point are measured from images and calculated. Percentage errors between the measured and calculated distance were less than 5%, except for one case. Conclusion : The simulation was successfully performed when the metal cylinder was skewed at an arbitrary tilted angle relative to the main magnetic field. This method will make it possible to monitor and guide both biopsy and surgery with real time MRI.

• Ocean Systems Engineering
• /
• v.8 no.4
• /
• pp.381-407
• /
• 2018

This paper presents a hybrid numerical approach, which combines a two-phase Navier-Stokes model (NS) and the fully nonlinear potential theory (FNPT), for modelling wave-structure interaction. The former governs the computational domain near the structure, where the viscous and turbulent effects are significant, and is solved by OpenFOAM/InterDyMFoam which utilising the finite volume method (FVM) with a Volume of Fluid (VOF) for the phase identification. The latter covers the rest of the domain, where the fluid may be considered as incompressible, inviscid and irrotational, and solved by using the Quasi Arbitrary Lagrangian-Eulerian finite element method (QALE-FEM). These two models are weakly coupled using a zonal (spatially hierarchical) approach. Considering the inconsistence of the solutions at the boundaries between two different sub-domains governed by two fundamentally different models, a relaxation (transitional) zone is introduced, where the velocity, pressure and surface elevations are taken as the weighted summation of the solutions by two models. In order to tackle the challenges associated and maximise the computational efficiency, further developments of the QALE-FEM have been made. These include the derivation of an arbitrary Lagrangian-Eulerian FNPT and application of a robust gradient calculation scheme for estimating the velocity. The present hybrid model is applied to the numerical simulation of a fixed horizontal cylinder subjected to a unidirectional wave with or without following current. The convergence property, the optimisation of the relaxation zone, the accuracy and the computational efficiency are discussed. Although the idea of the weakly coupling using the zonal approach is not new, the present hybrid model is the first one to couple the QALE-FEM with OpenFOAM solver and/or to be applied to numerical simulate the wave-structure interaction with presence of current.

• Journal of Convergence for Information Technology
• /
• v.7 no.3
• /
• pp.73-81
• /
• 2017

The generalized Hough transform (GHough) is a useful technique for detecting and locating 2-D model. However, GHough requires a 4-D parameter array and a large amount of time to detect objects of unknown scale and orientation because it enumerates all possible parameter values into a 4-D parameter space. Several n-to-1 mapping algorithms were proposed to reduce the parameter space from 4-D to 2-D. However, these algorithms are very likely to fail due to the random votes cast into the 2-D parameter space. This paper proposes to use internal gradient information in addition to the model boundary points to reduce the number of random votes cast into 2-D parameter space. Experimental result shows that our proposed method can reduce both the number of random votes cast into the parameter space and the execution time effectively.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.37 no.8A
• /
• pp.679-688
• /
• 2012

In this paper, performance improved hole-filling algorithm including boundary noise removing pre-process which can be used for an arbitrary view synthesis with given two views is proposed. Boundary noise usually occurs because of the boundary mismatch between the reference image and depth map and common-hole is defined as the occluded region. These boundary noise and common-hole created while synthesizing a virtual view result in some defects and they are usually very difficult to be completely recovered by using only given two images as references. The spiral weighted average algorithm gives a clear boundary of each object by using depth information and the gradient searching algorithm is able to preserve details. In this paper, we combine these two algorithms by using a weighting factor ${\alpha}$ to reflect the strong point of each algorithm effectively in the virtual view synthesis process. The experimental results show that the proposed algorithm performs much better than conventional algorithms.

• Journal of Intelligence and Information Systems
• /
• v.22 no.2
• /
• pp.127-142
• /
• 2016

Deep learning model is a kind of neural networks that allows multiple hidden layers. There are various deep learning architectures such as convolutional neural networks, deep belief networks and recurrent neural networks. Those have been applied to fields like computer vision, automatic speech recognition, natural language processing, audio recognition and bioinformatics where they have been shown to produce state-of-the-art results on various tasks. Among those architectures, convolutional neural networks and recurrent neural networks are classified as the supervised learning model. And in recent years, those supervised learning models have gained more popularity than unsupervised learning models such as deep belief networks, because supervised learning models have shown fashionable applications in such fields mentioned above. Deep learning models can be trained with backpropagation algorithm. Backpropagation is an abbreviation for "backward propagation of errors" and a common method of training artificial neural networks used in conjunction with an optimization method such as gradient descent. The method calculates the gradient of an error function with respect to all the weights in the network. The gradient is fed to the optimization method which in turn uses it to update the weights, in an attempt to minimize the error function. Convolutional neural networks use a special architecture which is particularly well-adapted to classify images. Using this architecture makes convolutional networks fast to train. This, in turn, helps us train deep, muti-layer networks, which are very good at classifying images. These days, deep convolutional networks are used in most neural networks for image recognition. Convolutional neural networks use three basic ideas: local receptive fields, shared weights, and pooling. By local receptive fields, we mean that each neuron in the first(or any) hidden layer will be connected to a small region of the input(or previous layer's) neurons. Shared weights mean that we're going to use the same weights and bias for each of the local receptive field. This means that all the neurons in the hidden layer detect exactly the same feature, just at different locations in the input image. In addition to the convolutional layers just described, convolutional neural networks also contain pooling layers. Pooling layers are usually used immediately after convolutional layers. What the pooling layers do is to simplify the information in the output from the convolutional layer. Recent convolutional network architectures have 10 to 20 hidden layers and billions of connections between units. Training deep learning networks has taken weeks several years ago, but thanks to progress in GPU and algorithm enhancement, training time has reduced to several hours. Neural networks with time-varying behavior are known as recurrent neural networks or RNNs. A recurrent neural network is a class of artificial neural network where connections between units form a directed cycle. This creates an internal state of the network which allows it to exhibit dynamic temporal behavior. Unlike feedforward neural networks, RNNs can use their internal memory to process arbitrary sequences of inputs. Early RNN models turned out to be very difficult to train, harder even than deep feedforward networks. The reason is the unstable gradient problem such as vanishing gradient and exploding gradient. The gradient can get smaller and smaller as it is propagated back through layers. This makes learning in early layers extremely slow. The problem actually gets worse in RNNs, since gradients aren't just propagated backward through layers, they're propagated backward through time. If the network runs for a long time, that can make the gradient extremely unstable and hard to learn from. It has been possible to incorporate an idea known as long short-term memory units (LSTMs) into RNNs. LSTMs make it much easier to get good results when training RNNs, and many recent papers make use of LSTMs or related ideas.