• International journal of advanced smart convergence
• /
• v.8 no.1
• /
• pp.141-146
• /
• 2019

These days, there are many applications using neural networks as parts of their system. On the other hand, adversarial examples have become an important issue concerining the security of neural networks. A classifier in neural networks can be fooled and make it miss-classified by adversarial examples. There are many research to encounter adversarial examples by using denoising methods. Some of them using GAN (Generative Adversarial Network) in order to remove adversarial noise from input images. By producing an image from generator network that is close enough to the original clean image, the adversarial examples effects can be reduced. However, there is a chance when adversarial noise can survive the approximation process because it is not like a normal noise. In this chance, we propose a research that utilizes high-level representation in the classifier by combining GAN network with a trained U-Net network. This approach focuses on minimizing the loss function on high representation terms, in order to minimize the difference between the high representation level of the clean data and the approximated output of the noisy data in the training dataset. Furthermore, the generated output is checked whether it shows minimum error compared to true label or not. U-Net network is trained with true label to make sure the generated output gives minimum error in the end. At last, the remaining adversarial noise that still exist after low-level approximation can be removed with the U-Net, because of the minimization on high representation terms.

• ETRI Journal
• /
• v.18 no.4
• /
• pp.315-338
• /
• 1997

We extend our previous theoretical analysis of electronic and optical properties of p-type quantum well structures based on the two heavy- and light-hole system to include all the three valence bands. These theories are then used to clarify the origin of the normal incidence absorption and photo current at photon wavelengths of 2 - 3 ${\mu}m$, which was observed in addition to the absorption around 8 ${\mu}m$ by a recent experimental investigation with heavily doped p-type GaAs/AlGaAs multi-quantum well (MQW) structures. In the theoretical analysis, the Hartree and exchange-correlation many-body interactions are taken into account within one-particle local density approximation, and it is shown that normal incidence absorption occurs in two wavelength regions over the transition energy range higher than barrier height for p-type GaAs/AlGaAs superlattices with well doping of $2{\times}10^{19}\;cm^{-3}$; one region has broad absorption peaks with coefficients of about 5000 $cm^{-1}$ around 8 ${\mu}m$, and the other has two rather sharp peaks at 2.7 ${\mu}m$ and 3.4 ${\mu}m$ with 1800 $cm^{-1}$ and 1300 $cm^{-1}$, respectively. The result indicates that the theory explains the experimental observation well, as the theoretical and experimental results are in close agreement in general absorption features.

• Transactions of the Korean Society of Mechanical Engineers
• /
• v.8 no.6
• /
• pp.591-598
• /
• 1984

The results of measurements showing normal vibrations and rubbing noise generated during unlubricated smooth sliding between metal surfaces are presented. The measurements were made on pin-on-disc type apparatus instrumented with piezoelectric acceleration transducers and microphones. Spectral analysis of the both signals up to frequency of 10kHz indicates that they are closely correlated. The major components of both signals in this frequency range are primarily associated with the normal contact vibrations which are excited by surface irregularities being swept through the contact region during sliding. As an approximation to the seismic input of surface irregularities, an effective surface wavenumber spectrum was assumed in the form of an inverse vibration and noise measurements for a number of surface finishes and mean loads. The predominant frequency component of which levels of the normal vibration and noise are close to overall levels of the both signals is induced by contact resonance between the two bodies and its frequency can be calculated from the Hertzian theory.

• Smart Structures and Systems
• /
• v.3 no.4
• /
• pp.423-437
• /
• 2007

The empirical mode decomposition (EMD) method is well-known for its ability to decompose a multi-component signal into a set of intrinsic mode functions (IMFs). The method uses a sifting process in which local extrema of a signal are identified and followed by a spline fitting approximation for decomposition. This method provides an effective and robust approach for decomposing nonlinear and non-stationary signals. On the other hand, the IMF components do not automatically guarantee a well-defined physical meaning hence it is necessary to validate the IMF components carefully prior to any further processing and interpretation. In this paper, an attempt to use the EMD method to identify properties of nonlinear elastic multi-degree-of-freedom structures is explored. It is first shown that the IMF components of the displacement and velocity responses of a nonlinear elastic structure are numerically close to the nonlinear normal mode (NNM) responses obtained from two-dimensional invariant manifolds. The IMF components can then be used in the context of the NNM method to estimate the properties of the nonlinear elastic structure. A two-degree-of-freedom shear-beam building model is used as an example to illustrate the proposed technique. Numerical results show that combining the EMD and the NNM method provides a possible means for obtaining nonlinear properties in a structure.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.8 no.11
• /
• pp.4050-4067
• /
• 2014

In wireless sensor networks, link scheduling is a fundamental problem related to throughput capacity and delay. For a given set of communication requests $L=\{l_1,l_2,{\cdots},l_n\}$, the MLS (maximum link scheduling) problem aims to find the largest possible subset S of Lsuch that the links in S can be scheduled simultaneously. Most of the existing results did not consider bidirectional transmission setting, which is more realistic in wireless sensor networks. In this paper, under physical interference model SINR (signal-to-noise-plus-interference-ratio) and bidirectional transmission model, we propose a constant factor approximation algorithm MLSA (Maximum Link Scheduling Algorithm) for MLS. It is proved that in the same topology setting the capacity under unidirectional transmission model is lager than that under bidirectional transmission model. However, compared with some work under unidirectional transmission model, the capacity of MLSA is improved about 28% to 45%.

• The Journal of the Acoustical Society of Korea
• /
• v.34 no.3
• /
• pp.184-191
• /
• 2015

In this paper, a normal mode reverberation model for a range-independent environment of shallow water is proposed to calculate the reverberation level in the low-frequency range. Normal mode is used to calculate the acoustic energy propagating from the source to the scattering area and from the scattering area to the receiver. Each mode is decomposed into up and down going waves to consider scattering strength at the scattering area. The scattering functional form combines Lambert's law with a Gaussian-like term near the specular direction based on Kirchhoff approximation considering bottom condition. For verification of the suggested model, the result is relatively compared to several solutions of the problem XI and XV in the Reverberation Modeling Workshop I sponsored by the US Office of Naval Research.

• Transactions of the Korean Society of Mechanical Engineers
• /
• v.12 no.5
• /
• pp.944-951
• /
• 1988

Global-Local Laminate Variational Model is utilized to investigate the characteristics of interlaminar stresses in thick composite laminates under uniform axial extension. Various laminates with different fiber orientation and stacking sequences are analyzed to observe the behavior of interlaminar normal stresses. From this result, the interlaminar normal stress distribution along the laminate interfaces is examined and discussed with an existing approximation model. The repeated stacking of Poisson's ratio symmetric sublaminates is found to be the best stacking method of thick composite laminates to reduce the interlaminar normal stresses for the prevention of the free-edge delamination.

• Journal of Mechanical Science and Technology
• /
• v.14 no.1
• /
• pp.93-102
• /
• 2000

The objective of the present study is to investigate the pressure-strain correlation terms of the Reynolds stress models for the three dimensional turbulent boundary layer in a $30^{\circ}$ bend tunnel. The numerical results obtained by models of Launder, Reece and Rodi (LRR) , Fu and Speziale, Sarkar and Gatski (SSG) for the pressure-strain correlation terms are compared against experimental data and the calculated results from the standard k-${\varepsilon}$ model. The governing equations are discretized by the finite volume method and SIMPLE algorithm is used to calculate the pressure field. The results show that the models of LRR and SSG predict the anisotropy of turbulent structure better than the standard k-${\varepsilon}$ model. Also, the results obtained from the LRR and SSG models are in better agreement with the experimental data than those of the Fu and standard k-${\varepsilon}$ models with regard to turbulent normal stresses. Nevertheless, LRR and SSG models do not effectively predict pressure-strain redistribution terms in the inner layer because the pressure-strain terms are based on the locally homogeneous approximation. Therefore, to give better predictions of the pressure-strain terms, non-local effects should be considered.

• Transactions of the Korean Society of Mechanical Engineers
• /
• v.18 no.8
• /
• pp.2113-2122
• /
• 1994

The 3-D FEM analysis for simulating the stamping operation of planar anisotropic sheet metals with arbitrarily-shaped tools is introduced. An implicit, incremental, updated Lagrangian formulation with a rigid-viscoplastic constitutive equation is employed. Contact and friction are considered through the mesh-normal, which compatibly describes arbitrary tool surfaces and FEM meshes without depending on the explicit spatial derivatives of tool surfaces. The consistent full set of governing relations, comprising equilibrium equation and mesh-normal geometric constraints, is appropriately linearized. The linear triangular elements are used for depicting the formed sheet, based on membrane approximation. Barlat's non-quadratic anisotropic yield criterion(strain-rate potential) is employed, whose in-plane anisotropic properties are taken into account with anisotropic coefficients and non-quadratic function parameter. The planar anisotropic finite element formulation is tested with the numerical simulations of the stamping of an automotive hood inner panel and the drawing of a hemispherical punch. The in-plane anisotropic effects on the formability of both mild steel and aluminum alloy sheet metals are examined.

• Communications for Statistical Applications and Methods
• /
• v.18 no.3
• /
• pp.391-401
• /
• 2011

Consider a cluster, so called a 'son cluster', whose mean and variance is composed of the means and variances of both clusters called as a 'father cluster' and a 'mother cluster'. In this paper, a method for identifying each of three clusters is provided by modeling the relationship with father and mother clusters. Under the normal mixture model, the parameters are estimated via EM algorithm. We were able to overcome the problems of estimation using ECM approximation. Numerical examples show that our method can effectively identify the three clusters, so called a 'family of clusters'.