• 한국CDE학회논문집
• /
• 제10권3호
• /
• pp.162-167
• /
• 2005

A 3D solid element mesh generation algorithm was newly developed. 3D surface points of global rectangular coordinates were supplied by a 3D laser scanner. The algorithm is strait forward and simple but it generates hexahedral solid elements. Then, the surface rectangular elements were generated from the solid elements. The key of the algorithm is elimination of unnecessary elements and 3D boundary surface fitting using given 3D surface point data.

• Management Science and Financial Engineering
• /
• 제6권1호
• /
• pp.1-19
• /
• 2000

This papeer proposes a general approach of the multivariate expontially weighted moving average(MEWMA) chart, in which the smoothing matrix has full elements instead of only diagonal elements. The average run length (ARL) properties of this scheme are examined for a diverse set of quality control environments and the information to design the chhart is provied. Performance of the scheme is measured by estmating ARL and compared to those of two group cumulative sum (CUSUM) chats. The comparison resullts show that the MEWMA chart can improve its ARL performance in detecting a small shifts out-of-control in the start-up stage, the general MEWMA chart of a full smoothing matrix appears to offer an exceptional protection aginst departures from control in the process mean.

• 소성∙가공
• /
• 제13권4호
• /
• pp.334-341
• /
• 2004

In this study, a remeshing algorithm adapted to the mesh density map using the Delaunay mesh generation method is developed. In the finite element simulation of forging process, the numerical error increases as the process goes on because of discrete property of the finite elements and distortion of elements. Especially, in the region where stresses and strains are concentrated, the numerical error will be highly increased. However, it is not desirable to use a uniformly fine mesh in the whole domain. Therefore, it is necessary to reduce the analysis error by constructing locally refined mesh at the region where the error is concentrated such as at the die corner. In this paper, the point insertion algorithm is used and the mesh size is controlled by using a mesh density map constructed with a posteriori error estimation. An optimized smoothing technique is adopted to have smooth distribution of the mesh and improve the mesh element quality.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2001년도 ICCAS
• /
• pp.109.3-109
• /
• 2001

The so-called curse of dimensionality arises when Gaussian mixture is used on high-dimensional small-sample-size data, since the number of free elements that needs to be specied in each covariance matrix of Gaussian mixture increases exponentially with the number of dimension d. In this paper, by constraining the covariance matrix in its decomposed orthonormal form we get a local PCA model so as to reduce the number of free elements needed to be specified. Moreover, to cope with the small sample size problem, we adopt BYY data smoothing learning which is a regularization over maximum likelihood learning obtained from BYY harmony learning to implement this local PCA model.

• 대한기계학회논문집B
• /
• 제25권6호
• /
• pp.853-859
• /
• 2001

This paper describes the extended topological clean up procedures to improve the quality of unstructured triangular meshes. As a postprocessing step, topological improvement procedures are applied both for elements that are interior to the mesh and for elements connected to the boundary and then Laplacian-like smoothing is used by default. Previous clean up algorithms are limited to eliminate the nodes of degree 3,4,8,9,10 and pairs of nodes of degree 5. In this study, new clean up algorithms which minimize the triple connection structures combined with degree 5 and 7 (ie ; 5-7-5, 7-7-5, 7-5-7 etc) are added. The suggested algorithms are applied to two example meshes to demonstrate the effectiveness of the approach in improving element quality in a finite element mesh.

• 융합신호처리학회 학술대회논문집
• /
• 한국신호처리시스템학회 2003년도 하계학술대회 논문집
• /
• pp.141-144
• /
• 2003

This paper proposes a design method of a CDBC(Continuous-time Deadbeat Controller)system that takes into account the response between the sampling instant and using second-order smoothing elements. The continuous deadbeat controller is composed of a serial integral compensator and a local feedback compensator introduced into the state feedback loop. A DC servo motor is chosen for implementing CDBC algorithm. Especially according to the variable input and disturbance, corresponding CDBC design method is suggested. A Matlab Simulink is used for simulation with the Motor parameter. By computer simulations, control inputs and system outputs are shown to have desirable property such as smoothness.

• 대한토목학회논문집
• /
• 제12권3호
• /
• pp.163-171
• /
• 1992

수치표고모델에서 좁은 격자 간격은 정확도에 효과적이지만, 지나치게 조밀하게 격자화된 표본을 만들 수는 없기 때문에 유형화가 요구된다. 수치표고모텔에서 유한 요소법의 적용은 표고 취득과정에서 발생된 원하지 않는 오차를 유형화하기 위하여 효과적이며, 지형의 불연속에 대한 문제점을 직접적이고도 간단하게 처리할 수 있는 이상적인 기법이므로 본 연구에서는 유한요소법이 적용되었다. 유한 요소법에 의한 수치표고모델은 표고 경중률이 유형화에 특히 큰 영향을 주는데, 본 연구에서는 표고 경중률과 유형화의 표준편차 사이에 Y = 1900625 $X^4-312987.6$ $X^3+20330.72$ $X^2-578.6029$ X+12.63772의 상관식을 얻었으며, 이 식으로 부터 최적의 표고 경중률은 0.05 임을 알 수 있었다.

• 한국정밀공학회지
• /
• 제17권3호
• /
• pp.136-148
• /
• 2000

In this work, a general method for the mathematical description of three-dimensional trimmed surface is proposed by introducing the base parametric surface and boundary curves. Since mesh density distribution for the analysis may vary by cases, a grid-based mesh generation algorithm using quadtree is proposed in the present work. For the assurance of connectivity of generated meshes among surfaces, a method for the pre-cleaning of boundary curves has been developed to be used in the automatic generation of the finite elements. In addition, mesh-smoothing algorithm is suggested which can be used in the general trimmed surface. In this algorithm nodes are moved on the original surface by the normal projection in each iterative smoothing procedure.

• 한국인공지능학회지
• /
• 제11권1호
• /
• pp.17-24
• /
• 2023

With the continuous growth in the amount of data collected and analyzed, deep learning has become increasingly popular for extracting meaningful insights from various fields. However, hardware limitations pose a challenge for achieving meaningful results with limited data. To address this challenge, this paper proposes an algorithm that leverages the characteristics of convolutional neural networks (CNNs) to reduce the size of image datasets by 20% through smoothing and shrinking the size of images using color elements. The proposed algorithm reduces the learning time and, as a result, the computational load on hardware. The experiments conducted in this study show that the proposed method achieves effective learning with similar or slightly higher accuracy than the original dataset while reducing computational and time costs. This color-centric dataset construction method using image smoothing techniques can lead to more efficient learning on CNNs. This method can be applied in various applications, such as image classification and recognition, and can contribute to more efficient and cost-effective deep learning. This paper presents a promising approach to reducing the computational load and time costs associated with deep learning and provides meaningful results with limited data, enabling them to apply deep learning to a broader range of applications.

• Structural Engineering and Mechanics
• /
• 제39권6호
• /
• pp.767-793
• /
• 2011

An efficient edge-based smoothed finite element method (ES-FEM) has been recently developed for solving solid mechanics problems. The ES-FEM uses triangular elements that can be generated easily for complicated domains. In this paper, the complexity study of the ES-FEM based on triangular elements is conducted in detail, which confirms the ES-FEM produces higher computational efficiency compared to the FEM. Therefore, the ES-FEM offers an excellent platform for adaptive analysis, and this paper presents an efficient adaptive procedure based on the ES-FEM. A smoothing domain based energy (SDE) error estimate is first devised making use of the features of the ES-FEM. The present error estimate differs from the conventional approaches and evaluates error based on smoothing domains used in the ES-FEM. A local refinement technique based on the Delaunay algorithm is then implemented to achieve high efficiency in the mesh refinement. In this refinement technique, each node is assigned a scaling factor to control the local nodal density, and refinement of the neighborhood of a node is accomplished simply by adjusting its scaling factor. Intensive numerical studies, including an actual engineering problem of an automobile part, show that the proposed adaptive procedure is effective and efficient in producing solutions of desired accuracy.