• Proceedings of the KSME Conference
• /
• 2001.11a
• /
• pp.324-329
• /
• 2001

A novel method for non-matching interfaces on the boundaries of the finite elements in partitioned domains is presented by introducing interface elements in this paper. The interface element method (IEM) satisfies the continuity conditions exactly through interfaces without recourse to the Lagrange multiplier technique. The moving least square (MLS) approximation in the present study is implemented to construct the shape functions of the interface elements. Alignment of the boundaries of sub-domains in the MLS approximation and integration domains provides a consistent numerical integration due to one form of rational functions in an integration domain. The compatibility of displacements on the boundaries of the finite elements and the interface elements is always preserved in this method, and the completeness of the shape functions of the interface elements guarantees the convergence of numerical solutions. The numerical examples show that the interface element method is a useful tool for the analysis of a partitioned system and for a global-local analysis.

• 한국전산유체공학회:학술대회논문집
• /
• 2003.08a
• /
• pp.63-68
• /
• 2003

The Grid[1] is a communication service that collaborates dispersed high performance computers so that those can be shared and worked together. So, the Grid enables a researcher to analyze a huge-sized problem which was impossible by using local resources. However, diverse communication speeds among computing resources and heterogeneity of computing resources can reduce parallel efficiency in the Grid, The present paper focuses on the development of an efficient load balancing algorithm suitable for the Grid. Proposed algorithm classifies the whole processors into several groups with relatively faster communication speeds. Computational domain is firstly partitioned to each group and then to the processor level considering the performance of each processor. Developed algorithm is validated in the homogeneous system by comparing the present result with the result of equally partitioned meshes and then applied to the heterogeneous system. Additionally, the present algorithm is expanded to be able to solve the decomposed domains and applied to some problems.

• Proceedings of the KSRS Conference
• /
• 2005.10a
• /
• pp.571-574
• /
• 2005

In this paper, we propose an efficient image segmentation technique for large-sized monochrome images using a hybrid approach which combines threshold and region-based techniques. First, an image is partitioned into fixed-size blocks and for each block the representative intensity is determined by averaging pixel intensities within the block. Next, the neighborhood blocks that have similar characteristics with respect to a specific threshold are merged in order to form candidate regions. Finally, those candidate regions are refined to get final target object regions by merging regions considering the spatial locality and certain criteria. We have performed experiments on images selected from various domains and showed that our technique was able to extract target object regions appropriately from most images.

• Journal of Korean Society for Quality Management
• /
• v.30 no.3
• /
• pp.186-194
• /
• 2002

In this paper we introduce formulae for estimating the failure rate of a large scaled software by using the Bayesian rule when a black-box random testing which selects an element(test case) at random with equally likely probability, is performed. A program or software can be treated as a mathematical function with a well-defined (input)domain and range. For a large scaled software, their input domains can be partitioned into multiple subdomains and exhaustive testing is not generally practical. Testing is proceeding with selecting a subdomain, and then picking a test case from within the selected subdomain. Whether or not the proportion of selecting one of the subdomains is assumed probability, we developed the formulae either case by using Bayesian rule with gamma distribution as a prior distribution.

• Proceedings of the Computational Structural Engineering Institute Conference
• /
• 2009.04a
• /
• pp.139-145
• /
• 2009

Reduction scheme is remarkably useful in the case requiring the repeated calculation procedure. Recently, the efficiency of the reduction scheme has been improved by combining scheme of sub-domain method. But, when the global domain is partitioned into a few sub-domains, sub-domains without constraints can be produced. it is needed to extract the ritz vector from each sub-domain to construct the reduced system of each sub-domain. it is easy to extract the ritz vector from sub-domain with constraint. on the other hand, pseudo inverse method should be employed to extract the ritz vector from sub-domain without constraint. generally, the pseudo inverse takes a large number of computing time to obtain a reduced system of a sub-domain without boundary condition. This trouble can be overcome by the reduced pseudo inverse scheme which proposed in this study. This scheme is based on the static condensation that is not related with selection of the primary degrees of freedom. Numerical examples demonstrate that present method saves computational cost effectively and predicts the accurate eigenvalues.

• Journal of the Computational Structural Engineering Institute of Korea
• /
• v.22 no.2
• /
• pp.173-179
• /
• 2009

Reduction scheme is remarkably useful in the case requiring the repeated calculation procedure. Recently, the efficiency of the reduction scheme has been improved by combining scheme of sub-domain method. But, when the global domain is partitioned into a few sub-domains, sub-domains without constraints can be produced. it is needed to extract the ritz vector from each sub-domain to construct the reduced system of each sub-domain. it is easy to extract the ritz vector from sub-domain with constraint. on the other hand, pseudo inverse method should be employed to extract the ritz vector from sub-domain without constraint. generally, the pseudo inverse takes a large number of computing time to obtain a reduced system of a sub-domain without boundary condition. This trouble can be overcome by the reduced pseudo inverse scheme which proposed in this study. This scheme is based on the static condensation that is not related with selection of the primary degrees of freedom. Numerical examples demonstrate that present method saves computational cost effectively. In addition, it is shown that the reduced system based on the proposed scheme predicts the accurate eigenvalues of global system.

• Korean Journal of Computational Design and Engineering
• /
• v.2 no.2
• /
• pp.77-84
• /
• 1997

This paper is concerning the development of multiple neural networks system of problem domains where the complete input space can be decomposed into several different regions, and these are known prior to training neural networks. We will adopt oblique decision tree to represent the divided input space and sel ect an appropriate subnetworks, each of which is trained over a different region of input space. The overall architecture of multiple neural networks system, called the federated architecture, consists of a facilitator, normal subnetworks, and tile networks. The role of a facilitator is to choose the subnetwork that is suitable for the given input data using information obtained from decision tree. However, if input data is close enough to the boundaries of regions, there is a large possibility of selecting the invalid subnetwork due to the incorrect prediction of decision tree. When such a situation is encountered, the facilitator selects a tile network that is trained closely to the boundaries of partitioned input space, instead of a normal subnetwork. In this way, it is possible to reduce the large error of neural networks at zones close to borders of regions. The validation of our approach is examined and verified by applying the federated neural networks system to the configuration design of a midship structure.

• Smart Media Journal
• /
• v.7 no.3
• /
• pp.57-63
• /
• 2018

With many types of data generated in the shift of business environment as a result of growth of an organization or enterprise, there is a need to improve the data-processing efficiency in smarter means with a single domain model such as Data Lake. In particular, creating a logical single domain model from physical partitioned multi-site data by the finite resources of nature and shared economy is very important in terms of efficient operation of computing resources. Based on the advantages of the existing Data Lake framework, we define the CDA-Concept (connected data architecture concept) and functions of Data Lake Framework over Abyss Storage for integrating multiple sites in various application domains and managing the data lifecycle. Also, it performs the interface design and validation verification for Interface #2 & #3 of the connected data architecture-concept.

• Smart Media Journal
• /
• v.7 no.4
• /
• pp.30-36
• /
• 2018

Single domain model like DataLake framework is in spotlight because it can improve data efficiency and process data smarter in big data environment, where large scaled business system generates huge amount of data. In particular, efficient operation of network, storage, and computing resources in logical single domain model is very important for physically partitioned multi-site data process. Based on the advantages of Data Lake framework, we define and extend the concept of Connected Data Architecture and functions of DataLake framework for integrating multiple sites in various domains and managing the lifecycle of data. Also, we propose the design of CDA-based AI service and utilization scenarios in various application domain.

• Journal of the Korean GEO-environmental Society
• /
• v.25 no.4
• /
• pp.13-20
• /
• 2024

In this paper, a parallel analysis algorithm for Smoothed Particle Hydrodynamics (SPH), one of the numerical methods for fluidic materials, is introduced. SPH, which is a meshless method, can represent the behavior of a continuum using a particle-based approach, but it demands substantial computational resources. Therefore, parallel analysis algorithms are essential for SPH simulations. The domain decomposition algorithm, which divides the computational domain into partitions to be independently analyzed, is the most representative method among parallel analysis algorithms. In Discrete Element Method (DEM) and Molecular Dynamics (MD), the Cartesian coordinate-based domain decomposition method is popularly used because it offers advantages in quickly and conveniently accessing particle positions. However, in SPH, it is important to share particle information among partitioned domains because SPH particles are defined based on information from nearby particles within the smoothing length. Additionally, maintaining CPU load balance is crucial. In this study, a highly parallel efficient algorithm is proposed to dynamically minimize the size of orthogonal domain partitions to prevent excess CPU utilization. The efficiency of the proposed method was validated through numerical analysis models. The parallel efficiency of the proposed method is evaluated for up to 30 CPUs for fluidic models, achieving 90% parallel efficiency for up to 28 physical cores.