• Nuclear Engineering and Technology
• /
• 제54권10호
• /
• pp.3864-3877
• /
• 2022

We propose a new approach called PESA (Prioritized replay Evolutionary and Swarm Algorithms) combining prioritized replay of reinforcement learning with hybrid evolutionary algorithms. PESA hybridizes different evolutionary and swarm algorithms such as particle swarm optimization, evolution strategies, simulated annealing, and differential evolution, with a modular approach to account for other algorithms. PESA hybridizes three algorithms by storing their solutions in a shared replay memory, then applying prioritized replay to redistribute data between the integral algorithms in frequent form based on their fitness and priority values, which significantly enhances sample diversity and algorithm exploration. Additionally, greedy replay is used implicitly to improve PESA exploitation close to the end of evolution. PESA features in balancing exploration and exploitation during search and the parallel computing result in an agnostic excellent performance over a wide range of experiments and problems presented in this work. PESA also shows very good scalability with number of processors in solving an expensive problem of optimizing nuclear fuel in nuclear power plants. PESA's competitive performance and modularity over all experiments allow it to join the family of evolutionary algorithms as a new hybrid algorithm; unleashing the power of parallel computing for expensive optimization.

• 대한기계학회논문집A
• /
• 제30권11호
• /
• pp.1447-1454
• /
• 2006

The evaluation of parallel performance of a high speed impact simulation is not an easy task because not only the development of parallel explicit code is difficult but also a large number of processors is not easily accessible. In this paper, the parallel performance of a new Lagrangian FEM impact code carried out on cluster supercomputer has been described in high speed range. In the case of metal sphere impacting to oblique plate, the overall speed-up continuously increases even up to 128 CPUs. Investigation of elapsed time of each part reveals that most of the inefficiency comes from the load imbalance of contact.

• Computers and Concrete
• /
• 제18권5호
• /
• pp.1019-1039
• /
• 2016

In this paper, a parallel algorithm of nonlinear dynamic analysis of three-dimensional (3D) reinforced concrete (RC) frame structures based on the platform of graphics processing unit (GPU) is proposed. Time integration is performed using Newmark method for nonlinear implicit dynamic analysis and parallelization strategies are presented. Correspondingly, a parallel Preconditioned Conjugate Gradients (PCG) solver on GPU is introduced for repeating solution of the equilibrium equations for each time step. The RC frames were simulated using fiber beam model to capture nonlinear behaviors of concrete and reinforcing bars. The parallel finite element program is developed utilizing Compute Unified Device Architecture (CUDA). The accuracy of the GPU-based parallel program including single precision and double precision was verified in comparison with ABAQUS. The numerical results demonstrated that the proposed algorithm can take full advantage of the parallel architecture of the GPU, and achieve the goal of speeding up the computation compared with CPU.

• 디지털콘텐츠학회 논문지
• /
• 제10권2호
• /
• pp.249-258
• /
• 2009

최근 무선 인터넷 시장의 발전으로 모바일 단말기를 이용한 응용 개발이 늘어나고 있다. PDA 같은 모바일 장치는 유비쿼터스 컴퓨팅이라는 장점으로 인해 컴퓨팅을 요구하는 다양한 환경에서 필수적인 요소로 자리 잡고 있다. 본 논문의 목적은 PDA 클러스터 시스템을 이용해 PDA 단독으로는 실행시킬 수 없었던 대용량 소프트웨어를 PDA 상에서 실행시키는 것이다. 구체적인 구현 방법으로는 기존 워크스테이션 클러스터 컴퓨팅 엔진 JPVM을 PDA로 이식한 버전인 PDA-JPVM을 이용하였다. PDA 클러스터 상에서 병렬 응용 프로그램들을 실행시킨 결과, 이식된 PDA 클러스터 시스템을 이용해 대규모 소프트웨어를 PDA 상에서 실행시킬 수 있음을 확인하였으며, 아울러 그 성능 평가 결과도 보인다.

• 대한수학회논문집
• /
• 제10권3호
• /
• pp.735-750
• /
• 1995

We present a domain decomposition algorithm for solving large sparse linear systems of equations arising from queuing networks. Such techniques are attractive since the problems in subdomains can be solved independently by parallel processors. Many of the methods proposed so far use some form of the preconditioned conjugate gradient method to deal with one large interface problem between subdomains. However, in this paper, we propose a "nested" domain decomposition method where the subsystems governing the interfaces are small enough so that they are easily solvable by direct methods on machines with many parallel processors. Convergence of the algorithms is also shown.lso shown.

• 대한용접접합학회:학술대회논문집
• /
• 대한용접접합학회 2003년도 추계학술발표대회 개요집
• /
• pp.156-158
• /
• 2003

An implicit finite element implementation for Leblond's transformation plasticity constitutive equations, which are widely used in welded steel structure is proposed in the framework of parallel computing. The implementation is based upon the multiplicative decomposition of deformation gradient and hyper elastic formulation. We examine the efficiency of parallel computation for the finite element analysis of a welded structure using domain-wise multi-frontal solver.

• 대한용접접합학회:학술대회논문집
• /
• 대한용접접합학회 2002년도 춘계학술발표대회 개요집
• /
• pp.161-163
• /
• 2002

An implicit finite element implementation for Leblond's transformation plasticity constitutive equations, which are widely used in welded steel structure is proposed in the framework of parallel computing. The implementation is based upon the updated Lagrangian formulation. We examine the efficiency of parallel compuatation for the finite element analysis of a welded structure using multi-frontal solver.

• 대한산업공학회지
• /
• 제1권2호
• /
• pp.51-56
• /
• 1975

A computer program for computing complex system reliability is described. The program is composed of three phases : Phase I program reduces all series, parallel and series-parallel components and subsequently obtains an irreducible non-series-parallel system. Phase II program enumerates all the possible paths from the source to the sink of the graph. Phase III program then computes system reliability based on the information obtained by the Phase II program. The program is based on a modified version of the algorithm published in [6]. An example of the use of the computer program is given.

• 한국전산유체공학회:학술대회논문집
• /
• 한국전산유체공학회 2006년도 PARALLEL CFD 2006
• /
• pp.185-187
• /
• 2006

• 한국통신학회논문지
• /
• 제41권11호
• /
• pp.1490-1501
• /
• 2016

본 논문에서는 인메모리(In-memory) 병렬 분산 처리 시스템 Apache Spark(이하 Spark)를 활용하여 사용자에게 실시간 측위 정보를 제공할 수 있는 영상 기반 실내 위치인식 시스템을 제안한다. 제안하는 시스템에서는 사용자에게 실시간 측위 정보를 제공하기 위해서, Spark를 이용한 영상 특징점 추출 알고리즘의 병렬 분산화를 통해 알고리즘 연산 시간을 단축시킨다. 하지만 기존의 Spark 플랫폼에서는 영상 처리를 위한 인터페이스가 존재하지 않아, 영상 처리와 관련된 연산을 수행하는 것이 불가능하였다. 이에 본 논문에서는 Spark 영상 입출력 인터페이스를 구현하여 측위 연산을 위한 영상 처리를 Spark에서 수행 가능하게 하였다. 또한 무손실 압축(lossless compression)기법을 이용하여 특징점 기술자(descriptor)를 압축된 형태로 데이터베이스에 저장하여, 대용량의 실내 지도 데이터를 효율적으로 저장 및 관리하는 방법을 소개한다. 측위 실험은 실제 실내 환경에서 수행하였으며, 싱글 코어(Single-core) 시스템과의 성능 비교를 통해 제안하는 시스템이 최대 약 3.6배 단축된 시간으로 사용자에게 측위 정보를 제공 할 수 있다는 것을 입증하였다.