Using parallel programming paradigms to reduce errors of correlated simulation in estimation and scheduling

• Main Authors: Pin-Yi Lee, 李品毅
• Other Authors: I-Tung Yang
• Format: Others
• Language: zh-TW
• Published: 2011
• Online Access: http://ndltd.ncl.edu.tw/handle/mn5w38

碩士 === 國立臺灣科技大學 === 營建工程系 === 99 === The success of a construction project can be determined by aspects of its scheduling , costs and quality. Its quality standards have already been specified the moment the contract is signed. Therefore, how to plan the schedule and control the costs have become an important issue in project management. In a construction project individual operations often have some correlation with one another. For example, the delay in operation time will cause the costs to increase.However, when material cost increases,it influences the costs of all the related activities. Therefore, if the correlation between time and cost is ignored, there is a possibility in giving a inaccurate evaluation. It is necessary to quantify the correlations between operations to obtain more accurate estimation. fter the correlations are determined,they can be used o conduct simulations through NORmal To Anything (NORTA) and Iman and Conover(IC) to develop a cost estimation or schedule prediction.During the process,Cholesky factorization is needed to conduct a correlation simulation.But in situations when the original correlation matrix is not positive definite,the Cholesky factorization will generate imaginary roots,thus causing the simulation to fail.Evan thought many scholars have already presented methods to adjust the correlation matrix,the modified matrix causes inaccuracy is cost estimale and schedule prediction. Therefore, this study uses particl swarm optimizstion (PSO)to search for a feasible correlation matrix, which after correlation simulation will lead to minimum error.Since PSO and cooelation simulation are both computational expensive,this study investigates the use of a computer cluster and three parallel progrmming strategies in reducing computatial time.