欢迎来到天天文库
浏览记录
ID:36785764
大小:3.42 MB
页数:133页
时间:2019-05-15
《过程系统优化的分布式并行计算》由会员上传分享,免费在线阅读,更多相关内容在行业资料-天天文库。
1、浙江大学博士学位论文过程系统优化的分布式并行计算姓名:张帆申请学位级别:博士专业:系统工程指导教师:钱积新;邵之江2002.5.1旦———————一塑垩查兰苎主堂竺兰奎群系统进行并行计算。在一个精馏塔算例的仿真计算中显示了该算法的有效性:提出了一种合适并行计算的大规模分解协调算法,研究用SQP算法进行底层优化、拟牛顿法作为协调算法开展大规模过程系统优化计算。此种并行算法体系所具有的粗粒度特征使之非常适合于用机群系统来实现。通过一个换热器的算例,表明该算法的计算优越性及良好的并行性能。提出了并行优化软件平台的总体设计框架,并设计了初始化函数,发送接收任务函数、进行任务函数和回
2、送并综合结果函数四大类的并行基础函数。在这些并行基础函数的基础上,实现了并行SQP与并行分解协调算法。从伪代码的分析中表明,利用四大类基础函数实现并行计算是简单而有效的。y塑垩奎兰堡主兰垒丝苎!!!AbstractProcessoptimizationhasemergedasoneofthemostvaluabletechniquesforsystemdesign,analysisandoperation.Currently,therapidtrendstowardincreasedmodeldetailandrigor,dynamicoptimization,on-lin
3、eoptimization,andschedulingacceleratetheneedtooptimizeverylargesystemsofequationswithmanydegreesoffreedom.Thelarge-scaleoptimizationofprocesssystems,however,continuestopresentamajorchallengebothinacademiaandinprocessindustries.Evenwiththehighperformancecomputersnowadays,therestillexistsman
4、ydifficultiesforasinglecomputertosolvelarge—scalechemicalprocessoptimizationproblems.Parallelcomputingwithclusterofworkstations,isahighperformance/pricesolutiontOsolvelarge·scaleprocessoptimizationproblems.Thisdissertationdetailstheinvestigation,developmentandimplementationofefficientalgor
5、ithmsandtechniquesforparalleloptimizationofprocesssystems.Themaincontributionsareasfollows:1)Thetheoryandobjecttoparallelcomputingtechnologyforprocessoptimizationproblemsarereviewed.Manyparallelalgorithmsandparallelmodelsarediscussed.Detailsofclusterofworkstations。arelativelyrecentlydevelo
6、pedtechnology,aregiventohi曲lightitsadvantagescomparedwithotherapproachesinparallelcomputing.Anewsolutionwithclusterofworkstationstosolvelarge-scaleprocessoptimizationpmblamsisproposed.TheplantCanuseclusterofworkstationstogainstrongcomputationpoweratlowcost.Observatiomindicatethatthedegreeo
7、fgranularityplaysamajorroleinthisappmaeh.Itshouldbecarefullyschemedtobalancetheloadofcommunicationandthedistributedcalculationsteps.Howtoimprovecomputationefficiencyinclusterofworkstationsisalsodiscussed.2)Severalparalleloptimizationalgorithmsarediscussedandev
此文档下载收益归作者所有