QQ客服 官方微博 反馈留言

题名	长周期运行管控的GPU计算实验研究
作者	郎小超
学位类别	工学硕士
答辩日期	2011-05-27
授予单位	中国科学院研究生院
授予地点	中国科学院自动化研究所
导师	郭大蕾
关键词	人工系统建模 计算实验 GPU 并行计算 Artificial systems modeling computational experiments GPU parallel computing
其他题名	GPU Computational Experiments for the Control and Management of Long-period Operation
学位专业	控制理论与控制工程
中文摘要	现代长周期工业安全运行的控制与管理，与计算机技术与智能信息控制技术的发展密切相关，近年来获得了许多有益的进展。同时，大量在实践中获取成功管控的长周期企业，迫切需要根据自身系统的特点，结合复杂系统智能控制与管理的最新理论与技术，开创大型复杂工业系统在数据处理、信息控制以及人员管理等方面的控制管理与决策体系。论文基于化工行业在长周期运行管理方面的成功实践，在设备状态监测、人员行为管理与制度考核等信息处理方面引入具有高性能并行计算能力的GPU（Graphics Processor Unit）方法，对大规模复杂系统的控制与管理进行计算实验分析与研究。 论文根据真实的生产系统建立起人工系统，并对人工系统进行计算实验。首先对长周期安全运行控制与管理系统进行建模分析，将其分为社会性模型和工程型模型，主要涉及操作人员、管理规则、设备以及工艺流程等。其中，人员和管理制度基于行为分析进行模糊逻辑评价；工艺流程模型结构单元包括了管道、分离塔、流程控制和流程故障四类基本单元模型；在设备的可靠性分析方面，使用蒙特卡洛模拟来计算，作为评价管理规则和掌握设备状况的依据。 论文选取了模糊逻辑规则计算和裂解炉失效率的蒙特卡洛模拟两个实例，将GPU的计算应用于人员管理和设备可靠性的模拟。对比传统CPU串行计算，对GPU计算任务进行了划分和分析，分别采用了单片和双片GPU，对企业级管理规则与设备运行状况进行了数万次模拟计算，获得了与传统CPU的串行计算相比最高47倍和22倍的加速效果，双片GPU最高达到85倍。实验结果验证了GPU并行计算对管理规则和设备评估的作用。 大型连续流程工业的控制与管理涉及人员、管理、设备、技术等信息与数据，其长周期安全运行同经济与社会效益密切相关，GPU计算的大规模密集运算将为人员管理、制度评估、设备监测与安全运行提供并行计算处理平台。
英文摘要	Increased with the development of computer technology and intelligent information control technology, recently the control and management of modern long-period operation industries have made great advances in the field of security production. Simultaneously, a large number of enterprises succeeded in long-period operation, eagerly develop the theories and techniques in intelligent control and management for the complex industrial environments, and construct a decision system of large complex industrial system in data procession, information and staff management by the terms of their own special characteristics of enterprises. Based on the successful practice in the management of long-term operation in chemical industry, this article apply Graphics Processor Unit (GPU), which has high performance in parallel computing capability to equipment condition monitoring, staff behavior management and institution examine. Firstly, an artificial system is established based on the analysis of the actual production system, which provides the model for the next step of computational experiment. In context, the intelligent computing model of control and management in long-cycle safe operation is analyzed, and mainly divided into social models and project models including staff behaviors, management and equipment rules. Staff and management models were classed into four calculating rules, which is evaluated by using fuzzy logic，and process model four basic units including pipelines, separator, control and faults of equipments and so on. Furthermore, for equipment reliability, model with Monte Carlo method is adopted to simulate and evaluate the management rules and master the equipment status. In the article, both of computational experiments by fuzzy logic and Monte Carlo simulation for ethylene cracking furnace failure are adopted to simulation to personnel management and equipment reliability by the means of GPU. Compared with the traditional CPU serial computation, for single and double pieces GPU, millions calculation of management rule show 47 times and 22 times acceleration effect in maximum, and dual-chip GPU up to 85 times in maximum. The experimental results indicated the GPU parallel computing could play an important role in rules and equipment evaluation for long-term industrial operation. The control and management of large-scale continuous process industry contains staff, management, equipment, technology and many other complex factors, the long-term o...
语种	中文
其他标识符	200828014628010
内容类型	学位论文
源URL	[http://ir.ia.ac.cn/handle/173211/7579]
专题	毕业生_硕士学位论文
推荐引用方式 GB/T 7714	郎小超. 长周期运行管控的GPU计算实验研究[D]. 中国科学院自动化研究所. 中国科学院研究生院. 2011.

个性服务

查看访问统计

相关权益政策

暂无数据

收藏/分享

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。