【作者】 韩春华；

【导师】 易思蓉；

【作者基本信息】 西南交通大学 ， 道路与铁道工程， 2008， 博士

【摘要】 铁路选线设计是整个铁路设计中事关全局的总体性工作,是集创造性、知识性与实践性于一体的多目标决策过程。其基本流程是:从社会经济需求出发,结合项目所处自然、生态环境,大面积着手,逐步细化,确定建设项目的技术标准、空间位置,协调布设各种建筑物。该决策过程受地形、水文、地质、土地利用等因数约束,需满足各种技术规范和运输要求,处理各种字符、数值、图形、图像资料,要考虑各种定性、定量、确定性和不确定的信息,加之评价指标不一致,使得该决策过程异常复杂,难以建立统一的数学模型。以图形交互为主要特征的计算机辅助设计的普及以及铁路勘测勘测设计一体化的推进,选线设计工作的速度、质量都有很大提高,但未能从根本上解决选线设计中方案有限、决策周期长、评价指标单一、设计劳动强度大以及采用方案难以有效论证等问题。铁路选线设计任务的复杂性、创造性、重要性决定对设计方法进行深入研究是必要的,而常规软件方法未能有效解决问题,需要探索智能化方法。因此铁路选线设计的智能化研究对变革铁路选线方法具有重要意义。借鉴国内外人工智能在铁路、公路选线设计中的应用经验,结合铁路选线设计的特点,铁路选线设计智能化的实现应该:1)有效识别和利用空间地理信息;2)协同人和计算机在智能选线中的智能职能;3)联合使用数值和符号的方法;4)从较高抽象层次和领域应用角度统一规划选线设计知识。为此论文提出了基于GIS铁路选线系统智能环境(GBGBRLIE)的综合解决方案,即基于知识工程和GIS技术,构建具备专家级问题求解能力,追求知识综合运用,集辅助性、创造性、教育性于一体,具有沉浸感的人机协同设计复杂智能信息系统。论文主要进行了以下几个方面的工作:1)提出了GBRLIE的本体建模方法并建立了铁路选线设计领域本体。借鉴知识系统开发经验,根据铁路选线对知识模型的共享、交互以及互操作等要求,提出使用本体建立铁路选线知识模型,并分析了本体在GBRLIE的地位和作用;结合GBRLIE特点,提出使用改进迭代法、UML语言和微软的Visio构建本体;提出并详细描述了采用面向对象技术和UML轻量级扩展的领域本体形式化表示方式;最后构建了铁路选线设计的领域本体,包括线路本体、选线地理环境本体和线路评价本体。2)智能环境中人机协同的知识表示、获取与利用技术。a)将选线系统知识分为面向领域人员和面向计算机知识,基于本体构建面向计算机知识模型和概念关系词典,基于概念关系词典建立面向领域人员的知识模型,面向计算机和面向领域人员的知识通过概念关系词典映射。b)面向领域人员知识概念上用语义网络表示,实现时采用面向对象和流文档技术表示为带元知识的知识森林。c)用面向对象技术表示面向计算机知识:一般知识类表示事实知识,规则采用层次模型和双层形式化模型,把问题求解知识表示为推理机控制知识,实现推理机与控制策略的分离。d)提出并实现了提案一目标验证通用嵌入式推理,即通过提出可能方案,经过一系列推理,然后再验证其是否满足目标条件的方式解决问题。e)提出了GBRLIE中符号推理和数值计算的集成方法。通过采用统一的知识模型,利用面向对象技术表示知识对象,并把这些知识内部表示为可执行形式,实现符号推理时直接使用数值方法中的对象;以及推理机按组件形式开发,在常规软件系统中直接调用推理机。3)提出了铁路选线系统智能环境中地理信息相关知识与信息的获取方法。首先讨论GBRLIE中地理信息获取与利用的任务层次,结合人机不同的信息利用方式提出了人机协同的地理信息利用模型。然后阐述基于本体以及遥感图像建立铁路选线地理信息数据库的方法。接着介绍GBRLIE中地形分析与可视化技术以及基于地理信息的线路评价知识获取方法。最后提出了基于知识与栅格路径分析的线路初始平面生成方法:通过采用16单元的骑士邻域单元模式,在费用模型上考虑桥梁、隧道、线路长度相关费用以及线路坡度、速度目标值,改进了Dijkstra算法,在多个控制点间生成栅格最优路径;利用平面设计相关知识和提案一目标验证推理机,实现自动曲线配置;最后通过二分渐进和自动曲线配置,依据最优路径生成线路初始平面。4)构建了基于地理信息系统的铁路选线智能平台。通过统一规划铁路选线设计知识,采用人机协同设计理念,基于地理信息系统,利用人机协同的知识模型,联合软件工程与知识工程的知识建模方法,借助提案——目标验证通用推理,建立了基于地理信息系统的铁路选线智能平台。5)提出了铁路选线系统智能环境的选线方法。a)采用方法本体表示与领域无关的问题求解方法,并用UML和面向对象技术描述方法本体;根据铁路选线设计任务需要,提出并描述了多目标决策问题求解的多目标决策方法本体,提出并描述了采用“生成—筛选—改进”方法解决的,追求在多个目标值上全局最优或最满意的创造性多目标(IMO)设计方法本体。b)通过改造方法本体适应领域本体和应用的方法,把多目标决策和IMO设计方法本体用于解决新建单线铁路局部方案和局部方案纵断面的自动设计问题。c)利用局部方案自动设计方法,借助GBRLIE的地形可视化技术和人机协同地理信息利用模型,基于IMO设计的“生成—筛选—改进”思想,提出了既能发挥传统技术标准选择和技术标准综合优化方法优点,避免各自缺点,又能实现技术标准多目标比选的人机协同技术标准比选方法。6)实现了新线铁路快速方案比选原型系统。为验证所提方法的科学性以及可行性,提出并实现了新线铁路快速方案比选原型系统,描述了其体系结构、实现方法和主要功能,并给出了实验验证结果。结果表明论文理论正确,方法切实可行。论文的研究结果可为铁路选线自动化或智能化的实现提供理论依据和可操作方法。研究是针对新建铁路设计的,研究成果经改进可用于增建二线及既有线改建设计,也可用于公路路线设计系统开发。研究成果还可供其他研究具有铁路选线设计类似特征的空间决策问题或土木工程设计问题的人员参考。更多还原

【Abstract】 Railway location is a synthetic work concerning railway design,and a process of multi- objective decision making that is characteristic of creativeness,practicalness and much knowledge.The process is:according to community economy,natural and ecological environment of concerning project,by studying at a large area and then gradually focus on specific problem,to determine railway main technical standard and spatial location,locate all kinds of constructions.The decision making process is restricted by terrain, hydrological,geologic factor and land use,meet technical codes and traffic needs,need to handle information that is qualitative,quantitative,deterministic or probabilistic,with form of character,numeric value,graph and image.The process also does not has consistent evaluating indicator.So the process is very difficult and cannot be solved by numeric model.With the development and use of CAD,and push of the integration of railway survey and design,the speed and quality of railway location has make great stride forward. But some basic problems of traditional method,such as few scheme,long decision period, intense labor,and lack of valid justification to adopted scheme,has not conquered.The importance,complexity and creativeness of railway location determine that it is necessary to further study the design method.Common software methods cannot tackle existing problems effectively,so it is significant to study intelligentizing railway location.After analysis of the situation of artificial intelligence application in road and railway location,and the character of railway Location,a conclusion can be drawn that,in order to actualize automatization of railway location,intelligent design method should:cognize and utilize geographic information,combine the intelligent function of human and computer, integrate method of character and numeric value,construct a uniform railway location knowledge model at a top abstract level.So,this thesis put forward GIS-based Railway Location Intelligent Environment(GBRLIE) as a synthetic solution.GBRLIE is a compound human-machine collaborative intelligent information system,based on GIS and knowledge engineering and with situated interface,which is capable of handling complex problem as experts,which pursues synthetic knowledge use and can act as assistant and educator.The main contributions of this thesis are:1) Ontology modeling method of GBRLIE is put forward and railway location domain onthologies is given.In light to lessons of knowledge system developing,the conclusion that railway design knowledge should be modeled is arrived at.After analysis of the requirement of railway location knowledge model in knowledge sharing and interoperability,and the essence of Ontology,Ontology is put forward as railway knowledge model method.In light to GBRLIE characters,GBRLIE anthologies is modeled with UML by Visio and improved iterative software engineering method,which include rail line ontology,railway location geographic environment ontology,rail line evaluation ontology.Also,in order to describe formal ontology,a method based on O-O technique and lightly extended UML is come up with.2) Human-machine collaborative knowledge representation,acquisition and utilization technology in GBRLIE.a) Knowledge of railway location is divided into Computer Oriented Knowledge(COK) and Domanial Person Oriented Knowledge(DPOK),both of which are erected on ontology,and connected to each other by Concept-Relation Dictionary(CRD).b) The conceptual model of DPOK is represented by semantic network, and is described by knowledge forest with meta-Knowledge in real use by O-O and Microsoft flow-document technology,c) COK also is represented by O-O technology: factual Knowledge is depictedby ordinary class with special fields;rule is depicted by a model of hierarchy and bi-formalization;problem solving knowledge is described as control knowledge which act as inference strategy of inference machine and can separate from inference machine,d) A Propose and Goal Test Generic Inference machine(PGTGI) is proposed and realized,which solve problem by approach as:coming up with a possible scheme,then reasoning and test if the goal is met.e) character reasoning and numeric calculation can be integrated by adopting uniform model and developing component inference machine.3) Technologies of geographic knowledge and information acquisition in GBRLIE. Firstly,According to task levels of Geographic Information Acquisition and Utilization (GIAU) in GBRLIE,a human-machine collaborative GIAU model is proposed.Then method of construct geo-database scheme based on ontology and remote sense image is introduced.Afterword,approach of terrain analysis,terrain visualization and rail line evaluation knowledge acquisition based on geographic information is put forward. Eventually method of automatically generating initial horizontal alignment according to raster least cost path and railway location knowledge is bring forward.In order to generate feasible path among multi control Points,the Dijkstra algorithm is improved by using Knight Neighbor mode and cost model considering bridge,tunnel,profile grade and length concerning cost.With PGTGI,horizontal alignment curve auto setting can be easily fulfilled.By gradual dichotomy approaching and alignment curve auto setting,least cost path can be convert to initial horizontal alignment that meet design code.4) An intelligent railway location platform based on GIS is set up.Based on GIS,with unified railway location knowledge,human-machine collaborative design thought and model,knowledge modelling method of unifying software and knowledge engineering knowledge,and PGTGI,the intelligent railway location platform established.5) Railway location methods in GBRLIE is bring forward.a) Problem solving method (PSM) is represented by method ontology described by UML and O-O technology. According to actual needs of railway location Multi- Objective Decision Making(MODM) and Innovative Multi-Objective Design(IMOD) method ontology are brought forward.The former is the reusable method tackling MODM task,and the latter use "Propose-Filter-Improve(PFI)" method to solve IMOD.b) By adapting method ontology to domain use,MODM and IMOD method ontology are put into use to handle railway local scheme and accordingly profile auto design.c) A new approach of railway Main Technical Criteria Decision(MTCD) is put forward,based on PFI design thought,terrain visualization and Human-machine collaborative GIAU model in GBRLIE.The approach takes the advantages both of traditional MTCD and synthetic MTCD method,also avoids shortcoming of the both. 6) For testing the possibility and scientificity of proposed method in this thesis,Rapid Location Prototype System of New Railway Alternatives(RLPSNRA) is put forward and fulfilled,with its architecture,main function and implementing method.The test results of this system manifest that the theories are correct and methods are practical.Results of this thesis provide theories and operatable method for realization of railway location automatization and intelligentization.Though the study of GBRLIE is aimed to new railway design,the results can be used to solve design problem concerning reconstruction or double-tracking existing railway by improving.They also can be used to road location.Furthermore,those theories and methods can be referred by other people if they hope to tackle problems with regard to spatial decision or civil engineering design which has the same characteristics as railway location.更多还原