【作者】 朱明娟；

【导师】 刘玉身；

【作者基本信息】 清华大学 ， 软件工程（专业学位）， 2016， 硕士

【摘要】 子模型提取与模型合并是BIM(Building Information models,建筑信息模型)研究领域的基础性问题,在协同设计、版本管理、数据交换、规范检查中具有广泛的应用。现有的子模型提取方式主要有3种:基于数据库、本体和IFC(Industry Foundation Classes,工业基础类)文件,前两种方式都有信息转换的代价,而基于IFC文件的提取方式,现有算法对冗余实例很敏感,也不能消除无效实例和无关实例。现有的模型合并算法往往是基于GUID(Globally Unique Identifier,全局唯一标示符)进行实例匹配,而GUID在IFC文件导入导出过程中常常会发生变化。针对上述问题,本文提出了基于IFC文件实例的BIM子模型提取和模型合并算法,主要工作包括:(1)提出了一种基于实例构建的IFC子模型提取算法。该算法以待提取的IFC文件和用户的子模型提取需求为输入。根据原始IFC文件实例之间的层次引用结构,自底向上遍历所有实例,在去除冗余实例过程中既考虑了与需求相关的物理实例也考虑了关系实例,最终递归构建出所有相关实例并输出子模型。与现有方法相比,本文算法在子模型提取过程中能够去除所有冗余实例,并考虑了Ifc Owner History和无效实例的影响,使得提取的子模型更加紧致,能更好的体现用户需求。为了验证算法的有效性,本文收集整理了一个包含292个IFC模型的Benchmark数据集,并在该数据集上进行了算法验证和比较,实验结果表明本算法能够得到更紧致,实例数更少,正确率更高的子模型文件。(2)提出了一种基于实例匹配的IFC模型合并算法。主要应用于具有一致文件实例和坐标系的两个或者多个IFC模型的合并。该算法对输入文件中建筑构件的物理实例进行匹配,其中,利用基于内容匹配的策略来匹配非几何信息,利用几何展现形式对比和Hausdorff距离度量来匹配物理实例的几何信息。完成物理实例匹配后,再递归删除无关实例,最终得到合并的IFC模型并输出。与现有的基于GUID匹配的合并方法相比,本文提出的算法对GUID变化不敏感,合并算法更加鲁棒,合并后的IFC模型更加紧致。在本文实验测试中,与基于GUID匹配的合并方法相比,合并后的实例数平均可以多减少54%。更多还原

【Abstract】 Partial model extraction and model merging are fundamental problems in the research field of BIM(Building Information Models),which is widely used in collaborative design,version management,data exchanging and rule checking.There are mainly three methods for partial model extraction,including database-based methods,ontology-based methods and IFC(Industry Foundation Classes)files based methods.The former two methods often lose information in file conversion.As for IFC files based methods,the existing implementations are sensitive to redundant data instances,and the obtained results include many syntax errors and irrelevant data instances.The previous method of model merging for BIM is mainly based on GUID(Global Unique Identifier)of data instances.However,this method is not stable since GUID is often changed when importing and exporting from different design platforms.To address the above issues,this paper proposes the algorithms of partial model extraction and model merging for BIM based on file instances.The main work of this paper is as follow.(1)An algorithm about instances construction based partial model extraction for IFC files.It begins with a classification for the extraction building elements in terms of user intent and a comprehensive analysis of the structure and content for the IFC file.Then establishes a hierarchical structure with bottom-up traversing.Elimates all the redundant instances and constructs a partial model with physical and nonphysical instances.In addition,a performance benchmark is provided by 292 IFC files in the expreriment part.In contrast with previous method,the presented algorithm takes the Ifc Owner History,the redundant and the syntax error instances into account which can get a more accurate,more compact and smaller partial model.(2)Another algorithm is proposed for instances matching based on the model merging for IFC files.This algorithm aims to match the physical instances in two IFC files.To detail,the non-geometrical information is first compared using content-based automatic comparison algorithm.Then,the geometrical information is matched based on the ananlysis of 3D shape representations,where the Hausdorff distance is used to measure the distance between the discretized point sets.Finally,based on the matched physical instances,all the repeated physical and irrelevant data instances are recursively deleted to form the merged file.The proposed approach ismore robust and insensitive to unstable GUID.Our method is superior to the GUID-matched approaches with 54% less instances on average.更多还原