基于语义相似性匹配的BIM模型智能审查系统  

Intelligent Compliance Checking System for BIM Model Based on Semantic Similarity Matching

作　　者：黄隆盛 林晓东 徐盛取 熊朝阳 丁志坤[2,3,4,5] HUANG Longsheng;LIN Xiaodong;XU Shengqu;XIONG Zhaoyang;DING Zhikun(Guangdong Constructing Engineering Supervision CO.,LTD.,Guangzhou 510010,Guangzhou,China;Shenzhen University,College of Civil and Transportation Engineering,Sino-Australia Joint Research Center in BIM and Smart Construction,Shenzhen 518060,China;Key Laboratory of Coastal Urban Resilient Infrastructures(Shenzhen University),Ministry of Education,Shenzhen 518060,Guangdong,China;Guangdong Laboratory of Artificial Intelligence and Digital Econonmy,Shenzhen 518060,Guangdong,China;Shenzhen Key Laboratory of Green,Efficient and Intelligent Construction of Underground Metro Station,Shenzhen University,Shenzhen 518060,Guangdong,China)

机构地区：[1]广东省建筑工程监理有限公司,广东广州510010 [2]深圳大学,土木与交通工程学院、中澳BIM与智慧建造研究中心,广东深圳518060 [3]滨海城市韧性基础设施教育部重点实验室(深圳大学),广东深圳518060 [4]人工智能与数字经济广东省实验室(深圳),广东深圳518060 [5]深圳市地铁地下车站绿色高效智能建造重点实验室,广东深圳518060

出　　处：《建筑科学》2024年第6期138-145,共8页Building Science

基　　金：国家自然科学基金(71974132);深圳市科技计划资助(JCYJ20190808115809385);深圳市科技计划资助高等院校稳定支持计划重点项目(No.20220810160221001)。

摘　　要：本研究提出了1种基于语义相似度匹配的BIM模型智能审查Web系统。采用基于机器学习的自然语言处理技术,自动化提取和分析BIM模型的语义信息,并将其与基于IFC标准表达的规范知识图谱本体进行匹配。通过结合Levenshtein距离和Cosine Similarity余弦相似度2种相似匹配度量算法,对BIM模型的语义信息进行相似度匹配审查。此外对精确匹配与相似匹配算法进行比较评价,结果表明,相似匹配在整体审查性能上优于精确匹配,并且匹配准确性得到提高。该智能审查系统具备强大的通用性和适应性,能够支持多种行业规范的模型审查,扩大了BIM模型匹配研究的应用范围。研究成果将有助于推动建筑行业更高效地应用BIM技术,从而提升工程项目的管理水平和质量。This study proposed a web-based intelligent inspection system for BIM models based on semantic similarity matching.The application of machine learning-based natural language processing technology enables automated extraction and analysis of semantic information from BIM models.This information is then matched with a normative knowledge graph ontology expressed based on the IFC standards.A combination of two similarity matching metrics,namely,Levenshtein distance and Cosine Similarity,was employed to perform the similarity matching inspection of the BIM model′s semantic information.Additionally,a comparative evaluation was conducted between precise and similarity matching algorithms,the results of which indicate that similarity matching outperforms precise matching in overall inspection performance and enhances matching accuracy.The developed intelligent inspection system demonstrates substantial versatility and adaptability,capable of supporting the inspection of models following diverse industry standards and extending the application range of BIM model matching research.The findings of this study are expected to facilitate the more efficient application of BIM technology in the building industry,thereby elevating the management quality and efficiency of engineering projects.

关 键 词：BIM模型 合规性审查 自然语言处理 本体 语义映射 相似匹配 

分 类 号：TU831[建筑科学—供热、供燃气、通风及空调工程]