Journal
COMPUTERS IN INDUSTRY
Volume 116, Issue -, Pages -Publisher
ELSEVIER
DOI: 10.1016/j.compind.2020.103186
Keywords
Information retrieval; Attribute similarity; Domain-specific retrieval; Building information modeling (BIM); Industry foundation classes (IFC)
Funding
- National Key R&D Program of China [2018YFB0505400]
- major R&D plan of China Railway Group [K2018G055]
Ask authors/readers for more resources
Building information modeling (BIM) has played a central role in architecture, engineering, and construction (AEC) industry, which also becomes an active research direction in smart buildings and smart cities. With the rapid development and popularization of BIM technology, online BIM resource libraries have grown rapidly. Fast and effective retrieval of BIM components from such great amount of resources has become an urgent demand. Traditional methods such as catalog browsing, keyword matching and shape matching are not capable of delivering satisfactory results, since they cannot extract the domain-specific information carried by BIM components. To resolve the aforementioned issue, we propose a novel similarity measurement and a new retrieval method, and integrate them into the BIMSeek system. The main contributions of our work are as follows. Firstly, we propose a novel algorithm for measuring the similarity between two BIM components based on their attribute information and Tversky similarity. Our proposed algorithm yields the best result in terms of Precision-Recall, F-measure and DCG compared to the traditional Tversky similarity measure and geometry similarity algorithm. Secondly, based on our proposed similarity measurement algorithm, we propose a novel retrieval method of BIM components called query-by-model. We integrate both our proposed similarity measurement algorithm and retrieval method into the BIM retrieval system, named BIMSeek, to greatly improve its retrieving speed and accuracy. Furthermore, we combine the query-by-model and query-by-keyword methods to refine the retrieval results iteratively. Finally, we conduct extensive experiments that compare our proposed method against previous retrieval methods. Results show that our method outperforms previous methods. (C) 2020 Elsevier B.V. All rights reserved.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available