Graph-based retrieval of building information models for supporting the early design stages
Introduction
Our research focuses on providing support for the early-stages of the architectural design process using information technology. Hillier [1] describes the design process as follows: It begins as an idea for the building, then becomes an idea of the building, then a more formalized concept, then a series of more and more refined representations, then a set of instructions and finally a building. The design process is an iterative process of searching for a plausible solution and involves a continual back and forth between creative and analytical and evaluative steps in which potential solutions or variants are developed by various means before narrowing down the selection to the most promising candidates. For this, traditional design tools are most commonly used in the early design stages such as model-making, sketches and the use of reference examples. Using reference examples is an acknowledged method in both education and architectural practise and helps in learning design principals and guiding the design process as well as for inspiration or even as an explicit solution.
Digital equivalents have already been devised for several traditional design tools, however many of these do not fully translate the original strengths of the analog methods into the digital world and fail to make maximum use of digital possibilities. The approach discussed in this article aims to overcome these shortcomings and proposes a new paradigm for designing with reference examples that takes into account the typical way a designer works in the early design phases and makes full use of current methods and technologies.
Current electronic search methods use textual rather than graphical representations for search queries. However, the configuration of spaces and the relations between rooms are hard to represent using keywords, in fact transforming these spatial configurations into verbally expressed typologies tends to result in unclear and often imprecise descriptions of architecture. Finding relevant reference examples is therefore often a tedious task and at present there is a lack of appropriate IT support for this task. A central problem in this context is finding a way of determining similarity between the designer’s query and the data stock. The solution we are proposing is based on graph representations that capture the topological relationships between spaces.
As BIM and interoperable standard for semantic building models such as the Industry Foundation Classes (IFC)1 become more widely adopted, information on buildings will increasingly be recorded in building information models, which can then be used for other purposes. Currently, however, there are no adequate means of using such building information models and the information they contain for supporting the early stages of the architectural design process. To retrieve and use this information, we propose the workflow depicted in Fig. 1.
To be able to search efficiently for relevant examples, we developed a sketch-based query language [2] with which it is possible to submit imprecise or fuzzy queries that may be vaguely formulated or that represent only part of the problem. In addition, we want to design a system that aligns with traditional ways of working using hand-drawn sketches to allow the user to explore and assess the retrieved information in the building information model.
For indexing and determining similarities we propose the use of semantic fingerprints [2] as a way of describing the arrangement of buildings in a manner analogous the way that fingerprints can be used to identify a person. In addition, to make searching easy and natural for the user, we propose a user interface which allows users to sketch floor plans and define qualitative relationships. From this sketch a graph is derived that forms the basis for the search query in a database containing reference solutions that are represented as building information models. Corresponding graphs therefore have to be derived from these models. The query graph and model graphs are compared on the basis of a similarity measure, and corresponding building configurations that are similar to the sketch are found and presented to the user.
In this article, we present methods for digitally combining two traditional design tools: sketches as a means of externalizing ideas, and reference examples as a source of inspiration and guidance for the design process with the aim of finding an appropriate solution. Important terms used throughout the paper can be found in the following publications: Spatial Configuration [1], Visual Language [3], Vague Query [4], Space Connectivity Graph [5], Sketches [6] and Labeled Graph [7]. Where necessary, further definitions are provided in the text. The term Semantic Fingerprint was proposed by the authors in [2] and is used to explain the basic idea of identifying buildings for search purposes without implying any specific data model, method or technology. Semantic Fingerprints of buildings can contain a variety of different types of building information such as topology and geometry, the orientation of spaces, energy performance and geo-reference to identify a building. These and other definitions are noted in the text. Mathematical definitions are given in Section 4.1.
In the following section, we discuss approaches from the field of case-based reasoning (CBR), sketch-based interaction, and graph theory. The concepts and approaches of the system and components are discussed in Section 3, and in Section 4 we present an approach for determine similarity (a modified approach based on Messmer’s method [8]). Finally, in Section 5 we present our experiments on similarity analysis and conclude with a summary of our findings and their interpretation in Section 6.
Section snippets
Related work
Based on the description of the problem as discussed in Section 1, in the following we discuss CBR-approaches in architecture for deriving similarity assessment strategies. Related work on sketch-based interaction and information representation focus on the generation of semantic fingerprints using graph theory approaches. For similarity assessment, we elected to use the decision tree approach outlined by [8] in an optimized form (see Section 4) which we tested experimentally in Section 5.
System concept
Based on the aforementioned requirements we have developed a concept where information technology is used to support the design of buildings by retrieving reference examples using vague sketch-based queries. By supporting we mean a strategy that focuses on the needs of the user and leverages the strengths of information technology to find similar reference examples and let the user evaluate and decide how to use the information. It is essential to be able to determine similarity to the problem
Graph matching
Spatial properties, such as how rooms are connected, are topological relationships that can be represented as graphs and compared using computational methods. The process of querying a larger repository of building information models can be interpreted as formulating a query as a sub-graph and searching for this structure in the repository using a semantic fingerprint index. As spatial structures are generally complex structures, we are confronted with the problem of the complexity of the
Experiments
For the experiments we used IFC step models (one floor only) created by students of architecture using mostly Graphisoft ArchiCAD, Autodesk Revit and Nemetschek Allplan to examine the aforementioned methods for deriving the inner spatial structure. This information was used to derive semi-automatic labeled graphs using Open IFC Tools3 and to store GraphML files. The nodes were named after the manually equalized respective name of the
Conclusion
Through the appropriate administration and archiving of building information models and their corresponding semantic fingerprint, it is possible to call up relevant reference examples that can serve as a source of information. Using knowledge management techniques together with sketch-based query strategies and graph theory methods, building information models can be retrieved.
As the approach relies on IFC models, the quality of the models is critical in order to obtain high quality semantic
References (52)
- et al.
A decision tree approach to graph and subgraph isomorphism detection
Pattern Recognition
(1999) - et al.
A case base of case-based design tools for architecture
Computer-Aided Design
(2001) - et al.
Cadre: case-based geometric design
Artificial Intelligence in Engineering
(1996) Case-based design in the seed system
Automation in Construction
(1994)- et al.
Supporting communication and cooperation in distributed representation for adaptive design
Advanced Engineering Informatics
(2013) - et al.
New visual languages supporting design of multi-storey buildings
Advanced Engineering Informatics
(2012) On a relation between graph edit distance and maximum common subgraph
Pattern Recognition Letters
(1997)- et al.
Topological analysis of 3d building models using a spatial query language
Advanced Engineering Informatics
(2009) - et al.
Faster subgraph isomorphism detection by well-founded total order indexing
Pattern Recognition Letters
(2012) Space Is the Machine: A Configurational Theory of Architecture
(1999)
The Electronic Design Studio: Knowledge and Media in the Computer Era
The room connectivity graph: shape retrieval in the architectural domain
Tools for Ideas: An Introduction to Architectural Design
Case-based reasoning: foundational issues, methodological variations, and system approaches
AI Communications
Looking back to the future – an updated case base of case-based design tools for architecture
Knowledge Modelling – eCAADe
Spatial composition using cases: IDIOM
Case-Based Reasoning
Precedents: memory structure in design case libraries
Cited by (72)
Design-condition-informed shear wall layout design based on graph neural networks
2023, Advanced Engineering InformaticsSemantic-driven Graph Transformations in Floor Plan Design
2023, CAD Computer Aided DesignIFC-graph for facilitating building information access and query
2023, Automation in ConstructionIntelligent design of shear wall layout based on graph neural networks
2023, Advanced Engineering InformaticsIntelligent beam layout design for frame structure based on graph neural networks
2023, Journal of Building Engineering3D building interior modelling for navigation in emergency response applications
2022, International Journal of Applied Earth Observation and Geoinformation