American Society of Civil Engineers

Hierarchical Clustering for Interpretation of Spatial Configuration

by Asad H. Udaipurwala, (Graduate Research Assistant and Ph.D. Candidate, Department of Civil Engineering, University of British Columbia, 6250 Applied Science Lane, Vancouver, BC, Canada V6T 1Z4 E-mail: and Alan D. Russell, (Professor, Department of Civil Engineering, University of British Columbia, and Chair, Computer Integrated Design and Construction, 6250 Applied Science Lane, Vancouver, BC, Canada V6T 1Z4 E-mail:
Section: Knowledge Representation, pp. 1-11, (doi:

     Access full text
     Purchase Subscription
     Permissions for Reuse  

Document type: Conference Proceeding Paper
Part of: Construction Research Congress 2005: Broadening Perspectives
Abstract: Ever since the introduction of computers—particularly the field of Artificial Intelligence—construction researchers have endeavored to develop systems that capture and encode the knowledge of seasoned construction practitioners with the goal of at least partially automating tasks such as construction methods selection, equipment selection, constructability reasoning and cost estimation. However, efforts to address these tasks to date have suffered from the lack of a mechanism for automatically inferring conditions such as uniformity/similarity in a facility’s spatial configuration. Identifying these conditions is crucial for evaluating the suitability of a construction method as they affect criteria such as reuse and achievable production rates. In the absence of such a mechanism researchers have relied on either statistical techniques that can be biased by outliers, or simply put the onus on the user by querying them about the number of reuses, etc. which undermines the usefulness of the system itself. In this paper, we introduce an algorithmic technique based on hierarchical clustering that can be used to infer the similarity of part or all of a construction facility with respect to any measure of interest—such as length, area, volume, and so on. The advantage of this technique is that it is immune to outliers in the data set, and it can accommodate the intuitive notion of acceptable variability. For example, in an expert’s judgment, a six percent variability in the dimensions of slab-bays is acceptable for use of flying truss formwork as it can be accommodated with infill panels or hinge panels. We start by providing a motivating example from the domain of building construction, and illustrate how the techniques adopted by researchers to date fail in the case of various spatial configurations. We then provide the hierarchical clustering algorithm in detail after a short discussion of our technique for representing the project’s physical context. Finally, we illustrate how the algorithm has been integrated with a project management system that provides a hierarchical representation of the physical view of a facility, and a production rule based expert system to aid in the selection of construction methods.

ASCE Subject Headings:
Artificial intelligence
Construction management