A Predictive Semantic Inference System using BIM Collaboration Format (BCF) Cases and Machine Learning

Vincent Kuo, Jyrki Oraskari

Research output: Chapter in Book/Report/Conference proceedingConference contributionScientificpeer-review

Abstract

Building Information Modelling (BIM) has been hailed as an artefact of collaboration, where all project participants are able to create, modify and implement design/construction configurations within the same virtual environment codified by International Foundation Classes (IFC). To facilitate communication between human participants during project development, the BIM Collaboration Format (BCF) was developed, enabling users of BIM applications to communicate issues and refer those to specific objects. A single BCF entity holds a textual description of the issue, a status, links to a BIM IFC model and objects, a picture of the issue, and a camera orientation. Therefore, BCF poses a rich repository of complex fuzzy semantic knowledge concerning design risks, such as change requests and rework proposals, in a readily accessible format based in the XML schema. This research investigates how data in BCF files can be extracted, and processed using a linear algebra method known as singular value decomposition (SVD). SVD is extensively used, along with dimensionality reduction, for pattern recognition applications and has been shown to infer semantic correlations amidst a variety of different unstructured data (e.g. texts, images, signals etc.), comparable to that of human cognitive
associations. Consequently, a dynamic knowledge-base is established of past BCF cases, whereby designers can flexibly query and systematically retrieve most relevant past issues and related objects, given any known current parameter – reminiscent of how a human may recall past experiences upon inquiry or design review. This paper introduces the BCF structure, describes the machine learning steps taken to extract and process BCF data, and presents the conceptual
framework of a queryable knowledge-discovery system. This allows for relevant past issues to be recalled and the knowledge integrated in future designs as problem- and change-prediction. It is of particular pertinence for users of BIM, in sight of the ever-growing masses of BCF data generated from project to project.
Original languageEnglish
Title of host publicationWBC16 Proceedings : Volume III
Subtitle of host publicationBuilding Up Business Operations and Their Logic Shaping Materials and Technologies
EditorsArto Saari, Pekka Huovinen
PublisherTampere University of Technology
Pages368-381
Number of pages14
Volume3
ISBN (Electronic)978-952-15-3743-1
Publication statusPublished - 2016
MoE publication typeA4 Article in a conference publication
EventCIB World Building Congress - Tampere Hall, Tampere, Finland
Duration: 30 May 20163 Jun 2016
http://wbc16.com/

Publication series

NameRaportti / Tampereen teknillinen yliopisto, rakennustekniikan laitos, rakennustuotanto ja -talous
PublisherTampereen teknillinen yliopisto
Number18
ISSN (Print)1797-8904

Conference

ConferenceCIB World Building Congress
Abbreviated titleWBC
CountryFinland
CityTampere
Period30/05/201603/06/2016
Internet address

Keywords

  • BIM collaboration format
  • singular value decomposition
  • latent semantic analysis
  • change management
  • knowledge management

Fingerprint Dive into the research topics of 'A Predictive Semantic Inference System using BIM Collaboration Format (BCF) Cases and Machine Learning'. Together they form a unique fingerprint.

Cite this