Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Assessing implicit knowledge in BIM models with machine learning

Assessing implicit knowledge in BIM models with machine learning

http://link.springer.com/chapter/10.1007%2F978-3-319-24208-8_33

The adoption of Building Information Modelling (BIM) constitutes a radical shift in the way models in the building and construction industry are described. Traditional representations for architectural knowledge, such as architectural drawings, 3d models, technical descriptions and spreadsheets are transitioning into semantically-rich information models. Building related information no longer exists within discrete entities, but is kept in an interlinked context. BIM authoring tools for Design and Construction and Facility Management systems provide now semantic information about elements and spaces within buildings, their constituting elements, as their interrelation and performance. Currently huge efforts are put in place to create links between buildings, on local, regional and international level, through e.g. standardisation committees (buildingSMART 2014) or the Geospatial communities (Geospatial Media 2014), but as well through international research project, such as DURAARK (Durable Architectural Knowledge) (DURAARK 2015), a three year EU funded project on the creation and maintenance of semantic links between representations of buildings.
Building Information Models, in formats such as a Revit database or IFC, have become the bond that connects disciplines by streamlining data exchange and connecting the construction with the operational phases of a building lifecycle. Building related knowledge is herein represented in an object oriented way, holding building element geometry, properties, and its interrelation to other objects. These objects can be part of the described building, but in addition relate to external objects or other sources of information, including building element libraries. Information can be related to physical entities, like a wall, as well as to intellectual or organisational constructs, for example spaces or organisations. Hence, the model can support many facets of the construction phase, and in addition guide the building’s operation with Facility Management tools or the planning of retrofitting tasks.
The new class of information is directly machine-interpretable, as it conforms to a structured schema. The use of BIM models in current practice is however predominantly focussed on explicit information, such as property values, augmented with aggregate functions for the extraction of quantity information and clash detection based on geometrical inference (Tamke et al. 2014a). BIM models hold however information that is not explicitly stated, but lies implicit in the interrelation between the entities within a single model or in the interrelation of a large variety of models. And while years of practice train a building professional to immediately apprehend the workings of a space by means of merely symbolic two-dimensional representations, this information can currently not be assessed by machines. We ask, how these implicit second order descriptors can be assessed and whether this approach holds the potential to describe the qualitative aspects of a building.

Thomas Krijnen

October 02, 2015
Tweet

More Decks by Thomas Krijnen

Other Decks in Research

Transcript

  1. 1 / 20 2 / 26 / 13
    Assessing implicit knowledge in
    BIM models with machine learning
    Thomas Krijnen Technical University Eindhoven, NL
    Martin Tamke Centre for Information Technology and Architecture, DK

    View full-size slide

  2. 2 / 20 2 / 26 / 13
    Background
    Building Information Modelling (BIM):
    Recent paradigm shift in the construction industry.
    Buildings structured as semantically rich models,
    with relational semantics and (parametric) geometry.
    IFC is an open standard to exchange such building models.
    Models are being used for design coordination and
    automatic (but imperative and hard coded) building code
    and requirement checking.
    Background

    View full-size slide

  3. 3 / 20 2 / 26 / 13
    DURAARK
    Durable Architectural Knowledge
    EU funded project
    started in 2013
    Concerns the long-term digital preservation
    and semantic enrichment of building models
    Context

    View full-size slide

  4. 4 / 20 2 / 26 / 13
    Aims
    Offer insights into
    implicit information stored in building models
    As an alternative to
    explicit querying and rule checking mechanisms.
    Apply both supervised and unsupervised techniques.
    Aims

    View full-size slide

  5. 6 / 20 2 / 26 / 13
    Unsupervised learning
    Outlier detection based on geometrical
    descriptors of building elements segmented
    by element type {wall, window, column, …}.
    Unsupervised learning

    View full-size slide

  6. 7 / 20 2 / 26 / 13
    Example model
    Unsupervised learning

    View full-size slide

  7. 8 / 20 2 / 26 / 13
    Wall geometries
    Unsupervised learning

    View full-size slide

  8. 9 / 20 2 / 26 / 13
    Anomalies
    Unsupervised learning

    View full-size slide

  9. 10 / 20 2 / 26 / 13
    Results
    The algorithm has been able to identify the
    geometrical essence of building elements.
    Elements that deviate from this definition
    are identified and turned out to be
    misclassified by the authoring tool.
    Unsupervised learning

    View full-size slide

  10. 11 / 20 2 / 26 / 13
    Supervised learning
    Apply neural networks to categorize floor
    plans according to their use
    {non-residential, residential}.
    Supervised learning

    View full-size slide

  11. 12 / 20 2 / 26 / 13
    Features
    Supervised learning

    View full-size slide

  12. 13 / 20 2 / 26 / 13
    Space – wall - ratio
    Supervised learning

    View full-size slide

  13. 14 / 20 2 / 26 / 13
    Space – slab ratio
    Supervised learning

    View full-size slide

  14. 15 / 20 2 / 26 / 13
    Wall – column ratio
    Supervised learning

    View full-size slide

  15. 16 / 20 2 / 26 / 13
    Number of doors per space
    Supervised learning

    View full-size slide

  16. 17 / 20 2 / 26 / 13
    Average space gyradius and variance
    Supervised learning

    View full-size slide

  17. 18 / 20 2 / 26 / 13
    Dataset
    35 IFC files of varying quality and complexity
    amounting up to 109 distinct floor plans
    manually classified 46 residential 63 as non-
    residential
    Supervised learning

    View full-size slide

  18. 19 / 20 2 / 26 / 13
    Results
    Successfully classified 90% of unseen samples
    [ training set (70%) cross-validation set (20%) test set (10%) ]
    Supervised learning

    View full-size slide

  19. 20 / 20 2 / 26 / 13
    Conclusions
    Data collection difficult
    impedes learning more interesting things
    Even well-established models contain flaws
    The quality of samples varies and exporters
    adhere to an individual IFC ‘dialect’
    learn architectural features rather
    than picking up exporter idioms

    View full-size slide

  20. 21 / 20 2 / 26 / 13
    Further research
    ML an alternative perspective to explicit
    querying mechanisms and model checking.
    Autonomous quality assurance (typical
    clearance areas, confluence of specific
    types) by learning from conforming models.
    More ambitious classifications

    View full-size slide

  21. 22 / 20 2 / 26 / 13
    Assessing implicit
    knowledge in BIM models
    with machine learning
    Thomas Krijnen Technical University Eindhoven, NL
    Martin Tamke Centre for Information Technology and Architecture, DK

    View full-size slide