back to list

Project: Deep Learning and AI for BIM Clash Classification

Description

Building Information Modelling (BIM) or Building Information Management, is a highly collaborative process that allows architects, engineers, real estate developers, contractors, manufacturers, and other construction professionals to plan, design, and construct a structure or building within one 3D model.

The most difficult part of BIM project designing lies in the proper positioning of
elements and entities. Clashes emerge when designs of two or more entities
show elements colliding in space or time sequence. If the clashes are not
resolved at the design phase, they lead to rework, wastage, inevitable delays
and budget overruns during the construction stage.

Modern AI and Deep Learning is capable of detecting specific patterns in complex structures, like 3D geometry and temporal or meta data. It is expected that neural networks are able to fully automatically label clashes (relevant, ignore) from a (localized) BIM geometry with its metadata. Input would be a list of thousands of potential clashes as calculated by a Bim-Collab Zoom run. Output would be a classification for each clash, including prediction confidence.

The objective of this project is to create and extend tools to introduce AI into the BIM software domain.

Details

Student

OR

Oana Radu

Supervisor

Andrei Jalba

Secondary supervisor

KK

Kubus