Building the old Carlsberg

Building the old Carlsberg

SHARE

Laser scanning technology was extensively used to transform the historical Brewery of Carlsberg into a sustainable area for living, recreation and business.

brewery of carlsberg point cloud
High density point-cloud in RGB, consisting of billions of XYZ-points.

The historical buildings of the Carlsberg Brewery were founded in 1847 on the outskirts of Copenhagen. Today these buildings stand as a national heritage with 150-year old architecture. Production continued until 2008 and the area is now being transformed into a sustainable area for living, recreation and business. During the transformation process, it was a top priority to preserve and protect the unique qualities of the old architecture. Hence, knowing the exact geometry of the existing structure became vital.

In 2012, Landmålergården I/S (now LE34 A/S) performed a building registration of the historical Brewery of Carlsberg. The registration was delivered as a BIM model that was constructed on top of an extensive point-cloud produced by laser-scanning. The project was carried out for the building owner, Carlsberg Brewery A/S, with Wilkinson Eyre Architects as professional advisors. Magnasoft Consulting India performed as a sub-contractor constructing the parametric BIM-geometry on top of the point-cloud.

Performing the task

The job was carried out using a high speed laser scanner. Scanning was performed with a density of 5-15 points per sq cm of the building surface and took about roughly 400 individual laser scans to cover the 6,000 sqm building floors, including a total cover of the building surface both inside and outside. The resulting point could count 15 billion RGB-colored points.

To obtain a high level of absolute accuracy, the pointcloud was calculated on top of a reference network counting about 200 fixed reference points and numerous moving targets. The reference network was constructed by observations from a high precision total station using the lesser square method, to obtain and estimate the level of precision down to a few millimetre on each reference point.

Points clouds were then filtered using both automatic and manual methods. Automatic filtering removed stray points and dark points, and manual filtering was performed to remove the obstacles and unwanted objects from the pointcloud. Finally, the refined point-cloud was delivered for Magnasoft to build up the BIM-geometry. This geometry was built using a detailed list of specifications on object types and accuracies. The final model that was delivered in the IFC-standard and was evaluated through extensive quality assurance procedures.

collecting laser scans
Collecting the laser scans

Challenges
As the job was performed during the winter season, covering the exterior of the construction involved a number of challenges. The project suffered weeks of delay because of continuous snowfall and rainfall making laser scanning impossible. Snow and ice-cover on the surfaces made scanning and any other kind of registration difficult to the level of impossible. The weather conditions also made access to the construction a risky business with icy and slippery surfaces.

Precision of geometry
The precision of the BIM-geometry is affected by three factors:

  • The precision of the initial survey.
  • The desired level of precision of the BIM.
  • The precision of the definition of the building object.

When this problem was somehow overcome, the point clouds turned out to have another disadvantage. When the weather finally changed from clouds to sunshine, the low sun of the winter produced very deep shadows and overexposed bright areas in the sunshine. This not only affected the BIM geometry, but also was a big disadvantage for visualisation purpose.

Another challenge is the interpretation of the pointcloud, especially when BIM geometry is classified and defined by the staff who don’t have on-location understanding. Using the digital images produced during laser- scanning, offers a great support in this task. However, mistakes were still present. An illustrative example was the presence of a steel thermos on top of a boiler that was mistaken for a valve. To minimise this kind of mistakes, first of all, highly trained staff was put on the task and, secondly, all geometries and classifications were quality assured by the local staff. Errors did occur even with such precautions. As an example, openings closed by light material and then covered by one sealing together with the original walls were missed. The only way to catch such errors was to examine each section of every constructive object on-location and comparing the resulting geometry with the old building drawings. On this project, this kind of quality assurance kept the errors of classification down to less then 0.1%. Still with thousands of objects, a number of errors were discovered.

Soren Aage Normolle,
 Partner, LE34 A/S,
san@le34.dk
Martin Tamke
, Asst Professor, The Royal Danish
Academy of Fine Arts, Schools of Architecture, Design and
Conservation,
martin.tamke@kadk.dk

The Big 3 Challenges
Visual quality: Laser scanning automatically picks up colorisation, with only small possibilities to improve the quality. For the production of the BIM geometry this has little impact, but it should be considered a high priority for the forward going usability of the point-cloud.
Quality of classification: Once the BIM-geometry is produced, it will be regarded as the final truth during the design and planning phases. Hence, the classification must be literally free of errors.
Accuracy: The geometry will be considered as the final truth about the building as well. However, one would want the BIM-geometry to be parametric and general rather than exact to the last millimeter. Walls should be vertical and dimensions should be fixed.