At the last building smart International Summit, held in March in Düsseldorf, Germany, Mark Enzer from Mott MacDonald delivered a keynote speech. He used the UK as an example saying that 99.5 per cent of the infrastructure that will be available to the public in 2020 is already built. Nevertheless, the focus of the architecture, engineering and construction industry is still the development of new projects. This fact is striking, since the cost of managing, maintaining and repairing those already built structures is far higher than the budget spent on the construction of new projects.

These findings and the comments made by Enzer lead to Amberg taking up its RetroBIM project, which is a setup that develops a valid and precise model to work as a digital twin of an existing piece of infrastructure. Also, one that can be created as automatically as possible. Furthermore, it allows the client to make its planning process more straightforward and thus more economical. In addition to this, the client always has a digital version of his building at his disposal.

The challenges of modelling an existing infrastructure

The Penn State College of Engineering defines “Existing Condition Modelling” as follows: “A process in which a project team develops a 3D model of the existing conditions for a site, facilities on a site, or a specific area within a facility. This model can be developed in multiple ways including laser scanning, photogrammetry and or conventional surveying techniques, depending on what is desired and what is most efficient.”

This definition though is not considering that the modelling of an existing infrastructure is more complex than modelling new architecture projects. First, it takes much more time to model an infrastructure as it is, with all its imperfections, objects and deformations. Modelling a new infrastructure from scratch usually can be done much faster. Secondly, there is no specific software on the market to model existing infrastructure. This lack of a common tool for the entire industry makes the planning process more difficult. Another point is that older tunnels for roads or rail tend to have a more complex geometry than new ones which were probably built with newer working methods. That means that perfectly vertical walls or horizontal decks are not guaranteed and usually profiles are inconsistent. Lastly, building errors or deviations from the original plans can be quite significant. Only laser scanning provides an accurate source of geometrical information, the so-called as-built status.

The launch of the retro BIM project

Although the project focuses on tunnels, the findings and lessons learned can be used on any given type of infrastructure. The first structure which was used to model a digital twin was the Oelberg Tunnel in the heart of Switzerland. It is part of the Gotthard Railway on the international transit line between Basel and Chiasso. The owner of the 2km tunnel, the Swiss Federal Railways (SBB) was a major partner in this pilot project.

How the Modelling process works

The processes that have been developed vary, depending on the goals and future use cases to be implemented into the digital twin. But the main goals of the modelling processes are mostly the same. Thanks to the observation of the model, future adjustments of the infrastructure can be visualised better which makes the decision-making process more efficient. Another goal is to generate 3D graphical information in a fast and automated way. Last but not least, the processes support the development of maintenance solutions in the shape of a restoration BIM model.

The process begins with the gathering of information and base data in order to model the infrastructure. There are analogue sources like old blueprints as well as newer, digital sources like in-site measurements or GIS servers. The most useful and valid data comes from a laser scanning of the site which provides exact information of the as-built geometry.

Depending on the type of infrastructure, laser scanning can be done either with stationary laser stations or with dynamic systems like manually pushed trolleys, locomotives, remotecontrolled small robots or cars. In the case of road and rail tunnels, dynamic systems are almost mandatory given their length and the logistical complexity of scanning.

Laser scanning has many advantages, but also some constraints in the form of shadows in the point clouds because of the geometry of the site and from the system used for the scanning. Not only that, laser scanning generates some noise in the form of inaccurate points. For these reasons it is very important to process the point cloud properly.

The data processing is done in the point cloud processing software 3D Reshaper. First, the point cloud files are imported and a noise filter to delete any unwanted, inaccurate points is applied. After that, if available, the alignment of the tunnel (as a .dxf file from AutoCAD) is imported as well. If there is no alignment file, it is possible to generate one with the existing tools of the software. Once there is an alignment, the built in options to automatically detect the shape of the vault and isolate the rest of the objects can be used.

Even though the process is already automated to a certain extent, some manual control work is needed to ensure that everything has worked properly. Usually that check can be done in quite a short amount of time.

As a next step, the points of the vault are used to generate a polygon mesh with a given precision that can be exported either as a .dwg or an .ifc file. This highly detailed polygon mesh contains details of rock spikes or imperfections in the vault model. Since the Oelberg Tunnel from the pilot project is a rail tunnel, the mesh later can be used to gain clash control in the form of a clearance profile.

Next part of the process involves generating geometry out of the point cloud. The separate point clouds of the vault and objects are exported to the Rhino 3D design software. Then, with the help of the Grasshopper visual scripting plug-in, the vault can be sliced at regular intervals into sections. The form finding Python script is then used to generate lines and curves with the shape of the section of the tunnel.

From this process a polyline results, which helps to see the shape of the vault, as well as the shape of the shoulders, rails and sleepers. The biggest constraints to the new process can be found here: There is no way to tell the real thickness of the vault and walls without some sampling or on site inspection. Furthermore, a prediction of what lies exactly behind the visible surface of the wall is impossible.

To overcome these problems, two solutions are possible: Either the information can be retrieved from one of the original blueprints and the invisible structures of the tunnel can then be reconstructed using a mixture of theoretical information and on site laser scanning. The second possibility is to model only the visible surfaces as geometries without thickness. That would mean that the invisible pipes, cables, shafts and objects outside are left out of the model.

In any case, if the tunnel has niches or sideways structures, the shape of the vault needs to be manually edited and their geometry added to the model since it is not easily recognised on dynamic laser scans. If it is necessary to add objects located in the tunnel to the point cloud, two possibilities are suitable: When the object appears only once, it can be simply modelled manually with the correct position and size. But if objects or systems such as cables, pipes, rails, sleepers, etc., span along the whole length of the tunnel, they can be generated automatically with scripts either in the software Grasshopper or Python.

Once every element in the tunnel is properly defined, the plug in VisualArq is used to attribute the objects via scripts with the relevant information needed for the project: name, type of object, material, kilometric point, GUID, etc.

As a final step, the model can be exported as an .ifc file for further project coordination with the client and other project stakeholders.

Outlook

A 3D BIM of the existing conditions of an infrastructure is only the starting point of such a project. But it is already a strong, single source of information which facilitates the communication between all involved parties for further project steps like refurbishment measures.

BIM will play a vital role within the entire tunnelling industry in the future, we must always be focused on the goals of saving time, increasing cost-efficiency and ensuring the sustainability of infrastructure over its entire life cycle.