Digital infrastructure models: a necessary foundation for the next IT wave

University of Cambridge is working up inexpensive commercial DIM models. Laing O’Rourke lecturer of construction engineering Dr Ioannis Brilakis explains why.

Map projection of a point cloud depicting reinforcement bars of beams at the second floor of the James Dyson Building at the University of Cambridgeby Dr Manuel Davila, CSIC.

The construction/infrastructure sector has traditionally been slow in exploiting IT waves to their full potential.

This has resulted in an industry that significantly lags others in most performance indicators: value for money, safety, sustainability and quality.  Currently, the sector is experiencing the midpoint of the BIM IT wave. The next wave will be centred on Big Data and the Internet-of-Things.

Big Data has yet to properly take its place in the supply chain, while the Internet-of-Things is just starting to materialise in the form of Smart Infrastructure. The exploitation of both, however, is hampered by our lack of digital infrastructure models (DIM).

I will attempt to explain why, by briefly answering the what, why, when, who, which, where and how questions for DIM.

What is a DIM? A DIM is a digital representation of a physical infrastructure asset. It is effectively a comprehensive database with everything there is to know about the asset, from design, construction and throughout its lifecycle.

Why do we need DIMs? The next IT wave is data-centric and data-intensive. Big Data, for example, can only show its true value when applied to tremendously large amounts of simultaneously accessible diverse and interlinked data. Smart Infrastructure, by its nature, generates large amounts of data. Both need data repositories to send, receive and store data in a way appropriate to infrastructure assets. This is an essential role for DIM.

Big Data is not useful unless it is applied on hundreds or thousands of data-rich and up-to-date DIMs. Likewise, Smart Infrastructure data outputs are of little use unless they are properly linked or integrated to a DIM and consequently all of its attributes.

When do we need DIMs? Now. The next IT wave is already starting. Sensors are increasingly being placed in structures, and data analytics tools are being tested on infrastructure data. We will not be able to fully exploit this wave in a timely manner if we delay the creation of comprehensive DIMs as standard practice.

Who needs them? Who creates them? Who updates them? The answer is the same for all three questions: all involved stakeholders. The diversity of data in DIMs is such that all parties involved are needed to create and update them. Likewise, a DIM as the ultimate source of knowledge about our infrastructure is useful to all parties.

Which attributes should they contain? Most of the expected attributes (i.e. geometry, design data and lifecycle management data) have been included in data standards to a large extent, such as the Industry Foundation Classes. However the standards are extensible. The beauty of DIMs is that there is no upper limit of what can be included.

Where should they be stored/accessed? We need secure, centralised or distributed DIM repositories that give access to many DIMs from any internet-enabled location. This has not happened yet. Most companies have individual DIMs scattered across their offices with little data apart from design geometry.

How do we get them? This is a real challenge. Design BIMs are gradually becoming the norm for buildings, but not as much for horizontal infrastructure projects. These BIMs often contain no more information than a CAD file. Also, given the current rate of infrastructure renewal, it may take us a long time before we manage to generate DIMs for most of our existing infrastructure.

This is where work at the Centre for Smart Infrastructure and Construction (CSIC), at the University of Cambridge, may change the course of events. With help from AVEVA, DfT, and other stakeholders, CSIC is working on creating new methods that can automatically convert infrastructure spatial and visual data into a barebones as-is DIM with geometry data, and methods for enriching and updating the DIM with visible damage and defect data.

Once commercialised, these methods will enable users to rapidly and inexpensively generate as-is DIMs and create a new ecosystem for future tools that can enrich them. This will help build up a ‘Facebook’ of buildings, with vast datasets rich with geometry, condition, operation and other data attributes. While this comes with security and privacy challenges, it also brings unique opportunities for machine learning and data analytics tools to harvest best practice knowledge - and  produce a means that will help us to be  more efficient constructors and asset managers.

Dr Ioannis Brilakis is a Laing O’Rourke Lecturer of Construction Engineering at the Division of Civil Engineering of the Department of Engineering and a Co-Investigator at the Centre for Smart Infrastructure and Construction (CSIC) at the University of Cambridge. See