TO TOP

"Everyone has their area of expertise" - The "Bridge Inspect" project and the possibilities of bridge inspection with the help of AI

04.05.2021

Parkhaus Rub

Towards the end of 2020, the research project "Bridge Inspect" of the Bundesanstalt für Straßenwesen - BASt for short - started. In addition to the company HOCHTIEF ViCon GmbH, the Chair of Computer Science in Civil Engineering is also a project partner.

Sven Zentgraf is involved as a team member from the IIB as well. He is a research associate and his research is mainly concerned with the fields of AI and AR. In this interview, he explains what the "Bridge Inspect" project is about and what possibilities it opens up for the future.

What is the goal of the "Bridge Inspect" project and what is the chair's role in it?

Sven: In practical terms, it's about developing software that makes it possible to detect bridge damages. The central elements are AI- and AR-based prototypes that are intended to support building inspectors. Existing damage is to be automatically recorded and evaluated. Under the coordination of Prof. Markus König, we supported the project with five employees from different research areas. Here, the chair brings experience and some know-how to the project, especially in the areas of AI and AR. Two employees, for example, are mainly involved in AR and GPS-less positioning in their research.

What are the current problems in the inspection of (partially) damaged bridges? What types of damage can be detected by the software?

Sven: During a bridge inspection, the entire bridge is inspected purely visually by an inspection team consisting of one inspection engineer and two technicians in accordance with the specifications of DIN 1076. The quality of the inspection depends on the experience and due diligence of the inspector. This means that there is always a risk of damage being overlooked that could affect the load-bearing capacity or traffic safety of the structure. Within the scope of the project, we are initially focusing only on the detection of concrete damage, such as cracks, spalling and so on. In principle, it can be very time-consuming to inspect all areas of large bridges efficiently and safely, which may also require scaffolding or special inspection vehicles - this is to be simplified as far as possible, for example with the help of drones.

What about project planning - what is the current status?

Sven: At the moment we are still at the beginning, the project is designed for two years. Before we can start developing the software, we are conducting a requirements analysis. The start of the realization is then planned for the beginning of the second project year. The requirements analysis determines which camera resolution is needed, whether the device must be held in the hand or integrated into the construction helmet, for example. In addition, aspects such as battery life, the need for an Internet connection and ultimately the performance of the device used are also considered. For example, we are currently investigating whether the Google HoloLens can be used as a head-mounted display; this is already available at the chair. Of course, we are always in close contact with the BASt for all decisions.

How should the software be tested? Is there a specific sample bridge?

Sven: So far, we haven't decided on a bridge as a test environment. Of course, we want to test the applicability of our software and preferably in reality. Our project partner HOCHTIEF ViCon can give us access to bridge structures. The choice also depends on the time horizon, because such a test requires some preparation. However, we will also use data from others to train the AI methods. HOCHTIEF has a certain data set of damage images from similar tests, which can be used later to test the AI methods. At the chair, we have already gained some experience with neural networks and AI-based damage detection, especially with asphalt and concrete damage. The resulting and processed data are used to create our own data catalogs, which can be used to better train AI procedures in subsequent projects. This is also planned for "Bridge Inspect".

What future application possibilities arise from the results of the project?

Sven: In perspective, the results of the project should simplify bridge inspection and support the inspection engineer in his work. Another possibility would be the use of drones for data acquisition and evaluation of the data by the developed software. The "four-eyes principle" of inspector and AI system will make the inspection less error-prone and faster. It would also be conceivable to automatically transfer the detected damage into a BIM model, i.e. for the creation of digital twins. This does not mean that the inspection process on the structure is carried out faster, but data entry, processing and follow-up are significantly accelerated. In short, the goal of the project is process optimization.

Parkhaus Rub

Towards the end of 2020, the research project "Bridge Inspect" of the Bundesanstalt für Straßenwesen - BASt for short - started. In addition to the company HOCHTIEF ViCon GmbH, the Chair of Computer Science in Civil Engineering is also a project partner.

Sven Zentgraf is involved as a team member from the IIB as well. He is a research associate and his research is mainly concerned with the fields of AI and AR. In this interview, he explains what the "Bridge Inspect" project is about and what possibilities it opens up for the future.

What is the goal of the "Bridge Inspect" project and what is the chair's role in it?

Sven: In practical terms, it's about developing software that makes it possible to detect bridge damages. The central elements are AI- and AR-based prototypes that are intended to support building inspectors. Existing damage is to be automatically recorded and evaluated. Under the coordination of Prof. Markus König, we supported the project with five employees from different research areas. Here, the chair brings experience and some know-how to the project, especially in the areas of AI and AR. Two employees, for example, are mainly involved in AR and GPS-less positioning in their research.

What are the current problems in the inspection of (partially) damaged bridges? What types of damage can be detected by the software?

Sven: During a bridge inspection, the entire bridge is inspected purely visually by an inspection team consisting of one inspection engineer and two technicians in accordance with the specifications of DIN 1076. The quality of the inspection depends on the experience and due diligence of the inspector. This means that there is always a risk of damage being overlooked that could affect the load-bearing capacity or traffic safety of the structure. Within the scope of the project, we are initially focusing only on the detection of concrete damage, such as cracks, spalling and so on. In principle, it can be very time-consuming to inspect all areas of large bridges efficiently and safely, which may also require scaffolding or special inspection vehicles - this is to be simplified as far as possible, for example with the help of drones.

What about project planning - what is the current status?

Sven: At the moment we are still at the beginning, the project is designed for two years. Before we can start developing the software, we are conducting a requirements analysis. The start of the realization is then planned for the beginning of the second project year. The requirements analysis determines which camera resolution is needed, whether the device must be held in the hand or integrated into the construction helmet, for example. In addition, aspects such as battery life, the need for an Internet connection and ultimately the performance of the device used are also considered. For example, we are currently investigating whether the Google HoloLens can be used as a head-mounted display; this is already available at the chair. Of course, we are always in close contact with the BASt for all decisions.

How should the software be tested? Is there a specific sample bridge?

Sven: So far, we haven't decided on a bridge as a test environment. Of course, we want to test the applicability of our software and preferably in reality. Our project partner HOCHTIEF ViCon can give us access to bridge structures. The choice also depends on the time horizon, because such a test requires some preparation. However, we will also use data from others to train the AI methods. HOCHTIEF has a certain data set of damage images from similar tests, which can be used later to test the AI methods. At the chair, we have already gained some experience with neural networks and AI-based damage detection, especially with asphalt and concrete damage. The resulting and processed data are used to create our own data catalogs, which can be used to better train AI procedures in subsequent projects. This is also planned for "Bridge Inspect".

What future application possibilities arise from the results of the project?

Sven: In perspective, the results of the project should simplify bridge inspection and support the inspection engineer in his work. Another possibility would be the use of drones for data acquisition and evaluation of the data by the developed software. The "four-eyes principle" of inspector and AI system will make the inspection less error-prone and faster. It would also be conceivable to automatically transfer the detected damage into a BIM model, i.e. for the creation of digital twins. This does not mean that the inspection process on the structure is carried out faster, but data entry, processing and follow-up are significantly accelerated. In short, the goal of the project is process optimization.