BIM, Bridge and Civil Structures, Building Performance and Systems, Civil Engineering, Digital Environments, Environmental and Building Physics, Façades, Machine Learning, Resilience, Security and Risk, Site Development and Regeneration, Systems Engineering, Transport Planning
This team is developing an innovative approach to asset management, via a machine learning application that enables condition change detection. The tool identifies defects with machine learning, and records the location, type, extent and severity, as well as length. Complex, labour intensive processes, were streamlined with an automated Machine Learning “brain” that evaluates the photogrammetric data. As such, greater accuracy can be achieved
This application builds on the mature knowledge and capabilities of the asset management market, and integrates machine learning technical capability that has been developed within Arup’s project team over the past few years. Further collaborators included thought leaders CERN and UCL, who pioneer data management and optimisation innovation.
The aim was to develop an open source machine learning system, that suits the commercial need for automated assessment of infrastructure asset structural condition. This project planned to catalyse Arup’s leading technical market knowledge to enable us to keep up with the evolving market and client expectations. Such innovation provides a step change, enabling the monitoring of an entire pipeline network every day, without human intervention.
Now is a time when industry is calling out for innovative digital methods to improve, supplement, and essentially supersede existing outdated processes. The development of digital image processing libraries is experiencing an exponential growth. Moreover, computational power is now very affordable.
The traditional practice for condition detection in infrastructure assets is considered labour intensive, subjective, and generally lacking. Usually, this involves sending personnel to visually inspect assets. As such, there is a significant industry momentum to cultivate innovative techniques. The nature of condition detection lends itself well to the capabilities of machine learning, enabling smarter and faster work on a large scale. Furthermore, while machine learning has penetrated other industries, it is time to embrace this within civil engineering to provide a better and cheaper service to the client.
The projected outcome is a developed open source machine learning system, with intellectual property residing in Arup. This can be rolled out as a commercial decision aid asset management tool for automated assessment of infrastructure asset structural condition. It will be scalable and transferable across different infrastructure asset types and sectors. Additionally, a version of the machine learning for crack detection on steel, and automating the quantification of corrosion degradation, is under development.
Smart infrastructure asset management through machine learning holds particular advantages for the infrastructure and asset owner, for whom operation and maintenance accounts for 80% of the whole life cost. In addition, this tool has significant health and safety component, enabling the engineer to spend less time in hazardous areas, as well as removing the need for expensive road closures for inspections. Ultimately, this capability widens the knowledge pool, and offers new possibilities for Arup, specifically for Tunnels UK.