ABSTRACT
In April 2021 the Stavå Bridge, a main World War II era bridge connecting the northern and southern parts of Norway, was temporarily closed for traffic. Over a few days an until then unknown structural defect had developed to the point where it seriously compromised the bridge’s structural integrity. The owner of the bridge, the Norwegian Public Road Administration (NPRA), closed the bridge, arranged for a temporary solution to the problem and reopened with traffic restriction while awaiting a replacement bridge.The incident was detected by remote observation using what constitutes the digital twin of the bridge processing data from IoT-sensors mounted at key locations. It was some of the resulting indicators that rapidly increased to alert the situation. They were also crucial in the online and offline diagnostic work, as well as in the process of localizing the structural damage.The three year-long monitoring campaign is part of an ongoing NPRA initiative to test new technology in its transformation from corrective and periodic maintenance to risk- and condition-based maintenance. This takes place in collaboration with SAPs Enterprise Product Development – Connected Products, a strategic IoT effort offering online monitoring and digital twins to reduce costs while improving safety and asset health.The case described demonstrates the value of such technologies. In a rapidly developing situation, a critical damage was detected early enough to avoid a serious incident, though too late to allow for preventive actions. Subsequent investigations show that early signs existed weeks earlier. Notwithstanding the costs savings involved in early detection, it is also key to improved traffic safety.The paper positions online monitoring and digital twins into the larger context of risk- and condition-based maintenance. It describes the Stavå Bridge project, including the setup of the sensors and methods used. It details the situation that arose in April 2021, how it was detected and diagnosed, as well as work in the aftermath. The paper concludes with a summary of lessons learned and an outline of future work with improved early detection capability, documenting the benefits of using machine learning in parallel with direct, deterministic, analysis.
