Revolutionising Operational Safety and Economy for High-value Infrastructure using Population-based SHM (ROSEHIPS)

Healthy infrastructure is critical in ensuring the continued health of UK society and the economy. Unfortunately, monitoring and maintaining our buildings and transport network is expensive. Considering bridges, inspection is usually carried out visually by human experts. There are not the resources to carry out the inspections as often as desired, or to make any repairs as quickly as needed; in the UK a backlog of maintenance works, identified in 2019, will cost £6.7bn. When resources are stretched, mistakes can be made, sometimes with tragic consequences; in 2018, despite warnings about possible problems, the Morandi Bridge in Genova, Italy, collapsed at a cost of 43 lives. Collapse is not the only problem; extreme weather events driven by climate change can test the performance of infrastructure beyond its limits e.g. consider the cost and inconvenience caused by bridge closures forced by flooding.

Bridges are only one concern. The offshore wind (OW) sector has driven down energy costs and increased power output, and now pioneers a global change to clean energy. The UK leads globally in OW energy, with ~8 GW of capacity, expected to exceed 25 GW by 2030, providing almost one third of the UK's annual electricity demand and helping meet the Climate Change Act's (2008) difficult 2050 target for an 80% cut in UK carbon output. The drive for turbines in deeper water demands new ways of asset management, decision making and controlling and limiting operation/maintenance lifetime costs. As turbines increase in numbers, size, and capacity, these issues become even more important.
The issues highlighted above are common across all elements of our infrastructure network (this PG will also consider telecoms infrastructure; another key test bed) and can be mitigated by automating the health monitoring. Instead of expensive, error-prone, human inspections, diagnoses can be provided economically by permanently-installed sensors, collecting structural data continuously and interpreting it via computer algorithms. This aim has led to the research discipline of Structural Health Monitoring (SHM), a subject of academic activity for over three decades. Despite intensive effort, SHM has not transitioned to widespread use because of a number of barriers - technical and operational.
The main technological barriers are: optimal implementation of hardware systems; confident detection in the face of confounding effects for in situ structures e.g. wind, traffic, for bridges; lack of damage-state data limiting the potential of machine learning for SHM. The operational barriers are: inertia - over-reliance on conservative design codes; trust - the SHM system must be as reliable as the structure itself; transparency - complex technology must deliver interpretable, secure decision support. The key to progress is to shift from thinking about individual structures to thinking about populations.
Population-Based SHM (PBSHM) is a game-changing idea, emerging in the UK very recently, with the potential to overcome the technological barriers above and transform our ability to automatically infer the condition of a structure, or a network of structures, from sensor data; this depends on an ability to collect a broader range of data, enriched into knowledge.

ROSEHIPS will extend and exploit PBSHM, developing machine learning, sensing and digital twin technology for automated inference of health for structures in operation now, and drive new standards for safer, greener structures in future. The Programme brings together the perfect team, mixing complementary skills in machine learning and advanced data analysis with expertise in new sensor systems and insight into complex infrastructure systems.

ROSEHIPS will provide open-source software systems, illustrated by realistic demonstrators and pre-populated with real-world data. Owners/operators will be able to customise and protect/secure their own data, while exploiting the knowledge base given.