CAPTURE (Capture, Analyse, Predict, Monitor, Visualise, Record, Evolve): a smart digital twin solution with transfer learning to sustainable railway bridge maintenance.

The proposed project aims to develop an intelligent decision support system that will help in detecting defects in masonry railway bridges in the UK. Currently, relevant authorities (e.g., Network Rail in the UK) invest nearly £2 billion every five years in maintaining the existing railway infrastructure, including bridges and tunnels.

The project will use emerging surveying methods and non-destructive testing technologies, such as unmanned aerial vehicles and terrestrial laser scanning, to develop non-subjective methods that provide a clear and conclusive overview of the structure's condition. It aims to develop a custom AI-based solution that will be able to detect masonry defects in railway bridges. The system will utilise modern deep learning techniques such as transfer learning and GAN to deal with imbalanced data and reuse existing datasets. The project will also create a digital twin of the bridges, which is an emerging topic in infrastructure maintenance, and will be integrated into the smart bridge maintenance system. The digital twin will enable predictive maintenance, which will help in reducing future spending and ensure the long-term sustainability of the UK railway network.

Overall, the proposed project will develop a smart bridge maintenance system that incorporates the intelligent decision support engine and the digital twin into the industry standard process. The system will help in maximising the benefits of maintenance investments, reducing future spending, and ensuring the long-term sustainability of the UK railway network.

Artefact/prototype system: a smart bridge maintenance system
that incorporates the intelligent decision support engine and the digital twin
into the industrial standard process.

 

Computer science: a custom deep learning solution for detecting
masonry defects in railway bridges, especially with imbalanced dataset.

 

Built Environment/Infrastructure Maintenance: a
digital twin smart bridge maintenance system that enables predictive
maintenance.

 

Economic Solution: discuss the cost-effectiveness of
our proposed solution to implementation in the UK rail system.

 

Outputs: Three high quality journal papers (IF
>2 and H5 >50).

·      
One for over-sampling techniques development.

·      
One for railways bridge defect detection
algorithm improvement.

·      
One for predictive maintenance and digital twin,
with a cost-benefit analysis.

 

Impact: The proposed PhD can run in parallel with DoS’s
ongoing the "Panoptic Bridge Management" projects, which complement
each other (the PhD is more focused on theory and algorithms while the project
emphasises on practical and system development). It could potentially be
incorporated into Network Rail’s centre IT system.