Digital Twin-Empowered Intelligent Maintenance of Buried Infrastructure through Investigating Their Deterioration and Resilience Mechanisms

Objectif

The Europe Union faces an enormous task in maintaining its buried infrastructure (BI), since BI constantly deteriorates due to ageing, harsh environmental conditions, and nearby construction disturbances. It is imperative that innovative maintenance techniques are implemented to detect BI deterioration and schedule repair works based on BI resilience (i.e. the ability of a structure to be recovered under repair measures after it has failed) in the most cost effective manner. Although digital twin models have been employed for the maintenance of BI and demonstrated superiority, they still face challenges in achieving useful results because BI’s deterioration and resilience mechanisms are ignored in existing digital twin models. Moreover, existing digital twin models face challenges in synchronising with their physical BI twin on a regular basis, adversely affecting their ability to automatically advise the BI operator about maintenance schedules.
DigitalBI will: 1) Develop new adaptive denoising deep learning (DL) methods and image-based numerical simulation methods to unravel deterioration mechanisms; 2) Develop a new numerical coupling method and physics-encoded neural networks to decipher the resilience mechanisms; and 3) develop a large multimodal model (TunnelGPT) to integrate approaches developed for the investigation of deterioration and resilience mechanisms to form a digital twin model for optimizing maintenance measures. This will combine the researcher's experience in deep learning-based BI maintenance with the supervisors' expertise in physics-informed DL and large language models.
DigitalBI will be expected to set the researcher for a career in academia in Europe, while improving the comprehension of BI's deterioration and resilience mechanisms to reduce maintenance costs and consumption of materials for the remediation of BI. DigitalBI will also enable policymakers to use TunnelGPT to obtain suggestions on maintenance measures including plan parameters.