| KEYWORD |
Area Architecture
AI-powered satellite-based bridge collapse predictive model
keywords ARTIFICIAL INTELLIGENCE, BRIDGE COLLAPSE, DEEP LEARNING, RADAR INTERFEROMETRY, SATELLITE DATA
Reference persons MARCO CIVERA, FRANCESCO DELLA SANTA
Thesis type NUMERICAL AND EXPERIMENTAL
Description Satellite data, especially Differential Interferometry SAR (DInSAR), are at the forefront of Civil Engineering for failure and collapse analysis. In fact, with the availability of data from the Cosmo SkyMed constellation with a high spatial resolution (3 x 3 m) and extreme measurement accuracy (millimetric), the methodology, while still considered at the research level, is becoming more and more used by professionals as well.
However, the difficulties in analysing the extremely large amount of data produced by the satellite at each passage over a large area still limit the usage of the technique for predictive analysis. That is to say, DInSAR is currently only performed ex-post on collapsed infrastructure to investigate, in retrospective for forensic engineering purposes, their behaviour before the failure.
However, leveraging a dataset of known bridge collapses and the corresponding data, it is possible to train a Deep Learning (DL) model, to achieve Artificial Intelligence (AI)-based predictive monitoring.
This can be further refined by data fusion, using a comparison between data points belonging to the target infrastructure and the surrounding background, multiple data from environmental sources such as weather forecasts, river level, and periodic offline retraining on novel case studies.
The thesis work will focus on applications for hydraulic- and rock/landslide-induced bridge collapses in the Italian territory over the last decade.
See also slide hack day madonna del monte.pdf
Required skills Good knowledge of Matlab and/or Python
Deadline 05/10/2025
PROPONI LA TUA CANDIDATURA