Current students

Nowadays, transportation of hydrocarbons through long distances (in the order of hundreds of kilometers) is mainly undertaken through pipeline networks, which represent an efficient and widespread tool for the conveyance of oil and gas products. Potential damages or failures to such structures might have severe environmental, health, and economic repercussions: as a result, pipeline monitoring becomes an operation of paramount relevance.
The current technologies for monitoring the condition of fluid transportation systems make use of multi-domain data collected along the conduits, from which data-driven relations can be derived. Such constitutive relations are subsequently analyzed in order to predict the future operational status of a given pipeline network.
To this date, it is still not completely clear how to build an effective monitoring system: in particular, it has not been established yet a rigorous methodology to be systematically applied both for the derivation of data-driven relations and to evaluate the reliability of pipeline integrity techniques.
This work has the aim of overcoming such limitations: in particular, the purpose is to develop novel machine learning approaches in order to monitor in real-time the integrity of pipeline fluid transportation systems, by making use of multi-point vibroacoustic data measured within the conduits.