In the present study, we introduce a methodology based on graph theory into a High Performance Computing (HPC) environment developing a screening tool aimed at identifying the most critical units of water and electricity interconnected plants subject to fire and water casualties. The application of the methodology is exemplified via small-scale graphs representing in detail the topology of real-life plants and via large-scale random graphs reflecting only the main properties of real systems for benchmark purposes. Small graphs allow to show the practicality and efficacy of the methodology in representing the plant layout configuration before and after the casualty. Large graphs allow to estimate the improvement in terms of performance resulting from the parallelization of the code. In particular, our goal was to create a digital replica (”twin”) of the ship plants for predicting the state of health and the residual functionality of degradable systems and components after a fire or a flooding casualty, thereby improving passengers safety.
This work has been carried out in collaboration with CETENA S.p.A. (Centro per gli Studi di Tecnica Navale), a Fincantieri S.p.A. group society, in the framework of an internship supported by CETENA S.p.A. as one of the sponsors of the MHPC (Master in High Performance Computing).