Proceedings of the

The 33rd European Safety and Reliability Conference (ESREL 2023)
3 – 8 September 2023, Southampton, UK

An Importance Function to Generate Scenarios for Training a Grey-Box Model for the Computational Risk Assessment of Cyber-Physical Systems

Juan-Pablo Futalef1,a, Francesco Di Maio1,b and Enrico Zio1,2,c,d

1Energy Department, Politecnico di Milano, Italy.

2Centre de Recherche sur les Risques et les Crises, MINES Paris PSL, France.


The operation, control and maintenance of many systems rely on the signal communication functions provided by telecommunication systems. This generates Cyber-Physical Systems (CPSs). Computational risk assessment is being advocated to properly account for the complexities and interdependencies of CPSs. However, simulation times can be high for practical feasibility. Surrogate models are being explored to address computational issues. Among these, Grey-Box Models (GBMs) have recently been proposed to merge the physical knowledge embedded into a high fidelity White-Box Model (WBM) with the learned-by-data knowledge used to train a Black-Box Model (BBM). In this paper, we propose the use of a novel Importance Function (IF) within a Repetitive Simulation Trials After Reaching Thresholds (RESTART) approach to simulate accidental scenarios for training BBMs, ultimately embedded into a GBM. A case study is considered concerning an Integrated-Power and Telecommunication (IP&TLC) CPS of literature.

Keywords: Cyber-Physical System (CPS), Risk assessment, Rare events simulation, Black-box modelling, Grey-box modelling.

Download PDF