Evidence-based resilience engineering of dynamic space systems

This paper will present a method for the design for resilience of complex systems under epistemic uncertainty when the characteristics of the subsystems are time-varying. In this approach, the complex system is modelled as a network of interconnected nodes, each of which is characterised by one or more quantities of interest. The quantities of interest of each subsystem are dependent on a number of decision and uncertain variables that are strictly related only to that subsystem. A set of scalar quantities, called coupling functions, exchange information between pairs of subsystems. Each pairing function is dependent on a set of coupling uncertain parameters.
The uncertainty associated to all uncertain variables is modelled using Dempster-Shafer theory of evidence. Thus the network is called Evidence Network Model (ENM). This work in particular will consider the case in which the quantity of interest of each subsystem has a state that depends on the uncertainty and can change with time. In this way we can simulate continuous transitions between fully functioning and degraded states and the effect of disruptions and shocks that can perturbed the system. One of the quantities of interest is the mass of the subsystem that we will use as generic performance indicator of the overall system. Hence, the value of the ENM is the sum of the individual masses of each subsystem. The problem is, therefore, to minimise the system mass under uncertainty while all the other quantities of interest are concurrently optimised.