Digital Risk Twin Driven Simheuristic Framework for Maintenance Optimization

Discrete Event Simulation (DES) is a powerful tool to enable RAMS analysts to estimate future outcomes (utility and cost) across a range of RAMS domains. Nonetheless, the interaction between the user and a DES can be improved by incorporating meta-heuristics optimization into the simulation framework. This ‘Simheuristic” closes the loop between manually defined parameters (decision variables) and a simulated outcome. Contrary to traditional optimization approaches that generally present the user with a single “optimal” solution, Simheuristics inherently provide the user with a suite of potential solutions that can be contrastively assessed over one another not only in terms of utility and cost, but the associated risk (stochastic bounds). Although it is the role of the optimizer to find a set of appropriate solutions, a priori decision-making criteria must be specified by the user to guide the search. In this context, we are referring to the mathematical expressions that influence the behavior (search strategy) of the optimizer. Practically, this requires that domain expert knowledge must be encoded up-front -an abstraction that is in most cases difficult for the human to provide especially in the face of uncertainty across a large set of decision variables. In this paper, we work towards closing this loop between user expression and simulation-based optimization by providing a novel, Model-Based Simheuristic approach for system availability optimization. Using this approach, we demonstrate by using the MADe MBSE tool we can capture domain knowledge contained in a Digital Risk Twin (DRT), and intelligently inform and guide the Simheuristic framework to generate a suite of analyst-preferable optimal solutions.