RAMS1 is often required by Dutch law or other government regulations for safety-critical systems like railroad infrastructure and signalling systems. Fault tree analysis (FTA) is a widely applied industry standard for RAMS analysis. It is frequently used and preferred by ProRail: ”complexe projecten: bij voorkeur foutenboomanalyse (FTA) of simulatiemodel”. FTA yields system availability and reliability, and can be used for critical path analysis. It can however not yet deal with a pressing aspect of railroad engineering: maintenance.
Fault trees (FTs) currently only support elementary maintenance aspects, e.g., their simple repair rates model that component”s faults are all repaired with the same speed. The ProRail context, however, demands for highly advanced maintenance models supporting condition-dependent strategies for preventive, corrective, clock- and age-based maintenance, inspection, monitoring, etc. In case of multiple failures, decision strategies need to specify which component(s) to repair first. In addition, maintenance costs are pivotal. FTA cannot yet cope with these advanced maintenance aspects.
Equipping FTs with maintenance models yields a powerful framework that enables one to (1) compare the effect of different maintenance strategies on the availability, reliability, cost, recovery speed, robustness to changes, etc., and (2) automatically synthesize maintenance strategies that are optimal w.r.t. given criteria (availability, reliability) subject to a (cost) budget. Key to tackle these challenges is the exploitation of stochastic model checking, an innovative technique heavily used by hardware manufacturers. Scientific core of our work is the extension, refinement, and tailoring of SMC techniques to maintenance planning problems. These require solving optimization problems for stochastic decision processes, known as Input/Output Interactive Markov Chains.
1 RAMS stands for Reliability, Availability, Maintenance, Safety