Abstract:

Reliability engineering is essential in industries such as healthcare and energy sectors, where system failures can have severe consequences. Fault Tree Analysis (FTA) is a widely used method to evaluate system reliability, identify failure risks, and ensure compliance with safety standards. Adding redundancies to systems is one of the main means to increase system reliability. The redundancies give rise to lots of different system designs/configurations. Traditional approaches to analysing different system designs/configurations are limited, as they require evaluating each system configuration individually. This process becomes computationally infeasible as the number of configurations grows exponentially with the number of redundancies added.

This research presents a novel family-based approach for analysing redundancy systems modelled using Fault Trees (FTs). By combining individual FTs into a Family of Fault Trees (FoFTs), the method creates a representation that encapsulates multiple system configurations. A new logic gate, called the switch gate, is introduced to enable this a unified representation of FoFTs. Combined with symbolic representations like Binary Decision Diagrams (BDDs), this approach leverages shared features across configurations to efficiently identify optimal system designs, addressing state-space explosion challenges through symbolic algorithms.
Building on successful family-based methodologies from related fields, this project aims to design, implement, and evaluate the FoFT approach. By designing the switch gate and implementing this gate in Storm we were able to evaluate the FoFT approach.

Community benchmark tests are used to assess the performance of this method. The results of the family-based method demonstrate clear advantages over the one-by-one analysis on various community benchmarks. Specifically, we introduce two variants of a family-based approach: a naive and a bottom-up implementation. The naive family-based approach significantly outperforms the one-by-one approach in all tested cases. However, the performance of the Bottom-Up approach is less consistent. While it does not outperform the naive approach in most benchmarks, it shows superior results in specific cases.