Dependency-aware Fault Tree Analysis
Alexander Prohaska · 2021
The proposed method enables fault tree analysis (FTA) to handle specific dependencies between basic events innately. The supported dependencies are mutual exclusion, negation, implication and influencing dependency. Dependent events have unique event labels which are traced throughout the fault tree. Depending on the set of traced events included in a gate, several conditional probabilities are calculated to measure the influence of single events or event groups at that gate. Logical modeling errors, like the presence of two mutually exclusive events, are recognized and the occurrences of dependent and repeated events are taken into account during probability calculation. This also includes negated events, making the approach suitable for non-coherent FTA. In dependency-aware fault tree analysis (DaFTA), standard fault tree gates are calculated by splitting the solution space into several sub-results which are calculated using the conditional probabilities of including and excluding selections of labeled events. For a moderate fraction of traced events, DaFTA leads to a significantly lower number of terms during inclusion-exclusion calculation, compared to precise minimal cut set analysis, without losing precision. Optionally, any basic event of interest can be selected for tracing, making it possible to perform impact analyses for single events or event combinations.