Confidence-Aware Fault Trees

Fault trees are one of the most well-known techniques for safety analysis, which allow both quantitative and qualitative statements about systems. In the classical approach, deterministic failure probabilities for the basic events are necessary in order to obtain quantified results. Classical hardware-related failures and events can be obtained through testing, whereas software-dependent failures are harder to measure and identify, but are still possible to quantify when the implementation is given. In contrast, Machine Learning models lack this information, as their behaviour is not explicitly specified. Up to today, there are very few methods available to judge the worst-case performance of these models and predict their general performance. To encounter this problem, we will introduce confidence levels inside the fault tree analysis. This will allow the usage of failure rate bounds at basic events that only hold with given probability. We will present how this information can be used in the computations towards the top event. Our approach can be seen as a parallel or double application of a fault tree to include the confidence levels. Consequently, also basic events that depend on machine learning models can be included in a fault tree analysis. The proposed technique is compared to probabilistic fault tree analysis in an example.