Fault trees are a widely used formalism for modeling failures in safety-critical systems and analyzing their reliability. Traditionally, fault trees are constructed manually by experts, which can be time-consuming and error-prone.
In this line of research, we aim to automatically derive fault trees from inspection and monitoring data. Given data containing failures of individual components and the complete system, we aim to derive a fault tree that (1) represents the data as closely as possible, while (2) being of small size.

In this talk, I will give an overview of several approaches we have developed to learn fault trees from data. The approaches use multi-objective evolutionary algorithms, Boolean logic, and optimizations based on exploiting the structure in the fault trees.