Facilitating Fault Tree Analysis with Generative AI

Fault Tree Analysis (FTA) is a cornerstone safety and reliability engineering technique. However, the manual development of fault trees can be time-intensive, error-prone, and challenging for complex systems. This paper proposes a novel application of Generative AI (GenAI) to automate and enhance FTA. Instead of using LLMs to generate a complete fault tree, however, we believe that it is essential that the human analyst still drives the analysis and "only" gets support from an analysis co-pilot. By leveraging large language models (LLMs), our approach suggests new sub-causes for existing fault trees. The methodology will be applied to a Lane Keeping Assist System (LKAS) to demonstrate how GenAI can extend fault tree coverage and completeness.