Fault Trees vs. Component Fault Trees: An Empirical Study

Abstract

When dealing with structural safety analysis, one of the most popular methodologies is Fault Tree Analysis (FTA). However, one major critique is the rapid increasing of the complexity, and therefore incomprehensibility, when dealing with realistic systems. One approach to overcome this are Component Fault Trees (CFT), presenting an extension to standard FT, allowing the separation of the analysis into less complex parts on the level of system components. CFTs are proposed to be more structured and partly reusable and therefore also claimed to be more straightforward to use by engineers with little safety domain experience.

In this work, we aim at getting an idea of the validity of presented theses and started an initial experiment with 13 computer science students, being asked to execute CFT or FT method on a given case study. Due to the number of participants, we focused on their empirical statements, the analysis solutions, and empirical results collected using a questionnaire.

Although the empirical impression has been that the resulting CFT models are better to use and more comprehensible than the FT models, the qualitative results have not supported this. Moreover, the component-wise modeling seams to mislead the students such that they have overseen failures outside the component structure, e. g., Common-Cause, Cross-Component, or external failures.