1What is fault tree analysis?
L'fault tree (Fault Tree Analysis, FTA) starts from a top event (the feared event to explain) and decomposes it logically through AND/OR gates, down to basic events (basic component failures) that are not decomposed further. It is a deductive approach: unlike FMEA, which starts with components and works towards effects, FTA starts from the feared effect and searches for its causes.
It also supports probability calculation. The top-event probability is derived from basic-event probabilities combined using gate logic (AND = product, OR = union), while identifying minimal cut sets reveals the shortest combinations of failures leading to the feared event.
2Method steps
- Define the top event — the feared event to explain, stated precisely.
- Identify immediate causes and connect them through an AND gate (all are required) or an OR gate (any one is sufficient).
- Decompose recursively each intermediate cause until reaching basic events (component failure, human error or external condition).
- Assign a probability to each basic event.
- Calculate the top-event probability by working up the tree using gate logic.
- Extract minimal cut sets using MOCUS—the shortest combinations of basic events sufficient to cause the top event.
3Example completed in Polyodos
4Common mistakes
| Mistake | Why it is a problem |
|---|---|
| Confusing AND and OR gates | Using an AND gate instead of an OR gate severely underestimates the top-event probability. Gate logic must match the system's physical reality. |
| Decomposing to an irrelevant level | Excessive decomposition buries the tree in details with no reliable probability data. Stop where a credible probability can be assigned. |
| Ignoring repeated events | A basic event appearing in several branches is not independent of itself. The minimal-cut-set algorithm must account for it; a naive probability calculation does not. |