Fault analysis is the structured study of how a system can fail and what follows. Done before a design is built, it finds weak points cheaply; done after an incident, it finds causes and not only culprits. Its tools turn "what could go wrong?" into a list that can be ranked and acted on.
Key concepts
- FMEA: list each component's failure modes, their effects and their causes, and rate how severe, likely and detectable they are.
- Fault tree analysis: start from a top event and work down through combinations of causes with AND and OR gates.
- Common-cause failures: one cause defeating redundant parts at once, such as a shared power supply.
- Fault injection: deliberately causing faults to see how the system reacts.
- Post-incident analysis: timelines, contributing factors and fixes, in a blameless style.
Vocabulary
- Failure mode: the way in which something fails.
- Root cause: the underlying reason that, if removed, prevents recurrence.
- Cut set: a combination of basic events that together cause the top event.
- Detectability: how likely a fault is to be noticed before it does harm.
What to study, in order
- Learn to distinguish fault, error and failure.
- Do an FMEA of a small circuit or a software service, ranking the risks.
- Draw a fault tree for one serious outcome and find its smallest cut sets.
- Inject faults into a program you wrote, such as full disks and slow networks, and note what breaks.
- Write a post-incident review of a real failure you saw, with a timeline and actions.