Fault tree analysis

Fault tree analysis begins with a top event, the undesired system state, and decomposes it into intermediate and basic events using logical gates. AND gates indicate that all inputs must occur for the output to happen, while OR gates indicate that any input suffices. The resulting tree is a static representation of failure combinations, which can be analyzed qualitatively (minimal cut sets) and quantitatively (probability of top event).¹ The method assumes binary states (failed/working) and independence of basic events, though extensions handle dependencies and dynamic behaviors.

Fault tree analysis was developed in 1962 at Bell Laboratories by H.A. Watson and colleagues for the Minuteman missile system, under a U.S. Air Force contract to evaluate the safety of the launch control system.² It was later adopted by the nuclear industry, notably in the 1975 Reactor Safety Study (WASH-1400), which used FTA to assess the risk of nuclear power plants.³ Since then, FTA has been formalized in international standards such as IEC 61025 and is a cornerstone of probabilistic risk assessment.

FTA is used across industries to identify single points of failure, optimize maintenance strategies, and comply with safety regulations. In aerospace, it supports the safety analysis of aircraft systems; in chemical engineering, it helps prevent process accidents; and in software, it is applied to safety-critical systems like medical devices.⁴ Extensions include dynamic fault trees with sequence-dependent gates, fuzzy FTA for uncertain data, and integration with Markov analysis for repairable systems.⁵ The method also underpins the concept of minimal cut sets, which are essential for identifying the smallest combinations of failures that cause the top event.

Beyond its standard use, FTA has niche applications such as in cybersecurity risk assessment, where it models attack scenarios as fault trees. It has also been used in the analysis of medical errors, helping to design safer healthcare processes. A notable edge case is the handling of common-cause failures, where dependent failures can invalidate the independence assumption; techniques like beta-factor models are used to address this.⁶ Additionally, FTA can be combined with event tree analysis to form a complete risk assessment framework, as seen in nuclear safety studies. The method's graphical notation has influenced other diagrammatic techniques, such as reliability block diagrams and cause-consequence diagrams.

Top event

The undesired system state being analyzed, placed at the root of the fault tree.

Basic event

A leaf node representing a primary component failure or external event with a known probability.

Minimal cut set

The smallest set of basic events whose simultaneous occurrence guarantees the top event.

AND gate

A logic gate that outputs a failure only if all its inputs occur.

OR gate

A logic gate that outputs a failure if at least one of its inputs occurs.