The Dangers of Failure Masking in Fault-Tolerant Software: Aspects of a Recent In-Flight Upset Event

On 1 August 2005, a Boeing Company 777-200 aircraft, operating on an international passenger flight from Australia to Malaysia, was involved in a significant upset event while flying on autopilot. The Australian Transport Safety Bureau’s investigation into the event discovered that “an anomaly existed in the component software hierarchy that allowed inputs from a known faulty accelerometer to be processed by the air data inertial reference unit (ADIRU) and used by the primary flight computer, autopilot and other aircraft systems.” This anomaly had existed in original ADIRU software, and had not been detected in the testing and certification process for the unit. This paper describes the software aspects of the incident in detail, and suggests possible implications concerning complex, safetycritical, fault-tolerant software.