False Alarm Rates
False alarms usually occur in 1 to 45% of the detected faults if the definition is expanded to include can-not duplicate failures. Obviously, one would like to minimize this effect and this generally is accomplished by designing repeated or recycled tests for each fault detection or increasing the tolerance factors for performance parameters. Analyzing to determine the level of false alarms requires the use of detailed and costly programs such as the STAMP program. In 1981, Hughes Aircraft Company did a study on this condition. The report is called "Anaylsis of Built-In-Test False Alarm Conditions, RADC-TR-81-220 that has some interesting comments on built-in-test and its limitations.
|