Abstract: Based on the mechanism of the false alarm and built-in test (BIT) system information treating process, a model was established for calculating false alarm rate (FAR). Based on the Bayes probability, an analysis model for FAR in general BIT systems was established. For decreasing FAR, an improved model for BIT systems with embedded abnormal data detecting and renewing functions was proposed, and the effectiveness of the proposed model was proved. The factors affecting FAR were analyzed systematically, revealing that the main characteristic parameters include the FAR on diagnosis layer and the missed alarm rate and fault rate on sensor layer.