基于贝叶斯决策的机内测试虚警滤波技术

Built-in test false alarm filtering technique based on Bayesian decision

提出了一种将贝叶斯决策用于机内测试(BIT)系统的虚警滤波方法,诊断的BIT系统包括正常、瞬态、间歇和故障四类状态,利用最小最大决策法确定这四类状态的先验概率,并建立相应的贝叶斯风险决策表.以机载电源的机内测试系统为例进行研究,为每个诊断模块建立相应的贝叶斯决策器,通过支持向量机识别出有异常的可更换模块后,启动相应的决策器对该模块的BIT状态进行决策.结果表明,该方法能够识别出系统中的瞬态和间歇故障,可有效降低BIT虚警发生的概率.

A built-in test （BIT） false alarm filter technique based on Bayesian decision is proposed in order to reduce the high false alarm rate and improve the accuracy of BIT diagnostic system. Unlike the conventional BIT technique, the proposed method divides the states of a system into 4 states, that is normal state, transient state, intermittent state and faulty state, employs min-max principle to calculate the prior probabilities of the four states, and sets up the corresponding risk decision table. The proposed method is applied to the BIT system of aircraft electrical power system, and when an abnormal power module is identified by the support vector machine （SVM） , a corresponding Bayesian decision machine performs to determine the real state of this identified module. The experimental results have shown that the proposed method can identify the transient fault and intermittent fault effectively, which filters the BIT false alarm of aircraft electrical power system.