基于贝叶斯网络的复杂系统故障预测

Fault prediction of complex systems based on Bayesian network

针对复杂电子系统信号具有不确定性的特点,提出一种基于贝叶斯网络的故障预测模型。该模型通过对连续的信号特征进行量化处理,利用专家知识结合信号建立贝叶斯网络结构;对不同样本采用不同算法来进行网络学习,采用概率推理定量估计信号的区间预测概率,从而建立一个可推理的预测模型。将该方法应用于电源系统进行故障预测,针对不同数据样本进行实验,结果验证具有较高的区间预测率,为复杂系统的故障预测提供了新手段。

A fault prediction model based on Bayesian network is presented for the uncertainty characteristic of signals in the complicated electronic system. Through transforming continuous signals into discrete ones and using the expert knowledge, it builds a Bayesian networks for space prediction. Using the differ stylebooks and its correspond algorithms, the space prediction probability is rational estimated through reasoning, thus there sets up a rational Bayesian network for prediction. Applying this model to fault prediction of a power supply system, the experiment results show high space prediction probability and provide a new method for fault prediction of complicated electronic systems.