基于多传感器信息融合的贝叶斯网络故障诊断方法研究及应用

针对齿轮泵信号具有复杂性和模糊性的特点,提出了一种基于多传感器信息融合的贝叶斯网络故障诊断方法。分析了齿轮泵振动和压力信号特点,以此为基础提取了振动信号的能量特征、分形特征和压力信号的高频压力脉动3种特征属性,构建了多故障贝叶斯网络对特征进行融合,设计了贝叶斯分类器,通过最大后验概率准则识别故障类型。两次融合结果表明:多传感器信息完备了特征空间,提高了诊断正确率,能够有效实现齿轮泵多种故障的诊断,具有较好的应用价值。

Fault Diagnosis Approach of Bayesian Networks Based on Multi-sensor Information Fusion and Application

Aiming to the complex and fuzz nature of gear pump signal,a fault diagnosis approach of Bayesian networks was presented based on multi-sensor information fusion.Firstly,this study analyzed the signal characteristic of vibration and pressure,and extracted energy and fractal characteristis of vibration signal and high frequency pressure pulse of pressure signal.Then the multi-fault Bayesian network was built to fuse these characteristics.Finally,Bayesian classifier was designed to identify the fault pattern through the maximum posterior estimation.Twice fusion results show that multi-sensor information completes the characteristic space,raises the diagnosis exactitude rate.This approach can effectively carry out the gear pump multi-fault diagnosis and have better application value.