基于信息融合与神经网络的复合振动故障诊断

Compound Vibration Fault Diagnosis Based on Information Fusion and Neural Networks

依据复合故障特性 ,提出了一种基于信息融合与神经网络的复合振动故障诊断方法。进行了传感器级的时间跨度的特征融合 ,然后再实行全局的空间跨度的融合。融合过程基于神经网络式特征提取 ,该方法可以在系统状态未知的情况下 ,自适应地融合不同故障测点的信息 ,从而较全面、准确、及时地反映系统的振动故障状态。并以汽车动力系统复合故障的诊断事例详细说明了该方法的具体实现步骤。结果表明 ,经过多故障特征信息融合 ,诊断结论的可信度明显提高 ,不确定性明显减小 ,显示了该诊断方法的有效性。

According to the compound vibration fault attribute, a method of compound vibration fault diagnosis based on information fusion and neural networks is proposed. The feature data from sensor are preliminarily fused over an extended time period.Then the results are fed into overall fusion center,in which the final fusion is performed spatially.Because the process is based on neural networks pattern clustering,the various information sources can be fused adaptively when the characteristic of system is unavailable.The fusion output can give an all-round,exact and timely description of vibration fault. The compound vibration fault diagnosis of the auto drive systems is made. The results show that the reliability of the diagnosis is improved and the uncertainty decreases markedly. So the diagnosis method is effective.[