基于改进的LVQ神经网络的发动机故障诊断

Fault diagnosis for engines based on improved LVQ neural network

学习向量量化(LVQ)神经网络可以通过监督学习完成对输入向量模式的准确分类,提出了一种基于改进的LVQ神经网络的发动机故障诊断方法,介绍了LVQ神经网络及其改进的学习算法。以长城哈佛GW2.8TC型发动机为实验对象,让发动机在怠速状况下,对发动机进行故障设置,利用金德KT600电脑故障诊断仪采集发动机数据流,运用改进的LVQ神经网络建立诊断模型,诊断结果表明,改进的LVQ神经网络能对发动机故障做出正确分类,准确率比较高。

Since learning vector quantization（LVQ）neural network can classify input vector pattern accurately by supervised learning,the fault diagnosis method for engines based on LVQ neural network is proposed. LVQ neural network and its improved learning method are introduced. Taking Great Wall Harvard GW2.8TC engine as the experimental subject,faults are set for the engine under idle speed condition. The data stream of the engine is collected by using Kinder KT600 computer fault diagnosis tester. The diagnosis model was established by using the improved LVQ neural network. The diagnosis results show that the improved LVQ neural network can classify engine faults accurately,and the precision rate is relatively high.