基于KPCA的发动机失火故障诊断研究

Engine Misfire Fault Diagnosis Based on KPCA

针对国六排放标准对汽车发动机失火故障诊断提出的新要求,提出一种基于KPCA的发动机多传感器数据驱动的失火故障诊断模型。该方案选取了发动机传感器网络中6个与失火关联度较高的传感器数据,分别进行数据处理和特征提取;然后使用KPCA对特征矩阵进行降维融合,挖掘特征之间的非线性关系,从而形成对发动机运行状态相对综合的评价;再以T2统计量和Q统计量作为指标进行失火故障诊断;最后通过发动机仿真数据集验证了该方案进行失火故障诊断的准确率。

To meet the engine misfire diagnosis requirements specified by China's stage VI vehicle emission standards,a misfire diagnosis model based on KPCA is proposed in this paper.Firstly,six sensors’data with high correlation to misfire are selected for data processing and features extraction respectively.Then,KPCA is used to reduce the dimension of the feature matrix,meantime the non-linear relationship between these features detected,therefore a comprehensive evaluation of the engine state formed.Then,T2 Statistics and Q Statistics are used as indexes for misfire diagnosis.Finally,the accuracy of misfire diagnosis is verified by engine simulation data set.