中文摘要：

针对以往基于小波分析提取气体传感器故障特征的不足,文中提出一种基于样本熵和峭度的自确认气体传感器故障诊断方法。该方法从信息论角度出发,首先直接提取传感器输出序列的样本熵和峭度作为故障特征,再输入支持向量机实现分类诊断。经提取的故障特征仅二维,极大地减轻了分类器模型的复杂度以及总体的诊断耗时。经仿真实验证明,文中方法可有效地提取故障特征,在低诊断耗时下依然有高达97%的准确率。

英文摘要：

Aiming to the shortage of wavelet-based feature extraction method of sensor,a self-validating gas sensor fault diagnosis method based on sample entropy and kurtosis was proposed.From the view of information theory,this method directly extracted sample entropy and kurtosis of sensor＇s output sequence as the fault features first,and then input them to train SVM.With the number of feature＇s dimension was only 2,this method reduced complexity of the classification model and diagnosis time extremely.The simulation results show that our proposed method can significantly extract fault features of self-validating gas sensor and the fault diagnosis accuracy rate reaches up to 97% in a lot less diagnosis time.