柴油机振动谱图纹理增强的NMF特征提取方法

Feature Extraction Method of Diesel Engine Vibration Signal with Texture Magnified Non-negative Matrix Decomposition

针对柴油机振动加速度信号特征提取困难的问题,提出一种对时频图像纹理信息进行增强的矩阵非负分解方法,用于提取柴油机振动信号的时频特征。首先,利用匹配追踪算法(MP)结合Wigner-Ville分布获取时频分辨率较好的时频表征;然后引入局部二值模式(LBP)算子对时频图像的灰度矩阵重新编码,再利用非负矩阵分解算法(NMF)获取时频图像对应的低维特征参量,以增强时频图像的纹理特征,提高NMF的特征提取效果。通过对柴油机4种不同状态的振动信号进行分析研究,可得出结论:该方法能有效地表达柴油机缸盖振动加速度信号的时频特征,可用于准确地诊断柴油机的气门故障。

In view of the difficulty in diesel engine vibration acceleration signal feature extraction,a kind of vibration signal time-frequency characteristics extraction method,which joined the local texture feature information of nonnegative matrix decomposition method,is put forward. Firstly,decompose the diesel engine cylinder head vibration acceleration signal by the matching pursuit(MP） algorithm, and obtain the 2-d time-frequency distribution diagrams by superimposing the decomposed signal Wigner-Ville distribution. Then, use NMF to calculate the corresponding low dimensional time-frequency characteristic parameters of the time-frequency distribution image. Next,introduce local binary pattern(LBP） operator pair recoding,frequency of the image gray level matrix to improve the effect of feature extraction of NMF. By the analysis of four different states of diesel engine vibration acceleration signal,which is corresponding to the valve clearance is normal,the valve clearance is too small or too big,and valve leakage,the results showed that the combination of LBP and NMF can effectively express the time-frequency characteristics of diesel engine cylinder head vibration acceleration signal,which can be used for the diagnosis of diesel engine valve fault accurately.