Research on Fault Feature Extraction of Diesel Engine Cylinder Based on MMD and KICA

Aiming at the problem of fault features of cylinder extraction being difficult because the vibration sources on the cylinder head surface of diesel engine are non-linear coupled and interlaced. A diesel engine cylinder fault feature extraction method based on multi-scale morphological decomposition (MMD) and kernel independent component analysis (KICA) was proposed to isolate and identify a single vibration source and extract the fault features from it. Firstly, a new combined difference morphological filter is constructed by using three combination operators to decompose the cylinder head surface vibration signals into a series of morphological pattern components describing different levels of vibration feature information. Then, the original signal and morphological pattern components are used as the input of KICA to identify a single vibration source signal from the obtained independent components. Finally, the single vibration source signal is extracted from the single vibration source signal The fault feature vector is extracted from the source signal, which is used for diesel engine cylinder fault diagnosis. The results of simulation signal and diesel engine fault signal analysis showed that the proposed method can effectively process the nonlinear coupling vibration signal and extract the fault features.