Research On Typical Fault Identification System And Method For Internal Combustion Engine Based On FBG Accelerometer

Structural health monitoring and.fault identification.technology without disassemble of internal combustion engine(ICE)is one of the key measures to ensure itself and large machinery and equipment operating safely and stably,whose power source is ICE.As the ICE surface vibration signal contains abundant information,home and aboad scholars have aroused widespread concern and certain results have been obtained.However,a number of challenges still need to be addressed.For instance,the signal With non-stationary and non-linear characteristics obtained by traditional electric vibration sensors is difficult to be directly used as the basis of fault identification,which is caused by structure and working environment of ICEs,the fault identification method is still in the preliminary exploration stage,etc.To slove the problems above,the typical fault of ICEs is studied.The fault identification model with high reliability is established to realize the accurate identification of typical faults through the study of FBG accelerometer design,demosulation system construction,signal processing,feature extraction,fault identification method and so on.This research carried on as follows:(1)In order to improve the sensitivity,flat response area and anti-cross interference ability of FBG accelerometers,high sensitivity FBG acceleration sensing models based on flexure hinge structure were proposed on basis of theoretical analysis of its sensing mechanism.The sensor based on flexure hinge structure were developed after response characteristics simlation by ANSYS and structural optimization.Then,a tunable F-P filter was used to construct the high frequency demodulation system based on the edge filtering principle.(2)A method based on multi-class support vector machine was proposed,which pro vied a reliable solution for ICE valve fault identification.Based on the optimization for installation location of the FBG accelerometer,a monitoring system of the ICE valve operating status was constructed,which realizes the accurate detection of the valve fault signal.The relationship between signal characteristics and valve fault was established on basis of feature extraction by wavelet decomposition and reconstruction algorithm and oscillation energy.Then,by using the support vector machine(SVM)with good classification accuracy and generalization ability,the ICE valve fault identification was realized with high accuracy by using a small number of model training samples.(3)A fault identification method for ICE fuel injection advance angle based on variational modal decomposition(VMD)algorithm and multi-class support vector machine was proposed.The relationship between the signal characteristics and the fuel injection advance angle fault was established by using VMD algorithm and oscillation energy to extract charateristics from fuel injection advance angle fault signals,which were collected by FBG acceleration sensor.Then,by using the SVM algorithm with good classification accuracy and generalization ability,the fault identification of ICE fuel injection advance angle was realized with high accuracy by using a small number of model training samples.(4)A hybrid fault identification method for ICEs based on diffusion mapping algorithm and multi-class support vector machine was proposed.The signal characteristics was extracted by wavelet decomposition and reconstruction algorithm and VMD algorithm.Then,the feature vector reduction method based on diffusion mapping algorithm was proposed to reduce the dimension of feature vectors and remove redundant information.Finally,by using the SVM algorithm with good classification accuracy and generalization ability,the hybrid fault identification of ICE vacle and fuel injection advance angle was realized with high accuracy by using a small number of model training samples.