Research On Injector Fault Diagnosis Technology Based On PCA-RF

The electronically controlled high pressure common rail fuel injection system can independently control the fuel injection pressure and fuel injection law,which is a more advanced fuel injection system at present.Among them,the fuel injector as its core component,a direct i MPact on fuel injection,atomization and combustion process,to a great extent,directly determines the performance of the diesel engine,due to the complex structure of fuel injector,the working conditions,the probability of failure is higher,therefore,the research of the injector of fault diagnosis technology is of great significance.At present,when diagnosing the fault of injector,the result of diagnosis is usually only normal and fault.When the fault occurs,it is also repaired after the fault.It is difficult to provide valuable reference suggestions for the preventive maintenance of injector,and the diagnosis method is mostly based on expert experience,and the requirements for technical personnel are relatively high.Therefore,in view of the above existing problems,this paper proposes an injector fault state monitoring and fault diagnosis method based on principal component analysis(PCA)and random forest(RF).The main research work is as follows:Firstly,in order to monitor the status of fuel injector in real time during the normal operation of diesel engine,the method of principal component analysis is proposed to monitor the fault status of fuel injector.Principal component analysis does not rely on the accurate physical model of the system,but also can better deal with the correlation between the feature data,map the feature data set to two orthogonal subspaces,and effectively monitor the fault state by monitoring the changes of the feature data in the subspace.Based on the time domain,frequency domain and time-frequency domain characteristic parameter data set of the extracted common rail pressure signal,the principal component analysis model is established,combined with the comprehensive statistics Ψ,the on-line fault condition monitoring of fuel injector is effectively realized.Secondly,injector failure types,such as injection hole plugging,hysteresis,needle valve,needle valve card guide surface wear,etc.,for the fuel injector failure occurs the exact location of the fault type and fault judging,by co MParing the different performance of fault diagnosis methods,build the random forest combined classifier model,55 test sample,Only one test sample failed to predict the fault pattern,and the accuracy rate of fault pattern recognition reached 98.18%.Out of 40 test samples,only one test sample was wrong in fault location judgment,and the accuracy of fault location judgment reached more than 95%.Then,in order to quantitatively represent the current health status of the injector and provide more valuable suggestions for the preventive maintenance of the injector,the health status evaluation system of the injector is established,the health level of the injector is divided,and combined with the Mahalanobis distance algorithm,The health status of fuel injector and its corresponding health level are quantitatively expressed by health index.Finally,from the point of engineering practice,the on-line fault diagnosis software system of electronically controlled high pressure common rail injector is designed by using MATLAB and Lab VIEW joint programming method,which provides an effective idea for fault state monitoring,fault analysis and diagnosis of injector in practice.