Fault Diagnosis Of Air System Of Commercial Vehicle Diesel Engine Based On Data Drive

Under the China Ⅵ emission regulations,the emission standard of diesel engine is further improved,in which the air system has EGR and supercharger components at the same time,the structure and working conditions are complex,the probability of failure increases,which will have a great impact on the power,economy and emissions of the engine.At the same time,due to the addition of a number of sensors in the air system of the China Ⅵ diesel engine,the perceptible state parameters of the diesel engine are increased.Under the background of the Internet of vehicles,the running state parameters of the engine can realize real-time remote monitoring.It is feasible to use the big data method to carry out fault diagnosis and pre-diagnosis of the air system of the commercial vehicle engine.In addition,due to the problems of high cost and low efficiency,the existing fault diagnosis technology has not been able to meet the needs of the current market,and the diagnosis method of diesel engine is not perfect,so to achieve a more efficient and accurate fault diagnosis method to meet the current development needs has a very high application value and very important theoretical significance.This paper uses data-driven methods to diagnose the air system faults of a certain type of commercial vehicle diesel engine.First,preprocess the original highdimensional process data,using data standardization,local linear embedding algorithm,and local anomaly factor detection methods to standardize the data,reduce dimensionality,and denoise,and obtain a high-quality data set for subsequent algorithm analysis;Secondly,for the nonlinear problems existing in the diesel engine,the kernel method is introduced,combined with the Gaussian kernel function,the kernel principal component analysis method and the kernel Fisher discriminant analysis method are proposed,which can effectively process the nonlinear data.The kernel principal component analysis method is used to establish the kernel principal component fault diagnosis model based on historical data,which realizes the accurate detection of abnormal process variables in diesel engine fault,provides the basis for in-depth diagnosis.The kernel Fisher discriminant analysis method is used to establish a classifier for historical data.The classifier is used to classify the unknown data and complete the diagnosis of air system faults.In view of the limitations of kernel Fisher discriminant analysis as a supervised learning method in sample labeling,a semi-supervised learning is introduced to form a semi-supervised kernel Fisher discriminant analysis method for diesel engine fault diagnosis,which uses a large amount of unlabeled data and a small amount of labeled data to train together.Through continuous iteration to establish a more accurate classifier,to achieve a more efficient diagnosis of the typical faults of the diesel engine air system.It reduces the difficulty of sample marking,shortens the time of diagnosis and improves the efficiency of diagnosis.For the selection of kernel parameters in kernel method,particle swarm optimization(PSO)is introduced to optimize the kernel parameters,which ensures the processing ability of kernel method for nonlinear data,and makes it better applied to the fault diagnosis of diesel engine.The algorithm of fault diagnosis for commercial vehicle diesel engine proposed in this paper is verified by simulation with different data sets.The results show that the proposed method is effective for fault diagnosis of diesel engine,can accurately diagnose six kinds of faults in the EGR system of diesel engine,and improve the efficiency of fault diagnosis,which is of great significance to reduce the occurrence rate of faults and improve the operation reliability.The obtained method is extended to other systems of diesel engine and provides support for the establishment of online health management system of diesel engine.