Engine Knock Detection And Intensity Evaluation Based On Sparse Representation

The minor reinforcement technique of gasoline engine makes the engine face more severe working environment,which leads to the increase of engine knock tendency.Strong knock will affect the normal combustion of the engine.High frequency shock wave may even damage the engine structure,which can reduce the service life of the engine.However,slight knock will make the mixture combustion in the cylinder close to the ideal condition of constant volume combustion,which is conducive to improving the thermal efficiency of the engine.Therefore,accurate identification of engine knock conditions and accurate evaluation of knock intensity is conducive to optimizing engine combustion and reducing the damage caused by knock,which is of great theoretical research significance and engineering application value.Based on the sparse expression theory,this paper takes the knock of gasoline engine as the research object,takes the body vibration signal as the basis,and combines the experimental and simulation means to study the engine knock identification method based on the improved K-means Singular Value Decomposition algorithm.The evaluation index of detonation intensity based on the common logarithm of fourth-order cumulant is proposed.Specific include the following content(1)Build a certain type of engine knock test bench.The engine knock test scheme is designed with the rotation speed,load and ignition advance Angle as variables.The synchronous signal acquisition of cylinder pressure and body vibration acceleration under various working conditions was carried out.The time-domain and frequencydomain characteristics of pressure changes in cylinder under different detonation degrees were preliminarily analyzed,which laid a foundation for subsequent vibration signal processing and extraction of detonation characteristics based on body vibration.(2)According to the body vibration signal,the method of knock feature extraction and identification is studied.The improved sparse representation algorithm is applied to engine body vibration signal processing to extract knock characteristics in vibration signals,and the sparse representation theory is introduced into the field of engine knock feature identification.Firstly,an improved K-SVD method based on generalized orthogonal matching pursuit is proposed,which improves the efficiency of atoms updating in dictionary learning process and overcomes the low computational efficiency of the traditional sparse representation dictionary construction method.After that,the influence of main parameters in the sparse representation process on the feature extraction effect is analyzed,and the results show that the improved K-SVD algorithm has good parameter robustness and computational stability.Finally,the effect of the improved sparse representation algorithm is verified,which shows good noise reduction effect and feature extraction effect in both analog signal processing and actual signal processing.(3)An evaluation index of engine knock strength based on fourth-order cumulant is proposed.In this method,the body vibration signals are sparse decomposed to obtain a sparse dictionary covering knock characteristics and a sparse coefficient for a single signal.On this basis,the common logarithm of the fourth-order cumulant of the reconstructed signal is calculated and used as the evaluation index of knock intensity.Calculation results show that the proposed method shows good capability of noise reduction and feature extraction for mixed with strong background noise of the cylinder block vibration signal,and improve the efficiency of the operation.It can accurately distinguish three states of the strong knock,the slight knock and normal combustion,which shows the application value in the field of engine knock recognition.