Research On Knock Characteristics Analysis And Evaluation Method Of Gasoline Engine

Strong knock is an abnormal combustion phenomenon that can reduce the performance of gasoline engine,and even destroy engine components,while the power and fuel economy of gasoline engine will increase when operating at the edge of knock.So the purpose of knock control is to make the gasoline engine operate at the edge of knock,and the key to knock control is to extract the knock feature well and combine an accurate knock recognition algorithm.Based on the combustion test data of high-speed gasoline engines,this paper aims at finding effective approaches to extract knock features and accurate algorithms to identify knock.Specific tasks are as follows:Firstly,the knock characteristic frequency at different positions in the combustion chamber was studied by comparing the frequency of different resonant modes and the spectral characteristics of two kinds of cylinder pressure sensors signals.The comparison result indicates that the edge position of the combustion chamber is sensitive to the tangential vibration mode,and the center of the combustion chamber is sensitive to the first-order radial vibration mode.On this basis,a method for knock recognition considering different resonant modes was proposed.The first order radial resonance mode method is adopted in case the cylinder pressure is collected at the center of the combustion chamber,while the first three order tangential resonance model is adopted in case the cylinder pressure is collected at the edge of the combustion chamber.Finally,the consistency of knock identification result by the two methods was verified.With a view to the failure criterion of VDO knock recognition algorithm,knock diagnosis expansion algorithm based on Moving Window Method(MWD)was proposed.The effect of calculating window width on the knock recognition was studied and the size was determined,meanwhile the differences of the knock calculation starting point,knock factor and knock intensity between the two algorithms was mainly contrasted.The results show that the knock factor of the MWD algorithm is more sensitive to the knock,which can effectively avoid the missed diagnosis of the knock occurring before the peak of cylinder pressure.Based on Mallat algorithm's prominent time-frequency analysis function and its frequency folding defect,single-band discrete wavelet reconstruction algorithm was introduced to extract the knock characteristic components from the cylinder pressure signal under different ignition advance angle conditions.Firstly,the characteristics of the decomposed sub-band were studied,and the result shows that most of the knock energy of the prototype is distributed in the frequency range of 6.25 to 12.5 kHz.The energy distribution characteristics of each sub-band under different conditions were analyzed.Finally,the difference between DWT and filtering method for knock feature extraction was compared,and the result shows that DWT method can highlight and effectively extract the knock features.Based on using DWT to extract the knock features,the time domain and the frequency domain knock evaluation methods were proposed,then the time domain root mean square(RMS)and frequency domain integral energy were respectively selected as the knock evaluation indexes.In order to reduce the interference of background noise and determine a fixed knock threshold,the knock evaluation indexes were normalized,meanwhile the heavy knock threshold and the soft knock threshold of the two methods were determined respectively.Through the comparison of knock evaluation index and knock intensity for two methods,it was found that the frequency domain method can avoid the defect of the time domain method due to the knock energy exceeding the range of knock window,which is conducive to knock recognition.