Research On Stochastic Knock Control For Gasoline Engine

As the core component of the automobile,the knock control of one of its core technologies has been monopolized by a few developed countries,which has formed a barrier to the improvement of the technical level of the automobile industry in China.In this paper,two of the important factors affecting the combustion efficiency of automotive engines that knock diagnosis and knock control are studied.The research on the core theory method in the international front is based on the GSX-R600 of four cylinder gasoline engine as the test object.The engine test stand was improved.The signal detection system based on vibration sensor was selected.On this basis,engine knock tests were conducted under partial conditions,and necessary test data were obtained for the study of knock diagnosis and knock control.A more complete knock feature extraction,automatic identification and knock diagnosis algorithm is proposed:(1)Discrete wavelet transform is performed on the detected signal(the signal detected by the vibration sensor)to obtain the knocking feature signal;(2)the envelope of the knock feature signal in the envelope window is taken by the Hilbert transform;(3)Fitting the envelope;(4)a knock intensity index is defined,and the knock intensity index is used to determine knock or not.The method is applied to the test engine.The results show that the proposed method can realize automatic machine recognition of knock feature,and the accuracy of diagnosis is better than the conventional knock diagnosis method.Using the experimental data,the correlation and distribution laws of the knock intensity data at each ignition advance angle were studied.The results show that knock events is independent stochastic process.Therefore,the knock event should be controlled as an independent stochastic process.Normal distributions,lognormal distributions,gamma distributions,Log-logistic distributions,t Location-Scale distributions,and generalized extreme value distributions for comparing and analyzing the knock intensity data.The results show that the distribution of the knock intensity data is close to the t Location-Scale distribution when ignition advance angle is small.the distribution of the knock intensity data is close to the lognormal distribution when ignition advance angle is large.A simulation model of knock intensity data is established,and an adequate and reliable knock strength data method is simulated based on the limited knock intensity data.On this basis,the conventional knock control method and based on likelihood ratio knock random control method are established the simulation model,and the two controllers are optimized.Through the simulation method,the steady state and transient state control characteristics of the conventional knock control method and based on likelihood ratio knock random control method are compared and analyzed,and their advantages and disadvantages are given.Based on this,a stochastic knock stochastic control method that can take into account both the magnitude of knock intensity and the probability of knock events is proposed.It is referred to as the KI-LN knock stochastic control method.The KI-LN knock stochastic control method was applied to test engine knock data.The simulation results show that this method not only achieves random control of engine knock,but also can control the engine to avoid Highintensity knock occurrences according to the set knock intensity threshold.Thus realizing operated at the knock edge to obtain the best power and economy of engine under the condition of ensuring the safe operation of engine...