Combustion Diagnosis And Feedback Control Of Gasoline Compression Combustion Engine Based On Cylinder Pressure

With the increasing of emission regulations and economic requirements,the research of new engine technology and the improvement of traditional internal combustion engine technology have been paid more and more attention.In this paper,a hot topic in the field of engine research,the diagnosis of abnormal combustion based on cylinder pressure and the feedback control of load and combustion phase of HCCI are investigated.In the aspect of HCCI misfire detection,the combustion exothermic analysis method based on cylinder pressure(total exothermic rate method and net exothermic rate method)is studied,and the phenomena of misfire,partial combustion and hysterical combustion are analyzed according to the relevant parameters of each cycle exothermic rate solved.The mean indicating pressure misfire diagnosis model and neural network misfire classification model based on cylinder pressure at specific position were established respectively.The research shows that the IMEP diagnosis method can accurately identify complete misfire and partial combustion at 180°CA after top dead center,and has good working condition adaptability.The BP neural network model can accurately detect the normal combustion cycle at 20°CA after top dead center,and classify the delayed combustion,partial combustion and complete fire into one class.The detection is more timely to provide the possibility for the fire-compensating measures.In the aspect of HCCI knock diagnosis,ⅡR high-pass filter and Mallat wavelet algorithm are used to extract the knock characteristics of cylinder pressure signal.After comparative analysis,third-order ⅡR filtering is selected as the knock feature extraction method in this paper,and STM32 is used to simulate and verify the real-time performance of ⅡR filtering algorithm.After analyzing the geometric properties of the knock fault characteristics from multiple aspects,the amplitude class,integral class,ratio class,envelope class and other time domain parameters are established successively,and a total of 12 evaluation indexes are screened.According to the properties of the index and its knock correlation,six index parameters,namely peak value of oscillation pressure,peak value of oscillation pressure rise rate,oscillation pressure integral,relative energy,attenuation edge index and standard deviation,were selected for joint diagnosis.In Simulink,the optimal BP neural network based on genetic algorithm and the knock classifier model based on fuzzy C-means clustering algorithm were established respectively,both of which realized the knock intensity discrimination based on cycle,and the neural network model has higher diagnostic accuracy.In terms of feedback control,GT-Power is used to establish the HCCI single-cylinder engine model,explore the control law of load and combustion phase,and construct the pseudo-linear system by using the inverse model of neural network.Discrete state space models with high matching degree with IMEP control system and CA50 pseudo-linear control system were obtained by system identification method,and corresponding Kalman state observer and MPC controller were designed.Matlab/Simulink and GT-Power were coupled to build a co-simulation platform to realize the feedback control of HCCI load and combustion phase respectively.The results show that the MPC controller designed in this paper has a good control and tracking effect.After using the tracking differentiator to set the transition curve for the step reference trajectory,the MPC controller can still accurately control the index changes along the transition trajectory.The control accuracy was ±0.23 bar for IMEP and ±0.84°CA for CA50.