| Most of the current combustion closed-loop control reflects inter-cylinder and inter-cycle combustion by testing the current cycle cylinder pressure data for the regulation of the next cycle injection parameters,which is hysteresis control.However,the combustion process in the engine cylinder is rapidly changing,generating thousands of cycles per minute,and there are cyclic variations between cycles,so this hysteresis control is subject to error.To address this problem,this paper develops an algorithm for real-time online prediction of combustion parameters under full operating conditions of diesel engines with the aim of combustion state prediction control.Firstly,0-D model of a YC6K420LN-C31 high pressure common rail diesel engine was built as a virtual engine to replace the physical engine to provide a signal source for the real-time online capability verification of the algorithm.The diesel engine is divided into four subsystems:the common rail fuel system,the intake and exhaust system,the cylinder system and the crank and connecting rod dynamics system,and the subsystems are modelled separately.The simulation data of rotational speed,cyclic injection volume and cylinder pressure at different operating points were selected and compared with the actual operating data of the diesel engine to complete the accuracy check of the model,in which the relative errors of cyclic injection volume and rotational speed were less than 10%and the R~2 of cylinder pressure fit was greater than 99.5%.Secondly,in order to establish a more accurate combustion model,the difference in exothermic rate and the choice of Wiebe function form of the YC-6K diesel engine under full working conditions are investigated.The original modelling data was obtained on the basis of a diesel engine test rig,with a total of 100 different operating conditions over the full operating range.The combustion stages at full operating conditions are divided according to the morphological characteristics of the exothermic rate curves,and the Wiebe function is used to match different basis functions for the different stages of the exothermic rate curves.Thirdly,to improve the performance of the Wiebe based combustion model,the unknown parameters in the model are identified,the identification of m values is based on the least squares method and the selection of a values in the double Wiebe function is based on a differential evolutionary algorithm,and the offline and online exergy curves are trained with an accuracy greater than 95%.In order to achieve online prediction of combustion process parameters,the mapping relationship between combustion parameters and CA50 was established based on BP neural network.The goodness of fit R~2 of BP neural network was 99.72%,the root mean square error RMSE was 1.52℃A,and the relative prediction error percentage of CA50 were less than 10%,which could meet the modelling requirements of the prediction model.Finally,to verify the real-time online prediction capability of the algorithm,the algorithm was validated based on the NI rapid prototyping simulation platform.The combustion model prediction algorithm was deployed to the NI real-time simulator and the diesel engine real-time simulation model was deployed to another NI real-time simulator as a virtual engine to provide a signal source for the combustion parameter prediction model.The average relative error of CA50 prediction for both steady-state was less than 8%.The test results in the rapid prototyping platform showed that the average computation time of the 6-cylinder engine model was about 7.54μs,which was much less than the time required for a single cycle of a real diesel engine,meeting the requirements for real-time prediction of the diesel engine combustion process. |
PDF Full Text Request