Research on multi-objective control of PPCI diesel engine combustion process based on data driven modelling

Abstract

Control of combustion stability in partial pre-mixed compression ignition (PPCI) engine is one of the main issues facing its application. However, the multi-parameter coupling and nonlinear increase in the combustion process make the model and controller design more difficult. Therefore, this study proposed a diesel engine control method that combines neural networks and model-free adaptive control in the absence of model and controller structure, which can achieve real-time coordination control of crank angle at 50 % of total heat release (CA50) and indicated mean effective pressure (IMEP) in the PPCI combustion process. Through comparisons under different operating conditions, it was found that the adjustment of algorithm parameters needs to adapt to the sensitivity changes of control parameters. In addition, the study validated the real-time performance and control effect of the algorithm, the experimental results indicate that the execution time of the control algorithm is approximately 5.59 milliseconds, which satisfies the real-time control requirements for the combustion process. By adjusting the weight coefficient matrix of the control authority, CA50 and IMEP are effectively tracked within the constraints of maximum pressure rise rate. The control error for CA50 remains within ±2.7 %, while that for IMEP is confined to ±1 %. Furthermore, the root mean square error for CA50 is measured at 1.1 crank angle, and for IMEP it stands at 23.5 kPa, thereby achieving precise real-time control of the PPCI combustion process.

Keywords

• PPCI
• MFAC
• Neural network
• Multi-objective collaboration
• Real-time combustion control