Non-calibration Control Algorithm Research For Gasoline Direct Injection Engine Fuel Injection System

Increasing the injection pressure of the fuel injection system is the key to improve the power,economy and emissions of gasoline engines.Traditional control algorithms has problems such as too many parameters need to be calibrated、uncertain disturbances and aging、wear and carbon deposits during full life cycle operation.This put forward higher requirements for complex working condition,active disturbance resistance and full life-cycle adaptive ability of the control algorithm.In this paper,an non-calibration control algorithm is designed based on self-learning accurate model based feedforward control and active disturbance rejection feedback control.The control algorithm is verified by simulation platform and experimental platform.Firstly,the transient object model is established based on the deep understanding of fuel injection system mechanism.2000 operating point from a 1.3L GDI gasoline engine are used to fit and verify the model accuracy.The results show that over 95% of the key parameters fitting error is within 5%.Provide an effective simulation platform for control algorithm development.Set up a four-cylinder gasoline engine experimental platform,combined with a self-developed DEMS control platform to achieve control algorithm experimental verification environment construction.Secondly,in order to reduce the sensor sampling noise and provide stable and reliable data for the control algorithm operation,a rail pressure sampling filtering algorithm was designed.In order to reduce the workload of traditional MAP calibration,a high-precision mean value model for feedforward control of the common rail system is established.In order to verify the robustness and generalization of the model,2500steady-state and 600 transient operating points from 1.3L and 1.4L gasoline engines were used to fit the model.The fitting and verification R Square of all operating conditions were above 0.99.In order to cope with the characteristics of complex engine operating conditions and the entire life cycle,a segmented self-learner based on the recursive least square method is designed to modify model parameters in real time.In order to compensate the modeling error of the feedforward model and the uncertain external disturbance,an active disturbance feedback controller is designed to compensate the disturbance.In order to improve the operation efficiency of the algorithm,a multi-core collaborative processing mechanism of the algorithm is designed based on the multi-core embedded chip and the real-time requirement of the algorithm.Finally,the algorithm ran on Simulink platform.The simulation results show that the parameter self-learner can guarantee that all parameters would convergence within a limited number of iterations.After calibration,the non-calibration controller’s steadystate IAE,transient overshoot,and response time reduced by more than 32% compare to PID algorithm.The non-calibration algorithm can quickly observe and suppress disturbances such as speed and load steps.During this period,the rail pressure IAE reduced by more than 45% compare to PID.The fuel injection carbon deposition and high-pressure oil pump wear simulation experiments are designed.Due to the improvement of the model accuracy after parameter self-learning,the corrected rail pressure IAE is reduced by more than 36%.The algorithm uses automatic code upgrade and integration in the autonomous DEMS controller.Dynamic control experiments are performed on the Great Wall 2.0L four-cylinder gasoline engine experimental platform.In the continuous step-up and step-down experiments,the non-calibrated algorithm reduces overshoot and response time by more than 70%,and the rail pressure IAE decreases by 36% throughout the process compare to PID.In the experiment of speed and load disturbance,the non-calibration algorithm can quickly observe and suppress the disturbances.An algorithm control effect experiment was designed for the entire life cycle.The control effect deteriorated when the old high-pressure oil pump was replaced with a new high-pressure oil pump.After the high-pressure oil pump selflearning algorithm was turned on,the rail pressure IAE decreased by 57%.