In order to precisely control the GDI engine fuel injection system during it"s whole life and complex working conditions

a model-based non-calibrated control algorithm (NC-MBC) is proposed. Established a precise mechanism model for high pressure common rail system (HPCRS)

then used 2600 experimental conditions to fit and validate the model. The self-learning algorithm is designed to optimize the unknown physical and structural parameters in the model online. The controller is designed by combining the precise feedforward model and the active disturbance rejection control algorithm

this realized the non-calibration control of the HPCRS and has disturbance rejection capacity. The NC-MBC is verified by Simulink simulation platform. The results show that the model parameters can converge quickly by using self-learning algorithm. During steady condition

the parameter converged NC-MBC is declined over 40% than PID in integrated absolute error

and is reduced by more than 45% in response time and overshoot during transient condition. NC-MBC has a low computational complexity and greatly improves the control precision and robustness

which is suitable for embedded control.