Professional Manufacture 0 432 191 426 Diesel Injector Common Rail Injector Engine Parts Vehicle Parts 0432191426

Abstract:

Multi-stage fuel injection has become a key technology in modern diesel and gasoline direct injection engines, as it enables precise control of fuel atomization, combustion temperature, and emissions. However, due to complex coupling among injection timing, duration, and quantity in each injection stage, conventional control methods struggle to achieve optimal performance across different engine operating conditions. To address this challenge, this study proposes an intelligent optimization framework for multi-stage injection control strategies, integrating physical modeling, machine learning, and real-time feedback control.

A high-pressure common-rail injection system model was established using MATLAB/Simulink and GT-Power, incorporating injector dynamics, fuel compressibility, and cavitation effects. The model accurately reproduced pilot, main, and post-injection processes under varying rail pressures (80–180 MPa). Sensitivity analysis identified key parameters—such as injection interval, pilot injection ratio, and rail pressure fluctuation—as major factors influencing combustion efficiency and NOx–PM trade-offs.

To optimize control performance, a reinforcement learning-based controller (Deep Deterministic Policy Gradient, DDPG) was developed to autonomously adjust multi-stage injection parameters in response to feedback from combustion and emission indicators. Compared with traditional PID and model predictive control (MPC), the proposed controller achieved a 7.8% reduction in specific fuel consumption, 12.5% decrease in NOx emissions, and 9.3% lower soot concentration under transient conditions. The learning algorithm demonstrated strong adaptability to varying fuel properties and ambient temperatures, maintaining stable performance during cold starts and load transitions.

Experimental validation on a four-cylinder diesel engine test bench confirmed the effectiveness of the optimized strategy. Cylinder pressure and heat release analysis revealed that the proposed intelligent control reduced combustion noise and improved fuel–air mixing uniformity.

Overall, this research provides a data-driven and adaptive framework for optimizing injector multi-stage control, offering theoretical support for next-generation intelligent combustion systems and energy-efficient, low-emission engine technologies.