A rotor dynamics and machine learning-based approach for high-accuracy engine torque tracking

Under the global carbon neutrality strategy, the upgrading of traditional engines has garnered increasing attention. As a crucial front-end development tool, 1D engine simulation requires high torque tracking accuracy under transient conditions, which directly affects the evaluation of design solutions. To address the issues of response lag and tracking deviation in traditional steady-state MAP-based control methods during transient processes, this study proposes an intelligent fuel injection quantity control strategy based on a dynamic coupling equation for rotor power balance. By integrating parameters such as rotational speed, target torque, and multidimensional dynamic variables, this strategy establishes an intelligent inference mechanism for fuel injection quantity. A hybrid methodology combining mechanistic modeling and machine learning is employed to achieve high-precision characterization of key parameters. Validation results under high-power dynamic operating conditions demonstrate that, compared to conventional methods, the proposed strategy reduces cumulative torque tracking error by approximately 89.5%, significantly enhancing system dynamic response speed and control robustness. This research provides an effective solution for improving the predictive accuracy and control reliability of 1D engine simulations under transient conditions.