Indicated efficiency optimization by in-cycle closed-loop combustion control of diesel engines

This paper revises the design process for closed-loop combustion control based on in-cycle feedback. Methods for the necessary observers, detectors and virtual sensors are also discussed. The control problem tackles the indicated efficiency optimization, following two approaches, direct and indirect optimization. The argumentation provides a comprehensive overview on how to specify and integrate different methods available at the literature. The results presented are based on previous experimental and simulation work on a Scania D13 engine by the authors. A pilot and main injection scheme is used to increase the degrees of freedom and achieve a higher nominal indicated efficiency. An in-cycle controllability study indicated that the pilot mass is the most significant parameter to monitor for the in-cycle disturbance rejection. Based on the heat release analysis, feedback is obtained from the pilot mass estimation, with a real-time accuracy of ±0.5 mg/st. Still, to overcome the inherent delay between feedback and control action, predictive models are necessary. The required accuracy can be obtained by the multi-cylinder adaptation of the model parameters. The paper presents two approaches for the design of in-cycle controllers to improve the indicated efficiency under hardware and emission constraints. In the direct efficiency optimization, a model predicts the indicated efficiency change from the in-cycle pilot estimations to optimize the main fuel injection. Compared to in-cycle open-loop operation, the in-cycle controller increases the indicated efficiency by +0.42%unit. In the indirect efficiency optimization, the controller is designed to reduced the cyclic dispersion of the pilot SOC (down to ±0.4CAD), the pilot burnt mass (down to ±0.6 mg), the main SOC (down to ±0.3CAD) and the IMEP (down to ±0.2 bar). In case pilot misfire is detected, the CA50 shift of +2.6CAD is compensated and reduced to +0.5CAD. The indicated efficiency is maximized by the set-point optimization in a Monte Carlo simulation of a stochastic combustion model that captures the reduced cyclic dispersion of the controller. This approach increased the indicated efficiency by +0.6%unit under constraints on the maximum pressure, the maximum pressure rise rate, the maximum and the minimum exhaust temperature. Despite the controller effectiveness was quantified experimentally, the simulation results must be validated experimentally as well. Further investigations can focus on including constraints on emissions, how to handle the constraints on-line, the inclusion of additional injection pulses in the optimization, as well as variable valve timing actuation or injection of emissions reductants