Design and Optimization of In-Cycle Closed-Loop Combustion Control with Multiple Injections

Sammanfattning: With the increasing demand of transportation, biofuels play a fundamental role in the transition to sustainable powertrains. For the increased uncertainty of biofuel combustion properties, advanced combustion control systems have the potential to operate the engine with high flexibility while maintaining a high efficiency and robustness. For that purpose, this thesis investigates the analysis, design, implementation, and application of closed-loop Diesel combustion control algorithms. By fast in-cylinder pressure measurements, the combustion evolution can be monitored to adjust a multi-pulse fuel injection within the same cycle. This is referred to as in-cycle closed-loop combustion control.The design of the controller is based on the experimental characterization of the combustion dynamics by the heat release analysis, improved by the proposed cylinder volume deviation model. The pilot combustion, its robustness and dynamics, and its effects on the main injection were analyzed. The pilot burnt mass significantly affects the main combustion timing and heat release shape, which determines the engine efficiency and emissions. By the feedback of a pilot mass virtual sensor, these variations can be compensated by the closed-loop feedback control of the main injection. Predictive models are introduced to overcome the limitations imposed by the intrinsic delay between the control action (fuel injection) and output measurements (pressure increase). High prediction accuracy is possible by the on-line model adaptation, where a reduced multi-cylinder method is proposed to reduce their complexity. The predictive control strategy permits to reduce the stochastic cyclic variations of the controlled combustion metrics. In-cycle controllability of the combustion requires simultaneous observability of the pilot combustion and control authority of the main injection. The imposition of this restriction may decrease the indicated efficiency and increase the operational constraints violation compared to open-loop operation. This is especially significant for pilot misfire. For in-cycle detection of pilot misfire, stochastic and deterministic methods were investigated. The on-line pilot misfire diagnosis was feedback for its compensation by a second pilot injection. High flexibility on the combustion control strategy was achieved by a modular design of the controller. A finite-state machine was investigated for the synchronization of the feedback signals (measurements and model-based predictions), active controller and output action. The experimental results showed an increased tracking error performance and shorter transients, regardless of operating conditions and fuel used.To increase the indicated efficiency, direct and indirect optimization methods for the combustion control were investigated. An in-cycle controller to reach the maximum indicated efficiency increased it by +0.42%unit. The indirect method took advantage of the reduced cyclic variations to optimize the indicated efficiency under constraints on hardware and emission limits. By including the probability and in-cycle compensation of pilot misfire, the optimization of the set-point reference of CA50 increased the indicated efficiency by +0.6unit at mid loads, compared to open-loop operation.Tools to evaluate the total cost of the system were provided by the quantification of the hardware requirements for each of the controller modules.