Academic Awards 2025 booklet

41 Self-Learning Fuel Path Control for Heavy-Duty RCCI Engines Reactivity-Controlled Compression Ignition (RCCI) engines combine diesel- and gasoline-like fuels to achieve high efficiency and low emissions in heavy-duty vehicles, where electrification remains challenging. However, their sensitivity to external conditions, such as fuel quality or fuel type, complicates calibration, with increasing development time and costs. Due to stricter emission laws and the goal of reducing CO 2 emissions by 90% by 2050, I developed a self-learning control algorithm to optimize the efficiency of RCCI engines, minimize calibration effort, and ensure safe operation. The algorithm uses Extremum Seeking (ES), a model-free optimization algorithm that adjusts fuel injection to shape the in-cylinder pressure toward an ideal thermodynamic cycle, thereby maximizing Gross Indicated Efficiency (GIE). Via the Principal Component Decomposition (PCD) of the measured in-cylinder pressure trace, we guaranteed that safety-critical limits would not be exceeded during calibration. Simulations show that the algorithm corrects fuel misalignments in approximately 100 seconds while achieving 43.1% GIE, while meeting safety limits. The algorithm effectively handles air path disturbances, offering a practical solution for accelerating the deployment of low-emission engines. Future work includes real-world testing and air path integration to meet the environmental goals. − 42 − 40 − 38 − 36 − 34 SOI [CADaTDC] 10 11 12 m DI [mg/inj] 49 57 . 5 66 m PFI [mg/inj] m PFI m DI 35 40 45 50 GIE [%] 3 3 . 5 4 4 . 5 IMEP g [bar] 80 90 100 p max ( θ ) [bar] 2 . 59 3 3 . 41 dp dθ max [ bar CAD ] p max dp dθ max 0 20 40 60 80 100 120 140 160 180 200 0 2 4 6 Time [sec] cov(IMEP g ) [%]