Higher Efficiency with Model-based Predictive Knock Control

The article delves into the development of a predictive knock control system for gasoline engines, which addresses the stochastic nature of knocking phenomena. Traditional systems react reactively to knocking events, leading to inefficient combustion. The new approach, developed through the FVV project, utilizes a model that predicts knock probabilities, allowing for optimized ignition timing and increased efficiency. The model was validated through simulations and real-time tests on a single-cylinder research engine, demonstrating a substantial reduction in CO2 emissions and exhaust gas temperature. The study also outlines plans for further development, including the use of AI-based reinforcement learning for even more advanced control strategies.