Progress in knock combustion modeling of spark ignition engines

Knock combustion is one of the primary factors limiting thermal efficiency improvement in spark ignition (SI) engines. After a century of research, scholars have made significant progress in understanding knock phenomena. Numerical simulation techniques play a crucial role in engine research and development, and knock models are a vital component of engine combustion simulation. While various models with different complexities and predictive accuracies have been proposed, a comprehensive review addressing their evolution and performance remains necessary. This article endeavors to systematically evaluate the progress of knock models and offers a critical overview of up-to-date knock models, including their features, advantages, and limitations. It delves into problems in existing knock models and proposes potential solutions. Among the most widely used models, those based on the Livengood-Wu (L-W) knock integral and autoignition reaction mechanism are extensively developed and applied. They are relatively simple and can predict knock onset time and knock intensity reasonably well. The combination of complex chemical reaction kinetics analysis and large-eddy simulation may be the most promising in capturing various aspects of knock combustion, including knock location. However, this method demands high computational resources, and its prediction is greatly affected by the simulation accuracy of autoignition reaction fronts. Machine learning can assist in developing empirical knock models by learning knock combustion characteristics from detailed physics-based modeling results or experimental data. These models have poor interpretability but could be very useful in engineering applications with sufficient accuracy. While many features of knock combustion can be characterized numerically, some details such as autoignition reaction front propagation and its impact, influence of turbulence modeling, and the effect of external random factors on knock modeling, still call for future research.