Validation of 0D Knock Models through Different Knock Detection Methods 2021-36-0028

This paper applies numerical models to simulate the Onset of Knock (OK) in a commercial automotive spark ignition engine. Results from different models were compared with different knock detection methods to verify the best agreement. Despite being widely studied and reported in the literature, avoiding knock still poses a challenge to engine efficiency improvement and it is the focus of many new technologies. Knock detection methods rely on measurements of incylinder pressure or engine vibrations to determine the onset of knock, while some 0D predictive models are deduced from experimental data and correlations involving the critical parameters related to the knock phenomenon. Although 0D predictive models are easy to apply and calculate, the initial formulation and the different experimental conditions used to create each model often result in different numerical formulations. In the present work, different operational conditions of a spark ignition (SI) engine were used to determine the OK using knock detection methods and numerical simulation models. Knock detection was performed through Maximum Amplitude of Pressure Oscillations (MAPO), Integral of Modulus of Pressure Gradient (IMPG), Integral of Modulus of Pressure Oscillations (IMPO). The experimental data is used as input for knock prediction calculations through Arrhenius type correlations and knock detection results were used as a reference to evaluate knock simulations. Simplified 0D models are effective tools in engine analysis but their variety of results match differently while compared with different experimental methods of knock detection. The model of Douaud & Eyzat, widely used in engine simulations indicated knock in an instant close to the instant of maximum pressure oscillations, while other models indicated delayed knock or did not indicated knock at all, despite all knock detection indexes showing similar knock intensity for all studied cases. This work aims to provide a useful dataset to support future choices of knock models and detection methods for engine analysis.