Investigate the frequency and intensity of knock zone in hydrogen engine under different operating conditions through experimentation and machine learning method

This study focuses on determining the Knock Index (KI) to analyze the frequency and intensity of engine knocking. Additionally, it provides solutions for optimizing working conditions to limit the occurrence of knocking. The research employs the Gradient Boosting Regressor (GBR) algorithm model and data collected from experiments, using the following input parameters: engine speed, injection timing, and injection pressure. A GBR-based Knock Index prediction model has been successfully developed, achieving high accuracy with: Coefficient of Determination (R²) = 0.993, Mean Absolute Error (MAE) = 10.447, Root Mean Square Error (RMSE) = 13.506. The model has been used to investigate the influence of input parameters on the likelihood of knocking. The results indicate that the knocking phenomenon occurs with high frequency and intensity when the engine operates under the following conditions: engine speed less than 1000 rounds per minute (rpm), ignition advance angle of −15 to −25° (deg) After Top Dead Center (ATDC), and fuel injection timing between −80 and −90 deg ATDC and −100 to −150 deg ATDC. Furthermore, the research shows that the maximum KI reaches 750 (indicative of super knocking) under specific conditions: an engine speed of 600 rpm, ignition advance angle of −20 deg ATDC, low injection pressure, and a fuel injection angle of 110 deg ATDC.