A data-driven approach for jet fire prediction of hydrogen blended natural gas pipelines

Hydrogen-blended natural gas (HBNG) pipelines have the potential to achieve safe, efficient large-scale long-distance transportation of hydrogen to end-users. However, blending hydrogen can significantly affect the combustion characteristics of natural gas. In particular, a jet fire is more likely induced by the leakage of HBNG pipelines since hydrogen has a lower minimum ignition energy and is more prone to spontaneous combustion. Furthermore, the characteristic space and time dimension of HBNG jet fire can dramatically change. Nevertheless, little attention has been paid to the prediction of HBNG jet fire characters (e.g., in terms of length, temperature, and heat radiation) mainly due to the lack of experiment data. Besides, a large number of repeated HBNG jet fire experiments is more difficult since HBNG jet fire experiments are expensive and dangerous. To fill the research gap, this study proposes a data-driven methodology to quickly predict the characteristics of HBNG jet fire. First, a computational fluid dynamics (CFD) model is established and validated to obtain enough data for the prediction of jet fire. Then, a machine-learning model is developed based on jet-fire simulation data to predict HBNG jet fire characteristics. A comparison between the machine-learning model and an empirical formula is conducted to validate the effectiveness of the model. The result shows that the developed model improves both the prediction accuracy and enlarges the application range. Therefore, this prediction model can be easily used for consequence analysis of HBNG jet fire and support risk assessment of HBNG natural gas pipelines.