Big data analysis on flow characteristics according to welded penetration locations for fire sprinkler piping system design

Abstract

This study presents a big data-driven computational fluid dynamics (CFD) analysis of Tee‐type fire sprinkler pipelines, focusing on how different weld penetration depths affect flow behavior and overall system performance. Over 2000 simulation cases were generated by varying key parameters, including inlet velocity, base and branch diameters, and penetration depths. The results show that deeper penetration can create pronounced recirculation zones and significant local pressure drops, especially in smaller‐diameter main pipes. These undesirable flow disturbances may undermine sprinkler efficiency by causing uneven velocity distribution or increasing cavitation risks. A regression model based on both a power‐law empirical approach and a Deep Neural Network was developed to predict average velocities at multiple monitoring points. The DNN model, supplemented by reinforcement learning, achieved a high accuracy within ±10% error, surpassing simpler regression techniques. This integrated approach highlights how big data simulations and machine learning can guide penetration depth selection, thus improving fire suppression reliability and reducing long‐term corrosion risks in sprinkler systems.