Numerical Fire Modeling for Modern High Performance Computing Infrastructure

Ventilated façade systems have been widely used in high-rise buildings due to their functions of weather protection, aesthetics, thermal insulation, and energy efficiency. However, numerous building fire incidents worldwide have highlighted the great fire hazards associated with these systems because the cavity can considerably promote upward fire propagation, posing severe fire safety risks. Extensive experimental research has been performed to investigate how the geometry of ventilated façade systems is influencing the characteristics of the thermal-driven flow within the cavity. However, fire experiments are costly and time-consuming, especially for large-scale experiments. This makes conducting environment-controlled fire experiments for comprehensive data collection challenging. To address this, numerical approaches such as FDS has been extensively used for modelling the ventilated cavity fires. But the simulation results are generally underpredicted due to the insufficient mesh resolution and computational resource. This research aims to address this issue by parallelising the FDS modelling on high-performance computing (HPC) systems and developing guidelines to instruct and improve the reliability, accuracy, and efficiency of FDS in cavity fire modelling. The improvement will focus on the mesh decomposition strategies, domain dependency, and mesh resolution dependency. While the research centred on ventilated cavity fires, the findings are expected to be beneficial to a broader range of FDS applications in fire safety engineering.