Scoping study on the significance of mesh resolution vs. scenario uncertainty in CFD modelling of residential smoke control systems

Computational fluid dynamics (CFD) modelling is a commonly applied tool adopted to support the specification and design of common corridor ventilation systems in UK residential buildings. Inputs for the CFD modelling of common corridor ventilation systems are typically premised on a 'reasonable worst case', i.e. no specific uncertainty quantification process is undertaken to evaluate the safety level. As such, where the performance of a specific design sits on a probability spectrum is not defined. Furthermore, mesh cell sizes adopted are typically c. 100-200 mm. For a large eddy simulation (LES) based CFD code, this is considered coarse for this application and creates a further uncertainty in respect of capturing key behaviours in the CFD model. Both co-existing practices summarised above create uncertainty, either due to parameter choice or the (computational fire and smoke) model. What is not clear is the relative importance of these uncertainties. This paper summarises a scoping study that subjects the noted common corridor CFD application to a probabilistic risk assessment (PRA), using the MaxEnt method. The uncertainty associated with the performance of a reference design is considered at different grid scales (achieving different 'a posteriori' mesh quality indicators), with the aim of quantifying the relative importance of uncertainties associated with inputs and scenarios, vs. the fidelity of the CFD model. For the specific case considered herein, it is found that parameter uncertainty has a more significant impact on the confidence of a given design solution relative to that arising from grid resolution, for grid sizes of 100 mm or less. Above this grid resolution, it was found that uncertainty associated with the model dictates. Given the specific ventilation arrangement modelled in this work care should be undertaken in generalising such conclusions.