Field modelling is increasingly becoming the main form of fire modelling for design purposes. To reduce the computational running time of field models designers are sacrificing fine grid resolution without considering the consequences this could incur on the results. This report aims to provide some validation on the extent to which grid size can be increased in the field model Fire Dynamics Simulator (FDS) before the results are compromised and to determine whether there is a point at which zone model predictions become more reliable. FDS model predictions using a range of grid sizes were compared against two separate sets of experiments:

1. The University of Canterbury McLeans Island tests. These tests were performed using two isorooms, each measuring 2.4 x 3.6 x 2.4 metres. 55kW and 110kW tests were simulated in FDS.
2. The US Navy Hanger tests in Hawaii. The hanger measured 98 x 74 metres x 15 metres high at its apex. Two tests were simulated in FDS. These had Heat Release Rates (HRRs) of 5580kW and 6670kW respectively. The second test had a draft curtain situated centrally around the fire. This was modelled in two different FDS constructions; one simulated the entire hanger, the other only the area of the draft curtain. Simulations using the zone model CFAST were also performed for all the tests outlined above. The comparisons with the McLeans Island Tests showed that FDS models with grids of 150mm (H/16) made temperature predictions as accurate as 100mm (H/24) grid models, generally falling well within +/-15% of the experimental temperature measurements. The 300mm (H/8) grid models made much poorer predictions and it was shown that the zone model, although vastly limited in the data it provided was more reliable. The simulations of the Hawaii hanger tests gave very unreliable temperature predictions in the fire plume; the model with 600mm (H/25) grids over-predicted the temperatures by about 150°C. Over-predictions of as much as 60°C were also observed in the temperature predictions within the confines of the draft curtain. These large discrepancies were due to the poor modelling of the high degree of turbulence that occurred in these areas. Locations away from the fire plume or outside the draft curtain gave much better predictions because turbulence was less. In these regions grid sizes of up to 1800mm (H/8) still gave similar accuracy to the 600mm (H/24) grid models. The model using 3600mm (H/4) grids began to display some inaccuracy in the temperature gradients it predicted. The zone models made much better predictions for the temperatures within the draft curtain. This was due to the relatively steady state nature and uniform temperatures that existed there. It was difficult to compare the zone models to the main hanger space because of the limited experimental data that was provided. Generally, the comparisons showed that the extent to which gird size could be increased not only depended on fire diameter but also on the size and geometry of the enclosure. FDS models were reliable for far field temperature predictions when grid sizes of up to half the fire diameter were used. However, for near field predictions the models could not be relied upon unless very fine grid resolution could be prescribed. The study also showed that in certain cases zone models are a better option than FDS models, especially in turbulent well-mixed scenarios where a steady state period is observed and FDS grids size is limited by computational time.