Verification and calibration of ventilation network models

There is no doubt that simulated ventilation networks must be adequately verified before they can be extrapolated and used for predictive purposes. There are two general approaches as to how the verification can be established and these are summarised as follows; The first approach is to measure all individual branches with trailing hoses or barometers to get individual pressure drop and flow per branch and then calculate individual resistance factors for input into the software. The simulator is run, checking that the output makes general sense (check some sensitivities if possible). The network is now considered as verified. The pressure drop to be measured is relatively small and thus it is inherent that the measured set of individual K factors will be imprecise no matter how much care is taken in the measurements and the resultant simulations will inherit the same problem. In the second approach, the network is assembled using estimated (Atkinson) K factors (based on physical features of airway, past experience and catalogued typical K factors). Only the main primary flow rates are measured and hence the relative distribution of the primary ventilation in the main airways is determined (probably about 10 off primary air flow measurements). Main critical pressure differentials at main and booster fans and across critical doors are measured (probably about 6 off primary pressure differential measurements need to be made). The simulator is run and the computer model is used to iterate and select an updated suite of K factors so that the flow balances and pressure differentials are correctly simulated (typically only 2 or 3 iterations are all that is needed). The network is now considered as verified. The pressure differences to be measured will be substantial in absolute terms and can be measured with high precision. Also, because only primary flows are measured, these will generally be measured at higher air speed points and hence improved relative accuracy. This leads to more accurate simulation results. This paper examines the advantages and disadvantages of the two approaches. 1 Background