Advanced Building Simulation

Uncertainty in building simulation 47
happened to be available from a separate wind tunnel study that was dedicated to this
particular office building.
It can be seen from the figure that all median values of the combined expert are (in
absolute value) higher than the measured values. This indicates a bias, that is the
experts tend to overestimate the wind pressure coefficients in absolute value.
Furthermore, the figure shows that the combined expert’s central 90% confidence
intervals are exceeded by 1 out of the 12 measured values. Clearly, the experts are
well calibrated in this respect.
When both aspects of calibration are combined in one score according to the
method of Cooke (1991), it can be concluded that the combined expert is overall
fairly calibrated and the results of the expert judgment study are suitable measures of
the uncertainty in wind pressure coefficients, which are assessed on the basis of
generic wind engineering knowledge and data.
Question 2. Figure 2.8 shows that overall, the uncertainty assessments from the
expert judgment study are somewhat larger than the uncertainty estimates used in the
crude uncertainty analysis, especially for the wind angles where the wind approaches
over built-up terrain (angles 0–90� and 270–360�). This corroborates the assumption
that some sources of uncertainty were omitted in the initial estimates.
The impact of this enlarged uncertainty in the wind pressure coefficients on the
building performance is deferred to Section 2.4.4.
2.4.3 Uncertainty in indoor air temperature distribution
In most current simulation tools, the air volume in a building space is typically
lumped into one single node, to which a single temperature, that is, the mean air temperature
is assigned. Under the assumption that the air temperature is uniform, this
air node temperature can be used in the calculation of the ventilation heat flows and
the heat flows from the air to the room enclosure on the basis of (semi-) empirical
models for the convective heat transfer coefficients. Moreover, the uniform temperature
assumption is adopted in the assessment of the average thermal sensation of an
occupant in the room.
However, the temperature distribution in the room air will generally not be
uniform. Indeed, in naturally ventilated buildings, which are considered in this
study, there is limited control over either ventilation rates or convective internal heat
loads. This results in flow regimes varying from predominantly forced convection to
fully buoyancy-driven flow. In the case of buoyancy-driven flow, plumes from
both heat sources and warm walls rise in the relatively cool ambient air, entraining
air from their environment in the process, and create a stratified temperature profile.
Cold plumes from heat sinks and cool walls may contribute to this stratification.
Forced convection flow elements, like jets, may either enhance the stratification
effect or reduce it, dependent on their location, direction, temperature, and momentum
flow.
In theory, the flow field in a space is fully determined by the Navier–Stokes equations
plus the equation for energy conservation with their boundary and initial conditions.
When these equations for the flow are solved simultaneously with the other
equations in the building simulation model, the two sets of equations supply each
other’s boundary conditions, and the temperature field is dynamically calculated.