This system of 27 equations when combined with the system of Equation 2.9
can be solved in various ways to provide the six main airflow rates and
potentially ten parasitic flow rates. This global system of equations contains
more equations than unknowns. There are several ways to address this
situation, and we have found that some methods are better than others for
application to air handling units. Therefore we present the tested methods
below.
Least square solution
The system of equations from 2.9 to 2.19 is over-determined: there are 26
equations for calculating 16 airflow rates. In zones where
~
II
i
6¼ 0, the system
could be solved by least square fit:
~
QQ
i
¼½C
T
C
T
i
1
C
T
i
~
II
i
ð2:20Þ
where C
T
is C transposed. The resulting flow vector is the one that best satisfies
the set of equations. However, the injection rate vector
~
II
0
, back-calculated
using Equation 2.7 with the resulting flow vector
~
QQ and the measured con-
centration will not be equal to the actual one. This method always provides a
solution, but, depending on the condition of the system of equations, this
solution could be far from the reality.
At nodes where the tracer i is not injected, the system can only provide
linear combinations of airflow rates, as far as the determinant jC
i
j¼0.
Eliminating some equations
Combining some of the equations of system 2.9 to 2.19 allows avoidance of the
measurement of some concentrations. A system having as many equations as
unknown airflow rates can be solved using:
~
QQ ¼ C
1
~
II ð2:21Þ
Experience showed that this interpretation method often leads to poorly condi-
tioned systems of equations. Results are then very sensitive to slight changes of
input data.
Looking for the best conditioned system
A set of N equations (N being the number of unknown airflow rates, in this
case 16) can be selected to give the best accuracy, or the smallest sensitivity
to variations or errors of injection rates and concentrations. This set can be
theoretically selected by calculating the con dition number (see Chapter 3,
‘Condition of the model matrix’) of all possible sets of equations extracted
from the full system, and taking the set with the smallest condition number.
This selection could be tedious: there are 13,037,895 sets of 16 equations that
can be extracted from the system 2.9!
Airflow Rates in Air Handling Units 27
Kommentare zu diesen Handbüchern