Possible observations are selected to be
the torques and pressures, 
, and
the angular velocities and flows,
. Other measurement
points are derivatives of these.
From the temporal
causal graph (Fig. 9) it can be seen that n and 
have
similar temporal effects, i.e., they are not separated by
integrating effects. To distinguish between failure of
these two parameters the
difference in angular velocity of the driving motor and the
rotating pump, 
, is added to the set of possible measurements.
Finally,
observations that are presently available in the secondary cooling
system, indicated by the
boxed variables in Fig. 7, show the pressure
generated by the centrifugal pump, 
, combined with the intake pressure,
, as an additional
observation, 
. Adding this measurement to the set of possible
observations results in a completely diagnosable system
for all possible single faults (Table 2).
When predictions are considered to the 
order sets of 5 observations
can be used. If predictions are considered till the 
order
7 observations are required. Note that 
is never used.
Table 2: Minimum sets of measurements for completely diagnosable single
faults.
