Environmental Engineering Reference
In-Depth Information
the second method above, the controlled process can be handled as a continuous
operation with the changeover signal ignored. It is suitable for a moderate process,
as it is not influenced by local variations and can handle variations with a long
response time.
The last method also handles the controlled process as a continuous operation
similar to the second method, except that the signals are made to appear continuous
by the use of a model instead of ignoring the influence of disturbances. In regener-
ative combustion, a short batch process made up of combustion and cutoff is repeated.
If the components before and after the cutoff action can be forecast to form a
continuous process, it is not necessary to wait until the flow rate becomes stable.
This method has one advantage, i.e., a conventional control method can be applied
without modification.
Each of the three methods mentioned above is suitable for use with different types
of facilities or combustion methods. Each type of furnace should use the appropriate
optimal control method. The waveform and response of the operation terminal at
changeover and the third investigated method are examined in the following.
The flow rate in the flow-down block at the time of cutoff is held at the preceding
value immediately before cutoff because the response cannot be predicted. Correc-
tion is applied throughout the period from cutoff action to the convergence of the
step response model.
The correction equation is expressed as shown below:
f
(
t
)=
measurement signal
g
(
t
)=
correction signal
k
(
t
)=
model (first-order lag step response of gain 1)
Assuming
t
= 0, when the hold is released (cutoff valve: OPEN),
g
(0)
= hold value,
and
f
(0)
= lower limit.
g
(
t
) =
f
(
t
) + (
g
(0) -
f
(0)) (1 -
k
(
t
))
When
k
(0) = 0
is established, the right side of the equation above is
g
(0). This
means that the hold value is the start level. Assuming the convergence time of the
step response to be
T
,
k
(
T
) = 1
is established to result in
g
(
T
) =
f
(
T
). This indicates
a convergence to the measurement signal. The second term on the right side (
g
(0)
-
f
(0))
is the function to converge to 0, which is added to the measurement signal.
The error of the model is simply added eventually, converging to zero. The time
constant of the step response model is the tuning parameter. A function of the first-
order lag is usually incorporated in most controllers, and can easily be made with
a difference equation.
Figure 5.52
shows the COG flow data. The data appear to be
somewhat overresponsive because the original data are controlled by hold processing
only. The controller changed the output for no response in hold processing, as is
clearly evident from the graph. This proved that controllability improved with an
increase in the amount of information exceeding that in holding type processing
when the flow rate is at a low level, thereby improving controllability.
Search WWH ::
Custom Search