Design Guidelines for High Performance Industrial Furnaces - High Temperature Air Combustion: From Energy Conservation to Pollution Reduction - page 289

400
380
360
340
320
300
280
260
240
220
200 1

101 201 301 401 501 601 701 801 901 1001 1101 1201

Time (x0.2s)

FIGURE 5.52 COG flow rate correction signal (at 100% combustion).

The following problems are anticipated with online processing of in-furnace
pressure signals:

1. Correction of pressure drop at changeover

2. Control of peak pressure at ignition

3. Correction of burner-specific error

As for item 1, the immediately preceding value must be held, as it cannot be
controlled with the flow rate. Item 2 above is also difficult to control as in the case
of item 1. Since there is an obvious relationship between the peak level and con-
vergence value, the controllability may be improved by reflecting the value from
the relationship to the correction value. For item 3 above, some errors with a plurality
of burners used can be ignored and others in the case of a pair of burners can be
corrected by varying the combustion amount.
Accordingly, this type of problem can be excluded from this discussion. The
waveform of the peak at ignition is of a typical lag-process impulse response.
However, the impulse gain data cannot be determined using an online method and
peak timing is also difficult to set. Correction using a modeling method is not
practical. For the reason mentioned above, a method to forcibly attenuate the signal
at the peak was examined.
The continuity of response cannot be guaranteed when the signal is attenuated
as the response component is also attenuated. The attenuation does not occur in the
same way as with the hold processing method. Changing the gain from 0 to 1 in
proportion to the convergence of the peak component will result in the response
component being increased as the peak component weakens. Compared with a case
where hold processing only is applied, the controllability will increase on condition
that the convergence time is short.
In the flow-down block at the time of cutoff, the immediately preceding value
is held as the response cannot be predicted, and correction is executed from the
moment of cutoff 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 )= attenuation model (monotonic increasing function from 0 to 1)