Background: One factory makes lime (CaO) by burning limestones (CaO3)
with natural gas in kilns. The stack gas from the kiln goes through a scrubber where it is
washed by recycled water to reduce the concentration of waste gas emission including CO2,
O2, N2 and others.
Target
Reduce the dust amount in stack gas (the
emission) to under L gr/DSCF (grains/dry standard cubic foot, tons/day)
Feature selection
The target is related to two types of factors (observations):
Type 1 kiln operation factors including natural gas flow,
CaO ton/day, kiln outlet temperature
Type 2 scrubber operation factors including scrubber diff
in. H2O, scrubber recycle, makeup H2O
Primary Features:
- total reduced sulfur in PPM
- Scrubber diff in. H2O scrubber differential pressure in
inches of water, which is the pressure difference between the inlet and outlet of the
scrubber. The higher this difference, the more dust is removed.
- recycle - leaching (washing) rate of scrubber
- Feed of natural gas
- Flame temperature - it was used and we found it indeed has an
inverse correlation with dust. But we think that it may be an uncontrollable parameter
and is directly related to the "outlet temperature", so we omitted it.
- Stack gas = CO2 + H2O + N2
Derived features:
- Scrubber water amount/stack gas flow rate.
When this is large, dust is less than 0.06
When this is small, dust is above 0.06
When this is small, dust is less
Correlation of Dust with Individual Factors
Table 3-1 Correlation coefficients between Dust(t) and each
factor
Factor |
Coefficient |
Notes |
Stack gas flow |
0.488 |
Large flow blow
out more dust |
CaO yield |
0.618 |
High yield
requests large gas (dust) flow |
Recycle/(stack
gas) |
-0.598 |
Large recycle
leads to less dust |
Kiln outlet temp |
-0.365 |
High temp gives
less dust |
Result-1
Kiln operation vs. scrubber operation
The graph in Fig. 3-1 is the result obtained by the KNN (k-nearest neighborhood)
clustering method. The x-axis and y-axis of the graph represents the optimization of dust
control by kiln operations (kiln outlet temperature) and by scrubber operations
(recycle/stack gas flow), respectively. This graph describes the relationship between kiln
outlet temperature and the ratio of scrubber recycle to stack gas flow.
Result- 2. CaO yield vs. Recycle/stack gas
Figure 3-2 (above) was obtained from a factor analysis on the
given data. It shows the optimal zone (where dust is under 0.06) and the bad zone (where
dust is above 0.06). It verifies the fact that less dust is produced for less CaO
production. However, to maintain high CaO production while reducing dust, we have to
increase the factor recycle/(stack gas flow).
Table 3-2 Correlation coefficients between Dust(t) and each
factor
Factor |
Coefficient |
Notes |
[Stack gas
flow]/ngas |
0.402 |
Large flow blow
out more dust |
Scrubber
in. H2O |
-0.405 |
High yield
requests large gas (dust) flow |
Recycle/(stack
gas) |
-0.640 |
Large recycle
leads to less dust |
Moisture |
0.476 |
|
Table 4-1 Optimal
conditions for dust control in CaO Production
If Factor1 is |
Then Factor2 should be |
O2 low |
Recycle high |
O2 low |
Stack gas flow low |
O2 low |
(Stack gas flow/natural gas) low |
O2 low |
(diff in. H2O/stack gas flow) high |
Kiln outlet temp high |
Recycle high |
Kiln outlet temp high |
Stack gas flow low |
Kiln outlet temp high |
(stack gas/natural gas) low |
Kiln outlet temp high |
(diff in. H2O/stack gas) low |
Kiln outlet temp high |
(Recycle/stack gas) high |
(Stack gas/natural gas) low |
(recyle/stack gas) high |
Conclusions
- Our data analyses have given reasonably good results to indicate
the regularities/controllability of CaO production. Our methods are feasible and could
give very good results when more data are available.
- Besides data analysis on dust, our study also shows that the
regularity of TRS (total reduced sulfur) is rather obvious. But the SO2
data is too few that we could not find any reliable result. Does the Client want a
mathematical model for TRS? We also found "TRS' relationship with other features.
|