&EPA
United States
Environmental Protection
Agency
Industrial Environmental Research
Laboratory
Cincinnati OH 45268
Research and Development
EPA-600/S2-80-211 Feb. 1981
Project Summary
Process Spill Monitoring,
Control and Recovery in the
Pulp and Paper Industry
George W. Gove, James J. McKeown, and Albert J. Carlson
In order to develop strategies to
control intermittent spills associated
with the production of chemical wood
pulp, examinations of process efflu-
ents from kraft pulp mills and investi-
gations of existing loss control
systems were conducted. Dynamic
computer modelling, using data from
process effluent monitoring, was
employed to illustrate the utility of this
technique to arrive at various loss
control strategies for particular
process configurations. Examples
were presented, using the monitoring
data, of the economic benefit possible
from recovery of chemicals and
organic solids. A loss control strategy
for pulping, pulp washing, and
chemical recovery areas was
implemented in the existing spill
control system of a large kraft pulp
mill. Control was successfully
effected utilizing a digital computer.
In addition to managing process
losses, the direct digital control sys-
tem allowed gathering, processing
and managing data obtained from the
sensors monitoring the system.
This Project Summary was devel-
oped by EPA's Industrial Environ-
mental Research Laboratory.
Cincinnati, OH. to announce key find-
ings of the research project that is fully
documented in a separate report of the
same title (see Project Report ordering
information at back).
Introduction
This report outlines the examination
and characterization of process efflu-
ents and explains a loss control system
for a large kraft pulp mill which utilizes a
digital computer.
Effluent Characterization
Samples were analyzed from process
and liquor storage areas and sewers in
20 pulp mills - representing various
locations, ages, process types, and
wood species. Because compositions
varied widely, predictive correlation
equations could not be derived. The
absolute values of the parameters
varied relatively little within each class
of fluid, however, so reasonably consis-
tent ratios could be established to
describe different parameter pairs.
Table 1 gives representative ratios for
weak black liquor (WBL) samples from
the mills tested. As an example, the
data for 26 different liquor samples
showed that biochemical oxygen
demand (BOD) was approximately 23%
of the total dissolved solids (TS) content,
with a standard deviation of 4%.
Selection of variables for monitoring
was also influenced by the linearity of
the measured response with concentra-
tion. Figure 1 shows how total organic
carbon (TOC) changes in proportion to
weak black liquor concentration and
Figure 2 indicates how conductivity
varies with weak wash concentration.
The data were obtained by adding con-
-------
Table 1. Characteristic Parameter Ratios for Weak Black Liquor Samples.
Patio
BOD/TS
TOC/TS
BOD/ TOO
Na/ Conductivity
PCU/ Conductivity
TS/ Conductivity
Average
0.23
0.39
0.59
0.71
4.84
2.90
Standard
Deviation
0.04
0.03
0.08
0.34
1.92
0.84
No. of
Samples
26
28
22
24
34
30
700
600
\
.500-
H
400-
^300-
200-
700
TOC
Mill Effluent 129
Weak Black Liquor 55,400
TOC= 134 + 511X
r2 = 0.999
0.2 0.4 0.6 0.8 1.0
WBL - % v/v
Figure 1. Linearity of TOC response
weak black liquor added
to mill effluent.
centrated samples incrementally to
normal mill effluent.
These findings suggest that pH and
other specific ions, TOC, color, and
conductivity might be appropriate as
indicators of effluent conditions. pH
was eliminated as lacking the required
sensitivity. Specific ions, TOC, and color
measurements were considered too
expensive for large monitoring net-
works, in part owing to requirements for
sophisticated sample systems. Conduc-
tivity was found to be a suitable linear
predictor of parameters such as sodium
and TOC; this variable can also serve as
a direct indicator of spills because its
value increases with dissolved inor-
ganic solids concentration. Moreover,
conductivity instrumentation is widely
used, relatively simple, reliable, and
reasonably inexpensive.
Control Equations
, Developing the control equations for
the spill recovery system required that
u
I
I
60
50
40
30
201
70
Lime Mud Weak Wash
added to
' Mill Effluent
Cond- 1,08 X + 1.33
r2 = 0.998
0
Figure 2.
10 20 30
Weak Wash -
40
v/v
50
Linearity of conductivity
response lime mud weak
wash added to mill
effluent.
conductivity be related to mass flows of
solids, sodium, and organic carbon from
the various processes. This was
accomplished using data from process
sewers in pulping, pulp washing, liquor
recovery, and causticizing areas at two
mills. Samples were drawn using auto-
mated equipment at 1 to 12 times per
hour depending on process conditions.
Analyses were made in a laboratory for
accuracy.
Linear equations of the form y-mx + b
were sought, relating conductivity (x) to
other parameters (y). Many samples
were analyzed and data points like those
in Figure 3 were used with regression
techniques to evaluate slope (m) and
intercept (b). Table 2 shows values
obtained at the monitoring stations in
one of the mills sampled.
System Design
The loss control system gathers and
analyzes sensor data to detect and log
4000
3000
2000
;ooo
TOC = 482(cond) - 309
n = 50 r2 = 0.950
01234567
Conductivity - mmho/cm
Figure 3. Linearity of conductivit
related to TOC.
problems, responds to appropriate siti
ations by queuing remote devices an
notifying the operator, and provide
reports and summaries of occurrences
These functions are implemented usin
a system based on a central compute
with a keyboard printer for operate
action and a process input/output (I/C
interface. Sensors include conductivit
flow, and level probes; actuators incluc
pumps, valves, and samplers. In tr
system specified, the interface may t
6000 feet from the central processir
unit and the remote devices can t
1500 feet from the interface.
Operation
The system was installed in the larg
kraft pulp mill in Figure 4 and include
the monitored stations listed in Table
Sumps receive overflows from varioi
storage tanks as well as leaks fro
transfer pumps and heaters. Each sun
has a conductivity probe, a flow meter,
sampler, and a self-priming pump.
Effluent normally flows from tr
batch digesters, brown stock washer
evaporators, and recovery furnaces '
mill sewers, where flow and conduct!
ity are measured, and then to biologic
treatment. Spills are picked up by tr
sump pumps and transferred to
200,000-gallon tank. The fluid m«
then be bled to the sewer or pump(
back to the 20,000-gallon overflowtai
and into the vats on the first stages
the washers.
Ranges of parameters and values
ratios obtained from the prelimina
analyses were used to determii
conductivity setpoint. Mass flow se
points for Na, TOC, and TS are calc
lated using flow measurements in tl
equations relating these parameters
-------
Table 2. Parameters Derived for Use in the Linearized Control Equation (y = mx + b)
Slope
No. of (m)
Station Variable Observations fjmho/cm
1 Na 97 0.249
TOC 50 0.482
TS 59 1.655
2 Na 96 0.308
TOC 68 0.599
TS 33 1.743
3 Na 95 0.261
TOC 55 0.462
TS 40 0.565
4 Na 60 0.258
TOC 40 0.492
TS 42 2.120
SN 7
SN 4 • C
/^~~~\ Spill I
\200K gal\Tank ' > 7 ~ f
\ /S/V 70 I EvapArea |
\s. iX^ r\' f/f\,Af\ /""""""N. A - * ^
& i _[rt^ O"
•*± .0- S/V 3 5/V2-T-1
,/U 77 <7 fN "^ 6/V J >->
A f *»i
T \ C^) <
j Recovery ^ lg
1 V/tf"Jace ( ) l^± ) \
' ' 1 i
Overflow* — s.
, v Tan* ^f 20K\
>\9^J
-T-. c«; jr 1 \ ) 1^(*^\
T SN 5 V_V L J
Dump A O'flow
Tank /— v ' f~\ /Recovery
I \--lSBL \ \ c,,rr>ar-a
V 7 V ; \fu»w<.« Loss control
•iguro 4.Schematic of Brown Co., loss control strategy s
Observed Observed
Conductivity Variable Multiple
Intercept Range Range Correlation
(b) /jmhos/cm mg/l Coefficient
-95 702-11730 134-2730 0.951
-309 1 140 - 81 10 144 - 4090 0.950
-884 900 - 18800 440 - 34520 0.943
-87 165 - 9560 18 - 3046 0.995
-101 165-9560 66-5730 0.985
-550 730 - 7160 1860 - 19790 0.977
-81 1166-15300 142-4772 0.974
-363 67-11700 94-5415 0.979
+863 460 - 5800 530 - 6230 0.458
•- 39 238 - 3351 43 - 822 0.988
+ 91 338-9630 249-4950 0.985
-845 350-8200 720-21400 0.928
conductivity. These mass targets rather
— +-To WTP than conductivity are used to initiate
action, so that control is exercised on
Overflow the basis of material being lost, the
~^-Z7\^~^ capacity of the waste treatment plant,
~~t f ) ( ) compatability of new spills with the
* \ J \ ) present contents of the tank, and
A |_ i7i0i ctnrano minimum total solids to be economical-
1_ WBL Oll»oj/c7 ... i.
(j- - „.. j „., j , ly recovered. As an example, high
• L — -. ( J small amount of material is passing
^g"N ' /^T\ ' ^X^ through the system, neither of which
Jj 1 In one instance, a conductivity
1 /^~>\C:i setpoint of 2 mmho/cm at 200 gpm
| \_J corresponded to 484 Ib/day Na, 786
! ^^ Ib/day TOC, and 2700 Ib/day TS. If the
1 V.J SttlpuuUs fot llieac CunSlilumUs weie
O established for reasons of economy or
environmental protection at 900 Ib/day
Na, 4000 Ib/day TOC, and 2900 Ib/day
System TS materja| wou|d not be pumped to
the spill tank unless flow or conductivity
ystem. increased.
-------
Table 3, Variables Measured at the Monitoring Stations for the Loss Control
System.
Station
Device 1
Device 2
Device 3 Device 4
Digester Sump
Brown Stock Washer Sump
No. 8 Recovery Sump
Main Kraft Sewer
No. 1 1 Recovery Sump
Caustic/zing Sewer
Total Caustic Sewer
Spill Tank
Spill Tank Valves
Overflow Tank
Weak Black Liquor Tanks
Conductivity
Conductivity
Conductivity
Conductivity
Conductivity
Conductivity
Conductivity
Conductivity
Valve to
Recovery
--
...
Flow
Flow
Flow
Flow
Flow
...
Flow
Level
Valve to
Sewer
Level
Level
Pump
Pump
Pump
...
Pump
...
...
Pump
...
...
...
Sampler
Sampler
Sampler
Sampler
Sampler
Sampler
Sampler
Sampler
...
---
...
Control Strategies
The sump pumps and the spill tank
valves may be operated manually,
under analog control using conductivity
setpoints, or in a direct digital mode
with targets stored in computer
memory. Digital control offers operators
the most immediate and comprehen-
sive information and also offers advan-
tages such as rapid and easy setpoint
changes. In addition, the computer
provides the best base of data for
management information.
The real-time executive for the
computer system has the logic shown in
Figure 5. The program restarts when the
computer is powered, and may be
initialized at any time for data averag-
ing. The sensors are read in a burst
every 10 to 15 seconds so all readings
are obtained in the same relative time
frame. Each reading is converted to
appropriate units, and running averages
are created and stored. Values are
checked to verify whether the sensor is
out of service or if action is already being
taken at the particular station. If not, the
readings are compared to parameters
such as low and high limits, allowable
rate of change, a confidence period to
determine if a condition has been out of
tolerance too long, and the number of
| Start Program ]
No
Initialize
Program
Batch Read
Sensors
Check One
Sensor Reading
Reading Within
Set Points?
All Readings
Checked?
Perform Data
Analysis
Store Results
In Memory
I Real Time ,
No' Appications '
Program |
• Modules i
Free Computer for
Off Line Programs
Figure 5. Real time executive
program module.
times the system must detect £
condition out of tolerance during the
confidence interval. If readings are no
within the setpoints, control is
transferred to a real-time application:
module like that in Figure 6, comprising
routines for each station and sensoi
which activate pumps or valves am
notify the operator of situations anc
actions taken.
Enter From
Executive
*
Confidence
Interval
Exceeded?
No
Problem
Exist?
Yes
Initiate
Action
/ Print
( Inform
\ Operator t
-—
Figure 6.
On line applicatio,
module general structurt
The executive and application
modules run in the foreground mod<
The executive completes a cycle in les
than 5 ms if no action is required, and i
about 500 ms if control is transferred t
the applications modules for ever
station. Since the sensors are rea
every 10 s, time is available to run othc
programs in the background. Fc
example, the operator ma
communicate with the system throug
the printer to monitor selected sensor
change tolerance parameters c
setpoints, open or close valves, tur
pumps on or off, activate samplers, <
mark sensors or control devices in <
out of service for maintenance.
Performance
The loss control strategy has bee
monitoring, intercepting and recoverir
process spills and losses in the pulpin
pulp washing and chemical recove
areas of a large kraft pulp mill for ov<
two years. During that time all materi
-------
that has been sent to the spill tank has
been returned to the chemical recovery
system, which has resulted in the spill
tank being generally less than 20% full.
Because of this, the spill tank also
served as extra weak black liquor stor-
age, but only during times of high liquor
inventory, as during mill shutdowns or
when evaporators were out of service.
During a representative 18-day period
the system intercepted and returned
nearly 900,000 gallons containing more
than 12 tons of sodium, 15 tons of BOD
and 67 tons of dissolved solids.
George W. Gove. James J. McKeown, and Albert J. Carlson are with the North-
east Regional Center, Department of Civil Engineering, Tufts University,
Medford, MA 02155.
D. L. Wilson is the EPA Project Officer (see below).
The complete report, entitled "Process Spill Monitoring, Control and Recovery in
the Pulp and Paper Industry," (Order No. PB 81-131 971; Cost: $17.00,
subject to change) will be available only from:
National Technical Information Service
5285 Port Royal Road
Springfield. VA22161
Telephone: 703-487-4650
The EPA Project Officer can be contacted at:
Industrial Environmental Research Laboratory
U.S. Environmental Protection Agency
Cincinnati, OH 45268
-------
United States Center for Environmental Research Fees'paid"
Environmental Protection Information Environmental
Agency C.ncinnati OH 45268 Protection
Agency
EPA 335
Official Business
Penalty for Private Use $300
ST
CHICAGO IL 60604
------- |