Kevin Starr, Timothy F. Murphy Reducing variability in the thick … · 2018. 5. 9. · Value Paper Authors: Kevin Starr, Timothy F. Murphy. 2 Reducing variability in the thick stock
Post on 15-Mar-2021
3 Views
Preview:
Transcript
Reducing variability in the thick stock system
Value Paper Authors: Kevin Starr, Timothy F. Murphy
2 Reducing variability in the thick stock system | ABB Value Paper
Reducing variability in the thick stock system
Abstract Increasing production and quality have always been key factors in the profitability of making paper. There have been advances in technology and equipment that have been funded based on improving both of these goals. However, as we get more technically advanced there are a few basics that can derail the most thought out process improvement plan. Those basics deal with making sure the fiber going into the headbox is stable. As simple as that sounds, we have found through our consulting business that more than 50% of the time the problem with quality and production can be tied to stability issues in the dry fiber flow. In this paper, we will review the first principles relating dry fiber to dry weight and then point out common sources of variability related to thick stock consistency and thick stock flow.IntroductionVariations in a machine can come from a variety of sources. In order to help in the troubleshooting of process variations, separation of the stock approach system and the paper machine is often useful. Typically there can be 50 to 100 control loops in the stock approach and there can be close to that many loops controlling the paper machine. Source detection is even more complicated when these two systems have different DCS systems controlling them. Often the dividing line is based on vendor rather than process. This is unfortunate because the machine does not care where a vendor starts and stops. Rather than getting bogged down with DCS system types, it is suggested that the bottle neck between these two important process can be defined as the Machine Chest Level, Total Head, Thick Stock Flow, and Thick Stock Consistency. Our experience has shown that if these are oscillating, then there will be oscillations present in the weight or moisture at the reel. Therefore, source detection has to move back in the process rather than to try and redirect those problems into control valves. Controllers can only do so much with disturbance rejection. However, if you can find the source and correct the problem, then 100% of the disturbance can be rejected.
The diagram in Figure 1 illustrates the overlap in these two parts of the machine.
Paper Machine(Profile, High speed,
MD, etc.)
Stock Approach(50 to 100 loops)
Osc
illatio
n
Ree
lO
scilla
tion
Disturbances to the machine
Machine Chest LevelMachine Chest Consistency
Thick Stock FlowHead
Figure 1: Process Area Definitions
ABB Value Paper | Reducing variability in the thick stock system 3
Thick Stock SystemThere are many combinations of stock to machine delivery mechanisms, but for the purpose of this paper, the following delivery mechanism shown in Figure 2 will be considered.
This is a very important part of the machine and can also be a source of considerable variability contributions to the weight and moisture. Evaluation of the mass balance equation for this system shows that oscillations in stock flow and consistency directly impact the weight of the machine.
(SliceWidth)(WireSpeed)
(FirstPassRetention)(UnitConversion)(reamsize)
WetEndDryWeight =
⎜⎝
⎛⎜⎝
⎛StockConsistency
100(StockFlow)
When troubleshooting oscillations in weight, we often assume that first pass retention and wire speed are constant. That is not always the case. However, for this discussion we will focus on common sources of problems related to Stock Flow and Thick Stock Consistency.
Closer evaluation of this mass balance equation shows that the fractional changes in stock, consistency, and speed directly impact dry weight.
ΔCC
ΔBDBD
=ΔQQ
ΔBDBD
=
Where:C = Thick Stock ConsistencyQ = Stock FlowBD = Bone Dry (Basis weight less moisture weight. Some call
oven dry weight)
These fractional changes have pretty dramatic interpretations. For example, consider a steady state reading of consistency 3.5% and bone dry of 40 pounds. Then if the consistency changed to 3.6%, this would result in a change in weight of 1.14 pounds. Care should be given to the type of consistency meter, installation, signal conditioning, and control of this important parameter. Unfortunately I have seen many cases where the consistency was oscillating in the 100th decimal point, but the measurement was only being displayed to the 10th decimal point. This resulted in significant weight variability that did not show up in any historical or process trending data. The following example was taken from a machine that was experiencing significant cyclic disturbances in the bone dry weight, but the source could not be seen anywhere in the process. It was not until the signal conditioning of the thick stock consistency was corrected to show additional decimal points that the problem could clearly be seen (see Figure 3).
The top plot on the left is consistency being shown with one decimal point. The plot on the right shows consistency being
Figure 3: Raw Data Trend
3.17
3.18
465
470
475
22.0
22.1
22.2
22.3
2000 2500 3000 3500 4000
Raw Data Trend
Sample Number, Ts = 5 , Total Samples = 9876
3.160
3.165
3.170
3.175
3.180
452.5
455.0
457.5
460.0
21.9
22.0
22.1
4000 4500 5000 5500 6000 6500
Raw Data Trend
Sample Number, Ts = 5 , Total Samples = 9997
Con
sist
ency
Sto
ckFl
ow
Bon
e D
ry
Con
sist
ency
Sto
ckFl
ow
Bon
e D
ry
MachineChest
StuffBox
HeadBox
Consistency
Flow
FanPump
Weight
Dilution
Stock Pump
Wirepit
Figure 2: Stock to Machine Components
4 Reducing variability in the thick stock system | ABB Value Paper
shown to 3 decimal points. Notice that with nothing more than a simple decimal point change, the source of this problem could be clearly seen with a simple process trend. Notice the strong cyclic correlation between the consistency -green (Top plot) and the bone dry – red (Bottom plot). This problem had been going on for at least 5 years.
Dry FiberThe primary goal of the stock approach system is to deliver a constant flow of fiber to the paper machine. This flow is often referred to as Dry Fiber Flow. The equation for dry fiber is as follows:
The strict multiplication of these two signals will be off by the delay between the consistency meter and the stock flow meter. However, this multiplication can provide a good estimate. The example in Figure 4 shows how dry fiber flow and bone dry weight match up very well.
In this example, the stock flow and consistency were multiplied together and then divided by 100 to get dry fiber flow. The dry fiber flow and bone dry weight were then normalized so they could be put on the same Y axis. Notice how well these two signals correlate. Differences can be attributed to retention, sensor calibration, or sensor accuracy.
Dry Weight Control OverviewThe stability of the weight at the reel is an important concern with regards to fiber saving, run ability issues, and general quality issues. Many control systems will work to stabilize the weight by adding automation to the consistency, stock flow, and the weight measurement. An overview of this configuration can be seen in Figure 5.
The control configuration shown here can be referred to as a 3 element control strategy or a Cascade control with a feedforward. The control loops are described as follows:
− Consistency Control – PID control loop − Bone Dry Control – Primary Control or Outer loop Controller − Stock Control – Secondary Control or Inner Loop Controller − Consistency Feedforward – Adjusts stock flow based on
consistency changes
Figure 4: Bone Dry and Dry Fiber Trend
3.17
3.18
465
470
475
14.8
14.9
15.0
15.1
Raw Data
Con
sist
ency
Sto
ckD
ry F
iber
Flo
w
-1.0
-0.5
-0.0
0.5
1.0
1.5
1500 2000 2500 3000 3500 4000Sample Number, Ts = 5 , Total Samples = 9876
Normalized Bone DryNormalized Dry Fiber Flow
Dry
Fib
erB
one
Dry
Wei
ght
MachineChest
StuffBox
HeadBox
Wire pit
Consistency
Flow
StockPump
FanPump
Weight
Dilution
ConsistencyControl
Feedforward Stock FlowControl
ValvePositionerΣ
WeightControl
StockControl
StockFlow
WeightProcess
WeightControl
-
-
SP2 MV2SP1MV1
Outer Loop
Inner Loop
ConsistencyControl Consistency
-
SPL
Load
MVL
Dillution
Feedforward
Stock to Weight (Process)
Consistencyto weight (Load)
Figure 5: Dry Weight Control Overview
Figure 6: Control Block Diagram
ABB Value Paper | Reducing variability in the thick stock system 5
Note – Some mills have gone away from consistency control due to issues with dilution water stabilization and sensor correlation. Also, some mills are moving away from a thick stock valve to a variable speed drive.
A classical block diagram of how this control strategy might look can be seen in Figure 6.
The goal of the feedforward is to cancel the impact of a fractional change in bond dry caused by a fractional change in consistency by making the same fractional change to the stock flow. This looks as follows:
ΔQQ
ΔCC
= -
Expansion of the earlier consistency change example to include a stock flow with an intial value of 2500gpm shows that the 0.1 negative change in consistency would result in a positive change in stock flow of 71.4 gpm. This positive change in stock flow would result in a constant dry fiber flow.
Troubleshooting TableNo matter how this area of the machine is controlled, the objective is to stabilize dry fiber flow. There are several issues that need to be considered when troubleshooting dry fiber oscillations. In order to help narrow down the search, several years of diagnostic reports associated with this area of the machine were reviewed. The table in Figure 7 shows common problem areas and the shaded areas represent the most common problem areas.
After putting this table together, it became clear that dry fiber problems can come from a wide range of places. Defining each of these areas would result in more of a book than a paper. Rather than go through each x in this table, it was decided that if the fundamental physics of what the control is suppose to be doing can be understood, then recognizing problems becomes much easier. As a result the remainder of this paper will review the concepts associated with the control fundamentals involved with maintaining a constant dry fiber flow.
Dry Fiber StabilizationThe primary means for stabilizing the dry fiber flow is with a consistency feedforward. The goal of this feedforward is to immediately change the stock flow based on changes in consistency. A simple check to determine if this feature is present on a control system is to evaluate a trend of thick
stock consistency and thick stock flow. If they are mirror images of each other, then this control feature is on. However, if the stock flow is being moved by weight to adjust the disturbances caused by consistency, then this control feature is not on and should be.
Notice in the example shown in Figure 8 how consistency compensation did not work. In this case, the feature was on, but not properly set up. The weight feedback control is what caused the weight to come back to the setpoint.
PID Values
xDelayTransient
xxSteady StateInner or OuterDeadtime
xTuningxxDeadband
xExecution Rate
xDisturbancePositioner
xPrecisionxType
xSizexxDeadband
xDecimal PointsxFilterxQuantization
xFilterxxCalibration
xxInstallationxType
PID Values
xDelayTransient
xxSteady StateInner or OuterDeadtime
xTuningxxDeadband
xExecution Rate
xDisturbancePositioner
xPrecisionxType
xSizexxDeadband
xDecimal PointsxFilterxQuantization
xFilterxxCalibration
xxInstallationxType
outerinner
WeightStockConsistency
XX
XX
X
X
X
X
X
Area
Feedforward
Cascade
Feedback
Actuator
Transmitter
Signal Conditioning
38.25
38.50
38.75
39.00
4.60
4.65
4.70
1375
1380
1385
Raw Data: 2 Ply Wet End – Base System
Bon
e D
ryC
onsi
sten
cyS
tock
Flo
w:
SP
1250 1500 1750 2000 2250 2500 2750
Sample Number, Ts = 2 , Total Samples = 2754
Bone Dry
Consistency
Stock Flow
Correction
Disturbance
Deviation from Setpoint
Figure 7: Troubleshooting Table
Figure 8: Poorly setup Consistency compensation
6 Reducing variability in the thick stock system | ABB Value Paper
In order to understand the steps needed to properly check out and configure this portion of the machine, evaluation of feedfoward principles are needed in order to understand how this feature is suppose to work. Figure 9 is a detailed feedforward diagram.
This block diagram shows that a change in consistency is reflected into the stock flow setpoint with the goal being no change in the bone dry weight. Typically, the ‘s’ or frequency domain representation of the weight to consistency and weight to stock flow models can be classified as a first order process plus delay (see Figure 10).
Where: K is the process gainTd is the delayτ is the time constant.
In this case further annotation is needed to represent the load and the process. L is the load representing the consistency to weight model and p is the process representing the stock to weight model.
Solving for the Feedforward compensator, produces the following s domain equation. This shows the gain, delay, and lead/lag components associated with feedforward compensators (see Figure 11).
Note: Vendors or system vintage will determine what three components of the feedforward are used. Many times the steady state gain and delay will be the only components considered.
In order to properly determine these feedforward parameters, two bump tests will be needed. Both are to be done with the weight control and moisture controls turned off. The first bump requires a change to the consistency either by a change in consistency control set point or dilution valve. The second bump involves a change in the stock flow setpoint. The next three sec-tions will be divided based on the three areas of the feedforward.
Steady State Gain DeterminationIf the calibration of the stock and consistency were correct, then the same fractional change in either stock or consistency would give the same fractional change in weight. However, this should not be assumed. Rather than assume the relationship to be equal, the addition of a new term, R is used.
ΔC1
C1
ΔBDBD
=Rci
ΔQ1
Q1
ΔBDBD
=Rqi
Where:R = Multiplierc = Consistencyq = stock flowBD = Bone Dryi = Ply number.
Evaluation of this factor can aid in troubleshooting problems. If these values are not one, then one of the following is likely causing the problem.
− Stock Calibration is off − Consistency calibration is off
Weight
Consistency (MV)
Stock Flow(SP)
Weight wrtConsistency MV
Model
Weight wrt Stock SPModel
e−sTdLKL
e−sTdLKp
GL
Gp
τL
=
=
Gff
GL
Gp
= −
s + 1
τps + 1
Figure 9: Feedforward Block Diagram
Steady State(Gains)
Transient(Time Constant)
Delay(Deadtime)
Gff=
KL
Kp
τLs + 1
τLs + 1e−s(TL-Tp)
Figure 11: Goals of the Feedforward Compensator
Input Processe−sTdLKG =
τ s + 1
Figure 10: S Domain First Order plus Delay Model
ABB Value Paper | Reducing variability in the thick stock system 7
− Weight or moisture sensor calibration is off − Retention aid or chemical change occurred during test − Speed changed during test − Multi-Ply stock system
In the last case, multi-ply stock, it turns out that the R factor is actually equal to the stock distribution ratio for each ply.
R has can be solved for as follows:
ΔBD/BDΔC1/C1
ΔBD
ΔC1
= C1
BDKci=
Cinitial
BDinitial
• •Rci =
ΔBD/BDΔQ1/Q1
ΔBD
ΔQ1
= Q1
BDKci=
Qinitial
BDinitial
• •Rqi =
Where:
ΔBD
ΔC1
=Kci
ΔBD
ΔQ1
=Kqi
Notice that the K term is simply the process gain identified from the bump test of weight with respect to consistency and weight with respect to stock flow. Once R has been determined for both stock and consistency, the fractional change relationship between stock and consistency becomes the following:
Rci
Rci
= ΔQ1
Q1
ΔC1
C1
-
This makes the steady state gain equal to the ratio of Rc to Rq.
Rci
Rqi
= -Gi
If the transmitters are properly calibrated, then this gain will calculate out to be one. If full correction is not needed with the feedforward, then a value of less than one could be used as the gain. This means that the feedback control will have to resolve any errors that the feedforward was not able to resolve.
Delay DeterminationThe delay component of the feedforward is designed to match the stock correction with the change in consistency.
e-s(TL-TP)=GDelay
If the delay component is not correct, then a small transient upset will be seen in the weight as a result of a consistency upset. The time delay between the consistency reading and the stock valve will depend on the volume of stock between these points and the stock flow.
nVnQ
=Delay Time
Where: − V is the volume of stock between the consistency transmitter
and the stock valve − Q is the stock flow
As a result, a fixed time delay is not valid over the entire stock flow range. Since this delay is not constant, a constant time entry will result in coordination errors over the range of stock flows used.
In order to resolve this problem, a dynamic delay time needs to be calculated on a regular basis. The same delay equation is used as before, but the Volume entry is now the only setup parameter. The system then takes this volume entry and divides by the stock flow to determine dynamically what the delay in time will be.
There are two methods for determining this volume. − Physical volume measurements − Bump Tests
Physical volume measurements involve measuring pipe diameters, lengths, and all storage components between the consistency transmitter and the stock valve. The problem with this method is that it rarely produces a valid volume estimate. This is often related to pipes not being full, cross flows in stuff boxes, and incorrect pipe measurements.
Out
Lead ( τp > τL)
Lag ( τp < τL)
In(τp = τL)
Load Actuator
Time constants:τp= Weight wrt StockτL= Weight wrt Consistency
τ
τp
L
s
s
Lead
Lag
+
+=
1
1
Figure 12: Lead Lag compensation
8 Reducing variability in the thick stock system | ABB Value Paper
The second method can be calculated from the same bump test used to determine the steady state gain. Simply determine the time delay from consistency to weight and subtract that from the time delay of stock to weight. This is the delay in time between these two points. If the system uses a volumetric entry, then take the stock flow and multiply it by the time delay. This will give a volumetric entry that can be used as the setup parameter.
Q * (TL-TP)/60=Delay Volume
Transient CompensationTransient compensation is useful when there are significant differences in the time constants of consistency to weight and stock to weight. This portion of the feedforward compensator is called a lead lag compensator and is shown in Figure 12.
Often industrial consistency compensators use only a steady state (tp = tL ) or a simple lag compensation (tp < tL ) being a filter on the load MV. If a lead configuration is used (tp >
tL ) there will need to be a feedforward filter to reduce large instantaneous lead corrections.
Dry Fiber Flow Stabilization ProjectRecently dry fiber stabilization work was completed on a two ply board machine. Figure 8 shows the problem this customer was having with Bone Dry upsets resulting from consistency issues. This mill ran a 60/40 split between the top and bottom ply. However, the dry fiber was so poorly set up that this distribution was something the operators had to manually adjust on a regular basis.
Top and Bottom ply bump test resultsIn Figure 13, the bottom four pictures show the bump tests applied to stock and consistency. The left side is the top ply and the right side is the bottom ply. The top two graphs are weight with respect to stock flow and the bottom two represent weight with respect to consistency.
Identification: First Order Model Actuator Process Model 1
4.904.955.00
23.1
23.2
23.3
23.4
23.5
1300
1325
23.00
23.25
23.50
23.75
850
875
900
23.25
23.50
23.75
24.00
4.45
4.50
4.55
23.30
23.35
23.40
23.45
23.50
23.55
23.60
23.65
0 10 20 30 40 50 60 70 80 90 100 110
100 150 200 250 300 350 400Kp = 1.210934E-02, taup = 49.00092, Tdt = 30, Points = 349, Ts = 5
Kp = 2.749114, taup = 15, Tdt = 45, Points = 130, Ts = 5
0 50 100 150 200 250 300 350 400 450 500Kp = 1.079013E-02, taup = 54.55465, Tdt = 30, Points = 509, Ts = 5
start
70 80 90 100 110 120 130 140 150 160 170 180 190Kp = 2.186698, taup = 15.8898, Tdt = 72.5, Points = 112, Ts = 5
Rq2=0.01079*867.5/23.624=0.396
Rc2=2.186*4.434/23.38=0.414 Rc1=2.81*4.95/23.21=0.61
Rq1=0.01186*1324.39/23.476=0.67
Top Ply Consistency
Top Ply Stock setpoint
start
start
start
Figure 13: Top ply and Bottom Ply Step tests
ABB Value Paper | Reducing variability in the thick stock system 9
Tuning and Setup CalculationThe bump test information was analyzed and used to determine the following tuning and setup information.
Steady State GainTop Ply: G = 0.414/0.396 = 1Bottom Ply: G = 0.61/0.67 = 0.91Note: Consistency transmitters had just been calibrated. So values of 1 should be expected.
Base DelayDelay Time = 45 – 30 = 15 secondsVolume Delay = (15 sec/60 sec/min)) *1325 gallons/min = 331 gallons
Top DelayDelay Time = 72.5 – 30 = 42.5 secondsVolume Delay = (42.5 sec/60 sec/min)) * 875 gallons/min = 234 gallons
Base TransientTime Constant Consistency = 15Time Constant Stock = 49Lead Action: 49/15 = 3.3
Top TransientTime Constant Consistency = 16Time Constant Stock = 54Lead Action: 54/16 = 3.37
Evaluation TestingThe check out involves enabling the feedforward with the weight control turned off. The dilution water was changed. This caused the consistency to change. Notice how the stock flow reflects the consistency change (se Figure 14). In this test, the feedforward gain was set to 0.9. Notice that only a small change can be seen in the bone dry. In order to show the dramatic impact this has, a bump test of stock was done with no change in consistency. Notice that the change is significant in weight.
Once the dry fiber was stabilized for both plies, the operators were able to adjust the ply loading as they needed. Figure 15 shows the significant change in fiber distribution stability once the consistency compensators were properly configured and tuned.
Statistical analysis done before and after the changes showed significant improvements in dry fiber which resulted in significant improvements in bone dry at the reel (see Figure 16).
1000
1025
1050
4.3
4.4
4.5
33.0
33.5
3500 4000 4500 5000 5500 6000 6500 7000 7500
Raw Data
Sample Number, Ts = 2 , Total Samples = 7537
Impact on weight with nofeedforward compensation
Bone Dry
Consistency
Stock Flow Setpoint
New Tuning Original Tuning
0.000
0.025
0.050
0.075
0.00
0.25
0.50
0.75
Two Sigma ComparisonPoints = 151
Bon
e D
ry
Dry
Fib
er
1 2 3 41ME1_CW11MV 0.08597301ME1_CW11MV 0.0595841
TotalDryFiber 0.9076759TotalDryFiber 0.6706053
30% improvement
26% improvement
Figure 14: Feedforward Evaluation
Figure 15: Ply loading stabilized
Figure 16: Variability Improvements
10 Reducing variability in the thick stock system | ABB Value Paper
ConclusionSteps taken to stabilize the dry fiber flow will pay off in many ways. A few of them are as follows:
− Reduced Bone Dry variability − Improved run ability − Increased Operator Confidence − Improved formation − Improved flocculation on the wire − Opportunity for fiber savings through target shifting − Improved tail strength after sheet breaks − Stable ply distribution in multi-ply systems
References
Analysis Software, ABB MD400 Loop tuning and analysis software
Timothy F. Murphy, “Quantifying Paper Machine Transition Performance”,
TAPPI 2008 Process Control, Electrical & Info. Conference Proceedings
Gerald Ostroot, The Consistency Control Book, Tappi Press, June 1997
N. Sell ed.,Process Control Fundamentals for the Pulp and Paper Industry,
Tappi Press, ISBN 0-89852-294-3.
K. Starr, Single Loop Control Methods, Skyline Publishing, ISBN 1-882811-
08-9.
K. Cutshall, “The Nature of Paper Variation”, Tappi Journal, Jun-90, pp. 81-
90.
Contact us
3BU
S09
4978
AB
B U
S C
reat
ive
Ser
vice
s 16
97ABB LtdAffolternstrasse 44 CH-8050 Zurich, Switzerlandwww.abb.com
top related