Practical process controlExamples from a chemical industry

Krister Forsman

2017-11-08

Typical tasks for the control group

• Improve productivity by decreased variation and increased automation.

– Smarter control structures, e.g. feedforwards, mid-ranging, split range, cascades, maximizing control, ratio-in-cascade

– PID control parameter tuning

– Plant-wide control issues

– Introduce new controllers

– Support in design and commissioning of new plants

• Process historian (database) ownership; applications and development in Industrial ICT

• Training seminars

K. Forsman, 2017-11-08, No. 3

Tools are centered around Matlab

• The most important tools in the optimization work is Matlab and IP21.

• We have developed our own libraries for

– Data collection from IP21

– Data analysis

– Simulation of controllers and control structures

– Identification

– Assessment of control performance

• Some examples of tools below

K. Forsman, 2017-11-08, No. 5

Process model structures supported by our Matlab library

sT

KesL

+

−

1KLT

Integrating( )sTs

eksL

v

+

−

1

IPZ( )

( )1

2

1

1

sTs

esTksL

v

+

+−

( )( )( )31

2

11

1

sTsT

esTKsL

++

+−

PPZ

( )31 sT

Ke sL

+

−

P3

~80% of all loops

~15% of all loops

K. Forsman, 2017-11-08, No. 6

Representation of control structures: PID (piping and instrumentation diagram ☺☺☺☺)

PC

v1 , q1 v2p q2

FC

Constantpressure = 1

Atmosphere

Simple example: gas pipe

FC = flow controller, PC = pressure controller, LC = level ctrl, TC=temperature ctrl

FT = flow transmitter, etc

Why not use block diagrams, as in most control textbooks?

PT FT

K. Forsman, 2017-11-08, No. 7

Block diagram for the same process

P1

P2

P3

v1q1

v2

q2

p-

C1

C2

K. Forsman, 2017-11-08, No. 8

Control optimizationA set of practical examples

K. Forsman, 2017-11-08, No. 9

Ex: Level control with improvement opportunity

• The level in the tank varies too much, because there are pressure variations in the line for the incoming flow.

• We can’t tune the controller more aggressively - then it becomes unstable.

• Can we still improve control performance?

LC

FT

K. Forsman, 2017-11-08, No. 10

Solution: Control the flow too!

= Cascade control

LC

FC

SP

PV

K. Forsman, 2017-11-08, No. 11

Cascade control block diagram

• Which disturbances motivate the use of cascade control?

C1 C2 P2 P1+

d1

r1 y1

u1 = r2 u2 y2

+

d2

+

d3

Answer: d1

K. Forsman, 2017-11-08, No. 12

CT

FT

Stock

Dilution water header

Consumers

Example: Dilution process

• Task: Control the concentration measured by the transmitter CT, by manipulating dilution water valve FV.

• How would you do that?

FV

K. Forsman, 2017-11-08, No. 13

CC

CT

FC

FT

FC.sp = CC.op

Stock

Dilution water header

Consumers

Structure for concentration control

• Concentration (consistency) control in cascade against dilution water flow.

• Scenario: There is a pressure drop in the dilution water header, because one of the other consumers suddenly increases its demand.

• We will now see how the cascade controller reacts to this, depending on the tuning of master and slave controller.

0 20 40 60 80 100 120 140 160 180

4.8

5

5.2

5.4

5.6

5.8

0 20 40 60 80 100 120 140 160 18045

46

47

48

49

50

51

52

0 20 40 60 80 100 120 140 160 180

50

52

54

56

58

60

62

y1

r1

y2

r2

u2

Master SP and PV

Slave SP and PV

Valve

Slave controllermuch fasterthan master.

Rule of thumb forcascade control

to work:Time constantin slave 10

times faster thanin master.

The disturbanceis handled by

the slave controllerbefore the mastercontroller reacts(in principle).

0 20 40 60 80 100 120 140 160 180

50

52

54

56

58

60

62

0 20 40 60 80 100 120 140 160 18045

46

47

48

49

50

51

52

0 20 40 60 80 100 120 140 160 180

4.8

5

5.2

5.4

5.6

5.8

y1

r1

y2

r2

u2

Master SP and PV

Slave SP and PV

Valve

Difference betweenslave and mastersmaller than inprevious slide.

The disturbanceis handled by

the slave controllerbut the master

controller also reacts,later on.

0 50 100 150 200 250 300 350

50

55

60

65

0 50 100 150 200 250 300 35040

45

50

55

60

65

0 50 100 150 200 250 300 3503

4

5

6

7

y1

r1

y2

r2

u2

Master SP and PV

Slave SP and PV

Valve

Master and slavecontroller almostequally fast.

The disturbanceis thrown

back and forthbetween master

and slave.

Almost unstable.

K. Forsman, 2017-11-08, No. 17

New case

K. Forsman, 2017-11-08, No. 18

Evaporator with poor level control

Problem:This level control worked poorly, because

this valve is very nonlinear.

K. Forsman, 2017-11-08, No. 19

Solution:Manipulate the flow incascade against level,using the flow meter.

More stable level and smoother flow using cascade control

LC

.PV

FI.

PV

LC

.OP

LC

.PV

FC

.PV

FC

.OP

In this case, cascade control is used to linearize the slave process,rather than for disturbance rejection.

K. Forsman, 2017-11-08, No. 21

Ex: Level control with improvement opportunity

• The level in the tank varies too much because of variations in output flow.

• We can’t tune the controller more aggressively - then it becomes unstable.

• Can we still improve control performance?

LC

FT

K. Forsman, 2017-11-08, No. 22

Feedforward: Give early information to the controller

( ) ( ) FF

t

i

c ueT

teKtu +

+= ∫

0

1

Feedback term Feedforward term

LC

FT

+

KFF

FF

Density control in a dissolver

Dissolver

Centrifuge 1 Centrifuge 2

FT20

FC19

FC21

FC22

Issue: Sometimes one of the flows FC-19 or

FC-21 is closed for cleaning the centrifuge.

Then there will be a large deviation in density

in the dissolver.

solids

solidsmother liquor mother liquor

This controller is the mostimportant one!

DC24

FF

+

water

SP

K. Forsman, 2017-11-08, No. 24

Dissolver: FF and PI tuning reduces variations

Feedforward, and some other improvements, introduced

K. Forsman, 2017-11-08, No. 25

New case

K. Forsman, 2017-11-08, No. 26

Evaporators: Ratio control issue

Raw material feed is master

Steam flow in ratio against feed

Process: Raise the concentration by boiling water away.

K. Forsman, 2017-11-08, No. 27

Traditional ratio control structure

C1 P1

r1

C2 P2

r2

α

12 yr α=

y1

y2

x

K. Forsman, 2017-11-08, No. 28

Disadvantage with this structure

Why do we gets humpsin concentration?

Feed; Setpoint

Concentration

Time [min]

K. Forsman, 2017-11-08, No. 29

Same scenario: comparison feed flow – steam

Feed; SP and PV

Steam flow; SP and PV

Time [min]

K. Forsman, 2017-11-08, No. 30

The ratio steam / feed not constant during SP-changes

What should we doin order to reduce

the deviation in ratio?

Ratio steam / feed

Time [min]

K. Forsman, 2017-11-08, No. 31

Speed up steam flow controller! Result:

Time [min]

Concentration

Feed; Setpoint

K. Forsman, 2017-11-08, No. 32

Feed; SP and PV

Steam flow; SP and PV

Time [min]

K. Forsman, 2017-11-08, No. 33

Current control structure

C1 P1

r1

C2 P2

r2

α

12 yr α=

y1

y2

x

Is there a different structure that would reduce deviations even more?

K. Forsman, 2017-11-08, No. 34

Alternative structure

12 rr α=

C1 P1

r1

C2 P2

r2

α

y1

y2

x

The SP of the slave is calculated from master SP rather than master PV!

What are the disadvantages of this scheme?

K. Forsman, 2017-11-08, No. 35

Test run: Ratio against SP

Feed; Setpoint

Concentration

Time [min]

K. Forsman, 2017-11-08, No. 36

New case

K. Forsman, 2017-11-08, No. 37

Reactor control; Residual oxygen

• Process: Two phase oxidation reactor (liquid-oxygen)

– On-line measurement of residual O2 in reactor gas; must be controlled.

– Which different control structures are feasible, and what are their respective advantages?

R-101

Liquid

Oxygen

FT101

AT101

FT102

LC

Liquid feed is master = TPM

(determines production rate)

Reactor gas

Product

K. Forsman, 2017-11-08, No. 38

Alternative 1: O2-feed in cascade against residual-O2

R-101

Liquid

Oxygen

FC102

Cascade control

FC101

AC101

K. Forsman, 2017-11-08, No. 39

Alt 2: O2 feed in ratio against liquid, cascade against residual O2

R-101

Liquid

Oxygen

FC102

FC101

AC101X

This structure is superior ifliquid feed varies, e.g. during a start-up.

K. Forsman, 2017-11-08, No. 40

New case

Crystallizer: Level disturbance from wash sequence

• A crystallizer is automatically flushed with water, once every second hour. The water flow is large enough to affect level.

• It’s important to keep a steady level.

Flushwater

Crystallizer feed

Improvement suggestions?

K. Forsman, 2017-11-08, No. 42

Crystallizer: Feedforward reduces level variations

Flushwater

Crystallizer feed

• A FF from on-off-valve to level controller reduces level variations.

• Thus, it’s ok to make a FF from a discrete (binary) variable.

+

K. Forsman, 2017-11-08, No. 43

Smaller level variations in crystallizer

Without feedforward from on-off valve With feedforward

K. Forsman, 2017-11-08, No. 44

Why don’t we do the full compensation?

There us a residual level disturbance,because the feedforward gain used wasonly 0.8 of that calculated theoretically.

Why didn’t we use the fullgain for compensating?

K. Forsman, 2017-11-08, No. 45

New case

K. Forsman, 2017-11-08, No. 46

Sewer pH-control process

pHT

NaOH

There are two valves for feeding caustic to the pit:a small, accurate one, and a larger coarse valve.

So we have one extra degree of freedom for controlling pH.How can we design a control scheme that utilizesthis freedom in a good way?

K. Forsman, 2017-11-08, No. 47

Solution: Mid-ranging (valve position control)

pHT

NaOH

pHC

VPCOP

PV

Let a pH-controller manipulate the small valve

Introduce a valve position controller (VPC) which controlsthe position of the small valve by slowly adjustingthe large valve (working through the process).

K. Forsman, 2017-11-08, No. 48

Block schedule for MR-control: Give setpoint for u1

C1

C2

r1 = SP for y

P2

+

P1

u1

u2

y

r2 = SP for u1

Fine valve

Coarse valve

VPC = valve position controller

K. Forsman, 2017-11-08, No. 49

pH control; Results

pH: Daily averages before and after new control structure

K. Forsman, 2017-11-08, No. 50

Improved pH-control gives fewer alarms

Before: 10 488 alarms in one month After: 418 alarms in one month

96% fewer alarms from this object.

K. Forsman, 2017-11-08, No. 51

New case

Exothermic reactor temperature control

TC

TT

Incomingcooling water

TT

FT

The reactor solution is circulated through a heat exchanger (cooler).The reaction is very exothermic: it is important to control the temperature.Typical variations/disturbances: Cooling water header pressure, CW temperature

HEX

Circulation loop

New control structure: Power control

TC

TT

TT

FT

DT X EC

SP

PV

Incomingcooling water

Power controller

K. Forsman, 2017-11-08, No. 54

HEX power control reduces variations between batches

K. Forsman, 2017-11-08, No. 55

New case

K. Forsman, 2017-11-08, No. 56

Mixingchest

QT

Dilution water header

Two step dilution: how to use the extra DoF?

This consistencyshould be controlled

K. Forsman, 2017-11-08, No. 57

Mixingchest

VPC

QC

Dilution water header

Two step dilution = Typical case for mid-ranging

K. Forsman, 2017-11-08, No. 58

0 10 20 30 40 50 603.7

3.8

3.9

4

4.1

4.2

4.3

4.4

4.5

4.6

4.7

Time [min]

0 10 20 30 40 50 603.7

3.8

3.9

4

4.1

4.2

4.3

4.4

4.5

4.6

4.7

Time [min]

Consi

sten

cy [

%]

Before

After

Consi

sten

cy [

%]

K. Forsman, 2017-11-08, No. 59

HD tower(~1000 – 2000 m3)

VPC

QC

Pulp from pulp mill

Low consistency

stock

Dilution water Dilution water

Another example from pulping: HD tower

K. Forsman, 2017-11-08, No. 60

New case

K. Forsman, 2017-11-08, No. 61

pHT

pH control

• The purpose of this process is to stabilize the pH in the solution coming out of the tank.

– Manipulate the caustic added to the feed line. Don’t care about the level.

– All variation comes from the pH of the feed solution

• Which pH-meter should be controlled in feedback?

• Is there a control structure better than just PID control?

NaOH

pHT

K. Forsman, 2017-11-08, No. 62

pHC

The worst solution

• The solution below will give the highest variations in pH out of the tank.

• Since the meter is far away from the MV the process has a long deadtime, so we have to tune the controller very defensively.

NaOH

pHT

K. Forsman, 2017-11-08, No. 63

pHT

Better

• This solution is much superior to the previous one. The controller can be tuned aggressively, since there is only a short delay.

• However, the final pH is not controlled here.

NaOH

pHC

K. Forsman, 2017-11-08, No. 64

pHC

Even better: Cascade control

• Can you argue why this solution is better than the first one (only local control)?

NaOH

pHC

K. Forsman, 2017-11-08, No. 65

Thank you!