CHAPTER 17: Inferential Control

When I complete this chapter, I want to be

able to do the following.

Identify key aspects of conventional and

inferential sensors
Determine good applications of inferential
sensors
Determine how to evaluate potential
inferential sensors

CHAPTER 17: Inferential Control

Outline of the lesson.

Conventional and inferential sensors

An example for a flash process
Inferential design criteria
Distillation example
Reactor example
Special considerations for dynamics

CHAPTER 17: Inferential Control

BASICS OF INFERENTIAL CONTROL
C

ALL SENSORS ARE IN SOME SENSE INFERENTIAL

- ORIFICE P FOR FLOW
- THERMOCOUPLE m Volt FOR TEMPERATURE

CONVENTIONAL SENSORS MEASURE "TRUE" PROCESS VARIABLE

- REASONABLE ACCURACY & REPRODUCIBILITY
- LARGE RANGE W/ O CORRECTION
- NOT PROCESS SPECIFIC

INFERENTIAL SENSORS
-

PROCESS SPECIFIC
LIMITED RANGE
USUALLY PROVIDED POORER ACCURACY
INFERENTIAL ADVANTAGES: FASTER RESPONSE, LOWER COST

EXAMPLE: FLASH DRUM

tru e va ria b le

= xe

in feren tia l va ria b le = T

m a n ip u la ted va ria b le
d is tu rb a n ce
in fe re n t ial re lat io n s h ip

=
=
=
:

liqu id com p os ition of eth a n e to b e con trolled

a t 1 0 %"2 .0 %
tem p era tu re
h ea tin g m ed iu m flow
feed com p os ition (a s s u b s equ en tly d efin ed )
xe = T +

Lets discuss
this process
T6

Feed

T1

P1

T5

T2

Vapor
product

P 1000 kPa
T 298 K

Methane
Ethane (LK) F1
Propane
Butane
Pentane
F2

T4

T3

L1

F3
A1

Process
fluid

Steam

L. Key

Liquid
product

CHAPTER 17: Inferential Control

INFERENTIAL VARIABLES FOR CONTROL

C

CVt = "TRUE"
VARIABLE

CONTROLLED

CVi = INFERENTIAL VARIABLE

C

CONTROLLING CVi TO ITS SET

POINT DOES NOT GUARANTEE
ZERO OFFSET FOR CVt

WHAT CONDITION RESULTS IN

ZERO OFFSET FOR CVt?

WHAT CONTROL ENHANCEMENT

WOULD YIELD ZERO OFFSET?

Figure 17-1. Block diagram of a feedback

control system with a true controlled variable,
CVt(s), and an inferential controlled variable,
CVi(s).

STEADY-STATE EVALUATION OF INFERENTIAL VARIABLE

C
C

CAN THE COMPOSITION BE MAINTAINED " 2 % ?

ASSUME TEMPERATURE CONTROL IS "0 .5 0 EC

The relationship between the flash temperature and the concentration of ethane in the liquid at the base case pressure
(1000 kPa). Changes in methane are compensated by changes in butane of equal magnitude and opposite sign.

Inferential variable when pressure changes

CHAPTER 17: Inferential Control

INFERENTIAL CONTROL WITH CASCADE RESET
C
C

FAST INFERENTIAL, SLOWER PRIMARY

ONE EXPENSIVE ANALYZER MEASURES TWO KEY VARIABLES

INFERENTIAL CONTROL DESIGN CRITERIA

Table 17-2. Design Criteria For Inferential Control
I. NECESSARY SITUATION
Measurement of the true controlled variable not available in a timely manner
A)
B)
C)

Not measured: on-stream sensor not possible

Not measured: on-stream sensor too costly
Unfavorable feedback dynamics: sensor has poor dynamics, e.g., long dead time or analysis time,
or is located far downstream

II. INFERENTIAL VARIABLE FEATURES

A)
B)
C)

Good relationship to true controlled variable for changes in the potential manipulated variable
Relationship in (A) insensitive to changes in operating conditions, i.e., unmeasured disturbances,
over their expected ranges
Favorable dynamics for use in feedback control

III. CORRECTION OF INFERENTIAL VARIABLE

A)
B)
C)

By primary controller in automated cascade design

By plant operator manually based on periodic information
When inferential variable is corrected frequently, the sensor for the inferential variable must
provide good reproducibility, not necessarily accuracy

CHAPTER 17: Inferential Control

EXAMPLE: DISTILLATION TRAY TEMPERATURE
tru e va ria b le
in feren tia l va ria ble
m a n ip u la ted va ria ble
dis tu rba n ces
pa ra m eters

= x D = h ea vy key in dis tilla te = .01 m ole fra ction

= T = tra y tem p era tu re
=
d is tilla te flow ra te
=
reboiler du ty, feed com pos ition
=
tra y efficien cy, th erm odyn a m ics

in feren tia l rela tion s h ip : x D = T +

Given th e in form a tion in th e tem p era tu re vs . tra y n o. figu re, wh ich is th e
bes t tra y tem pera tu re?
Or, wh ich tem pera tu re correla tes b es t with x d ?

LC

Manp. variables

LC

Trays
numbered
from the top
XD= fraction
xylene

Trays
numbered
from the top

Changes in
temperatures and
distillate
composition for
5% change in
distillate flow rate

Trays
numbered
from the top

Changes in
temperatures and
distillate
composition for
5% change in
reboiler duty

xB = feed benzene = 6, 10, 14%

Trays
numbered
from the top

6%

feed toluene= 45 - ( xB - 10)/2

10%
14%

Changes in
temperatures and
distillate
composition for
changes in feed
composition

IMPROVED PERFORMANCE VIA INFERENTIAL CONTROL

C

MUCH REDUCED DEVIATION

COMPOSITION DISTURBANCES

zero offset

FROM

DESIRED

FOR

FEED

CHAPTER 17: Inferential Control

CHAPTER 17: Inferential Control

E X AMPLE : CHE MICAL R E ACTOR
C

WH AT IS A G O O D INF E R E NTIAL VAR IAB LE ?

A B

T = - H rxn

C Ain

Cp

X A=

- H rxn
CA
CP

T = T4 - T3

wh ere
CA

Cp
H r xn
XA

= c on c e n t r a t ion of A,
m ole s / volu m e
= d e n s it y, m a s s / volu m e
= h e a t c a p a c it y, e n e r gy/ (EC *m a s s )
= h e a t of r e a c t ion , e n e r gy/ m ole
= fr a c t ion of fe e d r e a c t e d
= (C Ain -C Aou t )/ C Ain = C A/ C Ain

REACTOR INFERENTIAL: DYNAMIC RESPONSE

C

WOULD YOU EXPECT

FEEDBACK SYSTEM?

GOOD

PERFORMANCE

FROM

THIS

Figure 17-14. Plot of key variables for packed bed reactor inferential control. Note the significant
inverse response of the instantaneous temperature difference.
C

WHAT IF WE WANTED TO INFER B/ C, WHEN A6B a n d A6C ?