Engineering Encyclopedia

Saudi Aramco DeskTop Standards

ADAPTIVE CONTROL

Note: The source of the technical material in this volume is the Professional
Engineering Development Program (PEDP) of Engineering Services.
Warning: The material contained in this document was developed for Saudi
Aramco and is intended for the exclusive use of Saudi Aramco’s employees.
Any material contained in this document which is not already in the public
domain may not be copied, reproduced, sold, given, or disclosed to third
parties, or otherwise used in whole, or in part, without the written permission
of the Vice President, Engineering Services, Saudi Aramco.

Chapter : Process Instrumentation For additional information on this subject, contact

File Reference: PCI-102.10 PEDD Coordinator on 874-6556
Engineering Encyclopedia Control Loop Elements & Automatic Control Systems

Adaptive Control

Section Page

INTRODUCTION............................................................................................................ 2

DESCRIBE ADAPTIVE CONTROL AND ITS USES ...................................................... 3

DESCRIBE APPROACHES TO ADAPTATION ............................................................. 6

Programmed Adaptation ........................................................................................... 7
Example of Programmed Adaptation Using Process Variable
Information .......................................................................................................... 7
Feedforward Adaptation.......................................................................................... 10
Feedback Adaptation .............................................................................................. 15

IMPLEMENT ADAPTIVE CONTROLLERS .................................................................. 17

Adaptive Hardware ................................................................................................. 17

GLOSSARY ................................................................................................................. 19

Saudi Aramco DeskTop Standards i

Engineering Encyclopedia Control Loop Elements & Automatic Control Systems

Adaptive Control

INTRODUCTION

Adaptive control is a relatively new approach to control that is

continuously gaining more acceptance in the process control
industry. The concepts involved in adaptation are not new.
They have been known for some time and have been
researched academically with considerable success.
Industrially, however, in process control applications, their
successes have been limited mainly because of hardware
(software) limitations.

In the last few years the introduction of new products has

changed the picture considerably making adaptation much more
feasible today. Some of the concepts required for adaptation
are not new to us. These have been introduced in the various
modules of PCI 102 and form our knowledge base. We know,
for example, that the controller settings of a well tuned loop
depend on the process characteristics. If these characteristics
change either in the steady state or dynamic sense, the loop
has to be retuned to satisfy the new conditions.

The first step in adaptation requires a better understanding of

our process, followed by a better communication between the
process and the controller, and finally, a more intelligent or
smart controller able to adapt its parameters to meet the needs
of the process. Thus adaptation requires a continuous
monitoring of the relevant process information in order to adapt
the controller parameters to the needs of the process.

Today, microprocessor based products along with process

computers can deal with the process information much more
effectively. This allows the practical use of adaptation
techniques in industrial process control applications. The
effective use of adaptive strategies requires better user
understanding of the process.

Saudi Aramco DeskTop Standards 2

Engineering Encyclopedia Control Loop Elements & Automatic Control Systems

Adaptive Control

DESCRIBE ADAPTIVE CONTROL AND ITS USES

The word adapt means to change or fit by modification to new

conditions. An adaptive control system may be defined as a
system whose parameters automatically change in response to
changing process characteristics. The automatic change of the
control parameters allows compensation for the changes in the
process characteristics and the maintenance of a constant loop
gain. The benefit of this adaptation is the ability of the loop to
maintain stability and an optimum response of the controlled
variable while being subjected to load upsets. We have
effectively forced the control system to change its parameters to
satisfy the process requirements through adaptation.

Adaptation goes beyond the steady state gain linearization or

compensations in analyzing steady state gains that we
discussed in PCI 102.05.

A simple linearization to achieve constant steady state is not

considered adaptation since all the controller functions remain
the same. This is no more than linearization of a loop element
such as in an orifice plate application, where a square root
extractor is used to linearize.

Saudi Aramco DeskTop Standards 3

Engineering Encyclopedia Control Loop Elements & Automatic Control Systems

Adaptive Control

A nonlinear controller typically used in a pH application operates

at different gains based on the loop operating point. This
controller, even with its gain a function of the pH measurement
input, is not considered adaptive since its controller functions
are fixed.

Consider a pH process controlled by a nonlinear controller.

Applying the controller to the process in some cases linearizes

the loop for constant gain. The intent here is to get a linear
overall response by compensating the nonlinearity of the
process with the nonlinearity of the controller. This works for
well-behaved titration curves, where the high gain portion of the
pH curve coincides with the dead band of the controller.

Saudi Aramco DeskTop Standards 4

Engineering Encyclopedia Control Loop Elements & Automatic Control Systems

Adaptive Control

If the titration curve drifts (changes shape) the linearization

loses its effectiveness and there is nothing the controller can do
to take care of the problem. Therefore, this is strictly nonlinear
control. Remember, to be adaptive, the controller must change
its parameters in order to accommodate the changing process
parameters. To accomplish this requires a more capable
controller as well as additional communication between the
process and the controller.

Saudi Aramco DeskTop Standards 5

Engineering Encyclopedia Control Loop Elements & Automatic Control Systems

Adaptive Control

DESCRIBE APPROACHES TO ADAPTATION

A few approaches have been used to implement adaptive

control strategies. The following is a summary of some of these
approaches.

• Gain scheduling or programmed adaptation - based on a

change in a process variable i.e. the set point.

• Feedforward adaptation - based on a change in load.

• Feedback adaptation - based on a change in the controlled

variable (measurement.)

Summary of Adaptive Approach

Saudi Aramco DeskTop Standards 6

Engineering Encyclopedia Control Loop Elements & Automatic Control Systems

Adaptive Control

Programmed Adaptation

If it is possible to determine a relationship in a given process

between a process variable and its effect on the gain of that
process than a program can be set up to adapt the controller
adjustments (tuning settings) to accommodate the particular
process gain. Large capacity processes such as a heat
exchanger or a critical level application are examples of this
approach. Both of these processes have dynamic gains that
vary with load. They are also in most cases a once through-
process which may be difficult to control with a simple feedback
loop.

Example of
Programmed
Adaptation Using
Process Variable
Information
Let us proceed to investigate the level application for
programmed adaptation.

Saudi Aramco DeskTop Standards 7

Engineering Encyclopedia Control Loop Elements & Automatic Control Systems

Adaptive Control

A test of this process shows that the gain of the process is a

function of the load as well as the set point or operating point.

Where loads: f3 < f 2 < f 1

Notice that for a given load (f1, f2, or f3) the gain of the process
is a function of the set point.

→
G = dout din

→
∆h
i.e. G=
∆fi

→
and for a given load as r↑ G↓

Saudi Aramco DeskTop Standards 8

Engineering Encyclopedia Control Loop Elements & Automatic Control Systems

Adaptive Control

If we could analytically define the load curves for the process,

we could define the gain of the process for various set points at
the particular load. A computer is programmed to schedule
various controller gains, PID as a function of the set point and
the particular load. The tuning constants of the controller are
thus modified to compensate for the changing process gain and
to maintain optimum response.

Gain scheduling as seen above is designed to handle changes

in process gain.

The scheme looks like an open-loop compensation that uses

process variable information, to compensate for feedback
controller gains.

To design a gain schedule system it is necessary to determine

the operating variables that affect the product being made and
to implement the schedule with this information. The controller
PID settings will have to be tuned for the various process
conditions in order to handle the process gain variations and
maintain stability at all operating points.

Since the information basically is of the open-loop type, purists

may argue whether gain scheduling or this type of programmed
adaptation is truly an adaptive system.

Saudi Aramco DeskTop Standards 9

Engineering Encyclopedia Control Loop Elements & Automatic Control Systems

Adaptive Control

Feedforward Adaptation

We can improve on gain scheduling type of adaptive open-loop

approach by arranging an adaptive program based directly on
load information. This makes the strategy a feedforward
adaptation.

This is also a once through capacity process whose steady

state gain:

∆T2
Kp =
∆Fs

Saudi Aramco DeskTop Standards 10

Engineering Encyclopedia Control Loop Elements & Automatic Control Systems

Adaptive Control

The dynamic gain for this process can be evaluated as follows:

Where: Fw3 > Fw2 > Fw1

Recognize that as the load, Fw, increases through the heat

exchanger, the dead time decreases. That is, if we set the
steam flow, Fs, at a given flow and observe the outlet
temperature, T2, at a given cold water flow Fw3, we would see
some dead time and then a steady state outlet temperature. If
we decrease the load to Fw1 where Fw1 < Fw3, then, since the
flow through the process is less, the dead time is longer and the
steady state outlet temperature is higher.

The steady state gain for this process

∆T2
Kp = Load = Const
∆Fs

is a function of the load, as in the case of the level tank. As the

load, Fw, increases, the steady state gain decreases. Now
considering the dynamic gain, we know:

1 1
τd ∝ ∝ i. e. as load, Fw ↑, the Dead Time, τ d ↓
Fw Load

We also know that the period of oscillation and dead time are
related:

τ0 ∝ τd

Saudi Aramco DeskTop Standards 11

Engineering Encyclopedia Control Loop Elements & Automatic Control Systems

Adaptive Control

The dynamic gain of a self-regulating capacity is given by:

−1
  τ1  
2 2

G = 1 +  2π  
  τ 0  
 

For a given process, with a capacity time constant τ1, as τd

changes, τ0 changes and the dynamic gain G changes as
follows:

1 1
i.e. τd ∝ ; τ0 ∝
Fw Fw

1
G ∝ τ0 ; ∴G ∝
Fw

And as the load increases, τ0 decreases, the dynamic gain also

decreases since G is given approximately by:

τ0
G≈
2πτ1

Since we established that both the dynamic and steady state

gain of the process decrease with increasing load, we require
an increasing controller dynamic as well as steady state gain
with increasing load to maintain an optimum response. The
gain of a PID controller is given by:

→ 2
100  2 πD τ 
G= 1 +  − 0 
PB  τ0 2π I 

100
where PB is the controller steady state gain and the remaining
terms constitute the dynamic gain. For optimum response
under changing load conditions we require both the steady state
and dynamic gains of the controller to be a function of load.
Since the process gain is inversely related to load

1
Gp ∝
Fw

Saudi Aramco DeskTop Standards 12

Engineering Encyclopedia Control Loop Elements & Automatic Control Systems

Adaptive Control

we require a controller whose gain is directly related to load

Gc ∝ Fw

Steady state gain of the controller will affect damping of the

loop, and dynamic gain of the controller affects the period of
response as well as the damping.

Consider steady state gain adaptation:

1
Gp ∝ and Gc ∝ Fw
Fw

100 1
Gc = or Gc ∝
PB PB

Therefore PB relates to load as follows:

1
PB ∝
Fw

i.e., as the load increases we want the PB to decrease.

Consider the dynamic gain adaptation .

As derivative time, D is increased, τ0 decreases, and as integral

time I in minutes, is increased, τ0 decreases.

In order to maintain a constant response period, τ0, we must

adaptively change the controller gain setting of D and I.

We have already seen:

1
τ0 ∝ τd ∝
Fw
1
D∝ ∝ Fw
t0
1
I∝ ∝ Fw
τ0

∴ I, D ∝ Fw (where I and D are in min)

Saudi Aramco DeskTop Standards 13

Engineering Encyclopedia Control Loop Elements & Automatic Control Systems

Adaptive Control

Thus, as the load increases, we want I and D to increase in

order to maintain a constant damping period.

Putting all the information together we can now synthesize a

load adapted PID algorithm.

Consider a PID algorithm:

100  1 dc 
m=
PB  e+
I ∫
edt − 
dt 

We can convert this to the load adapted algorithm:

100  1 D dc 
m= f e +
PB  If ∫ edt − f dt 
Where f is the normalized (0 to 1) output of a flow transmitter
measuring the load Fw.

If we tune the controller at maximum flow (f = 1) the above

algorithm is a normal PID algorithm. At maximum load the
process gain is minimum and controller gain is maximum. As f
decreases, the process gain increases and we require the
controller gain to decrease and maintain the damped period, τ0
constant.

Summary of adaptation:

100
f as f decreases, Gp increases, and Gc decreases.
PB

1 as f decreases τd in the process increases and τ0

If increases. The effective integral time, I/F, increases
causing the period to decrease.

D as f decreases, effective derivative time increases,

f decreases τ0.

The proportional term serves to maintain constant damping with

the load change while the integral and derivative adapted terms
serve to maintain constant τ0 with decreasing load.

The gain compensation is accomplished by programming the

PID settings of the controller as functions of load (Fw).

Saudi Aramco DeskTop Standards 14

Engineering Encyclopedia Control Loop Elements & Automatic Control Systems

Adaptive Control

Feedback Adaptation

The titration curve shown below is extremely nonlinear.

The gain of this process varies significantly and is quite high

between 4 < pH < 11.

A nonlinear-controller has some possibilities regarding control of

this process if the titration curve is stable and of a fixed shape.
The titration curve however can change its shape or drift which
make the nonlinear solution ineffective. Other solutions became
available as electronic hardware was introduced to the process
industries. Electronic solutions became available in the early
1970's prior to the microprocessor age.

Saudi Aramco DeskTop Standards 15

Engineering Encyclopedia Control Loop Elements & Automatic Control Systems

Adaptive Control

One of the earliest industrial adaptive control strategies used

electronic hardware in a feedback type adaptation using two
controllers. One of these controllers had a nonlinear algorithm
with a dead band while the other controller performed the
adaptation by varying the dead band on the nonlinear controller.

The adaptive controller senses the period of the pH

measurement and changes the dead band as required to bring
measurement to the set point. The adaptive control strategy
involved the widening of the dead band if oscillations persisted
and the narrowing of the dead band when the response was
sluggish.

Saudi Aramco DeskTop Standards 16

Engineering Encyclopedia Control Loop Elements & Automatic Control Systems

Adaptive Control

IMPLEMENT ADAPTIVE CONTROLLERS

In the 1980s adaptive feedback-type controllers with self-tuning

capabilities attained significance in the process control industry.
Several commercial microprocessor-based products with
adaptive capabilities are available today.

Several hundred thousand industrial loops have some adaptive

strategy applied with varying degrees of success.

The adaptation capability offered by the adaptive controller, with

PID changing capabilities, still requires the control engineer to
have a good knowledge of the process and of the application of
this basic knowledge in setting up the loop for better control.
Linearization, process noise filtering, minimization of dead time
and lag in the instrumentation must be considered for proper
control.

The implementation of an industrial adaptive control strategy

requires more time and effort than a conventional PID controller.
Obviously, the adaptive strategies involve more difficult
processes.

An adaptive self-tuning type controller should not be used

arbitrarily on any process. If the process gain is fixed, even if it
is nonlinear, there may be other ways to handle the loop. For
example, as we saw in PCI 102.06, Tuning Feedback Control
Loops, we may only add some robustness to the tuning settings.
Also, it is not good practice to rely on adaptive control to take
care of an improperly designed loop. It is better to solve the
problem and apply conventional control.

Adaptive Hardware

There is a variety of adaptive or self-tuning controller hardware

available today. Each controller has a different implementation
requirement. Some of these controllers induce a load upset to
get information. Others wait for natural disturbance to make
PID changes.

Saudi Aramco DeskTop Standards 17

Engineering Encyclopedia Control Loop Elements & Automatic Control Systems

Adaptive Control

The products relate to tuning much like we did in PCI 102.06

with varying results as we did. One particular product uses a
pretune mode to get Ziegler Nichols type of information, than
through pattern recognition deals with gain or damping as well
as period adjustments. The likelihood of more advanced easier
to use hardware should enhance the adaptive control
capabilities in the future.

Saudi Aramco DeskTop Standards 18

Engineering Encyclopedia Control Loop Elements & Automatic Control Systems

Adaptive Control

GLOSSARY

adaptive control Control in which automatic means are used to change

the type and/or influence of control parameters in such a
way as to improve the performance of the control
system.

characteristic curve A graph (curve) which shows the ideal values at steady-
state, or an output variable of a system as a function of
an input variable, the other input variables being
maintained at specified constant values.

compensation Provision of a special construction, a supplemental

device, circuit, or special materials to counteract
sources or error due to variations in specified operating
conditions.

compensator A device which converts a signal into some function

which, either alone or in combination with other signals,
directs the final controlling element to reduce deviations
in the directly controlled variable.

dead band The range through which an input can be varied without
initiating observable change in output. (There are
separate and distinct input-output relationships for
increasing and decreasing signals.)

dead zone A predetermined range of input through which the output

remains unchanged, irrespective of the direction of
change of the input signal.

drift An undesired change in the output-input relationship

over a period of time.

feedforward control Control in which information concerning one or more

conditions that can disturb the controlled variable is
converted, outside of any feedback loop, into corrective
action to minimize deviations of the controlled variable.

linear The type of relationship that exists between two

variables when the ratio of the value of one variable to
the corresponding value of the other is constant over the

Saudi Aramco DeskTop Standards 19

Engineering Encyclopedia Control Loop Elements & Automatic Control Systems

Adaptive Control

entire range of possible values.

linearity The closeness to which the curve relating two variables

approximates a straight line. It is usually expressed as
the maximum deviation between the actual curve and
the best fit straight line.

linearization The process of converting a nonlinear (nonstraight-line)

response into a linear response.

microprocessor A large-scale integrated circuit that has all the functions

of a computer, except memory and input/output
systems. The IC thus includes the instruction set, ALU,
registers and control functions.

pH The symbol for the measurement of acidity or alkalinity.

Solutions with a pH reading of less than 7 are acid;
solutions with a pH reading of more than 7 are alkaline
on the pH scale of 0 to 14, where the midpoint of 7 is
neutral.

robust control loop A robust control loop is one whose controller tuning
settings can accommodate moderate changes in the
process parameters and perform well even in the
presence of these changes.

self-tuning Controllers with the ability to change their control

controllers parameters (PID) based on process information.

titration curve A plot with pH as the ordinate and units of reagent

added per unit of sample as the abscissa.

Saudi Aramco DeskTop Standards 20