PIDtuningsoftwareApracticalreview.pdf

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/224062500
PID Tuning Software: A Practical Review
Conference Paper · July 2006
DOI: 10.1049/cp:20060471 · Source: IEEE Xplore
CITATION
1
READS
3,095
1 author:
Some of the authors of this publication are also working on these related projects:
Modelling and Controller Design of a Pasteurisation Process View project
Developing Graduate Attributes View project
Tom O'Mahony
Cork Institute of Technology
52 PUBLICATIONS 419 CITATIONS
SEE PROFILE
All content following this page was uploaded by Tom O'Mahony on 13 September 2016.
The user has requested enhancement of the downloaded file.

PID Tuning Software: A Practical Review
Tom O’Mahony
Advanced Control Group,
Dept. of Electronic Engineering,
Cork Institute of Technology,
IRELAND
E-mail: tomahony@cit.ie
__________________________________________________________________________________________
A wide range of PID tuning software tools is currently available,
though the relative merits of each is not always clear. In this paper,
the key features of seven (mostly commercial) PID tuning programs
are summarised and subsequently the performance of the software
tools are compared on a laboratory-scale process. The results
indicate that tools which include IMC-based tuning, and support
practical PID controller structures (derivative filter) performed
especially well.
Keywords – PID design, IMC tuning, system identification, thermal
process
__________________________________________________________________________________________
I INTRODUCTION
With academic research maturing and entering the
region of “diminishing returns,” the trend in present
research and development of PID technology appears
to be focused on developing software that will
maximise the potential of PID control. Good
software tools enable a student or practitioner with
some control knowledge or plant information to be
able to tune a PID controller efficiently for various
applications. From an educational perspective, a
simple and efficient design process, can encourage
experimentation, ease the burden on (mathematically
weaker) students and provide a stimulating
environment for control engineering education.
The motivation for this paper arose from the reviews
[1], [2] where a list of currently available PID
software tuning packages were compiled and
categorised. However, little information on the
relative merits of these packages was imparted.
Indeed a literature search reveals a dearth of
information on this topic. Hence the contribution of
this paper is to redress this topic, in some small way,
by reviewing and evaluating seven software tuning
packages for PID control. While this review may be
of some use to the industrial practitioner, it is
anticipated that it will primarily interest the academic
teaching PID control technology and the student who
may be faced with the (daunting?) task of designing
a PID controller for a particular task.
II SELECTED SOFTWARE PACKAGES
Approximately 45 software packages were included
in the reviews [1, 2], and due to this breath very little
detail was presented on any of the tuning tools.
Therefore, in this paper, a more detailed review of a
selected number of these packages will be presented.
It must be emphasised that it is not the author’s
intention to suggest that the selected tools are the
best available. On the contrary, the choice was
motivated by the need to make a start, and this author
chose the following subset of PID tuning software:
RaPID [3], TOPAS [4], U-Tune [5], EZYTune [6],
IMCTUNE [7] and CtrlLAB [8]. The product, U-
Tune was not included in the reviews [1, 2] and, in
addition, the Expert Tuner for PID control (ET PID)
[9], was also included. The following review will
only consider those features that are directly related
to PID controller synthesis and analysis.
Furthermore, the PID nomenclature used is taken,
where possible, from [10].
Of the selected products RaPID is the most
expensive at €3,000. A single user license for
TOPAS costs €2,000. EZYTune and U-Tune are
both available for $200. A full version of EZYTune
was used in this paper, while the free trial version of
U-Tune was evaluated. IMCTUNE, CtrlLAB and ET
PID are all freeware MATLAB-based products.
III REVIEW OF SELECTED TOOLS
a) RaPID
Version 5.1.1 (evaluation version) of the RaPID
(Robust Advanced PID Control) software [3] was

used in this study. Of the tools reviewed, the RaPID
software has the best user interface; a really
professional look and is very intuitive to use. The
software has two main components – system
identification and controller design. Within the
system identification section, data can be easily
imported, displayed, and a number of pre-processing
functions are supported e.g. filtering, truncation, etc.
ARX models are estimated and the user can specify
the model order. By default, when the identification
button is clicked, the tool identifies a range of
models and the continuous-time model that yields the
smallest modelling error is automatically displayed
to the user.
The controller design problem begins by specifying
the desired controller algorithm which can vary from
the ideal PI to a two degree-of-freedom structure.
The controller can be designed to optimise either
servo or regulator performance and the PID
coefficients are determined by optimisation [11].
Performance objectives can be specified as
constraints on the percent overshoot, rise-time,
settling time, integral of square error (ISE) or
integral of time by absolute error (ITAE).
Robustness is enforced by requiring that the open-
loop Nyquist curve remain outside a parabola centred
on the critical point. This constraint can be
converted to guaranteed gain or phase margin
criteria. Furthermore, the high-frequency gain of the
controller can also be constrained to minimise the
impact of high-frequency noise. These two
constraints, the robustness factor (R) and the
sensitivity to the noise (Ns) represent the two
controller tuning factors. These ‘tuning knobs’ are, in
this author’s opinion, not intuitive. For the
application considered in Section IV, this author
found that it was easier to tune a PI controller by
trial-and-error than to use the optimisation based on
this robustness factor and noise sensitivity. For more
complex structures, Eq 2, it is expected that the
optimisation is a more efficient design paradigm.
( 1, 0)
j
−
With RaPID software the user can view the gain and
phase margin, settling time, rise time, percent
overshoot, the response to a command, load or
disturbance input. The controller, open-loop and
common closed-loop frequency responses are
available as well as the temporal responses to all
models within a stored set. Actuator constraints can
be included.
b) TOPAS
TOPAS [4] is a process control training and
optimisation tool that is available from ACT GmbH.
The version evaluated in this study is rather old
(version 1.0 © 1998). A core element of the TOPAS
software is the process training element where the
user can experiment with flow, level, temperature,
pressure, pH processes and examine the effects of
equipment size, valve technology, etc, and a wide
variety of control technology, including P, PI, PID,
ratio, feedforward, cascade, smith predictor, and
model predictive control. A user can specify generic
single-input single-output (SISO), multiple-input
multiple-output process models or identify a model
from open or closed-loop data.
The interface (on this version) looks its age and is
not nearly as intuitive as that associated with RaPID.
The Import data feature only supports text files
(maximum number of rows is ~ 2000) and each
column must be separated by exactly eight spaces!
Unfortunately none of this is mentioned in the Help.
A first or second order model with time delay can be
estimated from the data. The models can be
estimated from open-loop data, closed-loop data
(where the process is controlled by a proportional-
only controller) or a relay test. The Tuner enables
PI or PID controllers to be designed. In this version
the (automatic) tuning options are limited to either
set-point or load tuning where the proportional action
is on the error or the PV. Controllers are tuned to
give “very little overshoot”. The ACT website does
mention that a number of new PID tuning techniques
have been added to the most recent version of
TOPAS.
The effectiveness of the design can be evaluated by
visually examining the response to a step change in
the command or load disturbance. The mean, ITAE,
standard deviation, average deviation, resource
consumption, potential operation improvements and
resulting incentives can be calculated.
c) U-TUNE
Version 2 of U-Tune [5] was evaluated. Compared
with either TOPAS or RaPID this software is
virtually featureless and, no doubt, this accounts for
the price difference. The tool supports data entry via
a standard text-file (separated by commas, space or
tab). Self-regulating process can be identified from
an open-loop step response. The user does most of
the work by clicking and dragging a number of
sliders to fit an exponential curve to the data. A
model is then generated from this curve. Based on
this model, PI and PID (ideal, series and parallel)
controllers are automatically tuned to yield a fast or a
slow response. Details are not provided on the tuning
procedure. Analysis features are not provided.
d) EZYtune
EZYtune (Version 1.1.04) is a software package
which focuses exclusively on PID controller tuning.
A strength of the tool is its simple, intuitive interface.
First and second-order systems, with or without a
delay/integrator, can be tuned and a wide variety of
industrial PID control structures are supported; Allen
Bradley, Bailey, Concept, Fischer & Porter, GE
Fanuc, Honeywell, Modicon, Siemens, Yokogawa

and Foxboro. Considering the variety of industrial
controllers, the structures are surprisingly similar and
can be classified as either the series (20%) or the
parallel form of the PID controller with derivative
acting on the error (66%) or PV. In the majority, the
derivative action is filtered, and a variety of formula
are used to calculate the filter time constant, coming,
presumably, from manufacturers recommendations.
Tuning is achieved by specifying either a desired
closed-loop time constant, β , or a desired closed-
loop rise-time. In theory, this single-parameter,
intuitive tuning procedure is an advantage of the
EZYTune package. The design is not idiot-proof. As
an example, for the model
24
0.67
( )
1 112 1
d s s
m
m
k e e
M s
s s
τ
τ
− −
= =
+ +
, Eq. 3
choosing the Allen Bradley Logix 5550 Independent
PID controller with a desired closed-loop time
constant of 45sec yields the following parameter set:
{ }
1.5, 0.015, 10.59, 0.66
c i d
K K T γ
= = = − = −
where Kc and i
K are the proportional and integral
gains, Td is the derivative time constant and γ is the
derivative filter time constant (sec). The negative
value associated with this latter parameter implies
that the resulting controller will be unstable – even in
the absence of modelling errors. The EZYTune
software also comes with a PID Translator
which may be used to translate PID coefficients
between structures. Analysis features are limited to
viewing the set-point and load responses.
e) IMCTUNE
The IMCTUNE software is a collection of MATLAB
m-files developed for MATLAB 5.3 and requiring
the Control System and Optimization Toolboxes.
This author tested the software under MATLAB v6.5
and MATLAB v7.1. With minor debugging, and
aside from a multitude of warnings, the software
worked fine.
As then name suggests, IMCTUNE is based on the
internal model control tuning paradigm and enables
an exact IMC, state feedback, PID, feedforward or
cascade control structure to be computed. All of the
designs are SISO and a comprehensive description of
the design methodologies is available in the text [12],
which the software accompanies. The graphical user
interface enables a process, model and disturbance
transfer function to be entered. A two degree-of-
freedom (2-DOF) controller can be designed if a
disturbance model is available. In the 1-DOF design,
the user must specify the model dynamics that can be
inverted, 1
( )
M s
−
, and the order of the IMC filter,
. The IMC controller is then computed as
where the order of the
filter is usually chosen to ensure that is
proper. The IMCTUNE synthesis problem is then to
determine the smallest filter time constant,
n
1
( ) ( )/( 1)n
IMC
C s M s s
ε
−
= +
( )
IMC
C s
ε , that
satisfies the following (default) design constraints:
(i) the magnitude of all closed-loop frequency
responses between process output and set-point must
have a value less than 1.05, (ii) in the frequency
response curve with the maximum magnitude, all
oscillations must have a peak-to-peak amplitude of
less than 0.1, and (iii) the high-frequency gain of the
controller must not be more than 20 times its low
frequency gain. Controller tuning is achieved by
varying the optimisation parameters
{ }
1.05, 0.1, 20 or by directly specifying ε . A
nice feature of the IMCTUNE software is that
uncertainty can be catered for by specifying the
parametric variation. The IMCTUNE software then
guarantees robust stability [12]. The software derives
the following approximate PID controllers based on
the exact IMC controller: the ideal PID controller,
the ideal PID controller in series with a first-order
filter, the ideal controller in series with a second-
order filter, and the controller with filtered
derivative.
Once the design is completed, the closed-loop
frequency responses from set-point to output and
from disturbance to output can be displayed and the
closed-loop step response can also be generated.
However, the diagrams are not interactive and
quantitative measurements of performance are not
presented. Robustness can be assessed in the context
of upper and lower bounds of the process parameters
for which stability is guaranteed i.e. for the model
defined by equation 3 and a controller defined by
112 1
( )
0.67(45 1)
IMC
s
C s
s
+
=
+
the software computes that
stability is ensured provided 0. ,
495 0.845
m
k
≤ ≤
82.69 141.31
m
τ
≤ ≤ , 17.72 30.28
d
τ
≤ ≤ . A limited
non-linear analysis is possible by specifying bounds
on the amplitude of the MV.
f) CtrlLAB
Version 3.0 (designed for MATLAB 5.3 but works
fine with v6.3 and v7.1) of CtrlLAB was evaluated.
CtrlLAB is replete with features. A possible
summary is: linear and non-linear (via Simulink)
model entry; state-space transformations; model
reduction; system analysis in the frequency, complex
(root-locus) and time domains; controller synthesis
incorporating lead-lag, linear quadratic, pole-
placement, PID tuning, linear quadratic gaussian,
loop transfer recovery, and . The brief review
conducted here will limit itself to the scope of this
paper.
2
H ∞
H
Arbitrary order continuous-time models (delays are
approximated using a Padé approximation) are
supported and based on these P, PI, ideal PID or

ideal PID where the pure derivative term is on the
feedback. These controllers can be designed using
tuning rules (Ziegler-Nichols, Cohen-Coon, Refined
Ziegler-Nichols, Chien-Hrones-Reswick (CHR)
Tuning, Modified Ziegler-Nichols), the IMC method,
optimisation (the optimisation functions can be
integral of squared error, ISTE, IST2E, or a default
gain and phase margin). The CHR Tuning method
requires the user to choose between a design that will
achieve zero percent overshoot or 20% overshoot,
the Modified Ziegler-Nichols requires the parameters
(the gain margin is
b
r 1/
m b
A r
= ) and b
φ (the phase
margin 180
m b
φ φ
= − ) to be specified while for the
IMC method a filter time-constant, f
T , is required.
The optimisation methods do not require user
interaction and the documentation does not mention
the default values for the gain and phase margins.
Overall, the associated help is quite limited (for
example, no information is available on the Refined
and Modified Ziegler-Nichols tuning methods) and
the lecture notes (D. Xue and D. P. Atherton,
Feedback Control Systems Analysis and Design
using MATLAB) would be informative. While a
huge range of frequency- and time-domain responses
are presented in graphical form, quantitative
measures of performance are not available. The gain
margin, phase margin, and metrics are,
however, presented to the user.
2
H ∞
H
g) ET PID
This software package also focuses solely on tuning
PID controllers and is (currently) limited to systems
that can be modelled by a first-order lag plus delay
(FOLPD) model. The Expert Tuner for PID control
consists of 55 PID tuning rules, all of which are
applied to the process model to realise 55 control
designs [13]. The user is requested to enter desirable
performance or robustness constraints, and designs
which satisfy these constraints are presented to the
user. The PID controller coefficients, along with
performance metrics (rise-time, settling-time,
overshoot, IAE) for set-point and load step changes,
robustness metrics (gain margin, phase margin, delay
margin, peak of the sensitivity function), and
graphical responses (set-point, load, complementary
sensitivity function and input sensitivity function)
are all available. The tool is split into two (very
similar) components. The first corresponds to all of
the tuning rules for which user interaction is not
required and the second component corresponds to
rules in which a tuning parameter, α , (usually the
IMC filter time constant or some related parameter)
is required. In the latter case the user has the option
of specifying a range for the robustness tuning
parameter, PID controllers are calculated for that
range and the results (that satisfy the design criteria)
are presented. The user can then make an informed
decision based on the presented results.
IV PERFORMANCE EVALUATION
The software tools reviewed in Section III were
evaluated on a laboratory-scale thermal system
(CE103 from TQ Systems). The CE103 comprises a
duct through which air may be driven using a
variable speed fan. An electrically heated process
block is mounted in the air flow path, and
temperature equilibrium is attained by balancing the
heat gained through the heater coil and the heat lost
through convection/conduction. Platinum resistance
thermometers monitor the actual temperature of the
block. In this paper a SISO configuration was
assumed where the single input was the electrical
voltage to the heater coil and the single output was
the measurement from the insulated thermometer.
The fan speed was set at a constant 5V (50%). The
process was modelled by estimating a process
transfer function from an open-loop step change
from 3-4V. As all of the software tools supported
FOLPD models, and because FOLPD models are
commonly used to tune PID controllers, a FOLPD
was initially estimated using the MATLAB System
Identification toolbox and then converted to
continuous-time using the MATLAB d2c function,
yielding equation 4.
24
0.677
( )
1 111.9 1
d s s
m
m
k e e
M s
s s
τ
τ
− −
= =
+ +
, Eq. 4
An examination of the autocorrelation function of the
residuals suggested that a lot of deterministic
information was not being captured by equation 4.
This was supported by a visual examination of the
open-loop response which more closely followed the
classic ‘s’ shaped curve associated with a high-order
overdamped system.
The software evaluation proceeded as follows. The
open-loop data was imported into the tools that
supported a system identification process and a
FOLPD model was estimated. This model was then
used for tuning purposes. For consistency, the model
defined by equation 4 was also used in all of the
tools. The design criteria was to minimise the closed-
loop settling-time (in response to a step input)
subject to the following constraints (i) the overshoot
is less than 10%, (ii)
( )
OS 6 '
m
A dB s
(iii)
. The designs were initially culled using a
simple non-linear simulation that incorporated
actuator constraints. Real-time performance was
evaluated over the same set-point change as was
used to model the process. The OS, settling time
(
45o
m
φ
s
t ), variance of the PV ( PV
σ ), variance of the MV
( MV
σ ) and the IAE between setpoint command and
PV were calculated. A selection of results is
presented in Table 1.
Of the tools reviewed U-Tune is probably the least
useful. The limited tuning features of “slow” and

“fast” do not support the design of PID controllers
for generic processes. This is reflected in the results
listed in Table 1. A similar comment applies to the
version of TOPAS with respect to it’s PID tuning
capabilities which are limited to “set-point” and
“load” tuning. The system identification features are
quite good, though not intuitive and this author had
difficulties loading data into TOPAS and getting an
accurate estimate of the delay. Once these difficulties
were resolved the model accuracy was similar to that
from the MATLAB System Identification Toolbox.
Higher-order models were also identified, but did not
yield improved control system performance. The
remainder of the software products investigated gave
approximately similar results. This author found that
the design philosophy advocated in RaPID was a lot
less intuitive than the IMC philosophy adopted in
IMCTUNE and CtrlLAB. The system identification
component of RaPID is excellent and really good
results can be obtained with minimal knowledge. In
this respect, it probably surpasses the MATLAB Sys
ID, though RaPID has less features. A higher-order
model was also estimated using RaPID but designs
based on this model yielded no dividends. While
EZYTune is different in control tuning philosophy, it
is similar to IMCTUNE, CtrlLAB and ET PID in the
sense that the design is reduced to choosing a single
parameter. For the application considered, this author
found that less time was required to obtain good
results using the IMC tuning. This may be due to the
fact that the resulting PID design is only
approximately related to the closed-loop time-
constant ( β ) that is specified in EZYTune. Thus
good PI-based designs were obtained when 80
β = ,
while for PID designs the best results were obtained
for 25
β = . Furthermore the following choices of
{ }
45, 60, 80
β = yielded the following (95%)
settling times { }
240, 177, 123
s
t = , i.e. increasing
the closed-loop time-constant specification resulted
in faster closed-loop responses! This counter-
intuitive result did not help the design process. While
CtrlLAB supports a number of PID design options,
the IMC tuning procedure was the only one that
provided reasonable performance. IMCTUNE
provides an option to optimise the filter time
constant, ε , however, the resulting performance was
not acceptable. This is not altogether surprising as
perfect modelling was assumed i.e. parametric
uncertainty was not entered. The filter time constant
was therefore chosen by trial-and-error. In this
evaluation, the supported controller structures
differentiated IMCTUNE and CtrlLAB. In CtrlLAB
only the ideal form of the PID controller can be
tuned while IMCTUNE provides tuning for the PID
controller in series with first and second-order filters.
It was found that the PID controllers generated by
CtrlLAB resulted in large MV
σ . Hence a PI
controller was used. In contrast, the structures tuned
using IMCTUNE provided excellent immunity to
high-frequency noise. The point being, that while in
this application PI worked well, in other applications
derivative action may provide a significant
advantage. The ET PID software results in
(approximately) equivalent performance. The range
of tuning rules provided (each with a robustness
factor to be chosen) can perhaps distract the user
from the primary job at hand – to get a single
controller working well.
On a personal note, this author is an advocate of a
(simple) design methodology that enables students to
implement, experiment with and evaluate control
technology as quickly as possible. Especially at
introductory level courses, my philosophy is to
encourage students to capture the dominant process
dynamics (first or second-order models with a time
delay), design a PI(D) controller, try it out and see
what happens. Of the tools evaluated, RaPID comes
closest to satisfying this philosophy in that it has (i) a
really intuitive user interface and (ii) it integrates the
modelling, design and analysis components of a
typical design cycle. As mentioned previously, I was
particularly impressed with the modelling aspect of
the tool, generating (at least for this application), a
sufficiently accurate model (for control design
purposes) with literately the click of a button. While
the RaPID literature states that the software tool has
been used to tune ‘thousands of loops’ this author is
not convinced by the approach that is advocated. In
my opinion, it is just not sufficiently intuitive and
lacks simplicity. That being said, the concepts on
which the design is based - percent overshoot, rise-
time, settling time, integral of error, robustness, high-
frequency controller gain and the impact of high-
frequency noise – would typically be covered in a
traditional introductory-level course on process
control. Furthermore, the analysis features, which
emphasise the same concepts do support the
philosophy of “best practice”.
However, from a pedagogical perspective this author
would prefer the IMC philosophy, where the
requirement for a model is transparent and the
performance/robustness trade-off can be negotiated
with a single tuning parameter. The main limitation
of the tools that enable IMC-based PID controllers to
be designed (IMCTUNE, CtrlLAB, ET PID) is the
lack of an integrated modelling facility. Thus,
alternative (supportive) modelling tools must be
introduced, which places additional constraints (time,
learning, cost, etc) on existing courses. The tools
IMCTUNE, CtrlLAB, EZYTune and ET-Tune all
‘go beyond’ the traditional Ziegler-Nichols tuning,
which is a big plus in my opinion and, with the
exception of CtrlLAB, all provide alternatives to the
ideal PID algorithm. Tools to support an analysis of
the resulting closed-loop performance are limited in

EZYTune, IMCTune and, to a lesser extent,
CtrlLAB.
V SOME CONCLUDING REMARKS
This paper has endeavoured to review seven PID
controller design software packages. The reviewed
packages vary over the price/feature spectrum. The
software was evaluated on a laboratory-scale
process. This process, while relatively simple, does
represent a typical SISO process for which PID
control would be used. The process can be modelled
by a FOLPD system, though the actual dynamics are
at least second-order.
The results of the review and evaluation
demonstrated that, from a tuning perspective, some
of the software tools (e.g. U-Tune) are too limited to
be of generic use. Other tools (e.g. CtrlLAB) are
degraded by the limited control structures supported,
while most suffer from not having an integrated
modelling capability. From this author’s perspective,
a truly useful software tuning package should
incorporate a (simpler) method of tuning the
controller (than manually tweaking P, I and D), a
method of extracting dominant dynamics from
captured data, incorporate practical controller
structures and be simple to use. None of the
reviewed packages incorporate all four of these
features, though the RaPID tool comes closest.
Since all evaluations are biased by a-priori
experience, it must be noted that this author had
previously used the packages EZYTune and ET PID.
While every effort was made to devote the same
amount of time to each package this was not, in
general, possible as some tools (U-Tune, EZYTune)
are intuitively simple while for others (TOPAS,
IMCTUNE) the user will likely benefit from reading
the manual. The number of real-time trials provides
an indicator of the effort spent in obtaining the
results of Table 1. For RaPID ten trials were made;
TOPAS nine trials; U-Tune five tests; EZYTune
eleven tests; IMCTUNE eleven tests; CtrlLAB
twelve trials and ET Tune thirteen trials. It is not
intended to suggest that the results of Table 1 are
optimum in any sense, or that the performance could
not be improved if more time permitted. They do,
however, provide a practical indication of the
capabilities of the respective software packages.
V REFERENCES
[1] Ang, K.H., G. Chong, and Y. Li, 2005, ‘PID control
system analysis, design and technology’, IEEE Proc.
Control System Tech., Vol. 13, No. 4, pp 559-576
[2] Li, Y., K.H. Ang, G. Chong, 2006, ‘Patents, software
and hardware for PID control’, IEEE Control Sys. Mag.,
Feb 2006, pp. 42-54.
[3] RaPID, IPCOS, http://www.ipcos.be/products/generic/
rapid.html, (accessed March 2006)
[4] TOPAS, ACT GmbH, http://www.act-control.com
(accessed February 2006)
[5] U-Tune, Contek Systems, http://www.contek-
systems.co.uk (accessed Feb. 2006)
[6] EZYTune, Matrikon Inc., http://www.matri-kon.com,
(accessed February 2006)
[7] IMCTUNE, http://www.mathworks.com/matlabcentral/
fileexchange/loadFile.do?objectId=3369objectType=file,
(accessed March 2006)
[8] CtrlLAB, http://www.mathworks.com/matlabcentral/
fileexchange/loadFile.do?objectId=18objectType=file,
(accessed March 2006)
[9] ET PID, http://www.acg.cit.ie/software
[10] O’Dwyer, A., (2003), Handbook of PI and PID
Controller Tuning Rules, London, U.K., Imperial College
Press.
[11] Oviedo, J.J.E., T. Boelen P. Van Overschee, 2006,
‘Robust Advanced PID Control (RaPID) PID Tuning
Based on Engineering Specifications’, IEEE Control Sys.
Mag., Feb 2006, pp. 15-19.
[12] Brosilow, C. B. Joseph, (2002), Techniques of
Model Based Control, Prentice Hall PTR.
[13] Murphy, P. T. O’Mahony, 2004, ‘An evaluation of
PID controller tuning rules’, Proc. of ISSC ’04, Queens
University, Belfast, pp. 399-406
# Software Control
Structure
Design Summary OS% s
t (sec) PV
σ
(x10-4
)
MV
σ
(x10-3
)
IAE
1 RaPID Ideal PI Ns=2.5; R=60; tuning = 0.46; 2.3 126 5.35 1.71 15.1
2 TOPAS PID (Ideal) PID but with derivative
action on PV; set-point tuning.
1.5 228 6.82 1.97 22
3 U-Tune PI Ideal PI. Slow tuning 9.4 361 6.54 0.78 21.78
4 EZYTune PI Ideal PI. 70
β = 3.1 120 6.68 1.87 14.81
5 IMCTUNE PID Ideal PID in series with a
second-order filter; 40
ε =
1.89 116 5.79 1.44 13.31
6 CtrlLAB PI Ideal PI; 45
f
T = 3.91 123 5.00 1.37 14.87
7 ET PID PID Controller with filtered
derivative tuned using Leva
Colombo (2000) with 41
α =
2.87 122 7.03 3.05 13.64
Table 1: A summary of the closed-loop performance obtained from each of the seven PID tuning tools. Results
are based on an average of four 13minute tests.
View publication stats
View publication stats

PIDtuningsoftwareApracticalreview.pdf

More Related Content

Similar to PIDtuningsoftwareApracticalreview.pdf (20)

Recently uploaded (20)

PIDtuningsoftwareApracticalreview.pdf