Research Article

International Journal of Advanced

Robotic Systems
September-October 2017: 1–11
Adaptive proportional–integral ª The Author(s) 2017
DOI: 10.1177/1729881417728467
controller using OLE for process journals.sagepub.com/home/arx

control for industrial applications

Marco A Paz1, Tania A Ramirez-delReal1, Suselle C Garibo1,

Dejanira Araiza-Illan2, Carlos A DeLuna-Ortega1
and Marving O Aguilar-Justo2

Abstract
This article presents an implementation of an adaptive control architecture, which provides the combined advantages of
better dynamic performance compared to other conventional industrial controllers, and the use of widely available
hardware in process industry. Adaptive control architecture uses proportional–integral action and dynamic computation
of the controller’s gains (self-tuning regulator), to maintain performance specifications, even in the presence of parametric
disturbances. This architecture offers advantages over other advanced embedded control systems implemented on
industrial programmable logic controllers and other hardware platforms. Implementation of controllers on industrial
hardware platforms is possible through the Object Linking and Embedding (OLE) for process control communication
standard. The implementation for an adaptive controller here proposed was evaluated through experiments using first-
order and overdamped second-order systems emulated by hardware-in-the-loop, with a programmable automation
controller. Performance of the adaptive controllers was compared to that of conventional proportional–integral con-
trollers, and effectiveness of the former over the latter was demonstrated through the experiments carried out.

Keywords
Self-tuning regulators, PI controller, recursive least squares, programmable logic controllers, OPC, hardware-in-the-loop

Date received: 26 September 2016; accepted: 19 July 2017

Topic: Special Issue – Theoretical and Experimental Technologies for Advanced and Basic Machines
Topic Editor: Pedro Ponce

Introduction
In industry applications, more than 95% of the used control
loops are of the proportional–integral–derivative (PID)
type.1 PID controllers were introduced in industrial pro-
cesses by the end of the 1930s.2 The use of these controllers 1
Academic Secretariat, Polytechnic University of Aguascalientes,
is extensive due to the following main advantages: Aguascalientes, Mexico
2
Advanced Remanufacturing and Technology Centre, Agency for Science,
a. Their good performance in a wide variety of Technology and Research, Singapore, Singapore
applications within industrial processes and
Corresponding author:
manufacturing automation. Marco A Paz, Polytechnic University of Aguascalientes, Paseo San Gerardo
b. The fact that commercial PID controllers are 207 Fracc. San Gerardo, Aguascalientes, Ags. 20342, Mexico.
relatively inexpensive. Email: marco.paz@upa.edu.mx

Creative Commons CC BY: This article is distributed under the terms of the Creative Commons Attribution 4.0 License
(http://www.creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work without
further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/
open-access-at-sage).
2 International Journal of Advanced Robotic Systems

c. PID controllers can be tuned using heuristic meth- controllers and other control methods has received lot of
ods, although their performance is better when attention in literature.12–14
they are tuned with an analytic tool. Limitations of conventional control systems (PID
d. The work of existing technological platforms is to controllers) have brought about the development and imple-
aid in the implementation of PID controllers, thus mentation of advanced control strategies in industrial processes.
the adoption of different control strategies can Some authors propose innovative algorithms in order to tune a
represent technical and economic difficulties. PID controller; however, only simulation results are showed.15
Currently, the control of industrial processes has car-
The use of PID controllers in industrial processes and ried out through the use of computer-based platforms,
manufacturing is comparatively less effective than using such as distributed control systems (DCS), supervisory
other controllers systems in the presence of nonlinear control and data acquisition, and/or direct digital
dynamics, disturbances, and parametric uncertainties (in control. 14,16 In recent years, process control systems
sensors, in actuators, and in the process itself). That is to based on programmable logic controllers (PLCs) and pro-
say, when it comes to industrial processes, performance of grammable automation controllers (PACs) have become
PID controllers can be unsatisfactory, poor, or even unstable. more popular in Mexico. One of the reasons for this is that
Another problem related to this fact is that operators in PLCs and PACs have better reliability and lower cost
charge of supervising and adjusting the controllers are often compared to DCS. Implementation of control techniques
insufficiently and inadequately trained. Operators with years in computer-based platforms commonly requires digital
of experience are skilled at tuning PID controllers intui- signal processors, data acquisition cards, or closed archi-
tively, but it comes to working with more advanced control tecture PLCs. PID controllers with STR structures have
techniques they usually prove unable. Ender estimates that at been previously implemented in PLCs17,18 or another
least one-third of the PID control loops used around the advanced control techniques such as adaptive fuzzy con-
world operate in manual mode, also other portion of the trol in order to implement in industrial systems.19
loops, and the gains are tuned by “trial and error.”3 We propose an implementation of an advanced control
The use of an adaptive control has been seen as an technique: a PI with an STR structure. This type of adaptive
alternative to the use of control systems in the presence controller architecture allows the adjustment of parameters
of parametric variations and disturbances since the end of (the PI gains) based on dynamic changes in the process to be
the 1950s.4 The self-tuning regulator (STR) structure was controlled.6 The controller runs on a remote computer ter-
specifically designed concurrently by different authors, and minal, in communication with an industrial control platform
it is usually conceptualized, depending on its structure, as (PLC/PAC). This communication is performed using Object
being direct or indirect or as being explicit or implicit.5,6 Linking and Embedding (OLE) for process control (OPC)
The problems of performance due to changes in the standard, which is the most common architecture used in
process mentioned above and the need for manual adjust- industrial automation and control of processes.20,21 The
ment of the gains by operators can be solved with the implementation consists of two stages:
implementation of an advanced control technique in con-
junction with the use of a conventional structure such as a a. Developing an adaptive control algorithm that
PI controller. In this article, an STR-based approach is allows the dynamic computation of PI controller
employed to adjust the parameters in the control system, gains, based on the algorithm presented by Paz
and this system performs based on dynamic changes in the Ramos et al.,7 to maintain the required performance
process to be controlled, as proposed by Paz Ramos et al.7 criteria in the presence of parametric disturbances.
In the literature can be found mainly adaptive control The abstraction of systems of the control algorithm
applications in process control where the time constants are needs to be designed carefully to take the system
large; however, there are also applications of adaptive con- into first- or second-order transfer functions.
trol to control actuators used in robotic applications8; nev- b. Implementing the adaptive controller in a server
ertheless, they are implemented in specific purpose connected remotely to a PLC (model Allen Brad-
hardware platforms with which a closed architecture is ley ControlLogix 5561 model; Rockwell Automa-
usually used. Also, other adaptive control structures have tion, Wisconsin, USA). This stage includes
been used in manipulator robots.9–11 implementation of an interactive information
Academic interest in PID controllers has gradually transfer system using the OPC industrial commu-
decreased, giving space to other control strategies, for nication standard, based on the approach presented
example, fuzzy controllers, neuro-fuzzy controllers, sliding in literature.22,23 for conventional PID controllers.
mode controllers, nonlinear controller and adaptive con-
trollers, among others; however, these other control strate- Systems of different orders and their designed control-
gies have not been able to gain significant presence in lers were used to evaluate the proposal. The dynamics of
industrial processes. The gap between research conducted first-and second-order systems are emulated using hard-
by the industrial and the academic sectors on PID ware-in-the-loop (HIL) through a PAC; this emulation
Paz et al. 3

reduces the complexity, size, and cost of the test struc- It is evident, from equation (3), that the closed-loop
tures and increases the possibility of testing under con- control system from equations (1) and (2) yields a
trolled operation 24 in order to have repeatability 25 ; second-order system with two poles and one zero.
further, it allows to simulate different conditions, such The canonical form of a second-order transfer function is
as a load in robot manipulators.26 Control systems are
!n 2
evaluated using the value of the integral absolute error GðsÞ ¼ (4)
(IAE) and the settling time. s 2 þ 2!n s þ !n2

where  is the damping coefficient and !n is the undamped

In the following sections, the design and implementation of the
natural frequency.
adaptive PI controller architecture are presented. “PI con-
Two dynamic characteristics of the response of a pro-
trollers” section explains the computation of the PI gains for
cess, associated with second-order systems, are the maxi-
the first- and second-order systems. “Adaptive architecture”
section presents the adaptive architecture with dynamic gains mum overshoot and the settling time. According to the
based on the study by Paz Ramos et al.7 “Implementation” transfer function in equation (4), the maximum overshoot,
section presents the implementation using a PLC, as explained Mp , can be approximated to
above. Validation of the adaptive PI controller, when it is pffiffiffiffiffiffiffiffi2ffi
applied to a first- and second-order system emulated by HIL, Mp  ep = 1 (5)
is presented in “Experimental results” section. Finally, conclu-
sions are presented in the last section. for 0 <  < 1. The settling time, ts , with a 1% error band,
can be defined as
4:6
ts  (6)
!n
PI controllers
A wide range of loops can be successfully controlled with- The desired settling time and maximum overshoot, tsd
out derivative action. The use of derivative action is not and Mpd , respectively, can be specified for controllers. For
recommended when noise is present in the processing of closed-loop controllers, the use of a tsd of greater than 50%
signals of the inputs of a controller.27,28 Designed PI con- the settling time in open-loop is recommended. If no other
trollers work mainly with two types of systems: first order specific recommendation for tsd is available, an alternative
and second order and their transfer functions. Complex is the relation
systems can be decomposed into first- and second-order tsd ¼ 3t (7)
transfer function combinations. The procedure to compute
gains for PI controllers from the transfer functions via pole The closed-loop system maintains the stability if the
placement1 is reviewed in the following subsections. denominator of equation (3) complies with Stodola’s stability
criterion: All characteristic roots have real negative parts.29
PI controller gains for first-order processes. Assuming a math- The choice of the maximum overshoot must be sub-
ematical approximation of a system to first-order jected to the inherent constraints of the specific process
dynamics, where the reaction delay is small enough to be to be controlled. If no particular constraints exist, a maxi-
neglected, and the corresponding transfer function is mum overshoot can be opted in order to the relation
K Mpd  15% (8)
G 1 ðsÞ ¼ (1)
ts þ 1
Once Mpd and tsd have been selected, the location of the
where K is the gain in open-loop system and t is the time poles in the transfer function representing the desired sys-
constant. tem in closed-loop is proposed. The desired transfer func-
A dependent PI controller (or ISA)1 has the form tion becomes
Kc Ti s þ Kc
Gc ðsÞ ¼ (2) !nd 2
Ti s Gd ðsÞ ¼ 2
(9)
s 2 þ 2 d !nd s þ !nd
where Kc is the proportional gain and Ti is the integral time
constant. where !nd is the desired undamped natural frequency, and
The closed-loop control system (i.e. using a negative  d is the desired damping coefficient, according to equation
feedback) of a system (plant) described by equation (1) and (4). As Mpd and tsd are known, we can solve for !nd and  d
the controller in equation (2) produce the transfer function from equations (5) and (6), thus
  vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u  2
ðKKc Þ
ðTi s þ 1Þ u
u lnðMpd Þ
GLC 1 ðsÞ ¼
ðTi tÞ
(3) d  u
t  2 (10)
1 ðKK Þ
s 2 þ ð1 þ KKc Þs þ c p 2 þ lnðMpd Þ
t ðTi tÞ
4 International Journal of Advanced Robotic Systems

The second-order component in equation (16) can be

4:6
!nd  (11) proposed by choosing the maximum overshoot Mpd and
 d tsd settling time tsd , as described before for first-order pro-
Finally, the coefficients in the denominator, the closed- cesses, meaning that it is overdamped. Certain restric-
loop transfer function (3), can be computed through the tions need to be considered to propose the location of
equivalence of the terms in the denominator of equation the third pole.
(9), the second-order desired transfer function. The result- When it is constructing the second-order polynomial of
ing gains are now defined as equation (16), it is important to note that an arbitrary choice
of parameters Mpd and tsd could cause a configuration of
2t d !nd  1 pole locations that may turn the system unstable. There are
Kc ¼ (12)
K necessary and sufficient conditions to avoid incurring in
KKc this situation, therefore that the system is stable.29 These
Ti ¼ (13) include proposing the desired settling time as
tð!nd Þ 2
9:2ðt1 t2 Þ
The steps of the algorithm for first-order systems with- tsd > (17)
out delay can be summarized as follows7: t1 þ t2
Following the same procedure and design to calculate
1. Suggest Mpd and tsd , the desired maximum over-
the gains Kc and Ti , when the process has been modeled as a
shoot and settling time.
second-order transfer function, mentioned previously, the
2. Replace Mpd and tsd in equations (10) and (11) to
terms of the denominator in equation (15) and the expanded
obtain !nd and  d .
polynomial in equation (16) are related
3. Calculate the controller gains, Kc and Ti , applying
equations (12) and (13), !nd ,  d , and the system !nd 2 t1 t2 þ 2p1  d !nd t1 t2  1
Kc ¼ (18)
parameters K and t. K
Kc K
PI controller gains for overdamped second-order Ti ¼ (19)
t1 t 2 !nd 2 p1
processes
The algorithm for the computation of the design of a
A representative set of industrial processes can be approxi- second-order system, when  > 1, is summarized as
mated satisfactorily by an overdamped (i.e.  d > 1) follows7:
second-order transfer function
K 1. Approximate, where it is possible, the process
G2 ðsÞ ¼ (14) control structure to the form in equation (14).
ðt1 s þ 1Þðt2 s þ 1Þ
2. Propose the maximum overshoot such that
which in turn can be seen as a combination of two first- 0 < Mp < 1.
order transfer functions in serial order. 3. Propose tsd according to equation (17).
Assuming that the PI controller defined in equation (2) 4. Use equations (10) and (11) to calculate  d and !nd
is used to control a process represented by equation (14), with Mpd and tsd previously defined.
the following closed-loop (negative feedback) transfer 5. Calculate Kc from equation (18) and Ti from equa-
function results in tion (19).
ðKc KÞ ðKc KÞ
þ
ðt 1 t 2 Þs ðTi t 1 t 2 Þ
GLC2 ðsÞ ¼ ðt 1 þt 2 Þ ðKc Kþ1Þ Kc K
(15)
s3 þ þ þ
ðt 1 t 2 Þs 2 ðt 1 t 2 Þs Ti t 1 t 2 Adaptive architecture
The pole placement method is used to choose the gains A PI controller that incorporates an STR explicit design
Kc and Ti 1 in the controller, described by equation (2). The structure5 to compute its gains was proposed by Paz Ramos
denominator in equation (15), containing the gains and the et al.7 This STR contains an estimator of the dynamic
parameters that describe the system, is manipulated model of the system. The controller gains are calculated
through changing the geometrical position of the poles in by incorporating the estimated system’s parameters, during
the root locus. Controller gains should be chosen keeping in the same iteration.
mind a desired pole placement; therefore proposing a The model selected for the estimation of parameters is
closed-loop polynomial denominator. an approximation of the structure of a continuous-time
A third pole must be added to the polynomial proposed, auto-regressive moving average (CARMA) controller of
as the denominator in equation (15) is of third order the form
DðsÞ ¼ ðs þ p1 Þðs 2 þ 2 d !nd s þ !nd 2 Þ (16) AY ðzÞ ¼ Bz1 U ðzÞ þ CðzÞ (20)
Paz et al. 5

belong to the models and the design specifications in con-

tinuous time. Nevertheless, the implementation of the
adaptive version has carried out in discrete time. As part
of the transformation from continuous to discrete time,
the correct choice of the sampling period is very impor-
tant. In this article, we used the design methodology for
emulation from the study by Franklin et al. 31 The
following relation is suggested for choosing the sampling
period T , when the process has been modeled with the
structure presented in equation (1)
5t 5t
<T < (23)
50 20
If a second-order internal model in the form of equation
Figure 1. Architecture of the PI controller with STR structure.
PI: proportional–integral; STR: self-tuning regulator.
(14) is available, a suitable recommendation is
t1 þ t2
T¼ (24)
where Y is the system’s output, U is the system’s input, 10
and  is a disturbance modeled as a random sequence of
The controller gains can be computed from the RLS
uncorrelated zero mean. The coefficients A, B, and C are
estimated parameters in a direct manner, by means of the
defined as
following analysis based on the method presented in “PI
A ¼ 1 þ a1 z1 þ a2 z2 þ . . . þ an zn controllers” section. The discrete time equivalent of a
B ¼ b0 þ b1 z1 þ b2 z2 þ . . . þ bn zn (21) first-order transfer function obtained through a zero-
order hold (ZOH) is
C ¼ c0  
GðsÞ
The parameters A, B, and C are estimated in each itera- GðzÞ ¼ ð1  z1 ÞZ (25)
s
tion. It is also assumed that A^ ! A, B^ ! B, and C^ ! C,
that is, the estimated parameters converge to the real ones and expanding produces
over time. Figure 1 shows the architecture of the b0
controller.7 GðzÞ ¼ (26)
z þ a1
The recursive least squares (RLS) algorithm with expo-
nential forgetting factor is used for the estimation, due to its where b0 and a1 can be estimated by the RLS method.
extensive presence in literature and well-known advantages Applying a ZOH to equation (1), to establish a direct
and disadvantages.7 The computation of the estimated relation between the coefficients in equation (1) (contin-
parameters is described through the equations uous time) and the coefficients in equation (26) (discrete
time), produces
Ka ðkÞ ¼ Pðk  1ÞðkÞðlI þ T ðkÞPðk  1ÞðkÞÞ1
^ðkÞ ¼ ^ðk  1Þ þ Ka ðkÞðyðkÞ  T ðkÞ^ðk  1ÞÞ Kð1  eT =t Þ
GðzÞ ¼ (27)
z  eT=t
PðkÞ ¼ ðI  Ka ðkÞT ðkÞÞPðk  1Þ=l
(22) from which it follows that

where Ka is the adaptation gain, P is the covariance matrix, b 0 ¼ Kð1  eT =t Þ (28)
I is the identity matrix, ^ is the vector estimation,  is the and
regressor, l is the exponential forgetting factor, and k is the
sample in discrete time. a 1 ¼ eT =t (29)
In order to carry out parametric estimation processes,
Equations (28) and (29) are solved for the time constant
different methods can be used; still, the use of an RLS can
t and the open-loop gain K
offer a theoretical backup from linear algebra, which allows
for a supervision of the estimation process, through its T
t¼ (30)
covariance matrix and its inverse.6,30 lnða1 Þ
b0
K¼ (31)
Computation of the control gains 1  eT =t

As it was previously observed in “PI controllers” section, K and t can be calculated directly from b^0 and a^1 , in
the PI gains can be calculated using the parameters that equation (21), assuming that b^0 ! b0 and a^1 ! a1 .
6 International Journal of Advanced Robotic Systems

In the case of second-order systems, it is valid to assume

that the discrete ZOH computed equivalent is
b0 z þ b1
GðzÞ ¼ (32)
z2 þ a1 z þ a2
Applying a ZOH into the overdamped second-order sys-
tem represented in equation (14), and following the same,
the coefficients of the discrete form described in equation
(32) can be expressed as
K
b0 ¼ ðt2  t1 þ t1 eT=t2  t2 eT =t1 Þ (33)
t2  t1
!
K t1 eT =t 2  t2 eT=t1  Figure 2. Implementation of the STR in a PLC using HIL and
b1 ¼ (34) OPC-based communication. STR: self-tuning regulator; PLC:
t2  t1 t eT =t 2 eT=t1 þ t eT =t2 eT =t1
1 2 programmable logic controller; HIL: hardware-in-the-loop; OPC:
OLE for process control.
a1 ¼ eT=t1  eT=t2 (35)

a 2 ¼ eT=t1 eT =t2 (36) the main reasons for this is the fact that other control tech-
niques require a lot more knowledge and analysis ability of
In a complementary manner, the computations for the the system that is being put in operation.
parameters in the continuous overdamped second-order In order to close this gap and come up with a better-
system are expressed by performing and more adaptive strategy, the system here
T proposed goes more for an automation in the design of the
t2 ¼ ffi!
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (37) controller that won’t depend in its accuracy on the ability of
a 21 1 a1
In    the operator.
24a 2 a2 2a 2 Implementation of this structure was thought in such a
way that it could be useful in an industrial setting, since no
T t2
t1 ¼ (38) additions of hardware have to be made to the PLC-based
t 2 Inða2 Þ þ T control loops. The controlling software just needs to be
b 0 ðt2  t1 Þ installed in the logical programmable control server, and
K¼ (39) linked to these, through the OPC communication protocol.
t2 ð1  e t2 Þ  t1 ð1
T=  eT =t 1 Þ
In order to test the performance of the controller, a
K, t 1 , and t2 can be computed directly from b^0 , b^1 , a^1 , testing protocol was designed, that was based on an auto-
and a^2 from equation (21), assuming that b^0 ! b0 , matized implementation of HIL, which allows the avail-
ability of information from a real plant, to perform an
b^1 ! b1 , a^1 ! a1 , and a^2 ! a2 . T is a defined constant
evaluation and an analysis of the functionality of the sys-
in all cases.
tem working in closed loop, in the presence of induced
Finally, the algorithm for the PI controller with STR
parametric disturbances.
structure can be summarized as
The methodology to compute adaptive PI controllers,7
1. Select the order of the internal model. explained in “Adaptive architecture” section, was tested
2. Choose the sampling period T . with systems that could be satisfactorily approximated to
3. Propose initial parameters. linear first- and second-order system models. These systems
4. Apply a finite duration step, in order to assist the were emulated with HIL. Figure 2 shows the experimental
convergence of the estimation process. implementation for the STR and PLC, PI controller imple-
5. ^ B,
Estimate A, ^ and C.^ mented using HIL, and the OPC communication scheme.
6. ^ ^ ^
From A, B, and C, obtain K and t if nA ¼ 1 or K, t1 , A PLC from Allen-Bradley (Rockwell Automation),
and t 2 if nA ¼ 2. Where nA is the order of the ControlLogix 5561 model, was used for PI controllers
polynomial. implementation. The estimation process and gain computa-
7. Compute Kc and Ti . tion were implemented as MATLAB scripts within Lab-
8. Generate the control signal uðkT Þ. VIEW virtual instrumentation platform, communicating
with the PLC to control the (emulated) systems. LabVIEW
handles the process of signal acquisition and the industrial
communication (based on the OPC standard) the NI OPC
Implementation Servers tool.22,23 The HIL emulation was performed by a
As it was discussed in the first section, PID control has a real-time hardware platform, the National Instruments
preponderant presence in industrial process control. One of CompactRIO 9074 with analog modules.
Paz et al. 7

The objective of a test protocol is that of offering ideal 11. Put the PLC in run mode.
test conditions to evaluate the performance of the adaptive 12. Execute the VI file.
PI and its fixed gains.
A set point was established in the test protocol as a The performance of both adaptive and conventional PI
square signal, with a variable amplitude, which period is controllers was measured through the IAE, defined as
longer than twice the settling time in an open loop. X
n
The following step is to perform a connection between IAE ¼ j SPk  PVk j (40)
industrial platforms. The first connection is to be performed k¼0
between the PLC and the Personal Computer (PC) through
an OPC, with its ethernet communication port. The second where SP is the set point and PV is the process variable.
connection goes between the PLC and the PAC, through an Eighteen systems were tested in total, half first-order
analog output and analog input. and half second-order systems, which all of them emulated
Once all devices have been connected, the PI adap- by HIL. The systems were programmed as difference equa-
tive program of the fixed PI gains is carried out; after tions resulting from a ZOH discretization process over their
three complete cycles, the parametric disturbance can be transfer functions.
performed. First-order systems were formulated differently from
The OPC standard was developed by industrial manu- equation (1) as
facturers in order to have interoperability between different K
G1 ðsÞ ¼ (41)
automation vendors. The main objective of OPC-based ðt þ Þs þ 1
communication is the real-time interaction between differ-
ent devices.32 The procedure to get connectivity between using the following values: K ¼ f0:1; 0:5; 0:8g,
the control algorithm platform and the PLC, using the OPC t ¼ f2; 10; 20g, and the disturbance  ¼ f4; 10; 15g. The
standard, can be summarized in two steps: handling inputs/ combination of the three different values of K and t þ 
outputs and the manipulation of variables. To carry out the forms the first nine systems for the experiments.
implementation of the control system described on an Second-order systems were formulated as
industrial platform with the OPC communication standard, K
the following algorithms must be used. G2 ðsÞ ¼ (42)
½ðt1 þ  1 Þs þ 1½ðt2 þ  2 Þs þ 1
a. Operation of the inputs and outputs of the PLC: with values K ¼ f0:1; 0:5; 0:8g, t1 ¼ f4; 10; 20g,
t2 ¼ f10; 20; 30g, and the disturbances  1 ¼ f3; 5; 0g and
1. Set up the protocol of communication in the
 2 ¼ f0; 0; 10g. The combination of the three different
PC to the PLC with RSLinx.
values of K, t1 þ 1 , and t2 þ  2 forms the remaining
2. Register the racks of the PLC with RSLogix
systems for the experiments.
5000.
An advantage in the use of HIL is the repeatability
3. Release the ladder to the PLC with RSLogix.
of tests and the ability of testing the controller in
4. Shut down RSLinx.
similar conditions to those that can be found an
b. Handling variables in LabVIEW via the OPC industrial setting.
communication standard33: HIL allows the possibility of dynamics that go beyond
simple computational simulations, since the information
1.
Establish the communication port between that is shared between the controller and the plant is not
the device and the PC in the OPC server in simply transferred from one register to another in a given
NI OPC (developed by National Instruments) iteration; it is rather transmitted as analogical information
Servers software. between the analog–digital converter of the PLC and the
2. Include in NI OPC Servers the PLC tags. analog–digital port of the PAC.
3. Make the project in LabVIEW. It is also important to mention that the sampling
4. Select the OPC client in LabVIEW in the new periods of the different stages are not equivalent, not
server Input/Output (I/O). even in the analogical input module (1769-IF4 de
5. Configure the OPC client by selecting NI Allen-Bradley) of the PLC which has a band-pass filter
OPC Servers in the LabVIEW project. with the following frequencies: 50 Hz, 60 Hz, 250 Hz y
6. Create the bound variables. 500 Hz, with a magnitude fall of 3 dB, which can
7. Incorporate the variables of the OPC Server. be interpreted as an addition of colored noise to the
8. Allow the reading and writing for the bound closed loop.
variables. This way, the implementation here proposed is more
9. Aggregate a VI file. efficient in several aspects to computational simulation,
10. Program the VI in the block diagram handling and in itself is a test structure that can be utilized with other
the variables. control strategies.
8 International Journal of Advanced Robotic Systems

Table 1. Integral absolute error evaluation before and after the Table 2. Integral absolute error evaluation before and after the
disturbance for closed-loop first-order systems. disturbance for closed-loop second-order systems.

P
d1 P
n dP
1 P
n
jek j jek j jek j jek j
k¼0 k¼d k¼0 k¼d

Parameters Parameters
K; t;  PI PI PI PI K; t 1 ; t2 ; 1 ;  2 PI PI PI PI
STR STR STR STR
0.1, 2, 4 19.37 18.85 51.23 25.52 0.1, 4, 10, 3, 0 6.41 8.18 9.43 7.51
0.5, 2, 4 20.98 19.67 46.61 21.25 0.5, 4, 10, 3, 0 6.44 8.02 9.41 7.56
0.8, 2, 4 22.88 20.62 52.77 21.75 0.8, 4, 10, 3,0 6.43 7.91 9.50 7.97
0.1, 10, 10 28.41 28.32 55.18 33.27 0.1, 10, 20, 5, 0 23.27 11.23 13.47 11.72
0.5, 10, 10 28.92 28.34 52.97 33.07 0.5, 10, 20, 5, 0 23.29 11.03 13.55 11.43
0.8, 10, 10 29.42 28.59 55.65 33.41 0.8, 10, 20, 5, 0 23.04 11.30 13.49 11.69
0.1, 20, 15 10.45 11.43 15.29 12.37 0.1, 20, 30, 0, 10 10.90 10.64 12.06 10.47
0.5, 20, 15 10.45 11.39 17.78 12.38 0.5, 20, 30, 0, 10 10.89 10.47 12.04 10.44
0.8, 20, 15 10.48 11.38 17.89 14.30 0.8, 20, 30, 0, 10 10.89 10.41 12.08 10.68
PI: proportional–integral; STR: self-tuning regulator. PI: proportional–integral; STR: self-tuning regulator.

Table 3. Integral absolute error evaluation for closed-loop first

Experimental results order systems.
The implemented performance evaluation of PI control-
lers with an STR structure consisted of comparing their P
n
jek j
dynamic parameters (overshoot and settling time) cho- k¼0
sen at design (“desired” dynamic behavior by specifying
Mpd and tsd ), against the ones of conventional PI con- Parameters
K; t;  PI PI
trollers, when following a reference signal. A distur-
bance is introduced in one of their parameters, at a STR
specific time. Conventional PI controllers follow the 0.1, 2, 4 70.61 44.38
design presented in “PI controllers” section. PI control- 0.5, 2, 4 67.59 40.94
0.8, 2, 4 75.66 42.38
ler with an STR structure was implemented as explained
0.1, 10, 10 83.59 61.60
in “Adaptive architecture” section. 0.5, 10, 10 81.89 61.42
The expected consequence of the disturbance in the 0.8, 10, 10 85.08 62.01
systems controlled by a conventional PI is an increment 0.1, 20, 15 25.69 23.80
in the IAE and a change in the response time. In the case of 0.5, 20, 15 28.24 23.78
the PI with STR architecture controllers, the IAE is also 0.8, 20, 15 28.38 25.69
expected to increase but in smaller magnitudes than in the PI: proportional–integral; STR: self-tuning regulator.
conventional controllers. The IAE was computed in three
settings: before the disturbance (at the time interval
(0, d1)), after the disturbance (at the time interval l ¼ 0:985. For the conventional PI controller, with
(d, n)), and throughout the experiment (time (0, n)). K ¼ 0:8, the values of the controller gains were
Tables 1 and 2 show the IAE at the intervals of time before Kc ¼ 1:7330 and Ti ¼ 7:9635. The disturbance appeared
and after the disturbance, for the first- and second-order at 960 s.
systems, respectively. Tables 3 and 4 show the IAE through- Figures 3 and 4 show the behavior of the estimations
out the experiment, for the first- and second-order systems, over time. A small estimation error is observed. When
respectively. The results show that the magnitude of the the disturbance is triggered, the gains in the controller
IAE is smaller for the adaptive controller, compared to are also modified. This allows for a better approxima-
the conventional one, in most cases. tion to the reference signal, in closed loop, by modifying
The rest of this section shows an illustrative example the control signal.
using one of the second-order systems with parameters of The performance of a PI controller with an STR
t1 ¼ 4, t2 ¼ 10,  1 ¼ 3,  2 ¼ 0, and K ¼ 0:8 based on structure is shown in Figure 5(a). The internal structure
equation (42). The desired values for the response were of of the estimator computes a second-order CARMA
Mpd ¼ 10% and tsd ¼ 45 s. The selected sampling period difference equation. The output signal PI-STR is close
for the discretization was 2 s. For the estimation, the initi- to one chosen in design (desired response), according to
alization values were P0 ¼ 5000I,  0 ¼ ½0; 0; 0; 0, and the settling time.
Paz et al. 9

Table 4. Integral absolute error evaluation for closed-loop

second-order systems.

P
n
jek j
k¼0

Parameters
K; t 1 ; t 2 ; 1 ;  2 PI PI
STR
0.1, 4, 10, 3, 0 15.85 15.70
0.5, 4, 10, 3, 0 15.86 15.59
0.8, 4, 10, 3, 0 15.93 15.88
0.1, 10, 20, 5, 0 36.74 22.95 Figure 5. (a) Performance of PI controller with an STR structure
0.5, 10, 20, 5, 0 36.84 22.47 and (b) performance of PI controller. PI: proportional–integral;
0.8, 10, 20, 5, 0 36.54 22.99 STR: self-tuning regulator.
0.1, 20, 30, 0, 10 22.96 21.12
0.5, 20, 30, 0, 10 22.93 20.92
0.8, 20, 30, 0, 10 22.98 21.09 while the gains of the controller remained constant. After
PI: proportional–integral; STR: self-tuning regulator. the disturbance, the controller generates a signal with more
overshoot and even more settling time (longer time than
desired), thus the desired performance criteria is not
reached anymore.

Conclusion
In this article, an implementation of an adaptive (STR) PI
controller architecture was presented and evaluated. The
STR structure allows to adapt the controller parameters to
accomplish a good performance according to the settling
time desired and adjust in order to face changes in system
parameters. The implementation involves the use of avail-
Figure 3. Estimation of parameters a 1 and a 2 for a second-order able hardware platforms (PLCs), connected to a PC to
system. perform the adaptive computations, and communicating
through the OPC standard. This allows a more feasible
implementation of modern control techniques in different
applications in process control, manufacturing, and
robotics, employing general use hardware platforms
instead of application-dependent equipment.
The performance of the implemented adaptive PI
controller architecture was compared against conventional
PI controllers, also implemented in the same hardware. The
gains of the conventional PI controllers were tuned by pole
assignment, from a proposed desired dynamic response.
The adaptive PI controllers employ estimation of system’s
parameters to compute the gains. This experimental setup
Figure 4. Estimation of parameters b 0 and b 1 for a second-order served two main purposes: demonstrating the feasibility of
system. controlling systems in OPC-based architectures, teamed
with industrial hardware platforms (e.g. PLCs); and
Figure 5(b) shows the performance of a conventional PI demonstrating the performance of adaptive PI controllers
controller in closed loop, which conforms to the parameters against conventional PI ones, in the presence of distur-
specified in the design stage (desired behavior) before the bances. Nineteen controllers were designed and implemen-
disturbance occurs. The disturbance causes deterioration in ted, for the first- and second-order systems emulated by
the output PI when tracking the reference. The observed HIL. Their dynamic performance was evaluated, by mea-
change in the systems response when the disturbance suring the IAE with respect to the desired behavior from
appears is due to a migration of the poles of the closed- design, in the presence of a disturbance. Adaptive PI
loop system (i.e. the systems parameters have changed), controllers performed better than the conventional ones.
10 International Journal of Advanced Robotic Systems

Declaration of conflicting interests 14. Desborough L, Nordh P, and Miller R. Control system relia-
The author(s) declared no potential conflicts of interest with bility: process out of control. In: InTech, 2001, pp. 52–55.
respect to the research, authorship, and/or publication of this https://ww2.isa.org/intech/; http://www.loopscout.com/Info/
article. Intech.pdf
15. Anbarasi S and Muralidharan S. Enhancing the transient
Funding performances and stability of AVR system with BFOA tuned
The author(s) disclosed receipt of the following financial support PID controller. J Control Eng Appl Inform 2016; 18(1):
for the research, authorship and/or publication of this article: The 20–29.
author(s) received financial support of Polytechnic University of 16. Creus A. Instrumentacin industrial. Instrumentacion, Spain:
Aguascalientes for the publication of this article. Marcombo, 2005.
17. Kocian J and Koziorek J. An outline of advanced process
References control and self-tuning techniques on PLC background. In:
1. Astrom KJ and Hagglund T. PID controllers: Theory, design Conference on emerging technologies and factory automa-
and tuning. United States of America: ISA, 1995. tion (ETFA), Toulouse, France, 5–9 September 2011, pp. 1–8.
2. Bennet S. Development of the PID controller. IEEE Control IEEE. https://www.ieee.org/
Syst Magaz 1993; 13(6): 58–65. 18. Vijayalakshmi S, Manamalli D, and PalaniKumar G. Closed
3. O’Dwyer A. PI and PID controller tuning rules: an overview loop experimental validation of linear parameter varying
and personal perspective. In: Proceedings of the IET Irish model with adaptive PI controller for conical tank system.
Signals and Systems Conference, Dublin Institute of Technol- J Control Eng Appl Inform 2014; 16(4): 12–19.
ogy, Dublin, Ireland, June 2006, pp. 161–166. The Institution 19. Mendes J, Araujo R, Sousa P, et al. An architecture for adap-
of Engineering and Technology. http://www.theiet.org/ tive fuzzy control in industrial environments. Comput Indus
4. Astrom K. Adaptive control around 1960. IEEE Control Syst 2011; 62(3): 364–373.
1996; 16(3): 44–49. 20. Mahnke W and Leitner S. OPC unified architecture.
5. Astrom K and Wittermark B. Adaptive control. Boston: Addi- Springer, 2009.
son-Wesley, 1995. 21. Pendli P, Gorbatchev V, Schwarz M, et al. OPC and its
6. Wellstead P and Zarrop M. Self-tuning systems: Control and strategies for redundancy. In: Proceeding of the 10th WSEAS
signal processing. England: Wiley, 1991. international conference on communications, Athens,
7. Paz Ramos M, Garibo Esquivel S, DeLuna-Ortega CA, et al. Greece, 13–15 July 2006, pp. 163–168. World Scientific and
PI controller with dynamic gains calculation to comply time Engineering Academy and Society. www.wseas.org/
specs in presence of parametric disturbance. In: 8th Interna- 22. Paz Ramos M, Avila Alonso JL, Garcı́a Barberi SA, et al.
tional conference on electrical engineering computing Sintonización de controladores PID vı́a OPC. In: 6to
science and automatic control, Toluca, México, 10–13 Congreso Internacional sobre Innovación y Desarrollo
November 2009, pp. 322–327. IEEE. https://www.ieee.org/ Tecnológico CIINDET 2008, Cuernavaca, México, 8–10
8. Ponce Cruz P, Paz Ramos M, Fernández del Busto R, October 2008. IEEE Morelos. http://ieeemorelos.org/
et al. DTC scheme for an induction motor using general- index.html/
ized minimum variance controller with dynamic pole 23. Paz Ramos M, Velázquez Alvarez E, and Valencia Maza I.
assignment. In: Proceedings of the IEEE international Detector de algoritmos de control PID vı́a OPC. In: 6to
symposium on industrial electronics, Dubrovnik, Croatia, Congreso Internacional sobre Innovación y Desarrollo
20–23 June 2005, pp. 957–962. IEEE. https://www.ieee. Tecnológico CIINDET 2008, Cuernavaca, México, 8–10
org/. October 2008. IEEE Morelos. http://ieeemorelos.org/
9. Sharifi M, Behzadipour S, and Vossoughi G. Model reference index.html
adaptive impedance control in Cartesian coordinates for 24. MacNeal R and Belkhayat M. Standard tools for hardware-in-
physical human robot interaction. Adv Robot 2014; 28(19): the-loop (HIL) modeling and simulation. In: Electric ship
1277–1290. technologies symposium, Arlington, VA, USA, 21–23 May
10. Zhang D and Wei B. Design, analysis and modelling of a 2007, pp. 130–137. IEEE. https://www.ieee.org/
hybrid controller for serial robotic manipulators. Robotica 25. Munteanu I, Bratcu AI, Bacha S, et al. Hardware-in-the-loop-
2017; 35(1): 1888–1905. based simulator for a class of variable-speed wind energy
11. Tran T, Ge S, and Hei W. Adaptive control for an uncertain conversion system: design and performance assessments.
robotic manipulator with input saturations. Control Theory IEEE Trans Energy Convers 2010; 25(2): 564–576.
Technol 2016; 14(2): 113–121. 26. Martin A and Emami MR. Dynamic load emulation in hard-
12. Takatsy H and Itoh T. Future needs for control theory in ware-in-the-loop simulation of robot manipulators. IEEE
industry report of the control technology survey in Japanese Trans Indus Elect 2011; 58(7): 2980–2987.
industry. IEEE Trans Control Sys Technol 1999; 7(3): 27. Shinsky FG. Process control systems application, design, and
298–305. tuning. United States of America: McGraw-Hill, 1998.
13. Rhinehart R. The century’s greatest contributions to control 28. Bissel C. Control engineering. England: Chapman and Hall,
practice. ISA Trans 2000; 39(1): 3–13. 1994.
Paz et al. 11

29. Paz MA, Ramirez-delreal T, Garibo S, et al. Algoritmo didáctico 31. Franklin GF, Poweill JD and Workman M. Digital control of
para la asignación implı́cita de polos mediante la especificación dynamic systems. Addison-Wesley, 1998.
de caracterı́sticas temporales deseadas usando controladores PI 32. Santori M. OPC: OLE for process control. Real-Time Magaz
industriales. Formacio´n Universitaria 2013; 6(6): 23–40. 1997; 97(4): 78–81.
30. Isermann R and Lachmann K. Parameter-adaptive control 33. United States of America, National-Instruments. Help con-
with configuration aids and supervision functions. Automa- tents de NI OPC servers, 2009. http://www.ni.com/tutorial/
tica 1995; 21(6): 625–638. 4626/en/