Int. J. Engng Ed. Vol. 19, No. 4, pp. 586±592, 2003 0949-149X/91 $3.00+0.

00
Printed in Great Britain. # 2003 TEMPUS Publications.

OPC-driven Data Exchange between

MATLAB and PLC-controlled System*
STOJAN PERSIN, BORIS TOVORNIK and NENAD MUSKINJA
Faculty of Electrical Engineering and Computer Science, University of Maribor, Smetanova 17,
2000 Maribor, Slovenia. E-mail: stojan.persin@uni-mb.si
Many manufacturing, and mechatronics systems, are controlled by PLCs and SCADA systems.
These systems provide very limited possibilities for research, optimisation, testing and simulation.
The main advantage of this proposed solution, which uses OPC, is an integration of the numerical,
computational and graphical strengths of MATLAB1 with the robustness of industrial systems.
The experimental setup, the processing of the data and the results are presented with particular
emphasis on student use in a laboratory environment but the techniques employed in the
experiments will likely to be seen by the students in their subsequent employment after completion
of their college careers.

INTRODUCTION examined and optimised: regulation loops, optimal

production structuring, calculation of parameters
INDUSTRIAL AUTOMATION has reached the at various points, etc. The equipment presently
point where one cannot progress further without used in the industry, such as PLC and SCADA,
applying industrially tested, robust and market- is unsuitable for performing such data processing.
accessible automation products such as sensors, Instead, Matlab would provide much better results
actuators, programmable logic controllers (PLC) [6] if only it were possible to feed it with real-time
and supervisory control and data acquisition data from the manufacturing process. By applying
(SCADA) systems [1] which enable the acquisition this technology a wide spectrum of possibilities for
of large quantities of data. Process automation is testing and analysis emerges which can contribute
particularly important in isolated areas where the to the improvement of the whole process. This
physical presence of humans is undesirable, as for system can also be upgraded to perform a real-
example in the production of electronic compo- time parallel optimisation procedure. In this
nents [2] or in the pharmaceutical industry. Such configuration the manufacturing process would
manufacturing systems operate with various be running independently whereas Matlab would
degrees of success and are rarely modified [3]. be performing in parallel with all the necessary
Every alternation within any industrial process is mathematical operations, the results of which
followed by extensive documenting, testing and would be fed back into the process. There are
validating of the new system, sometimes even by two difficulties in implementing such real-time
certified authorities such as ISO or FDA. More- connections: the problem of connectivity, and the
over, in the case of modification quite often difficulty of presenting a sufficient universal solu-
production must be stopped entirely which gives tion that would not require altering our existing
rise to substantial costs for the entire operation [4]. manufacturing systems, especially when new
On the other hand there are a number of research requires updating or changing our
research laboratories where various approaches existing configurations.
are constantly the subject of change, development One possible solution would be to use the OPC
and testing in order to bring about improvements. (OLE for process control) standard [7] and the
This research work often implements mathemati- associated technology. This article presents a solu-
cal and simulation software such as Matlab tion by showing how to use an OPC client in a
(Matrix Laboratory) which enables the solving of Matlab environment. The former is used to
complex mathematical operations, simulations, exchange data as shown in Fig. 1. The OPC is
optimisations, etc. Their main disadvantage, being increasingly used [1, 8±10] which lends our
however, is that they are substantially limited solution the required level of universality and
in their connectivity to other systems and relevance to engineering education.
applications [5]. The following sections contain an outline of
Automated processes include many procedures some of the most frequently used techniques in
where large amounts of data could be processed, the real-time mathematical analysis of process
data, a summary of the main properties of the
OPC, and a plan and an implementation of
* Accepted 22 August 2002. the connection between Matlab and a PLC. In

586
 OPC-driven Data Exchange between Matlab and PLC-controlled System 587

applications which offer data exchange rather

than actual data sampling. These measurements
are continuously performed anyway for the
purposes of PLC and SCADA, which means that
they do not need to be further duplicated. The
simplest data exchange can be implemented
through an ASCII file which is fed by both
SCADA and Matlab. To a limited extent, this
application is universal, although some modifica-
tion of the SCADA programme is necessary, and
Fig. 1. The connection between Matlab and PLC using OPC. the data transfer is also slow.
By using OLE (object linking and embedding) it
is possible to connect Matlab to Microsoft Excel
addition, there is also performance testing of OPC which is available through the Excel Link toolbox
and descriptions of two examples for use in the in Matlab. Excel fully supports DDE (dynamic
education process. The final section contains an data exchange) which is also supported by
analysis and a summary of the main properties of SCADA systems and hence represents a solution
our approach. that is one of the most universal. However, DDE
never obtains a stronghold in the process control
since it is inefficient and unreliable in transferring
CURRENT SOLUTIONS FOR REAL-TIME large quantities of data [13]. Another difficulty in
ANALYSIS OF PROCESS DATA using DDE is that it is necessary to alter the
SCADA system used for sending DDE data,
The application of user-defined programmes assuming that DDE is not originally supported
in one of the higher programming languages by the SCADA.
available with the majority of SCADA solutions The application of OPC (OLE for process
appears, at first sight, to be quite a simple solution control) for data exchange between Matlab and
for additional mathematical analysis. Difficulties PLC represents a logical continuation in the devel-
only arise when complicated mathematical opment of the above techniques and is described in
operations must be programmed. detail in the following section.
One of the more common solutions for feeding
real-time data into Matlab is the use of special PC
I/O modules [11, 12], supported by Matlab, for EXPERIMENTS FOR OPC PERFORMANCE
example those by Burr-Brown or National Instru- TESTING
ments. Apart from product integration, the main
advantage of such configuration is the speed of The Matlab mathematical software and the
data sampling which can be faster than 1 ms per Simulink package were used in creating test
piece of data. Such systems, however, are purely environment for the experiment, as shown in
laboratory systems and are rarely used in indus- Fig. 2. The purpose of the experiment was to
trial automation because every hardware purchase accomplish an online transmission of process
is unavoidably assessed by its integration difficul- data from a PLC (Fig. 3) into Matlab. In the
ties in the manufacturing process. Every such majority of industrial environments PLCs and
interface card must be physically connected to SCADA systems are the only acceptable solution.
the existing sensory equipment by a parallel or However, the Matlab package with Simulink is a
serial linkage to the PLC. This presents the very powerful and efficient mathematical tool
problems already mentioned above, namely the which offers much better data analysis than any
costs that arise during production standstill, docu- SCADA system. Therefore, it appears sensible to
menting and validating the new system. A similar include it in the system of process control.
solution is offered by using measurement instru- Siemens S7±414 PLC was used in the experi-
ments with the GPIB and VISA communication ment, although the whole system was also tested
options which is also supported by Matlab. with General Electric, OPTO 22, Honeywell,
In Matlab it is also possible to access serial bus Omron and Allen-Bradley PLCs. The aim was to
which enables the construction of APIs (applica- read and write online process data from the PLC
tion programming interface) which listen to the into Matlab. PLC was connected to a computer
traffic on the PLC network and, if necessary, through MPI/RS 232 (Multi Point Interface) with
return some data. The main advantage of this a maximum transfer rate of 19.2 kbit/s. Even better
solution is that it does not directly interfere with results could be achieved by using a Profibus which
the process itself. Its main disadvantage is that the would yield significantly higher rates of data
construction of such an interface for an industrial transmission. The data was acquired using the
bus is costly in time and the solution is not OPC Server. It was then necessary to create a
universal [6]. communication channel, to select a device where
The development of suitable solutions leans data could be captured and to define the group of
more and more in the direction of software variables to be monitored. The OPC client for
588 S. Persin et al.

Fig. 2. A part of our testing environment in Matlab.

Fig. 3. Testing environment for the experiment.

Fig. 4. Matlab commands for communication with one input and one output.

Matlab was used which enabled communication order to configure this package and Fig. 4 contains
between the OPC and Matlab. This is a collection an example for a communication with one input
of additional commands which are available in the and one output.
Matlab environment and enable connection to any When testing the data transfer rate from a
OPC server. It is necessary to enter a few physical device into Matlab, a continuous
commands into the Matlab command window in sinus signal was used as shown in Fig. 5 where
 OPC-driven Data Exchange between Matlab and PLC-controlled System 589

of data. The testing in a closed loop PLC-Matlab-

PLC was performed by altering `sleep' time in
Matlab, as well as the load on the computer
being changed by executing various processes. It
appeared that the highest load for the system was
image manipulation. The processing of data in
Matlab was executed in fixed time intervals and
the relationship between completed and uncom-
pleted cycles was monitored. If a cycle was incom-
plete and hence a data transfer was unsuccessful, it
was assumed the computer was busy with other
processes and, as a result, it was overloaded. It is
important to emphasise at this point that even
industrial systems for process automation do not
Fig. 5. Signals for testing the data transfer rate. show 100% reliability in communication, and data
transfer failures can occur during their normal
operation. This problem is solved by re-requesting
discreteness as a result of sampling was also notice- the data for which transfer has failed (retry).
able. All measurements and tests were performed The test was performed at 20 ms and 60 ms time
under the most unfavourable conditions for the intervals between executions of data processing
entire system which yielded data on its marginal cycles in Matlab, with various loads on the operat-
capabilities. The server captured data from the ing system (OS), and by processing eleven pieces of
controller, checked its quality and evaluated it actual process data. The results presented in Table
for a change of state. It then forwarded it to the 1 show that with execution intervals of 60 ms and
client which used the same route to return the data more the communication reliability achieves an
once it was processed. Synchronous reading was acceptability rate which would satisfy industrial
used for the `worst case' scenario, where each piece environments. The speed of data transfer does not
of data was changed every time and the same represent any particular problem since in the
computer ran a server, two clients, as well as majority of industrial processes the required data
SCADA and Matlab software. Figure 6 shows transfer rate from PLC to the computer is in the
the signals in detail, whereby a is the sinus signal, range of seconds.
b is the server data and c is the client data. The The third part of the experiment was done in the
total transfer time is represented by t3 whereas the same way as the second one, with a closed loop
initial or the fixed part of the transfer time is given PLC-Matlab-PLC. The total time of data transfer
by the constant t1 ˆ 300 ms which is the shortest (t3 in Fig. 6) was measured because the speed of
time needed for continuous refreshing of our OPC. Matlab was known. It was concluded that its
The difference between t1 and the measured time is average was 720 ms which is still within the accep-
then due to the server load which was on average table interval for supervisory control of industrial
t2 ˆ 11 ms for 11 signals, i.e. 1 s for approximately processes.
1000 data. The results obtained are fully compliant In the fourth part of our experiment an alter-
with the specifications of the OPC standard as native method was sought for transferring data
well as with observations made in other experi- into Matlab. There exist a possibility of transfer-
ments performed on industrial SCADA systems [8, ring the data through the KEPServerEx interface
10, 14]. to Microsoft Excel where Matlab can access it in
The second part of the test was to establish real time using the Excel link which is, similarly to
communication reliability by rating the relation- the OPC Client, a group of additional commands
ship between successful and unsuccessful transfers for Matlab. A system as shown in Fig. 7 was
constructed and data transfer rates were measured.
It became apparent that data travels from the
controller to Matlab faster than in the opposite
direction. The reason was in the macro used to
write data to the OPC server or the controller. The
connection was also very susceptible to the

Table 1. Results from testing the execution of a Matlab

applicationÐthe percentage of incompleted cycles.

Cycle execution in Matlab

20 ms 60 ms

Unloaded OS 0.12% 0.02%

Loaded OS 1.69% 0.61%
Fig. 6. Signals in detail.
590 S. Persin et al.

Fig. 7. Interfacing Matlab and Excel.

computer load since data transfer rates ranged of mechatronics and electrical engineering are
from 600 to 1200 ms. Furthermore, the computer taught in this course.
load had a higher influence on reading data into The process, which is connected to the PLC, is
Matlab rather than writing it into the PLC. composed of a flow meter and an electrically-
Finally, the experiment was performed under driven control valve. The overall system is shown
conditions for which the proposed solution is the in Fig. 3 and was described in a previous section.
most suitable: a transfer of data into Matlab for The aim is to control the mass-flow and a PID
simulations, optimisations and testing. In all these algorithm was chosen because it represents simple
processes data had already been acquired by a classical control which can also be `upgraded'. In
SCADA system (see Fig. 2). Thus, a standard addition, at this point in the educational process,
industrial automation system, but now with a the students have had significant exposure to this
link to Matlab, was used. It is important to type of controller and should be relatively comfor-
emphasise that in the described solution there is table in applying it to the mass-flow control prob-
no time delay between the data in Matlab and that lem. Several tests were performed during the
in the SCADA system since both of them are course of this study. However, for brevity, only
connected to the same OPC server. Furthermore salient results are reported. Students began a study
no delays in the operation of the system as a whole of the process in detail and observe the nonlinear
were noticed due to the additional OPC Client. steady-state characteristic (Fig. 8) which was
This option can, therefore, be used for educational, obtained using Matlab and OPC connection.
experimental and research purposes through Then a PID controller was designed in Matlab
having all the advantages of reliable and verified and the controller parameters were downloaded
industrial solutions. directly into PLC. As shown in Fig. 9 the control
was not equally appropriate over all operating
points which indicates a need for gain scheduling
control. Gain scheduling is then a two-step proce-
EXAMPLES OF USING OPC IN A dure where firstly, local linear controllers are
TEACHING PROCESS designed based on linearisation of the nonlinear
system at several different operating points and
The aim of the first example is the implementa- then a global nonlinear controller for the nonlinear
tion of a gain scheduling and adaptive control system is obtained by interpolating or scheduling
system for use in undergraduate engineering among the local operations point design. The
education. The system was developed to demon- combination of PLC, OPC and Matlab was used
strate the benefits of using a connection between again, and Fig. 10 represents a control using gain
Matlab and PLC and this is what senior students scheduling PID. For advanced students there is

Fig. 8. A nonlinear steady-state characteristic of the process. Fig. 9. Classical PID control.
 OPC-driven Data Exchange between Matlab and PLC-controlled System 591

Fig. 10. PID control with gain-scheduling.

also a possibility to design and test an adaptive was changed every time and the same computer
PID controller using on-line identification proce- ran a server, two clients, as well as the SCADA and
dure in Matlab. We have to remember that this Matlab software. The measurements performed
process is controlled by PLC and realisation of show that, due to the fixed dead time, the worst
identification in PLC-code would be very difficult. results are obtained when transferring a single
Our second example of using OPC is also suit- piece of data only. A rate of approximately 1000
able for mechatronic educational purposes. We data per second can be expected with large quant-
have a six-axis Kuka robot at our faculty. The ities of data. The process data is available to
whole system is not very open and the students Matlab virtually at the same time as to the
have found experimenting, evaluating results and SCADA system which makes the application suit-
recording the data quite difficult. On the other able for industrial environments.
hand, there is a vendor-made OPC Kuka Server Two examples for use in undergraduate engin-
available. On-line connection between the robot eering education were also described. The first
and the Matlab could make these things easier. example describes a procedure where the OPC
Furthermore, in Matlab there is also on-line and connection between PLC and Matlab was used in
off-line analysis available for comparing the the educational process for teaching control. The
results, optimisation, etc. Moreover, we are real process was connected to the PLC and
seriously considering the feasibility of adding SCADA system which is quite common in indus-
some additional sensors (acceleration, force . . . ) trial practice. The OPC link to the Matlab was
and connecting these sensors via different PLCs then established and used for several purposes: to
and OPC servers onto the same Matlab/Simulink record the nonlinear steady-state characteristic of
window. In this case the amount of information the process, to download controller parameters
about the whole system would increase signifi- into PLC and also for On-line identification of
cantly, and there is also a possibility for OPC to the process. The second example describes a
become a universal platform, as was proposed by proposal for how to use the OPC connection
Mintchell [15]. between Matlab and Kuka robot to make
students' workload easier. The use of OPC in
laboratory experiments has generated many posi-
CONCLUSIONS tive results. Student reaction to the experiments
has been very good and interest has been increased.
In this paper a solution was proposed on how to The students seem to appreciate the `feel' that they
connect an existing system to the Matlab environ- gain from the laboratory experiments, as opposed
ment by using the OPC. Such a connection can be to a computer simulation.
used for monitoring industrial process data which The proposed solution demonstrates tech-
can then be used for the analyses and optimisation nologies that are just gaining widespread support
of procedures, early fault detection and diagnosis, today in industry and the examples illustrate that
further data processing or data documentation. OPC can form a flexible universal data platform
The proposed solution is particularly suitable for for factory floor automation. The techniques
systems where the industrial process must remain employed in these laboratory experiments will
unchanged, or must not be stopped. likely to be seen by the students in their subsequent
The performance testing system consisted of an employment after completion of their college
OPC server which captured data from the control- careers. The quality of graduates is improving
ler, it checked the data's quality, evaluated it for a and the industry is receiving engineers trained in
change of state, and then forwarded it to the OPC solving real-world problems. Our goal is to reduce
client who returned it along the same route once it the training period for new engineers when they
was processed. Synchronous reading was used for start work and we believe that our paper is a
the `worst case' scenario, where each piece of data contribution to this aim.
592 S. Persin et al.

REFERENCES

1. M. Janke, OPC-plug and play integration to legacy systems, Pulp and Paper Industry Technical
Conference, Atlanta, USA, 228±295 (2000)
2. Mitsugu Kishimoto, Optimized operations by extended X-Factor theory including unit hours
concept, IEEE Transactions on Semiconductor Manufacturing, 14(3), 2001.
3. T. G. Kirner Quality requirements for real-time safety-critical systems, Control Engineering
Practice, 5, 1997, pp. 965±973.
4. W. David Coit, Tongdam Jin, Prioritizing System-reliability prediction improvements, IEEE
Transactions on Reliability, 50(1), 2001.
5. Linda Wills et al., An open platform for reconfigurable control, IEEE Control Systems Magazine,
21(3), 2001.
6. J. Rodgerson, Teaching systems and control using Matlab, Int. J. Electical Eng Educ., Manchester
Univ. Press, Manchester, 6, 1992.
7. OPC Foudation, Data Access Automation Interface Standard 2.04, OPC Foundation (2001).
8. Wu Sitao, Using device driver software in SCADA systems, Power Engineering Society Winter
Meeting, Singapore (2000).
9. P. Studebaker. ALL For OneÐThe Top 10 Trends in Process Control. Control Magazine 1, (1998).
10. Al Chisholm, Intelluton Inc. OPC/OLE for Process Control Overview, World Batch Forum (1998).
www.intellution.com/opchub/opc_wbfoverview.asp
11. M. Stianko, Merz. OPC Server according to the Data Access 1.0 and 2.0 Specification: Technical
Descripton, Czech Avzomatizace Magazine, 11 (2000). www.merz-sw.com/articles/opc_da.php3
12. Mohamed Am, et al., Real-time implementation of a robust H-infinity controller for a 2-DOF
magnetic micro levitation positioner, Journal of Dynamic Aystems Measurement and Control
Transactions of the ASME, 12, 1995.
13. Al Chisholm., A Technical Overview of the OPC DataAccess Interfaces, Intellution Inc. (1998).
14. Frank Iwanitz, Jurgen Lange, OLE for Process Control, Huthig Gmbh Heidelberg, Germany
(2001).
15. G. A. Mintchell, OPC integrates the factory floor, Control Engineering, 1, 2001.

Stojan Persin received the B.Sc. degree in electrical engineering from the University of
Maribor, Slovenia, in 1995. From 1995 to 1999, he was with Metronik, Ljubljana, Slovenia
and was a senior engineer for industrial automation systems and afterward a manager at the
branch office. Since December 1999, he has been with Faculty of Electrical Engineering and
Computer Science, Maribor, Slovenia. He is currently working towards a Ph.D. in electrical
engineering. His research interests include industrial automation, building automation,
fault detection and diagnosis and intelligent systems.

Boris Tovornik received the B.Sc. degree in Electrical Engineering, from the University of
Ljubljana, Slovenia and M.Sc. and Ph.D. from University of Maribor, Slovenia in 1984 and
1991, respectively. He was a teacher of Electrical Engineering school, Maribor 1973,
engineer for automation, Ironworks Store 1975, engineer for automation at Drava river
Power plants 1977. Since 1978 he has been with Faculty of Electrical Engineering and
Computer Science, Maribor, Slovenia. He is Head of Laboratory of Process Automation
and holds the rank of Associate Professor. His fields of research interests are computer
control of industrial processes, modelling and process identification, fuzzy control,
intelligent systems and fault detection.

Nenad Muskinja received the B.Sc., M.Sc. and Ph.D. degrees in electrical engineering from
the University of Maribor, Slovenia, in 1988, 1992, and 1997, respectively. Since 1989, he
has been a faculty member in the Department of Electrical Engineering and Computer
Science, University of Maribor, Slovenia, where he currently holds the rank of Assistant
Professor. His research interests include industrial automation, adaptive control, sampled-
data control, fuzzy control, and intelligent systems.