Proceedings of STeP'96. Jarmo Alander, Timo Honkela and Matti Jakobsson (eds.),
Publications of the Finnish Artificial Intelligence Society, pp. 137-143.

Learning environment for a process automation system using computer networks

Juha Lindfors, Leena Yliniemi and Kauko Leiviskä
Control Engineering Laboratory
University of Oulu
PL 444
90571 Oulu, Finland
Phone: +358 81 553 2462
Fax: +358 81 553 2466
Email: juha.lindfors@oulu.fi
WWW: http://sun3.oulu.fi/~posaa/jlind.html


Index


Abstract
1. Introduction

2. Network systems
2.1 A pilot system
2.2 The system to be built

3 Hypermedia materials
3.1 Training material
3.2 The rotary dryer module
3.3 Hypermedia in the automation system

4. Conclusions

References

Abstract


Process automation has been one of the fastest developing fields of process industry in recent years. This makes for quite high demands on automation personnel. They should know and master the process they are supervising. The high level of automation can, however, reduce the skills of the personnel to cope with tasks the automation system cannot do. The recent development in automation systems and networks enables supply of information through nets to support learning during the work. This paper presents a learning environment to be developed in the Control Engineering Laboratory at the University of Oulu. The environment consists of a pilot scale rotary dryer, relating process automation system, and the network connections to local network and Internet. The paper will also describe how earlier, in a COMETT II project, developed hypermedia material is going to be utilized in this environment. It will also outline problems that may be confronted while integrating different systems together and building this kind of learning environment.

Index


1. Introduction

The fast development in process technology and automation in recent years has put certain pressure on the personnel of industry. The lifelong learning is one of the most popular words in Finland today. This means that personnel in industry should learn new things while they are working. Until now, learning has often occurred in various kinds of institutes. This is possible only by going off the work. However, there is always people who cannot leave their work or it is difficult. In any case, leaving the work can cause extra costs for the company.

Development of automation has been studied by Bainbridge [2] who found out that automation can reduce the operators' skills and knowledge, because the automation system is taking care of the normal control actions and the operators need to intervene to the control of the process more seldom. Furthermore, the need of additional knowledge and training simulators in pulp industry has been observed as early as 1981 [4] and in an other project conducted in our laboratory, the lack of even the basic knowledge was observed [9]. The better the level of education and the better practical skills the personnel have, the better is their ability to control processes optimally.

The recent development in the fields of hypermedia and networks could make it possible to build such a learning environment that could be installed at the person's working place. In our case this means that the integration of a hypermedia learning environment and a process automation system can be interesting. Until now, process automation systems have been closed, vendor-defined systems. At the present, development progresses towards more open systems where data is transferred through different kind of nets. Openness and nets make it possible to integrate different types of systems together. This means that it is possible to link information into a process automation system using networks or offer tailored courses available on internal nets of companies [7].

This paper presents a solution for building hypermedia based learning environment that includes hypermedia material obtained in earlier projects and hypermedia based simulation. These will be integrated in very special kind of an environment; in a process automation system. The integration gives process operators a possibility to get information for their acute needs. Furthermore, they can study the process through simulations and animations at a quiet time.

The learning environment will be built into the Damatic XD process automation system that exists in the Control Engineering Laboratory. The system is controlling the pilot scale rotary dryer that is used in several research projects.

Index


2. Network systems

2.1 A pilot system

In the Tiesu project [6], a pilot network was built in the Damatic XD automation system. In the project an operator support system based on an information model was studied. The pilot system consists of Damatic XD network, Internet and a simulator installed in a PC. The system in this case controls the ACSL-based simulator of the power plant. "Operators" of the simulated process are able to use a WWW browser installed in the HP computer. All the computer equipment of the system, except the workstation containing the WWW material, is in the control room. The WWW material consists of on-line documentation for the process configured in the automation system.

2.2 The system to be built

In the laboratory there is also a rotary dryer process, by which new methods of the dryer control are studied. For example, such methods as adaptive and fuzzy controls are available [1,15,16]. The control algorithms are programmed with Matlab and are installed in the HP computer. The process itself is a pilot scale dryer in the laboratory hall and it is controlled by the Damatic XD system.

A network server will be used to provide the learning. This server includes ToolBook and Matlab material, and a WWW server. Using a workstation, students studying the control of the dryer, can use hypermedia material on the server to support their studies. The rotary dryer simulator is located at the server and it will be connected to the automation system via a database so that the dryer model can update its parameters corresponding to the current dryer parameters in the automation system. The workstation will be a part of a network utilizing the ATM technology [17]. The ATM technology will enable the use of real-time video and sound to and from the workstation. Figure 1 shows the planned ATM network configuration.


Figure 1.Network configuration.

Index


3. Hypermedia materials

3.1 Training material

The basic hypermedia material was made initially in the COMETT II project during years 1990 - 1993. The target of the project was to develop training material for process automation so that the material can be used in continuing education in different fields of process industry, in universities, and in different technical institutes [8].

The training material developed [10,11] was modularly grouped into series of courses that could be used at different levels of education and in different fields of industry. The basic idea behind the system was that it would be possible to pick a tailor-made module with the main "textbook"-part, examples, problems, demonstrations, and audio visual material for some main topics of process automation.

3.2 The rotary dryer module

The material obtained in COMETT -project includes also a module for learning the basic theory of modelling and control of a rotary dryer. The behavior of the rotary dryer is simulated with Matlab. Both open-loop and closed-loop simulations are possible. The simulation is started from the hypertext site, however, the user interfaces are build in Matlab where the process and control parameters can be changed.

Later this module was filled in books concerning fuzzy control, neural networks, and genetic algorithms. Also some simple examples have been included. Asymetrix Multimedia Toolbook 3.0 and its OpenScript language was used for hypertext, simple animations, and simulations. Tables 1 and 2 show the contents of the hypermedia material.

Table 1: Summary of training material of the theory of the rotary dryer [12].
BooksIncludesPages
StartContents2
Rotary DryerTheory
Measurements
Control
Mathematical model
Fuzzy control
Simulation
8
10
9
11
5
4


Table 2: Summary of training material of fuzzy control [12].
BooksIncludesPages
StartContents2
Fuzzy controlIntroduction
The Mathematics of Fuzzy Control
Fuzzy Controller
Expert Systems and Fuzzy Control
Introduction to Adaptive Fuzzy Control
A Guide to Fuzzy Application
Example
Bibliography
41
Neural NetworksIntroduction
Types of Neural Networks
Bibliography
39
Genetic AlgorithmsIntroduction
A Top-Level View of the Genetic Algorithm
The Anatomy of a Genetic Algorithm
Concepts
Bibliography
13


Index


3.3 Hypermedia in the automation system?

Dillon and McAleese have identified four main strategies to browse hypermedia material [3,13]: scanning, browsing, exploring, and searching. The first three are not as problematic as the last one. Avoiding difficulties in the first three can be arranged providing a good structure of the information and a good searching tool. More problematic is the issue when the operator is using the simulator or running the process and uses the material as a supporting tool. It is believed that the best benefit of the system will become by practising with the simulator. The simulation will be based on a model of the rotary dryer in the automation system or in the network server. This model will be in the core of the hypermedia operating system. Thus, the model will have "an intelligent model" round it, which is surrounded by a manual and instructions. All the other layers, but the core is based on principles of hypermedia, that in this context is interpreted as linking pieces of knowledge together. Figure 2 shows an approach to this question.


Figure 2. Cross section of the environment.

During simulation the operator is running the model of the process. If the learning is kept in mind, the operator should get right information quickly for the actions he made. The next question is how to provide this information for the operator in the best way? Browsing the right information from a large hyperbase can be boring and if it lasts long, the operator may lose his interest. The answer may be an intelligent browser or an intelligent model that can give, in some way, the right and quick response to the result of the simulation.

On the other, if the process is used instead of the simulator the problem is more difficult. The static linking in the allready built hypermedia material is not proper for a dynamic system. Thus, an intelligent hypermedia system is needed to provide dynamic linking.

One possible solution could be to implement some fuzzy logic to hypermedia, especially to support linking between the event and the information. Mullier presents a dynamic linking system to produce links between nodes [14]. The system connects nodes, which are connected indirectly via semantic structure directly and weights them according to the strenght of the semantic relationship. In addition, a neural network is used to model the way the user is browsing through hypermedia material

. In our case, the dynamic linking has to be done according to the events in the process rather than the user of the domain. Thus, a model of the operator is not needed, but a neural network could be used providing linking between events and reasons.

On the other hand, linguistic equations could be used for dynamic linking. The linguistic equations are often applied to fuzzy-expert systems, where they are providing a flexible environment for combining expertise. The knowledge base of the expert system is represented by linguistic relations that can be changed into matrix equations. The reasoning is based on these equations or on the aggregated sets of linguistic relations obtained by solving the equations. The system is adaptive since the meaning of the linguistic values depends on the working point of the process. This presentation is easily generalized for finer fuzzy partitions and transferred between the programming systems. Membership functions are tuned by simulation experiments or by experiments with real systems [5]. The use of a knowledge base and expert system may be inflexible in our case, but it is believed that during the simulation or running the process, the events could be converted to linguistic equations that can be handled by a dynamic linking system.

Index


4. Conclusions

This project is challenging in many ways. On the one hand, there is the control engineering approach to the issue; on-line simulation in an automation system, load questions in the automation system without problems, model building and adjustment, and question of tools. On the other hand, it is not an easy task to integrate different kind networks and programs together, although the operating systems and networks have become more open recently. Furthermore there are difficulties in the hypermedia and dynamical linking in order to get the best benefits out of the learning system.

Although the ToolBook material is in principal part ready, the integration of these two different operating systems is expected to cause problems. In spite of the ToolBook is running in the local net, launching applications that are pieces of a larger system with system book, etc. can be difficult. The WWW material is more easily available, but can not meet, in this case, all the needs of the operators.

Since hypermedia is often semantic, the fuzzy-linguistic approach with neural networks could be applied to make the dynamic linking system, enabling an intelligent navigation between pieces of information. The project is in the beginning and all the questions are under research. It is not sure if dynamic linking during the simulation could be applied because it may take too much of computing time. Moreover, lack of tools can cause difficulties and can lead, more or less, to use of already existing tools.

Index


References

[1] Altavilla M., Koskinen J., Yliniemi L. Rumpukuivaimen säätö neuroverkolla. Report B, No 3, January 1996. Control Engineering Laboratory, University of Oulu, Finland . 15 p. (in Finnish).

[2] Bainbridge L. (1983). Ironies of Automation, Automatica, Vol. 19, No. 6, pp. 775-779,

[3] Dillon A. (1990). Designing the human computer interface to hypermedia applications. In: Artificial Intelligence and Human Learning (D. Jonassen, Ed.). Springer-Verlag.

[4] Jutila E., Leiviskä K. (1981). The use of computer simulation in the pulp and paper industry. In: Mathematics and Computers in simulation, Vol. XXIII, No 1, pp. 1-11.

[5] Juuso E. (1994). Application of Fuzzy Logic and Neural Networks. In: Annual Report of Control Engineering Laboratory 1994 (Leena Yliniemi, Ed). Control Engineering Laboratory, University of Oulu, Oulu , Finland. p. 14.

[6] Kaarela K., Oksanen J. (1994) Operator Support Systems Based on an Information Model. In: Proceedings of the '94 Symposium On Human Interaction With Complex Systems, September 18-20, Greensboro, NC, USA. pp. 156-165.

[7] Lallukka L., Sundström S. (1995) Metallurgical Knowledge System and Quality Management: Hypermedia in Education in Steel Industry. In: Proceedings of Hypermedia in Sheffield '95, July 3-5, Sheffield, UK, pp. 249-256.

[8] Leiviskä K., Yliniemi L., Lindfors J. (1991) Training Materials for Process Automation: Demonstration of Teaching Process Measurements. In: Proceedings of Nordic Conference on Computer Aided Higher Education, August 21-23, Espoo, Finland. pp. 73-77.

[9] Leiviskä K., (1995) Hypermedia in Simulation and Control. Presentation in SiE/WG Workshop, June 29-30. Brussels, Belgium.

[10] Lindfors J., (1993) Training Materials for Process Automation in Hypermedia Environment. Licentiate thesis, University of Oulu. p. 70.

[11] Lindfors J., Leiviskä K., Yliniemi L., Piirainen I. (1994) Experiences of Using and Building a Medium Large Hypermedia Material. In: Proceedings of Hypermedia in Vaasa '94 Conference, June 8 - 11, Vaasa, pp. 270-276.

[12] Lindfors, J. (1996) Integration of Hypermedia Material into a Process Automation System. In: Proceedings of First International ToolBook Virtual Conference '96, January 18. Internet. 4 p.

[13] McAleese J. (1989) Navigation and Browsing in Hypertext. In: Hypertext Theory into Practice; (R. McAleese Ed). Blackwell Scientific.

[14] Mullier D., J. (1995) Using a Neural Network to Model Hypermedia Browsing. In: Proceedings of Hypermedia in Sheffield '95, July 3-5, Sheffield, UK, pp. 115-122.

[15] Yliniemi L., Alaimo L., Koskinen J., Development and Tuning of a fuzzy Controller for a Rotary Dryer. Report A No 1, December 1995. Control Engineering Laboratory, University of Oulu, Finland. 14 p.

[16] Yliniemi L., Koskinen J., Rumpukuivaimen sumea säätö. Report B No 1, December 1995. Control Engineering Laboratory, University of Oulu, Finland. 17. p. (in Finnish).

[17] Yliniemi L., Lindfors J., Leiviskä K. Development and experiences on the teaching of process automation via network. In: Proceedings of the EAEEIE’96 Conference, June 12-14, Oulu, Finland, pp. 255-260
Index