Control 2004, University of Bath, UK, September 2004...

6
A METHODOLOGY FOR COMPONENTS BASED MODELLING: APPLICATION TO A HEAT TREATMENT LINE * N. Iriondo, F. Artaza, S. Calvo, I. Cabanes, D. Orive, M. Marcos Universidad del País Vaco, UPV/EHU (Spain) Abstract: This paper presents a methodology for hierarchical process modelling. One of its main goals is the re-use of generic components of the model at each level of the hierarchy. This reuse can be performed in the own model, but also in other application domains. To illustrate the methodology, a case study, an industrial application of a Heat Treatment Line, is described. This application has been hierarchically modelled using Matlab/Simulink software. Keywords: process modelling, simulation, hierarchical design. 1. INTRODUCTION 1.1 Component based modelling overview Many of today industrial manufacturers are faced with difficult challenges including reducing time to market and achieving right first time designs. To get this, they are making increasing use of "virtual prototyping" where a computer model is replacing the physical prototype. At this point, there are many benefits to this approach including faster development, lower costs, greater number of design alternatives evaluated and more optimal designs achieved. Mackulak and Cochran [1] say that 45% of a simulation project’s effort is in model building, formulation, and translation. Therefore, implementing a generic model that is applicable to many systems can save a significant portion of time in a simulation study. In addition, such models, once developed, are already debugged and can be optimized for quicker run times. So, this type of simulation development allows less expensive and quicker simulation analysis, and must promote model (or more correctly module) reuse [2,3,4,5]. The requirement for this approach, however, is that such a library of modules must exist. The role of a model library is to manage all the information and to make model fragments reusable. Usually, design tools are associated with libraries of reusable modelling components that will make the description of the models and also their validation much easier [6]. Ruppel and Wysor [7] analyze and compare some of these modelling tools. There are some tools that are very specific, like LabVIEW, for instrumentation systems, VHDL for electronic components. Between generic tools, it is worth emphasizing ACSL (Advanced Continuous Simulation Language) [8], that has the greatest advantage of its experience since it was the first tool introduced in the market, MATRIXx/RealSim [9], that has been acquired by The MathWorks and it tends to disappear, SCILAB-SCICOS [10], that has the greatest advantage of belonging to the free software, i.e., it is free of charge, and 20-sim [11], that allows for a close interaction with Matlab and Simulink, the most widely extended tool. This has been the main reason to select Matlab/Simulink as the modelling tool. Besides that, ACSL is more expensive and SCILAB-SCICOS has little experience in the market. Another tool not very extended yet but that deserve special mention is Modelica [12], a new modelling language very focused on components re-use. Modelica is meant as a new standard for exchange of models, model libraries and experiments. So, it is a textual language, not as user friendly as Matlab, 20-sim, ACSL or SCILAB-SCICOS. In this paper a hierarchical design of process model is presented using the Matlab/Simulink environment [13], a tool with features such as block libraries, hierarchical modelling, signal labelling, and subsystem customisation, that provide a powerful set of capabilities for creating, modifying, and maintaining block diagrams. These modelling features, together with Simulink comprehensive set of predefined blocks, make it easy to create concise representations of systems, regardless of their complexity. This work has been developed within IST-2001-37269 FLEXICON project, whose general goal is to develop methodologies that enable Commercial Off-The-Shelf (COTS) tools integration for the design and deployment of Distributed Control Systems (DCS) with high degree of flexibility, dependability and re-usability. FLEXICON project specifically addresses the development of the capability to produce open, high performance, dependable, distributed fault tolerant systems in reduced timescales and at lower cost. Within the toolset, the different phases of the design cycle are addressed using the appropriate tools. In this sense, the prototype that is being designed integrates ARTiSAN RtS UML, Matlab/Simulink/Stateflow and ISaGRAF Enhanced tools. ARTiSAN RtS UML tool is used to model the overall heterogeneous distributed control system. Matlab environment is used to model the process dynamics (during simulation phase). Finally, the control system is implemented in ISaGRAF Enhanced tool. A methodology for achieving parallel modular designs for the process model and the control system is proposed. The goal is to make easy the integrated tests of the control system. Control 2004, University of Bath, UK, September 2004 ID-243

Transcript of Control 2004, University of Bath, UK, September 2004...

A METHODOLOGY FOR COMPONENTS BASED MODELLING: APPLICATION TO A HEAT TREATMENT LINE

* N. Iriondo, F. Artaza, S. Calvo, I. Cabanes, D. Orive, M. Marcos

Universidad del País Vaco, UPV/EHU (Spain)

Abstract: This paper presents a methodology for hierarchical process modelling. One of its main goals is the re-use of generic components of the model at each level of the hierarchy. This reuse can be performed in the own model, but also in other application domains. To illustrate the methodology, a case study, an industrial application of a Heat Treatment Line, is described. This application has been hierarchically modelled using Matlab/Simulink software. Keywords: process modelling, simulation, hierarchical design. 1. INTRODUCTION 1.1 Component based modelling overview Many of today industrial manufacturers are faced with difficult challenges including reducing time to market and achieving right first time designs. To get this, they are making increasing use of "virtual prototyping" where a computer model is replacing the physical prototype. At this point, there are many benefits to this approach including faster development, lower costs, greater number of design alternatives evaluated and more optimal designs achieved. Mackulak and Cochran [1] say that 45% of a simulation project’s effort is in model building, formulation, and translation. Therefore, implementing a generic model that is applicable to many systems can save a significant portion of time in a simulation study. In addition, such models, once developed, are already debugged and can be optimized for quicker run times. So, this type of simulation development allows less expensive and quicker simulation analysis, and must promote model (or more correctly module) reuse [2,3,4,5]. The requirement for this approach, however, is that such a library of modules must exist. The role of a model library is to manage all the information and to make model fragments reusable. Usually, design tools are associated with libraries of reusable modelling components that will make the description of the models and also their validation much easier [6]. Ruppel and Wysor [7] analyze and compare some of these modelling tools. There are some tools that are very specific, like LabVIEW, for instrumentation systems, VHDL for electronic components. Between generic tools, it is worth emphasizing ACSL (Advanced Continuous Simulation Language) [8], that has the greatest advantage of its experience since it was the first

tool introduced in the market, MATRIXx/RealSim [9], that has been acquired by The MathWorks and it tends to disappear, SCILAB-SCICOS [10], that has the greatest advantage of belonging to the free software, i.e., it is free of charge, and 20-sim [11], that allows for a close interaction with Matlab and Simulink, the most widely extended tool. This has been the main reason to select Matlab/Simulink as the modelling tool. Besides that, ACSL is more expensive and SCILAB-SCICOS has little experience in the market. Another tool not very extended yet but that deserve special mention is Modelica [12], a new modelling language very focused on components re-use. Modelica is meant as a new standard for exchange of models, model libraries and experiments. So, it is a textual language, not as user friendly as Matlab, 20-sim, ACSL or SCILAB-SCICOS. In this paper a hierarchical design of process model is presented using the Matlab/Simulink environment [13], a tool with features such as block libraries, hierarchical modelling, signal labelling, and subsystem customisation, that provide a powerful set of capabilities for creating, modifying, and maintaining block diagrams. These modelling features, together with Simulink comprehensive set of predefined blocks, make it easy to create concise representations of systems, regardless of their complexity. This work has been developed within IST-2001-37269 FLEXICON project, whose general goal is to develop methodologies that enable Commercial Off-The-Shelf (COTS) tools integration for the design and deployment of Distributed Control Systems (DCS) with high degree of flexibility, dependability and re-usability. FLEXICON project specifically addresses the development of the capability to produce open, high performance, dependable, distributed fault tolerant systems in reduced timescales and at lower cost. Within the toolset, the different phases of the design cycle are addressed using the appropriate tools. In this sense, the prototype that is being designed integrates ARTiSAN RtS UML, Matlab/Simulink/Stateflow and ISaGRAF Enhanced tools. ARTiSAN RtS UML tool is used to model the overall heterogeneous distributed control system. Matlab environment is used to model the process dynamics (during simulation phase). Finally, the control system is implemented in ISaGRAF Enhanced tool. A methodology for achieving parallel modular designs for the process model and the control system is proposed. The goal is to make easy the integrated tests of the control system.

Control 2004, University of Bath, UK, September 2004 ID-243

The hierarchical design of the control system is described in [14]. This hierarchical design follows the recommendation of the IEC 61131 standard for programming PLCs. This paper focuses on the modular design of the process. Each module has a correspondence with a module in the control systems. Generic components can be used to develop incremental validation tests of the control system, developed using ISaGRAF tool [14]. Integrated tests are performed using co-simulation framework developed with FLEXICON project [15]. The co-simulation framework uses CORBA as middleware to achieve co-simulation between a process model developed in Simulink and a control system developed in ISaGRAF. This implies that the control system and the process model can be geographically distributed. The lay-out of the paper is as follows: section 2 describes the proposed methodology. In section 3 the methodology is illustrated through a case study, consisting of an industrial Heat Treatment Line (HTL). Finally, in the last section conclusions obtained are highlighted.

2. METHODOLOGY

The model design has been developed following the guidelines and recommendations of one of the industrial partners of FLEXICON project, and must be developed according to the hierarchical design of the control system. This hierarchical design allows the re-use of generic components at each level of the hierarchy and to implement the model hierarchy in a ‘top-down’ or ‘bottom-up’ structure. The system needs to represent mechanical aspects and properties balances involved in the application. Once a subsystem has been modelled as a generic component and its content masked, it can be used setting up the parameters and connecting it to other subsystems of the same level. Figure 1 shows a generic component.

Figure 1. Simulink model generic component.

All the generic components are described in the following way:

• Name. The generic component type that is being characterized.

• Tagname. The identifier assigned by the implementor.

• Category. The path of the generic component in the hierarchy.

• Functionality. If the generic component belongs to the lowest level, it describes the component functionality, otherwise, this section enumerates the components of the lower level.

• Internal Data. List of local variables to implement the generic component functionality, if needed.

• Inputs. List of all the incoming connections, from other components or from the outside (field signals).

• Outputs. List of outgoing connections, to other components or to the outside (field signals).

• Parameters. List of user configuration parameters. They can vary depending on the concrete application.

3. INDUSTRIAL APPLICATION: HEAT TREATMENT LINE Nowadays, there are a lot of industrial applications that have common elements that only differ in the value of some configuration parameters. That is the case, for example, of a heat treatment line, a finishing system of steel rolling line and an aluminium casting machine. So, only modelling these common elements once, they could be re-used in all applications that use them. In this paper, an industrial Heat Treatment Line has been used as case study. 3.1 Heat Treatment Line description The HTL application processes metallic pieces in order to get the necessary hardness and ductility. The pieces to be treated are screws and other small and slight pieces. So, the load is modelled as a continuous flow of pieces instead of considering them as individual objects. The whole treatment line is composed of a load system, an austenizing furnace, a tempering tank, a washing system and an annealing furnace, as Figure 2 shows.

Load system

Austenizing furnace

Tempering tank

Washing system

Annealing furnace

Figure 2. A typical Heat Treatment Line

Control 2004, University of Bath, UK, September 2004 ID-243

The load system introduces the pieces into the austenizing furnace in a uniform and continuous manner, putting them on the austenizing furnace conveyor. As a result of the austenizing, the material improves its hardness. After leaving the furnace, the pieces fall into a tempering tank, where they rapidly get cooled in an oil bath. Next, they are moved to a washing system where they have water showers in order to clean the surface from oil. Finally, the pieces are placed into an annealing furnace and they get the necessary ductility. This paper presents the modular design of the austenizing furnace and the load system. The other subsystems in the HTL are designed using components created for this two subsystems, or creating new ones, if needed, following the same design structure and looking for the re-use of the new components. 3.1.1 Load system (LS) The load system, is supposed again to be composed of different parts: pannier-elevator, shakers, weighting system and economizer (see Figure 3). These different subsystems are entirely mechanical.

Max. Level

Hopper

Economizer

f(PLC) SP x ShakerSp

Austeniziting Conveyor

Level sensor

FURNACE

Weightingmachine

Min.Level Shaker 1

Shaker 2

Shaker

Shaker

Figure 3. Load system The load system operates as follows: The elevator lifts up a pannier containing the pieces and empty out it into a hopper. The hopper feeds, through a shaker, a weighting system that drives the desired amount of load into an economizer. The economizer is a chamber between two gates whose opening and closing are sincronized, so both gates are never open both at the same time. This is necessary to avoid heat loss at the input of the furnace. At this point, the load is shaked again in order to extend it homogeneously over the conveyor belt inside the austenizing furnace. 3.1.2 Austenizing furnace (AF) The subsystem named austenizing furnace is composed of two subsystems: the austenizing furnace itself and the conveyor belt inside.

The austenizing furnace is divided in four zones, as shown Figure 4, having two burners per zone. The temperature regulation is performed in each one of the four zones, although the furnace itself is actually a unique volume. The inner temperature of the furnace must be around 850 ºC. Each zone is provided of a thermocouple for measure the temperature, a security thermocouple and two burners where the combustion process is performed via gas and air mixture. There is also a chimney for furnace fume evacuation.

Figure 4. Austenizing furnace. The other AF subsystem is a mechanical element, a conveyor belt, that can drives the pieces through the furnace at different speed, allowing them to get a certain residence time in the furnace depending on the heating treatment required. The model has to take into account all the different operation states of the furnace: empty with and without heat exchanged with the walls, entering load, full of load and load leaving. Different models will represent these situations. Each of the modules representing the process corresponds to a module in the control system. See [14] for a detailed description of the control system. 3.2 Hierarchical and modular design 3.2.1 Modular design of load system model The hierarchical design of load system is shown in Figure 5. As an example, the generic component used to model the shaker 1 is shown in Figure 6. The most common elements used to model the subsystems of the load system (all of them are mechanical components) are basically logical operators (AND, OR, NOT, …), switches, triggered and enabled subsystems, integrators, saturators, … The connection between the different subsystems is developed by means of charge variables that make the

Control 2004, University of Bath, UK, September 2004 ID-243

load go from one mechanical element to the following, until it arrives at the furnace conveyor.

Load System

Elevation System

Elevator

Hopper

Shaker 1

Charger

Hopper Gate

Weighting System

Charger

LEVEL 1 LEVEL 2 LEVEL 3

Weighting System

Economizer Charger

Economizer Gate

Shaker 2

Figure 5. Load system hierarchical design

Figure 6. The generic component named Shaker 1 3.2.2 Modular design of austenizing furnace model The continuous processing of the line is going to be modelled distinguishing among different operation modes representing the different possible states. These states are only for the first zone. The rest of the zones have only empty states because the load has already reached the reference temperature in the first zone [16], and there is no more heat absorbed by the load [17]. The generic component to model the temperature regulation in all the states is built using the same philosophy. Furnace inner fluid is assumed to be an element whose temperature changes uniformly through the time, in all points of the element. With this

assumption, the fluid temperature variation could be represented by the equation 1 and the Simulink model of the Figure 7:

gaslostcombustiongas qqq

dt

dTmc =−=⋅⋅ (1)

• c: constant volume specific heat (KJ/kg ºC) • m: gas mass (kg) • Tgas: gas temperature (ºC) • qgas: difference between the amount of heat that

enters and leaves the gas element ( KJ/h)

Figure 7. Beginning model for the austenizing furnace The energy balance applied is based on the diagram in Figure 8 [18]. In this figure, it can be seen that the only heat that enters the gas element will be the heat generated by the combustion process. The amount of heat that enters the gas element has been considered positive, and the amount of heat that leaves the gas element negative.

Figure 8. Energy balance diagram As the furnace is composed of four zones, there is a certain amount of heat that goes from one zone to the others. But more or less there is a compensation between these heat exchanges, and they can be rejected. So, now combustion, heat to the load, heat to the fumes, heat lost throw the walls and heat absorbed by/given to the walls generic components of level 3 have to be added to the model. The models for different states and zones will have the same structure with little changes. For example, the level 2 generic component for the state called Empty do not have the level 3 generic component called Load, because there is not heat transferred to the load. This generic component of level 3 will be different

Walls

Combustión air

FURNACE 850º

Fumes

29.3%

Load 15.64%

100%

1.19%

47.24%

Energy consumption

Control 2004, University of Bath, UK, September 2004 ID-243

depending on the amount of load in the zone (start of load entry, full of load or finish of load entry). In the same way, the level 3 generic component named Walls, representing the heat lost by walls, will have different parameters in the different level 2 generic components for the different zones. With these examples it gets very clear that the generic components re-using is very used along all the model, and it can save a lot of time and efforts. All the generic components have the same aspect as the generic block represented in Figure 1. The austenizing furnace hierarchical design for the energy balance can be seen in Figure 9. Besides, there is a mechanical generic component for the conveyor.

Austenizing Furnace

Full of load

Empty

Start of load entry

LEVEL 1 LEVEL 2 LEVEL 3

Empty with heat absorved

by walls

Combustion

Walls

Fumes

Load Finish of load entry

Figure 9. Austenizing furnace hierarchical design The connection between the different subsystems is developed by means of a state diagram, built using Stateflow [19], that handles the sequence in which the models are executed, depending on the system operation state. Transitions between states are driven by different conditions. The generic components of the different states are subsystems that only are enabled when their state is active. The Stateflow diagram in Figure 10 also contains states corresponding to the furnace stopped operation, secure operations to be performed before starting burners, like the purge operation in order to renovate the furnace inner air. But these states only have a transition time, and they are not considered as generic components.

Figure 10. Stateflow diagram of the first zone

4. CONCLUSIONS In this paper a methodology for process hierarchical modelling has been proposed and validated. This methodology has been illustrated in an industrial case study, a Heat Treatment Line. Generic components of different hierarchical levels have been designed. The connection between lower level components to create a higher level component depends on the type of components, and it has also been validated. The possibilities of re-use of generic components at all the levels of the hierarchy has been proved, as well as for other industrial applications. This modular design has been used in order to perform incremental validation tests of the control system, using a co-simulation framework based on CORBA middleware 5. ACKNOWLEDGEMENTS

The authors gratefully acknowledge the financial support of the European Union’s Information and Science Technologies programme for the FLEXICON project IST-2001-37269. 6. REFERENCES 1. Mackulak G.T., Cochran J.K., 1990, “The generic-

specific modelling approach: an application of artificial intelligence to simulation”, IIE Integrated Systems Conf. & Society for Integrated Manufacturing Conf., San Antonio, 82-87.

2. Diaz-Calderon A., Paredis C.J.J., Khosla P.K., 2000, “Organization and selection of reconfigurable models”, Winter Simulation Conference, 386– 392.

3. Kasputis S., Ng H.C., 2000, “Composable simulations”, Winter Simulation Conference, 1577 – 1584.

4. Davis P.C., Fishwick P.A., Overstreet C.M., Pegden C.D., 2000, “Model composability as a research investment: responses to the featured paper”, Winter Simulation Conference, 1585 – 1591.

5. Son Y.J., Jones A.T., Wysk R.A., 2000, “Automatic generation of simulation models from neutral libraries: an example”, Winter Simulation Conference, 1558 – 1567.

6. Breneuse A., Top J., Broenink J.F., Akkermans J., 1998, “Libraries of reusable models: theory and application”, Simulation, 71.

7. Ruppel F., Wysor W., 1997, “Guidelines for simulator-based control system testing”, Technology Transfer, SEMATECH.

8. McKay W., Monti A., Santi E., Dougal R., 2001, “A co-simulation approach for ACSL-based models”, Huntsville Simulation Conf., Huntsville, Alabama.

9. Wang B., Gong K., Yang D., Li J., 2003, “Fine attitude control by reaction wheels using variable-

Control 2004, University of Bath, UK, September 2004 ID-243

structure controller”, Acta Astronautica, 52,613-618.

10. Lacabanne B., Benjelloun-Dabaghi Z., Bouskela D., 2003, “Some examples of industrial problems modelling under SCILAB-SCICOS”, 22nd IASTED Int. Conf. on Modelling, Ident. and Control, Innsbruck, Austria.

11. Broenink J.F., 1999, “20-SIM software for hierarchical bond-graph/block-diagram models, Simulation Practice and Theory, 7, 481-492.

12. Broenink J.F., 1999, “Object-oriented modelling with bond graphs and Modelica”, Int. Conf. on Bond Graph Modeling and Simulation, San Francisco, California.

13. Fellini, R., Michelena, N., Papalambros, P., Sasena M., 1999, “Optimal design of automotive hybrid powertrain systems”, EcoDesign 99, Tokyo, Japan, 400-405.

14. Iriondo N., Calvo S., Orive D., Marcos M., 2004, “A methodology for hierarchical industrial control systems: application to a heat treatment line”, submitted to Control 2004, Bath, UK.

15. Marcos M., Gangoiti U., Estévez E., Portillo J., Calvo I., 2004, “A CORBA based co-simulation framework for integrating COTS tools”, submitted to CONTROLO 2004, Faro, Portugal.

16. Limpens M., Rindt C., Van SteenHoven A., 2002, “Coupled radiation and convection modelling in an industrial furnace”, Section Energy Technology, Faculty of Mechanical Engineering, Eindhoven University of Technology.

17. Astigarraga, J., 1994, “Hornos industriales de resistencias: teoría, cálculo y aplicaciones”, McGraw-Hill.

18. CADEM, otros, 1994, “Hornos industriales: fichas técnicas”, Ente Vasco de la Energía.

19. The Math Woks, Inc, 2003, “User guide of Matlab/ Simulink and Stateflow”.

Control 2004, University of Bath, UK, September 2004 ID-243