Easy Java Simulations - International Journal of Engineering
Contents
1. 03 Oulu Finland June 2003 9 N Tan D Atherton and S Dormido Systems with variable parameters Classical control extensions for undergraduates IFAC Symposium on Advances Control Education ACE 03 Oulu Finland June 2003 10 W Christian The open source physics project http www opensourcephysics org 11 ECP Educational Control Products Magnetic Levitation System manual for model 730 1999 12 K H Johansson The quadruple tank process A multivariable laboratory process with an adjustable zero IEEE Trans on Control Systems Technology 8 3 2000 pp 456 465 13 S Dormido and F Esquembre The Quadruple Tank Process An Interactive Tool for Control Education Proceedings of the European Control Conference Cambridge UK 2003 Jos Sanchez received his Computer Sciences degree from Madrid Polytechnic University in 1994 and his Ph D from UNED in 2001 Since 1993 he has been working at UNED s Department of Computer Sciences and Automatic Control as an Associate Professor Francisco Esquembre was born in Alicante 1963 and received a Ph D in Mathematics in June 1991 from the University of Murcia Spain where he has worked since 1986 holding a permanent job as an Associate Professor from 1994 Sebastian Dormido received his Physics degree from Madrid Complutense University 1968 and his Ph D from Country Vasque University 1971 In 1981 he was appointed Professor of Control Engineering at UNED He has sup2. gie include interactions and a transmission zero loca tion that can be tuned in operation With adequate tuning this system presents non minimum phase behavior that arises due to the multivariable nature of the problem For this reason the quad ruple tank has been used to show the results of different control strategies and as an educational tool in teaching advanced multivariable control techniques Although the setup is simple the process can still illustrate interesting multivariable phenomena The process flowsheet is displayed in Fig 15 The target is to control the levels AZ and h2 in the lower two tanks with two pumps The process inputs are the voltages v and v2 applied to the pumps The differential equations representing the mass balances in this quadruple tank process are dh ay es he a aay a3 ykiv1 2glhs a YEST gi dh d2 alee 7 u at a4 pkv pe Sea ea ae ah YE UT ga dh3 a3 A 7 a nN AR y k vn kad s h3 s3 h3 dh4 pr d4 laa ar dt s4 ha da y kvi kad 9 s4 h4 s4 h4 Readies ky en Pee er o a nna C n g3 a T O Ha S Tr Foon es TA ies m wn t Sac ake J ETTI Ja ib public void ChangeMaped EpGlexpal we ypme yma ifea eos eel nagadi weight2 nass2 g Fig 14 Elements for the view of the magnetic levitator Easy Java Simulations An Open Source Tool 809 hi Tank 1 Tank 3 Tank i h H
3. Models can be described in EJS applica tions either using its self contained facilities ODE solvers and Java code editors or Simulink block diagrams If using the latter EJS can wrap the Simulink model with a Java made layer of inter activity The final product is a Java application that efficiently interchanges data input para meters and states variables with the Simulink model through MATLAB s workspace Fig 3 The tool can be used as a timely first step for control educators as a way of adopting an inter active based approach in order to be at the cutting edge of innovative initiatives in education such as the different world wide open source initiatives J Sanchez et al that physics instructors are currently encountering 10 Because of its interest this paper describes EJS in more detail and presents two case studies of the development of control virtual labs The paper is organized as follows First Easy Java Simulations are briefly described with parti cular emphasis on the development of control virtual labs that use MATLAB Simulink as the simulation and mathematical engine The next section shows an example of this collaboration between tools to develop a virtual lab on a magnetic levitator Then a more advanced virtual lab is described the quadruple tank process which also uses Simulink to develop the model The paper ends with some conclusions on the combined use of EJS and MATLAB Simulink as a way to fos
4. Acknowledgements This work has been supported by the Spanish CICYT under grant DPI2001 1012 and DPI2004 01804 REFERENCES 1 S Dormido The role of interactivity in control learning Plenary Lecture IFAC Symposium on Advances Control Education ACE 03 Oulu Finland June 2003 2 S Dormido Control learning Present and future Annual Reviews in Control 28 2004 pp 115 136 3 Y Piguet SysQuake User Manual Calerga 1999 4 E Esquembre Easy Java Simulations A software tool to create scientific simulations in Java Comp Phys Comm 2002 See also http fem um es Ejs 5 S Dormido J Aranda J M Diaz and S Dormido Canto Interactive educational environment for design by QFT methodology Proceedings of 5th International Symposium on Quantitative Feedback Theory and Robust Frequency Domain Methods pp 223 230 Pamplona August 2001 6 S Dormido F Gordillo S Dormido Canto and J Aracil An interactive tool for introductory nonlinear control systems education IFAC World Congress b 02 Barcelona July 2002 7 P Albertos J Salt S Dormido and A Cuenca An interactive simulation tool for the study of multirate sampled data systems IFAC Symposium on Advances Control Education ACE 03 Oulu Finland June 2003 8 S Dormido M Berenguel S Dormido Canto and F Rodriguez Interactive learning of introductory constrained generalized predictive control IFAC Symposium on Advances Control Education ACE
5. Tank 2 ham Fig 15 Schematic view of the quadruple tank process P wi la ae Amt o_o z E A A Ee 8 2 Un I Ln 4 mss vi Lard om a cL i lt i Giss R a la ua aje am iy m 8 an ar eer a a r T k PTR im ai na a p na aj AH E ie Candee or mes v Persie Tanks r Geis iis faa Ti ai mei s CEO Bjal Fm des 17 PT y sp i ALLEEL Hh I th eran E i Of 9S mF ga Fhine skates Je GS Gi it na aa GF as Feriae Heme be Bo bi ii aa AF aa Fig 16 View of the quadruple tank virtual lab where A is the liquid level in tank i a is the outlet cross sectional area of tank i s h is the cross sectional area of tank i v is the speed setting of pump j with the corresponding gain k y is the portion of the flow that goes into the upper tank from pump j and d and d are flow disturbances from tank 3 and tank 4 respectively with corres ponding gains kg and kaz The process manipu lated inputs are v and v2 speed settings to the pumps and the measured outputs are y and y2 voltages from level measurement devices The measured level signals are assumed to be propor tional to the true level i e yy km h and y2 km2 2 The level sensors are calibrated so that k km 1 This process exhibits interacting multivariable dynamics because each of the pumps affects both of the outputs The linearized model of the quad ruple tank process has a multivaria
6. and numerical values However control engineering instructors are not true programmers and the addition of an extra layer of interactivity to their MATLAB Simulink applications would be very time consuming This statement also applies even more strongly to the development of virtual and remote labs 800 Two software products now exist that can help solve this problem SysQuake 3 and Easy Java Simulations 4 The first is a commercial tool oriented to the development of interactive applica tions using a language fully compatible with MATLAB Consequently MATLAB legacy code i e M files can be reused and benefit from SysQuake s high graphical capabilities to produce valuable interactive instructional control applica tions Additional information on this software can be found at http www calerga com and examples of advanced control applications developed with it are described in 5 9 Easy Java Simulations EJS is an open source free of charge software tool that helps create dynamic interactive scientific simulations in the Java language The tool is targeted for both science students and teachers with basic programming skills and is therefore very much suited to the situation one finds in most university classrooms Within EJS simulations are created by specifying a model for the simulated system and by building a view that continuously visualizes the state of this model and that readily responds to user inter action
7. block diagram of the magnetic levitator described in equations 1 to 8 with the modifica tions needed for it to be used by EJS There are three main steps to using a standard Simulink block diagram within EJS The first consists in making some changes to the original Simulink model These changes are necessary because Simulink models are obviously created to be run within Simulink But since the commun ication between Simulink and EJS goes through MATLAB s workspace we first need to change the Simulink model so that it sends receives the value of some of its variables and parameters to from MATLAB s workspace Fig 9 shows how some system parameters of the block diagram have been adapted to be read from MATLAB s work space magnets masses polarities and parameters of the magnetic configuration The block diagram has also been slightly modified so that it writes the simulation state in MATLAB s workspace time and positions of the two magnets after every integration step The reading of the values from MATLAB s workspace variables is done by including MATLAB function blocks in the diagram to evaluate variables and or expressions Fig 10a shows a Simulink block in which an expression using MATLAB variables is evaluated to generate one of the magnetic forces To write the model state back into MATLAB s workspace after each integration step it is necessary to include a To workspace block in the diagram for
8. each of the variables in the state vector of the model As an example Fig 10b shows the block used to send the position of magnet variable yl back to the workspace A second difference to a standard bock diagram comes from the fact that Simulink models are usually played and paused by the user through Simulink s user interface Hence the model must be changed so that it stops after every integration step For this it is necessary to include in the model diagram one MATLAB function block with the MATLAB command set_param gcs SimulationCommand Pause In this way EJS can control exactly when the model needs to play The second step needed to connect the Simulink model with EJS is the creation of a simple M file This file provides EJS with basic information about the model and about the variables and parameters that can be accessed through MATLAB s workspace Fig 11 shows an extract from the file that we created for this example The syntax for this text file is rather simple 806 J Sanche Hock Pararnerthers Fil i Hj HATLAR Fen Pow the nmp values lo a HATLAR uncom ion eaan The kunotion mus retum a angie vaio agreed of length Duio vajh sarees ari enjoh Peetu ult HATLAE hreion LH gaty uia e al Dujrd dh 1 Duig Sra by pee auo a z et al Block Parameters yl to workspace To Wipe j Velie input lo packed maii m MATLAD man woktapace Thema fa
9. last two options can be used to configure the virtual lab as a variety of single input single output SISO and multi input multi output MIMO systems By using a repulsive force from the lower coil to levitate a single magnet an open loop stable SISO system is created Attractive levitation via the upper coil configures an open loop unstable system Two magnets may be raised by a single coil to produce a SIMO plant If two coils are used a MIMO plant is produced These may be locally stable or unstable depending on the selection of the magnets polarities and the nominal magnet positions The plant has inher ently strong nonlinearities due to the natural properties of magnetic fields Thus this virtual lab provides a dynamically rich test bed for imple mentation of introductory and advanced control methods However the most relevant feature of this didactical application is that MATLAB Simulink Easy Java Simulations An Open Source Tool 805 Fati Fegi CE HETLAS Farrais m2 1 HATLAR jana Faih nJ aa et a AATAS Farti na Fmiz s ca e E Fig reat no Fur Fa MATLAS Fanion mf Fit h MATLAS E Suna CT Pour fA pitia PP gaoat i paptisn_i Tima bo ls krp no MATLAR Panchen Piia Birra bork pis tup 2 val ect 2 partinn 2 Fig 9 Modified Simulink block diagrams is running in the background Fig 9 shows the Simulink
10. representations of the system Let us now apply our discussion on interactivity to the real educational world where the de facto standard software for instructional purposes is MATLAB Simulink If we try to apply our concept of interactivity to teaching with Easy Java Simulations An Open Source Tool 799 bss Dest tet 5 rani E 4 Se Sete ed 210 8 40 Ao Fig 1 Virtual lab for studying control problems related to the Furuta pendulum lt lt a J u E rii F mtn EET rages EE ce a er g LESE i fr KE i impna So Biu Oe pia a Jam bees eme menee Rare H nei Hii r nmam i ep wa ee i bgm j aisle satan es 2 See Le Fes a fijian Feces EP Seam f ie ee imsi pe ee fe Da a far eee Fig 2 Interactive toolbox for analyzing and designing control linear systems MATLAB Simulink we will encounter serious difficulties because this software was not designed to develop graphical applications using this new approach The creation of rich graphical interfaces using only MATLAB Simulink is not an easy task and high programming skills are required to develop useful applications with even a basic level of interactivity When an instructor uses MATLAB Simulink s he is an expert in the subject that is to be explained and in the use of the mathematical facilities of the software to program algorithms S he is also capable of presenting the results by means of different types of plots
11. set of dripping holes pipes pumps and valves that configure the system We tried to make this first image as self explanatory as possible The black arrows indicate the set points of the lower tanks those under control These set points can be changed interactively by dragging the arrows up and down The lower part of the main window is occupied by primary controls that let the user specify how the system should run With the left hand column of buttons the user can perform the following actions e Play pause and reset the simulation Select to run the nonlinear or linear models or both for the system e Choose whether to control the system manually M or automatically A Select in the case of automatic control whether to use a P or PI controller this can easily be extended to other types of controllers A detailed explanation about the possibilities of this control application can be found at 13 However in that work the application was fully developed using EJS In this paper the four tank application presented simulates the nonlinear part of the model using a Simulink block diagram and the linear model and the view have been programmed under EJS The process required to create such a hybrid application is similar to the one used for the magnetic levitator example For the first step of this process we first had to implement the nonlinear differential equations corresponding to the tanks Equations 9 in a
12. the accuracy of the simulation However the creation of the necessary graphical user interface that is the view 0 Baep Jeo Siis Dei Prr bee Java ETS Bar introduction Model Varlables initialization Hier mimi f maj im ipai i Ber second miaj War 1 VIEW Evelution 801 is the part of the simulation that demands more knowledge of advanced programming techniques And to make the situation even worse the addi tion of interactivity to this interface involves mastering sophisticated software techniques which demand a big investment of time and effort EJS helps its users through both tasks The tool provides extensive scaffolding to define the model while still retaining the flexibility of a general programming language so that it is possible to specify almost any type of algorithm This is pedagogically important since the process of learning good control fundamentals consists to a great extent in understanding the basic principles required to build models In order to define the model in EJS the user follows a prescribed sequence declare the variables that describe the system initialize these variables to a correct initial state e describe how the value of these variables changes over time and establish how they affect each other when the user interacts with the system and modifies one or more of their values The last two steps consist in translating the math ematical equa
13. Int J Engng Ed Vol 21 No 5 pp 798 813 2005 Printed in Great Britain 0949 149X 91 3 00 0 00 2005 TEMPUS Publications Easy Java Simulations an Open Source Tool to Develop Interactive Virtual Laboratories Using MATLAB Simulink J SANCHEZ F ESQUEMBRE C MARTIN S DORMIDO S DORMIDO CANTO R D CANTO R PASTOR and A URQUIA Department of Computer Scien ces and Automatic Control UNED CI Juan del Rosal 16 28040 Madrid Spain Email jsanchez dia uned es fem um es This paper describes the make up of interactive computer simulations that implement virtual laboratories in the field of Control Engineering education We introduce Easy Java Simulations EJS a Java based tool designed to help control educators with a low profile in programming to create interactive scientific simulations This tool can be used on its own generating stand alone Java applications or applets or in conjunction with MATLAB Simulink using them as the internal engine that describes and solves the model This tool allows users to develop complete interactive simulations in three steps writing the mathematical model optionally using MATLAB Simulink building the graphical user interface GUI using off the shelf graphical elements and linking the GUI elements to the variables of the model In this way models of control engineering can be specified either using the programming support provided by EJS or by using MATLAB Simulink which
14. Simulink file Fig 17a Easy Java Simulations An Open Source Tool 811 i Lap eed cake ep i eel bee at aiis F roduciion Model View ET EN paie FSi Eapirperi Cea ar Cian fe Timea jana Laai heen a mee ee pee a Se a e O a O i j l i r BIT pees toe a a p T hams n wpa e Tammar inei in ik EW i te i a a AL 1 all tj Intraduriso Manel View eorebies PHigingten Eroivien Corateta Cuman ey Bias Lie Se til EHE cmd D mim bad ri x bia a z li ns E Fig 19 a Variable page illustrating the link between some EJS and MATLAB variables b Evolution page showing the EJS instruction to order Simulink runs one step of simulation shows the Simulink subsystem of tank 1 We also created an M function for the PID controller used to control the tank levels and included this M function in two Simulink blocks connected to the nonlinear models of tanks 1 and 2 Fig 17b shows the diagram of the Simulink subsystem used to implement the manual and automatic control system the PID M function is included in the PID1and PID2 blocks which are used to compute the control signals v1 and v2 the voltages of the two pumps two switch blocks named Switch1 and Switch2 allow us to select automatic or manual control depending on the value of a MATLAB variable auto that can be interac tively changed using EJS view In this way when the user selects automatic control in the g
15. ble OQutputdnoly dry integration step REJS Parameter maxstep Inputdnly ul Control ation 1 EIS Variable nputonly Udi Canteol action 2 EJS Variable nputonly A tHagnetic configuration EJS Variable nputonly b tMagnetic cant iguration EJS Variable Inputonly D tMagnestic configuration tEJS Variable InputoOnly d tMagneatic configuration 615 Variable InputOnly mi tMasa of magnet 1 EJS Variable InputoOnly Mit tMaea of magnet 2 EJS Variable nputonly Fig 11 Extract from the file evitator m used to connect Simulink and EJS Easy Java Simulations An Open Source Tool Geral lel esamples Ecernaigps Metlabiectator m koube t double Gout koube doubie doube doube doube douwe iow brie igibha digui Ajuba ume jute wagta magal He ty ya he jut GB BaN EV input Gutpul c input E Aini G a Fig 12 EJS page of variables showing the connection to MATLAB s variables To make the connection between EJS and MATLAB Simulink it is necessary to open a page of external variables in the EJS application Fig 12a In this page it is necessary to introduce into the External File text field the path to the information M file previously created for the levi tator Now whenever you create an EJS variable and right click on it you will be given a window with a list of the MATLAB variables to which you can connect your EJS variable Fig 12b By clicking on any o
16. ble zero which can be located in either the left or the right half plane by simply adjusting the throttle valves y and 2 Johansson showed that the inverse response non minimum phase occurs when 0 lt y 72 lt 1 and minimum phase for 1 lt y 72 2 The valve 810 J main 5 j 7 mam a _ l a ty bk i W Bo e M i B F T H rami T i ian i ET H wt eh Sanchez et al Fig 17 Simulink subsystems of tank 1 and the PID controllers tEja Ejs 4Eja tEjs diaml Ejs diam E S gammal Eja gamma E sS v1 tEja V2 Ejs controller E s controller2 Ejs auta E S Model Variable Gutputonly Variable Variable Variable Inputonly Variable Inputonly Variable Inputonly Variable Inputonly Variable Variable Variable Variable Variable Inputonly Fig 18 Extract of the M file fourtanks m to connect Simulink and EJS settings will give the overall system entirely differ ent behavior from a multivariable control view point Unmeasured disturbances can be applied by pumping water out of the top tanks and into the lower reservoir This exposes students to distur bances rejection as well as reference tracking Implementation of the quadruple tank with EJS and MATLAB Simulink The main window of the EJS application for this example can be seen at the left hand side of Fig 16 Schematically displayed in its upper part are the four tanks and the
17. can then be fully controlled from EJS In this paper we describe this particular feature in detail and provide some examples that show the advantages that this tool offers to the world wide engineering education community INTRODUCTION RECENT DEVELOPMENTS in computer hard ware and software now make it possible to provide students with interactive programs that can be considered as midway between regular labs and lectures and that allow us to display multiple view representations of a given dynamic system and some of its attributes on the computer screen 1 Visualization thus appears naturally in the origins of automatic control in the discovery of new relationships among mathematical objects and in the transmission and communication of know ledge These tools called interactive virtual labs can be used to explain basic concepts to provide new perspectives on and insights into a problem and to illustrate analysis and design topics 2 However as a natural consequence of the enhanced performance and graphical capabilities of modern computers the search for interactivity must also take us a step forward Nowadays the use of the word interactivity in the design of automatic control systems means to many people a process of repeating several times a loop whereby a user introduces some parameters in the software runs the algorithms of design and analyzes and compares the results graphically in static plots If the res
18. cilitate the student s understand ing of the virtual lab In our case this view using EJS terminology is divided into several elements as displayed in Fig 8 Because the simulation is to be used for didactical purposes it was very impor tant for us to provide a visualization that displays the physical system as realistically as possible but that also provides support for graphic tools commonly used in control plots diagrams input devices controllers etc A second requirement was to provide a set of interactive elements that could be manipulated by the students to make changes in a dynamic way during an experiment The view is therefore structured in one main window and three secondary dialog windows The main window is located at the left hand side of Fig 8 The big central part of the main window schematically displays the magnetic levitator and allows the user to change interactively the set points by dragging the arrows up and down The lower part of the main window is filled with buttons and check boxes that allow the user to choose the operating conditions of the system With these elements the user can select to e Play pause and reset the simulation e Choose the control system mode manual or automatic e Change the mass of the magnets change m e Change the magnets polarity polarity m 1 and polarity m 2 Select the magnets that will operate the levitator upper magnet and or lower magnet The
19. ck the classes in a compressed file and end up with a ready to use simulation Simulations created using EJS can either be used as stand alone applications under different operating systems for instance a BAT file is provided when running under Windows or can be distributed via the Internet and run as applets within HTML pages also generated by EJS by using any Java enabled web browser The tool also includes a simple GEE oiy Jara Sieun Cih dooementos lars eects E 1S a m Pi al Introduction Dive H Enema af Simulation view Gee frame enduum p O panei euth O preii O nBr bution by i butte saute huni 20 butiovstep Danek heck F cheias Fo ocheckBoePict EE gidariari E iiri sidai Model View D EE lala Liters he he eer Ciel ee Fig 5 Library of the graphical elements of EJS Easy Java Simulations An Open Source Tool J 803 L Pe De ee ibaa G5 G aa maa zi 1 Bi te ii ii E B i T f Lin Pact Lit i P Bi Lun E e iF AE Bi i Amg m Mirae Fig 6 View of an interactive application developed using EJS HTML editor to help the teacher enhance the generated web pages with pedagogical information and or instructions for the simulation At present the use of MATLAB Simulink is not supported when running simulations as applets although we are working on a future model that will allow EJS to generate distributed applications with Java interfaces and Simulink
20. ct ora cokan pat input alarmar anid ora iow par emule ep Dala ki rec evenly urti thes samba m sapped on paed Paea ki aiabie name ames ________ Piana mumis of oaa Sava jamat ham 0 C ea j e jool e b Fig 10 Source and sink blocks to communicate MATLAB s workspace and Simulink What enables EJS to read and access these vari ables is the special comment at the end of each line This comment always starts with the keyword EJS immediately after the comment character The keyword must be followed by one of the following words Model Variable or Para meter Model is used to tell EJS the name of the Simulink file that describes the model Variable provides EJS with a description of a variable of the Simulink model A variable can be declared to be accessible for reading writing or both If no modifier is indicated the variable can be freely accessed and EJS will synchronize its value with the associated EJS variable before and after every integration step of Simulink as for example happens with the positions of the magnets variables y1 and y2 If declared as InputOn1ly however EJS will understand that the variable is not to be changed by the Simulink model and will always force its value to what ever it is within EJS In our example the masses variables m1 and m2 and the control actions u1 and u2 are input only variables that will be set by the user through the simulation inte
21. del and the view is what turns the simulation into a dynamic interactive application This mechanism configures a simple though very effec tive way of designing and building advanced interactive user interfaces Fig 6 The reason for this is that linking is a two way connection When the variables of the model change the view is informed so that it immedi ately displays the new state of the model In return since the view elements have built in interactive capabilities any interaction between the student and the interface immediately affects the model variables that have a connection to it For ex ample let us imagine an EJS simulation of the control level of a tank The tank and the water inside it are represented in the view by means of two coloured rectangles If the dimensions of the tank are modified by dragging with the mouse the corners of the outer rectangle the variables repre senting the tank width and height in the analytical model will reflect this change thus affecting the dynamics of the system Similarly if the water level changes as a consequence of the model s evolution the visible height of the rectangle will change to display the current water level With all the high level information which only the user can provide EJS takes care of all the low level procedures needed to create the final simula tion It will generate the Java code that handles all the internal tasks compile it into Java classes pa
22. ervised 25 Ph D theses and co authored more than 150 conference papers and 100 journal papers Since 2002 he has been President of the Spanish Association of Automatic Control CEA IFAC His scientific activity includes various subjects in the control engineering field computer control of industrial processes model based predictive control robust control modeling and simulation of hybrid systems and control education with special emphasis on remotes and virtual labs Easy Java Simulations An Open Source Tool Carla Martin received her M Sc in Electrical Engineering in 2001 from University Complutense of Madrid She is currently completing her doctorate studies at UNED Spain S Dormido Canto received his M Sc in Electronic Engineering in 1994 from the ICAI and his Ph D in 2001 from UNED Since 1994 he has been working at UNED s Department of Computer Sciences and Automatic Control as an Assistant Professor R D Dormido received her M Sc in Physics in 1994 from University Complutense of Madrid and her Ph D in 2001 from UNED Since 1994 she has been working at UNED s Department of Computer Sciences and Automatic Control as an Assistant Professor R Pastor received his M Sc in Physics in 1994 from University Complutense of Madrid He is currently completing his doctorate studies at UNED Spain A Urquia received his M Sc in Physics in 1992 from University Complutense of Madrid and his Ph D in 2000 from UNED Currently he is
23. f these variables the connection between the EJS variable and the MATLAB vari able will be automatically established Establishing a connection means exactly that a the value of the EJS variable will be pushed to the variable of the model before running Simulink and b the EJS wv kava tiewlaleoes Lh Preyer toe ave ES he Siete eapi E Irr C Introduction Model View i Warables L nitialization Evolution variable will be given back whatever value the Simulink variable has after the model has run one integration step Our Simulink model is now ready to be used EJS will run the Simulink model whenever we insert the Java sentence external step 1 in a page of the evolution section the main loop of EJS Fig 13 A call to this method has the following effects 1 The value of every EJS variable that is con nected to a MATLAB variable except those declared as OutputOnly is pushed to the Simulink model Constraints C Custom Fig 13 Evolution page of the EJS to control an external model running in MATLAB Simulink 808 J Sanchez et al 2 The Simulink model is run exactly one integra tion step 3 The value of all MATLAB variables that are connected to EJS variables except those declared as InputOnly are retrieved from the Simulink model We can now operate EJS in the standard way in order to build our desired view and to link it to the variables of the model in a form that suits our pedagogica
24. facilitate their development An obvious way to achieve this is to select one of the well known software packages in the control engineer ing field such as MATLAB Simulink and to provide a tool that would wrap models with the necessary layers of interactivity without demand ing a deep knowledge of computer programming From the point of view of MATLAB s users the benefits of this approximation is twofold it lets 812 J Sanchez et al them reuse all the MATLAB legacy code and also takes advantage of their experience in the use of this de facto standard This paper shows two examples of the type of control applications that have been created using Easy Java Simulations a software tool that helps create in a very easy way dynamic and interactive scientific simulations in the Java language In the view of the two final developments it is clear that EJS takes a step forward in the use of interactivity and MATLAB in an instructional environment A more detailed explanation of the development of EJS applications using MATLAB Simulink can be found in the EJS documentation The software package and the examples shown in this paper are available for free at http fem um es EJS Finally we would like to emphasize that under no circumstances should instructors forget that their ultimate goal is to arouse students curiosity to discover question and wonder Interactivity is one of the inherent ways for human beings to perform this
25. he upper and lower actuators charac teristics may be assumed to be identical The interaction of the two magnets has a nonlinear force characteristic When two magnets are used in the configuration they are usually placed with like poles facing each other so that they repel When the lower magnet is stationary the upper magnet comes to rest in equilibrium between the upward repulsive magnet force and downward gravitational force The magnetic force 804 J Sanchez et al AS CTE TE E p A G oa Ji 1E 1g 20 Fy b am Wms i Win cee a ee TR ad ri E Wi 4 ca T u ll F j fi P i n L tl i 14 aa ti a0 i wie IH Pisgmei aceton imj Miegrent raparation icmj Scream H Mort adi hee va te Tit T ii f i Tinie T Pet tg in Cem r Cee Lais rpe mpri r irain u itip Perne im Bend F yew hepna mid pene Polarity itt Riia Fig 8 Magnetic levitation virtual lab characteristic may be measured by adding and subtracting weight from the upper magnet and measuring the equilibrium height Using this procedure we obtained the following values for the parameters a 1 65 N b 6 2 cm c 2 69 Nem y d 4 2 cm Implementation of the magnetic levitator with EJS and MATLAB Simulink When developing this type of simulation one of the most important things that the teacher needs to bear in mind is utilising the correct configuration of windows and operating controls of the simula tion in order to fa
26. l needs Fig 14a shows the tree repre sentation of the view elements that we chose for our view see also Fig 8 As an example Fig 14b displays the table of properties for the graphical element mass2 which allows users to resize magnet 2 and change its mass Fig 14c shows how a Java method called ChangeMass2 is linked to the On Drag action of this graphical object In this way when the user drags the object the magnet will be resized and the Java method will be invoked to compute the new value of the magnet mass according to its current dimensions Finally it is worth noting that EJS is capable of running more than one Simulink model from within one single simulation This is simply done by creating two or more separate pages of vari ables each with a different M file and connecting EJS variables to the corresponding MATLAB variables When the system runs EJS will take care of opening as many MATLAB sessions as are needed and of managing all connections An advanced virtual lab the quadruple tank process Our second example of an EJS MATLAB appli cation is the quadruple tank process Originally introduced by Johansson 12 it has received great attention because it presents interesting properties in both control education and research The quadruple tank exhibits complex dynamics in an elegant way Such dynamic characteristics lies hmi p dima pa The ees t E nae Coy P Cl pretin O pases GO EE ij
27. models running in differ ent computers using the Internet to interchange information A virtual lab for the magnetic levitator system The system modeled in this section is the ECP magnetic levitation system model 730 11 Fig 7 shows the free body diagram of two suspended magnets Each magnet is acted on by forces from either driver coil from the other magnet from gravity and from friction modeled as viscous The differential equations for both magnets are my y Fm2 Fiat Fai mg my2 C22 Fm2 F2 Fu mg Cail 2 Miechimkal Cail 2 rod limit of mation Curren Magret As Miana urreni Fig 7 Free body diagram of the magnetic levitator The magnetic force terms are modeled by 11 iy Fu ice 3 Fu Dr 4 Fai ra 5 Fin e 6 Fm2 aaa 7 Y2 Vet y2 N 8 where a b c d and N are constants which may be determined by numerical modeling of the magnetic configuration In this case N 4 and a b cand d have been determined by empirical methods in order to adjust the simulation and experimental results The identification of the parameters of the model was done by measuring the specific input output characteristics of the magnet coil actuators and the magnet magnet interactions as they vary with relative position The strong magnetic field nonlinearity however is inherent to this class of magnetic systems For most control design purposes t
28. non and e relationships among these variables corres ponding to the laws that govern the phenom enon expressed by computer algorithms 2 The control which defines certain actions that a user can perform on the simulation 3 The view which shows a graphical representa tion either realistic or schematic of the differ ent states that the phenomenon can have These three parts are deeply interconnected The model obviously affects the view since a change in the state of the model must be made graphically evident to the user The control affects the model because control actions can and usually do modify the value of variables of the model Finally the view affects the model and the control because the graphical interface can contain components that allow the user to modify variables or perform the predefined actions As mentioned above EJS further simplifies the construction of a simulation by merging the control half into the view half into the model The model is what conveys the scientific part of the simulation and is thus the responsibility and the main interest of the user Control teachers are used to describing their algorithms in terms of mathematical equations and to expressing these using a given computer language or by means of high level tools such as MATLAB Simulink The main target when creating the model of a simula tion is to concentrate on the analytical description of the phenomena the content and
29. raphical view the values of the control signals v1 and v2 are computed and the two PID blocks are selected to drive the Simulink subsystems for tanks 1 and 2 respectively however when the control is manual the values of the variables v1 and v2 are set directly by the user and sent to the MATLAB workspace from which the Simulink blocks v1 and v2 read and pass them to the subsystems for tanks 1 and 2 We also needed to adapt the block diagram for interaction with EJS The mechanism for this is similar to that used for the magnetic levitator example and requires the introduction of enough source and sink blocks as needed to exchange information between Simulink and MATLAB s workspace The second step of the process was to create the M file to inform EJS about accessible MATLAB variables Fig 18 shows an extract of the text file used for this example The third and final step was to open this file from EJS and to connect MATLAB variables with EJS variables Fig 19a shows how EJS variables are connected to the MATLAB variables defined in this M file Once all this was done we could introduce a simple sentence in the form _external step 1 in one of EJS s evolution pages to instruct Simulink to run the block diagram once This way at every evolution step of EJS the two models the linear and the nonlinear advance in a synchronized way generating the state of the system Note that both models share the same user inte
30. rface Finally if declared as OutputOnly EJS will read the variable after every integration step of the Simulink model is run but it will not try to change its value in the Simulink model For instance the time variable of the model is read only since it is the result of Simulink execution after each integration step Parameter declares variables that are defined inside Simulink blocks Ejs makes no distinction between variables and parameters for computa tional use However for technical reasons the comment for a parameter must include informa tion about the Simulink block or menu in which the parameter is defined and the function of this parameter within the block In our example the integration step dt is declared to be the general that is not defined within any block parameter maxstep since it is located in a dialog panel of the menu Parameters located in the Simulink menu Simulation The third and final step in this process consists of connecting the variables of the EJS model with MATLAB s workspace variables which are in turn associated with variables of the Simulink model Fig 12 shows the variables in EJS for our example and how they are connected to MATLAB s variables Notice that we have chosen similar if not identical names for both sets of variables jmadel levitatoar mdl Simulink file EJS Model Yl tFosition of magnet 1 REJS Variable yz tFPosition of magnet 2 EIS Variable Ea tCurrent time ETS Varia
31. rface the nonlinear model developed in Simulink and the linear model written using the features of EJS to define and solve differential equations The use of different models within the same application is completely transparent to the user EJS generates the necessary Java code to exchange the information between MATLAB and EJS variables CONCLUSION Using computer simulations in an instructional context usually implies using computers to build models of real world phenomena in order to help students grasp the fundamentals of the behavior of complex systems Interacting with an instructional simulation can enable learners to gain better en gineering understanding of a real system process or phenomenon through exploring testing hypoth eses and discovering explanations for the mechan isms and processes Furthermore students have an excellent opportunity to experiment with their own ideas in terms of engineering design by simple interaction with the tool Not only can interactive virtual labs be effective in presenting engineering concepts in the classroom they also can be bene ficial in extending students experience in analysis and design assignments This invitation to creativ ity can be most useful when it comes to specialized control engineering student projects In this context many control educators would be more willing to use interactive simulations in their lectures and labs if they had a set of tools that would
32. rovides a set of advanced graphical elements Fig 5 that build on top of both standard Java Swing compo nents containers buttons text fields sliders combo boxes etc and on specific scientific two and three dimensional visualization classes from the Open Source Physics project 10 particles vectors images vector and scalar fields etc These elements can be used in a simple drag and drop way to build the interface The user just needs to design the view so that it will offer a visual ization of the phenomenon appropriate to the desired pedagogical goals In particular the view should encourage students to explore the phenom enon from different engineering perspectives in order to gain a better insight into the system To complete the view its different elements have to be instructed to display on the screen according to the values of the variables of the model Every graphic element of EJS has certain internal values called properties that can be customized to make the element look and behave in a particular way change its displayed value position size etc The user can very easily connect the proper ties of the graphical elements of the view to the value of the different variables of the model Some of these properties can also be instructed to trigger user defined actions typical routines defined in the model when the user interactively changes them This procedure which is called linking the mo
33. ter the use of interactivity in control education and more precisely the construction of instruc tional control virtual labs A brief survey of Easy Java Simulations A quick description of Easy Java Simulations places it in the category of code generators This means that users need to provide only the most relevant core of the simulation s algorithm and the tool automatically generates all the Java code needed to create a complete interactive simulation including a wide range of sophisticated software techniques such as handling computer graphic routines communication protocols multithread ing and others What makes EJS very special within this category of software is that it has been conceived by teachers and for teachers and for students who are more interested in under standing the simulated phenomena than in the underlying computer specific aspects The tool uses an original simplification of the successful Model View Control software paradigm Simulation JMatlinkdass J n i Matlabfiles Ean a Simulinkmodels Fig 3 Building a simulation using EJS and MATLAB Simulink Easy Java Simulations An Open Source Tool MVC and structures simulations in two main parts model and view The MVC paradigm states that a simulation is composed of three parts 1 The model which describes the phenomenon under study in terms of e variables which hold the different possible states of the phenome
34. tions or physical laws that rule the phenomenon under study into computer algo rithms Sometimes though the actual implemen tation of these algorithms can be a difficult task For this reason EJS provides two extra facilities The first is a built in editor and solver for systems of ordinary differential equations ODE The user writes the equations in a way that is similar to how s he would write on a blackboard and the system automatically generates the code that numerically solves the equations using one of several of the standard algorithms provided Fig 4 The editor also includes support to handle simple state events of the ODE The second facility is a connection to MATLAB Simulink that lets users design and solve their models with the help of these tools In Canstralnts Custom O an m iHismami iT Hi REFS 2 Pee y T Velerance baal Fig 4 Edition panel to write the ODE and select the solver 802 J Sanchez et al this second case the model is fully defined by a Simulink block diagram and all that is needed is an M file that informs EJS about the names of the MATLAB Simulink variables so that it can create once the final application is generated a bi direc tional link between the Java coded interactive interface the view and the Simulink diagram for the model Two examples on the use of this option are further described in the following sections In order to help create the view EJS p
35. ults are not as expected the user must enter new parameters and repeat the execution analysis process until s he finds a configuration that is Accepted 2 July 2005 798 satisfactory Proceeding in this way however it is difficult for the user to recognize the relationship between the gradient of change in the results usually the system response in temporal and frequency domains and the values of the input parameters This perception would be considerably improved if the user could work with simultaneous graphical representations of the input parameters and the model s dynamic properties and if there were an instantaneous translation of any modifica tion of an input parameter to the system s response This situation could then be considered true interactivity since a change in the simulation s parameters has an immediate effect on the differ ent views of the system according to the mathe matical relationships that govern the dynamics of the system Figs 1 and 2 present two examples of introducing interactivity into virtual labs Fig 1 shows a virtual lab for testing different control strategies for a Furuta pendulum The next figure Fig 2 corresponds to an interesting application for the study of the theoretical control back ground In such applications the signals inside a plot can be dragged by the mouse thus behaving as input parameters and the new values have an immediate consequence for the other graphical
36. working as an associate professor at UNED Spain 813
Download Pdf Manuals
Related Search