Object-oriented Modeling of Mechatronics Systems in Modelica Using Wrapped Bond Graphs François E. Cellier and Dirk Zimmer ETH Zürich.

Slides:



Advertisements
Similar presentations
Start Presentation September 12, 2013 Teaching Physics by Modeling Prof. Dr. François E. Cellier Computer Science Department ETH Zurich Switzerland.
Advertisements

ARCHITECTURES FOR ARTIFICIAL INTELLIGENCE SYSTEMS
Start Presentation October 11, 2012 Thermodynamics Until now, we have ignored the thermal domain. However, it is fundamental for the understanding of physics.
Start of presentation Mathematical Modeling of Physical Systems © Prof. Dr. François E. Cellier September 20, 2012 Object Oriented Modeling The aim of.
Start of Presentation Mathematical Modeling of Physical Systems © Prof. Dr. François E. Cellier September 20, D Mechanical Systems In this lecture,
Lect.3 Modeling in The Time Domain Basil Hamed
Lecture 3: Signals & Systems Concepts
Start Presentation November 8, 2012 Bond Graphs for Mechanical Systems We shall look today in a bit more detail at the modeling of 1D mechanical systems.
Start Presentation September 27, 2012 The Structural Singularity Removal Algorithm by Pantelides This lecture deals with a procedure that can be used to.
PLANAR KINETICS OF A RIGID BODY: CONSERVATION OF ENERGY
System Analysis through Bond Graph Modeling Robert McBride May 3, 2005.
The robot structure model design 2 Curse 5. Modeling: the robot AcTrMStTk V(t) T(t)  (t) q(t) x(t)
Martin Otter: The New MultiBody Library, July 18, The New Modelica MultiBody Library Martin Otter Deutsches Zentrum für Luft- und Raumfahrt (DLR)
Start Presentation Conference 2008 © Prof. Dr. François E. Cellier March 4, 2008 World3 in Modelica: Creating System Dynamics Models in the Modelica Framework.
Start Presentation November 8, 2012 Planar Mechanics We shall now look at a first application of multi- bond graphs: planar mechanics. We shall notice.
Case Tools Trisha Cummings. Our Definition of CASE  CASE is the use of computer-based support in the software development process.  A CASE tool is a.
The Dymola Bond Graph Library
Start Presentation ICSC 2005: Beijing October 25, 2005 Object-oriented Modeling in the Service of Medicine François E. Cellier, ETH Zürich Àngela Nebot,
Impulse-Bond Graphs Authors: Dirk Zimmer and François E. Cellier, ETH Zürich, Institute of Computational Science, Department of Computer Science Bondgraphic.
Wrapping Multi-bond Graphs: A Structured Approach to Modeling Complex Multi-body Dynamics François E. Cellier and Dirk Zimmer ETH Zürich.
Quantized State System Simulation in Dymola/Modelica Using the DEVS Formalism Tamara Beltrame VTT, Industrial Systems PO Box 1000, VM Espoo, Finland.
Start Presentation Conference 2008 © Prof. Dr. François E. Cellier March 4, 2008 Support for Dymola in the Modeling and Simulation of Physical Systems.
Dirk Zimmer François E. Cellier Institute of Computational Science Department of Computer Science ETH Zürich A bondgraphic modeling tool and its application.
Start Presentation Eurosim 2007, Ljubljana September 10, 2007 Electronic Circuit Modeling and Simulation in Modelica François E. Cellier ETH Zürich, Switzerland.
Start of Presentation © Prof. Dr. François E. Cellier 3 rd MODPROD Workshop – Keynote February 4, 2009 Technologies for Creating Easily Maintainable Component.
François E. Cellier and Matthias Krebs
The Equivalence between Static (rigid) and Kinematic (flexible, mobile) Systems through the Graph Theoretic Duality Dr. Offer Shai Tel-Aviv University.
Start of Presentation Maple soft Conference 2006 The Need for Formulae Manipulation in the Object-oriented Modeling of Physical Systems François E. Cellier,
© Dirk Zimmer, February 2006, Slide 1 Master Thesis: A Modelica Library for Multibond Graphs and its Application in 3D-Mechanics Author: Dirk Zimmer Adviser:
Start Presentation November 15, th Homework In this homework, we shall exercise the model wrapping technique. To this end, we shall create a mini-library.
Start Presentation November 29, th Homework – Solution In this homework, we shall attempt the modeling of a planar mechanical system. We shall.
Definition of an Industrial Robot
MapleSim and the Advantages of Physical Modeling
T. K. Ng, HKUST Lecture IV. Mechanics of rigid bodies.
Start Presentation November 22, th Homework In this homework, we shall attempt the modeling of a planar mechanical system. We shall do so once.
GENERAL CONCEPTS OF OOPS INTRODUCTION With rapidly changing world and highly competitive and versatile nature of industry, the operations are becoming.
An Introduction to Software Architecture
Start Presentation November 15, th Homework - Solution In this homework, we shall model and simulate a mechanical system as well as exercise the.
Start PresentationSeptember 21, 2009 Modeling Chemical Reactions in Modelica By Use of Chemo-bonds Prof. Dr. François E. Cellier Department of Computer.
Bond Graphs II In this class, we shall deal with the effects of algebraic loops and structural singularities on the bond graphs of physical systems. We.
Building Tools by Model Transformations in Eclipse Oskars Vilitis, Audris Kalnins, Edgars Celms, Elina Kalnina, Agris Sostaks, Janis Barzdins Institute.
MathCore Engineering AB Experts in Modeling & Simulation WTC.
ECO-BOND GRAPHS An Energy-Based Modeling and Simulation Framework for Complex Dynamic Systems with a focus on Sustainability and Embodied Energy Flows.
SOFTWARE DESIGN.
Design engineering Vilnius The goal of design engineering is to produce a model that exhibits: firmness – a program should not have bugs that inhibit.
Introduction to SimCreator and Multi-Body Dynamics  April 25, 2005.
Start Presentation October 11, 2012 The Theoretical Underpinnings of the Bond Graph Methodology In this lecture, we shall look more closely at the theoretical.
Start Presentation December 6, th Homework – Solution In this homework, we shall model and simulate a 3D mechanical system with proper animation.
November 21, 2005 Center for Hybrid and Embedded Software Systems Example To illustrate how changes in DB can be used to efficiently update a block diagram,
ME451 Kinematics and Dynamics of Machine Systems Review of Linear Algebra 2.1, 2.2, 2.3 September 06, 2013 Radu Serban University of Wisconsin-Madison.
Start Presentation November 24, th Homework In this homework, we shall exercise the model wrapping technique. To this end, we shall create a mini-library.
Start Presentation October 25, th Homework - Solution In this homework, we shall exercise the modeling of a simple electrical circuit using bond.
Start of presentation September 27, 2012 Algebraic Loops and Structural Singularities The sorting algorithm, as it was demonstrated so far, does not always.
Start Presentation November 1, 2012 Treatment of Discontinuities II We shall today once more look at the modeling of discontinuous systems. First, an additional.
Modeling the ODP Computational Viewpoint with UML 2.0: The Templeman Library Example José Raúl Romero, Antonio Vallecillo Universidad de Málaga, Spain.
© Maplesoft, a division of Waterloo Maple Inc MapleSim and the Advantages of Physical ModelingMapleSim and the Advantages of Physical Modeling.
Engineering Self-adaptive Service Mashups Mahdi Bashari LS3 Colloquium May 7 th 2014.
Instructor :Kashif Mehmood
Finite State Machines (FSM) OR Finite State Automation (FSA) - are models of the behaviors of a system or a complex object, with a limited number of defined.
E-learning Some examples of new ICT tools in action - Alan Coop, July 2011.
Multi-bond Graphs We shall today look at vectors of bonds, called multi-bonds. Especially when dealing with 2D and 3D mechanics, the d’Alembert principle.
Lecture VIII: Software Architecture
Start Presentation October 4, 2012 Efficient Solution of Equation Systems This lecture deals with the efficient mixed symbolic/numeric solution of algebraically.
Canonical Equations of Motion -- Hamiltonian Dynamics
ME451 Kinematics and Dynamics of Machine Systems Start Position, Velocity, and Acc. Analysis 3.6 October 21, 2010 © Dan Negrut, 2010 ME451, UW-Madison.
Texas A&M University, Department of Aerospace Engineering AUTOMATIC GENERATION AND INTEGRATION OF EQUATIONS OF MOTION BY OPERATOR OVER- LOADING TECHNIQUES.
SysML and Modelica Integration Working Group Meeting 3/11/09 Peter Fritzson Wladimir Schamai.
Structured Modeling of Mechatronic Systems in which you meet the modest but talented multiport component.
An Introduction to Software Architecture
5th Homework In this homework, we shall exercise the modeling of a simple electrical circuit using bond graphs. We shall also model the same electrical.
Presentation transcript:

Object-oriented Modeling of Mechatronics Systems in Modelica Using Wrapped Bond Graphs François E. Cellier and Dirk Zimmer ETH Zürich

Outline Motivation Graphical Modeling Modeling Software Requirements Bond Graphs Electronic Circuits Multi-bond Graphs 3D Mechanics

Motivation In today‘s engineering practice, mathematical models of physical processes are frequently produced that are composed of thousands of equations. These models are difficult to create and even more difficult to maintain. A typical example of systems leading to highly complex models are mechanical multi-body systems. Tools are needed that enable us to keep the complexity of individual component models within limits. Model wrapping presents itself as a tool suitable for such purpose.

Graphical Modeling Graphical modeling is generally more suitable for the creation of models of complex systems than equation- based modeling. This is true because graphical models are naturally two-dimensional. Errors in hierarchically structured and topologically interconnected graphical models are usually discovered more easily and rapidly than errors in corresponding equation-based models. Evidently, the graphical models must be replaced by equation-based models at the lowest-possible level in the modeling hierarchy.

Example: 3D Mechanics

Example: 3D Mechanics II

Example: 3D Mechanics III

Example: 3D Mechanics IV

Software Requirements The semantic distance from the lowest graphical layer to the equation layer should be kept as small as possible. In this way, as much as possible can be modeled graphically. Mechanical multi-body component models are too complex to be used conveniently as building blocks of the lower-most graphical modeling layer. To this end, multi-bond graphs are considerably more suitable, as we shall demonstrate.

Bond Graphs: Example

Bond Graphs: Example II

Causal Bond Graphs u 0 = f(t) i 0 = i 1 + i L u L = u 0 di L /dt = u L / L v 1 = u 0 u 1 = v 1 – v 2 v 2 = u C i C = i 1 – i 2 du C /dt = i C / C u 2 = u C i 2 = u 2 / R 2 i 1 = u 1 / R 1

Advantages of Bond Graphs Bond graphs represent a generally usable approach to modeling physical systems of arbitrary types. They offer a suitable balance between general usability and domain orientation. The concepts of energy and power flows define a suitable semantic framework for bond graphs of all physical systems. The semantic meaning of each bond graph component model is sufficiently simple to afford easy maintainability of the equation layer below.

The BondLib Library of Dymola Bond graphs can be drawn graphically on the computer. The resulting model can be simulated immediately. The library affords application specific solutions, such as a sub-library for electrical circuits.

Multiple Modeling Interfaces in BondLib

Electronic Circuit Modeling in BondLib The Bipolar Junction Transistor Diagram window Icon window

Electronic Circuit Modeling in BondLib II

Electronic Circuit Modeling in BondLib III

Electronic Circuit Modeling in BondLib IV

Electronic Circuit Modeling in BondLib V

Electronic Circuit Modeling in BondLib VI

Electronic Circuit Modeling in BondLib VII Inverter Circuit

Electronic Circuit Modeling in BondLib VIII Initial number of equations Final number of equations Simulation Time

Electronic Circuit Modeling in BondLib IX

Bond Graphs For Mechanical Systems Mechanical systems are three-dimensional. Every mechanical body that can move freely has six degrees of freedom. For this reason, the dAlembert principle must be formulated six times for each mechanical body. Mechanical bond graphs have a tendency of quickly becoming very large. Holonomic constraints cannot be formulated directly in the bond graph.

Example: A Planar Pendulum Holonomic Constraint

Causality Conflict Example: A Planar Pendulum II dv x dt m = -F sin(φ) dv y dt m = -F cos(φ) + mg x = ℓ sin(φ) v x = ℓ cos(φ) φ. y = ℓ cos(φ) v y = -ℓ sin(φ) φ. 1 I :m vxvx vxvx F sin(φ) FxFx 1 I :m Se :mg vyvy vyvy vyvy F cos(φ) FyFy mg 0 F sin(φ) ℓ cos(φ) φ. 0 F cos(φ) ℓ sin(φ) φ. vxvx 0 φ = asin( x / ℓ ) Dq x TF φ

 We shouldn’t have to derive the equations first in order to be able to derive the bond graph from them.  The resulting bond graph didn’t preserve the topological properties of the system in any recognizable form. Mechanical Bond Graphs It has been possible to describe the motion of the planar pendulum by a bond graph enhanced by activated bonds for the description of the holonomic constraint. Unfortunately, the bond graph doesn’t tell us much that we didn’t know already.

Multi-bond graphs are a vectorial extension of the regular bond graphs. A multi-bond contains a freely selectable number of regular bonds of identical or similar domains. All bond graph component models are adjusted in a suitable fashion. Multi-bond Graphs } f3vf3v tt fyvyfyvy fxvxfxvx Composition of a multi-bond for planar mechanics

The MultiBondLib Library A Dymola library for modeling systems by means of multi-bond graphs has been developed. The library has been designed with an interface that looks as much as possible like that of the original BondLib library. Just like the original library, also the new multi-bond graph library contains sub-libraries supporting modelers in modeling systems from particular application domains, especially from mechanics.

Example: A Planar Pendulum III Multi-bond graph of a planar pendulum

Multi-bond Graphs: 2 nd Example Model of a crane crab Prismatic Joint Revolute Joint Rod Mass 2 Mass 1

Multi-bond Graphs: 2 nd Example II

WallPrismatic Joint Mass 1 Revolute JointRod Mass 2

Multi-bond Graphs: 2 nd Example III Wrapper models Mechanical connectors

Model Wrapping  On the upper mechanical layer, an intuitive and simple to use interface is being offered.  The lower multi-bond graph layer offers a graphical interpretation that makes it possible to decompose even complex mechanical component models graphically into much simpler subcomponent models. Model wrapping offers the best properties of two worlds:

3D Mechanics: Example Multi-bond graph model of an uncontrolled bicycle

3D Mechanics: Example II Multi-body diagram of an uncontrolled bicycle

Multi-bond Graphs for 3D Mechanics: Multi-bond graphs offer too low an interface to be used for modeling multi-body systems of 3D mechanics directly. The basic multi-bond graph component models are not at the right modeling level to carry meaningful multi-body system semantics. Consequently, multi-bond graphs of even fairly simple multi-body systems become quickly unreadable and therefore also poorly maintainable.

Multi-body Diagrams for 3D Mechanics: Multi-body diagrams are easily interpretable, as their component models carry semantics that can be mapped one-to- one to those of the underlying physical system to be modeled. The standard Dymola library offers a multi-body library that is user-friendly and therefore widely used. However, the component models of that library have been implemented using matrix-vector equations directly. These models are therefore difficult to understand and maintain. The multi-bond graph library of Dymola offers a sub-library for 3D mechanics that re-implements the standard multi-body library. Yet, each of its component models has been internally realized as a multi-bond graph.

3D Mechanics: Example IV State variables: FrontRevolute.phi RearWheel.phi[1] RearWheel.phi[2] RearWheel.phi[3] RearWheel.phi_d[1] RearWheel.phi_d[2] RearWheel.phi_d[3] RearWheel.xA RearWheel.xB Steering.phi 2 systems of 3 and 15 linear equations, resp. 1 non-linear equation Simulation 20 sec, 2500 output points 213 integration steps 0.7s CPU-time Plot window: Lean Angle

3D Mechanics: Example III State variables: FrontRevolute.phi RearWheel.phi[1] RearWheel.phi[2] RearWheel.phi[3] RearWheel.phi_d[1] RearWheel.phi_d[2] RearWheel.phi_d[3] RearWheel.xA RearWheel.xB Steering.phi 2 systems of 3 and 15 linear equations, resp. 1 non-linear equation Simulation 20 sec, 2500 output points 213 integration steps 0.7s CPU-time Animation Window:

Animation Dymola offers means for animating models of mechanical systems. This is another reason, why multi-body diagrams are important. It is not meaningful to try to animate a multi-bond graph. Multi-bonds don’t carry suitable semantics for connecting them with an animation model. In contrast, the basic building blocks of an animation model are exactly identical to the multi-body component models. Therefore, animation models can be easily associated with multi-body component models, and this is the approach that Dymola took. Animation models have also been associated with the component models of the multi-body sub-library of the multi- bond graph library.

Simulation Run-time Efficiency The run-time efficiency of the generated simulation code of a multi-body system model depends strongly on the selection of suitable state variables.

Simulation Run-time Efficiency II The run-time efficiency of the generated simulation code using the standard multi-body and the 3D mechanics sub-library of the multi-bond graph library is essentially the same. For simple models, the generated equations are identical. In more complex cases, the equations may differ slightly, because the connectors of the two libraries are not identical. Whereas the standard multi-body library carries for rotational dynamics only angles and torques in its connector, the multi- bond graph library carries angles, angular velocities and torques. This occasionally leads to slightly different constraint equations that will reflect upon the final set of generated simulation equations.

Multiple Interfaces in MultiBondLib

Conclusions Dymola offers a consequent and clean implementation of the principles of object-oriented modeling of physical systems. Dymola supports graphical encapsulation of models, topological interconnection of component models, and hierarchical decomposition of models. Model wrapping is essentially nothing new. It provides simply a systematic interpretation of the object-oriented modeling paradigm. Whereas object-oriented modeling provides no guidance as to how models should be encapsulated, the model wrapping paradigm, through the wrapper models, provides clean and consistent connectors at each layer of the model hierarchy.

Conclusions II Bond graphs (regular bond graphs, multi-bond graphs, thermo-bond graphs) offer the lowest graphical modeling framework that still carries physical meaning. The semantic distance from the bond graph down to the equation layer below is sufficiently small, so that the bond graph libraries are easily maintainable. The bond graph models can then be wrapped to carry the semantics of the component models up to a suitable level that specialists of the application domain are familiar with. All of the wrapping is done graphically, i.e., there is no equation modeling beyond the level of the basic bond graph component models.

The End