沖壓機(jī)床液壓控制系統(tǒng)設(shè)計(jì),沖壓,機(jī)床,液壓,控制系統(tǒng),設(shè)計(jì) mented in the AMESim simulation tool. Body and joint components are the basic components 1569-190X/$ - see front matter C211 2005 Elsevier B.V. All rights reserved. * Corresponding author. Tel.: +33 4 72 43 85 58. E-mail address: wilfrid.marquis-favre@insa-lyon.fr (W. Marquis-Favre). Simulation Modelling Practice and Theory 14 (2006) 25–46 of this library. Due to the library philosophy requirements, the mathematical models of the components have required a generic vector calculus based formulation of the constraint equa- tions. This formulation uses a set of dependent generalized coordinates. The dynamics equa- tions are obtained from the application of JourdainC213s principle combined with the Lagrange multiplier method. The body component mathematical models consist of di?erential equations in terms of the dependent generalized coordinates. The joint component mathematical models are based on the Baumgarte stabilization schemes applied to the geometrical, kine- matic and acceleration constraint equations. The Lagrange multipliers are the implicit solution of these Baumgarte stabilization schemes. The first main contribution of this paper is the expression of geometrical constraints in terms of vectors and their exploitation in this form. The second important contribution is the adaptation of existing formulations to the AMESim philosophy. C211 2005 Elsevier B.V. All rights reserved. A planar mechanical library in the AMESim simulation software. Part I: Formulation of dynamics equations Wilfrid Marquis-Favre * , Eric Bideaux, Serge Scavarda Laboratoire d’Automatique Industrielle, Institut National des Sciences Applique′es de Lyon, Ba?t. St Exupe′ry, 25, avenue Jean Capelle, F-69621 Villeurbanne Cedex, France Received 25 March 2003; received in revised form 17 December 2004; accepted 8 February 2005 Available online 17 March 2005 Abstract This paper presents the mathematical developments of a planar mechanical library imple- doi:10.1016/j.simpat.2005.02.006 domains. One can now carry out modeling, analysis and simulation for systems con- sisting of pneumatic, powertrain, hydraulic resistance, thermal, electromagnetic and 26 W. Marquis-Favre et al. / Simulation Modelling Practice and Theory 14 (2006) 25–46 cooling components for instance. The restriction to only one-dimensional motion for the mechanical components has motivated the development of a two-dimensional mechanical library. Keywords: AMESim; Planar mechanics; Dynamics equations; Constraint equations; Lagrange multi- pliers; Baumgarte stabilization 1. Introduction This paper, organized in two parts, presents a new library for the simulation tool AMESim [2]. The first part is dedicated to the theoretical developments of the library. The second part shows the composition of the library as it was primarily implemented in AMESim and illustrates it with an application example of a seven-body mechanism. This library proposes components belonging to the planar mechanical domain. The objective with this library was not to compete with multi- body system software tools that are better adapted to this domain. The objective was more to enlarge the range of industrial applications capable of being treated by AMESim. From a theoretical point of view the challenge of implementing this library was to fit existing mechanical formulations to the inherent requirements of AMESim philosophy. The solution has been found by adapting the dynamic equa- tions expressed from JourdainC213s principle and the Lagrange multiplier method together with BaumgarteC213s stabilization. Also a generic feature of the formulation has been researched over the library components (bodies and joints) and one key contribution of this paper is concerned with this generic feature. Basically the formu- lation consists of expressing the geometric constraints associated with joints in terms of vectors and carrying out the developments of this form. The result is the set up, for kinematic and acceleration constraints, of a unique expression that fits every joint presented in the library. The generic feature of the formulation proposed thus enables the derivation of joint contraints to be systematized. One can then imagine a new joint with its corre- sponding vector constraint and derive straightforwardly the corresponding mathe- matical model by applying the proposed formulation. Also, in the context of predefined component models, the given formulation clearly shows the frontiers of the di?erent mathematical models in terms of inputs and outputs. Therefore it also helps to define in which models output equations must be implemented. Also, the formulation proposed intrinsically enables closed loop structures to be dealt with. AMESim (for Advanced Modeling Environment for performing Simulations) is organized in component libraries. The components, represented by symbolically technologically suggested icons, can be interconnected exactly like the system under study. AMESim was first applied to electrohydraulic engineering systems with simple one-dimensional mechanical systems (like inertia, springs, and dampers in transla- tion or in rotation). It recently opened its libraries to a variety of other component ally reduced and a number of constraints are a priori symbolically eliminated. Like- W. Marquis-Favre et al. / Simulation Modelling Practice and Theory 14 (2006) 25–46 27 wise, tools based on bond graph (e.g. 20Sim [1] or MS1 [18]) can deal with multibody systems in a pluridisciplinary context (e.g. [4,7]). The essential feature of bond graph language is its ability to describe the energy topology of a model at an acausal level. This enables all the model variables to be globally assigned and all the equations to be globally organized. This also eliminates superfluous dependencies of the multi- body models. Section 2 presents an overview of some multibody codes and object-oriented tools, as well as the environmental requirements of AMESim. These requirements have some implications on how the 2D library is built. Section 3 details the theoret- ical developments that enabled the mathematical models of the library components to be set up. Section 4 concludes this first part. 2. Constraints of AMESim library philosophy After a brief overview of multibody code principles and some object-oriented tools, a presentation of AMESim requirements is given. Concerning multibody codes a state of the art is given by [23]. Details are not reproduced here and readers are referred to this book for a more profound presen- tation. Although more than a decade has passed and certain tools are no longer developed and others have changed, this state of the art book gives a good idea of the main principles that can be used as a basis for multibody codes. Also this over- view enables the library proposed to be positioned with respect to these codes. There are di?erent approaches for writing dynamic equations. The approaches most repre- sented in multibody codes are, the Newton–Euler equations applied to each body, the Newton–Euler equations applied to sets of bodies, LagrangeC213s equations and KaneC213s equations [13,14]. The variables, in whose terms the dynamic equations are written, are either absolute coordinates or relative coordinates. Also supplementary methods are used for reducing the index of the Di?erential–Algebraic Equations. The principal ones are the coordinate partitioning method, the projection matrix method, the Baumgarte stabilization and the penalty formulation [9]. The first two methods aim at working with a set of independent generalized coordinates while the Baumgarte stabilization enables the constraints, together with the di?erential equations, to be handled and the penalty formulation increases the di?erential sys- tem order by introducing extra dynamics into the model. In the domain of the object-oriented tools to which AMESim may be attached, certain enable multibody systems to be treated with a di?erent approach to the mod- elling. For instance Dymola [21] is, like AMESim, based on well-identified techno- logical components in a pluridisciplinary context but it sets up the mathematical model in a di?erent way. Basically each component model consists of equations not oriented in terms of variable assignments nor organized a priori. Then, at the component connection stage, all the mathematical models are gathered in an implicit form and the compilation carries out the variable assignments and the organization of the equations in a consistent manner. Thus the order of the whole model is glob- It is now important to show the key features of AMESim to justify how the planar mechanical library has been implemented. Its feature oriented towards engineering systems and its user friendliness make AMESim work with well-identified technolog- ical components, symbolically manipulated by means of icons. These icons are inter- connected, one to the other and identically to the engineering system architecture under study. Fig. 1 shows an example of a door locking system using a permanent magnet modelled in AMESim. The icons displayed here belong to the magnetic, mechanical and signal libraries. This simple example shows the coupling between mechanical and magnetic domains where one circuit, fed by a permanent magnet (right-hand side magnetic circuit), is forced to move with respect to another passive circuit (left-hand side circuit). The main components consist of a permanent magnet (rectangle with a compass needle inside), three magnetic circuit parts characterized 28 W. Marquis-Favre et al. / Simulation Modelling Practice and Theory 14 (2006) 25–46 by a certain reluctance (rectangles with C212squareC213 ports with a diagonal cross inside), two variable air-gaps (vertical twin rectangles), two mechanical nodes (both sides of the air-gap components), a signal generator with a signal-to-displacement converter (in the centre of the right-hand side circuit), and a component for the set of the mag- netic medium characteristics (B–H diagram in a circle). Each component can be associated with one model from a set of component compatible mathematical models. As soon as the model has been chosen the component conserves this mathematical model. Contrary to acausal tools, AMESim works with component models that have equations both a priori oriented in terms of variable assignments and organized. This feature requires implementing new models in a predefined calculus scheme. Also the mathematical formulation of a component model has to be organized in order to fit into other potential component connections. So each component associated with a mathematical model has a predetermined set of input and output variables. It can thus be considered as a causal model. The connection of the components enables the exchange of those variables on the way out a component for those variables that are calculated by its mathematical model (outputs) and, the exchange of those vari- ables on the way in a component for those variables that are calculated by a con- nected component mathematical model (inputs). This causal feature of AMESim philosophy is the main constraint when implementing new components. This di?ers Fig. 1. Example of an AMESim model representation. from other object-oriented tools, based on acausal component models or acausal phenomenon models, like Dymola, or tools with a bond graph input (e.g. 20-Sim or MS1). Fig. 2 gives an example of two components in the mechanical (a mass in transla- tion) and the hydraulic (a two way hydraulic pump) domains respectively. The con- necting ports of the components show the variables exchanged by them and especially the outputs (C212exitingC213 arrows) passed to the connected components and the inputs (C212enteringC213 arrows) received from the connected components. These con- necting ports are intimately associated with power ports since two of the variables W. Marquis-Favre et al. / Simulation Modelling Practice and Theory 14 (2006) 25–46 29 exchanged at these ports are power variables. Fig. 2 examples illustrate two key features of a library oriented simulation tool. The first one is the domain port concept. It shows how AMESim can deal with plu- ridisciplinary systems. The second feature is the connecting port constraints. Since one component mathematical model requires given inputs to then calculate its state and its outputs, not all combinations of the component connections are allowed. For instance the Fig. 2 examples cannot be connected one to the other by any port. How- ever a mass component may be connected to a spring component or a damper component. Another key feature of a library oriented simulation software tool is the modular- ity concept. This often results in symmetrical components with respect to their con- necting port. This symmetry property, though not generalized to all components in AMESim, has been adopted for the planar mechanical library. The reason will ap- pear obvious when components of this library are presented. In the context of planar mechanisms and rigid bodies the library is not restricted to any mechanical domain application. The library also accepts closed loop struc- tures. Although relative coordinates are generally more e?cient for dynamic equa- tion formulation, AMESim philosophy requires the use of absolute coordinates. The absolute coordinates of the mass center have been chosen for each body. Nev- ertheless the planar feature of the library does not require any specific variables for the body orientation. Thus the absolute angular position has been chosen for each body as well. Once again, due to AMESim philosophy, the equations of the compo- nents cannot be globally reorganized when the components are connected. This for- bids the use of the coordinate partitioning method or the projection method to decrease the index of the Di?erential–Algebraic Equation systems. For this reason the Baumgarte stabilization has also been used in the library. Fig. 2. Example of two AMESim components. 3. Theoretical developments of the library components As has been explained in the previous section the library must be organized in 30 W. Marquis-Favre et al. / Simulation Modelling Practice and Theory 14 (2006) 25–46 with m i the body mass and g the gravity acceleration. We consider here ~y 0 as the ascendant vertical axis. The star superscript indicates virtual quantities. The coe?- cients of the virtual velocities in A* are derived from the kinetic coenergy (e.g. [6]) of a body by the equation: A q ? doT dto_q C0 oT oq with q a generalized coordinate e3T 1 A nomenclature is given in Appendix A. well-identified technological components. It has been decided to base the planar mechanical library on a body component and on joint components. The body com- ponent is associated with a supposed rigid material item of a mechanism. Its behav- ior is essentially governed by its kinetic state. The joint components are associated with the abstract items that represent the attachment of bodies in a mechanism. They are supposed to be ideal and their mathematical model is based on the constraints that they impose on the connected bodies. 3.1. Body component mathematical model The mathematical model of the body component is based on JourdainC213s Principle formulation (e.g. [5,23]) 1 : A C3 ? P C3 e1T where A* is the virtual power developed by the acceleration quantities and P* the virtual power developed by the actions on the body. In the library philosophy there is no a priori privileged candidate for the role of the generalized coordinates. For a planar motion, the generalized coordinates, which have been chosen, are the absolute mass center coordinates projected onto the absolute frame of reference ex G i ;y G i T and the absolute angular position h i (Fig. 3). This choice enables the more general case of a body motion to be dealt with. The body motion restriction will be determined by the joint constraints, as shown later. With this choice of generalized coordinates ex G i ;y G i ;h i T Eq. (1) members may now be written A C3 ? A x _x C3 G i tA y _y C3 G i tA h _ h C3 i P C3 ? Q x _x C3 G i teQ y C0m i gT_y C3 G i tQ h _ h C3 i e2T W. Marquis-Favre et al. / Simulation Modelling Practice and Theory 14 (2006) 25–46 31 Applied to Fig. 3 body in planar motion these quantities are written simply: A x ? m i €x G i A y ? m i €y G i with I i the body moment of inertia around eG i ;~z 0 T A h ? I i € h i e4T Q x , Q y , and Q h are the generalized forces including the constraint actions resulting from the fact that x G i , y G i , and h i are not necessarily independent after the connection of a body component to a joint component. From Eq. (1) and by taking a compat- ible virtual transformation with the joints as they exist at time t, we can now write the three identities that constitute the formulation basis for the body components. These three identities are m i €x G i ? Q x m i €y G i ? Q y C0m i g I i € h i ? Q h e5T Fig. 3. Schema of a body in planar motion. This formulation requires that the expression of the three generalized forces Q x , Q y , and Q h be further developed in order to fit any potential connected joint component. First let us inspect the case of a body with only one connecting port at a point M. Let us also consider simply a given action on the body characterized by a wrench about point M (e.g. [17]): fWg : ~ F ? F x ~x 0 tF y ~y 0 force ~ MeMT?C z ~z 0 torque about point M ( e6T The virtual power developed by this action is P C3 ? ~ F C1 ~ V 0 C3 eMTt ~ MeMTC1 ~ X 0 C3 i e7T where ~ V 0 C3 eMT is the virtual absolute velocity of point M and ~ X 0 C3 i is the virtual abso- lute angular velocity of the body. The velocity transport (e.g. [11]) enables Eq. (7) to be written as: 32 W. Marquis-Favre et al. / Simulation Modelling Practice and Theory 14 (2006) 25–46 3.2. Joint component mathematical model First a general formulation is given for the joint component mathematical model. It is then illustrated in the example of a translational joint. Let us consider this time two bodies connected by a joint. By the only fact that both bodies are connected (a joint component between two body components) their generalized coordinates (x G i , y G i and h i for body i and x G j , y G j and h j for body j) are no longer independent. In the library philosophy the constraint equations are ex- pressed in the joint component, which in turn furnishes the constraint actions to the body components. These constraint actions correspond to the variables F x , F y , and C z previously presented and passed to each body component. The general expressions of these variables are now determined. The joints considered in the planar mechanical library generate only geometrical constraints. These constraints may be expressed in a general way in an implicit form by Eq. (10) (e.g. [15]). g k eq 1 ; ...;q n T?0 for k ? 1tom e10T with n the number of generalized coordinates involved in the constraints and m the constraint number. It is supposed here that the constraints are scleronomic [16], which means that time does not explicitly appear in the constraint equations. At the kinematic level these equations become P C3 ? ~ F C1 ~ V 0 C3 eG i Tt ~ F C1 ~ X 0 C3 i C2G i M C131C131! C16C17 t ~ MeMTC1 ~ X 0 C3 i ? ~ F C1e_x C3 G i ~x 0 t _y C3 G i ~y 0 Tt ~ MeMTtG i M C131C131! C2 ~ F C16C17 C1 _ h C3 i ~z 0 e8T From Eq. (8) we can clearly identify the generalized forces used in the dynamic for- mulation of a body component: Q x ? ~ F C1~x 0 ? F x Q y ? ~ F C1~y 0 ? F y Q h ? ~ MeMTtG i M C131C131! C2 ~ F C16C17 C1~z 0 ? C z t ~ F C1e~z 0 C2G i M C131C131! T e9T Since G i M C131C131! is a characteristic vector of the body, the variables F x , F y , and C z , char- acterizing the given force at point M, are the only variables passed to the body at the connecting port. The variables Q x , Q y , and Q h are calculated in the body component model on Eq. (9) basis. It is shown in the next section that th