Modeling of Electromechanical Systems

Werner Haas, Kurt Schlacher and Reinhard Gahleitner Johannes Kepler University Linz, Department of Automatic Control and Control Systems Technology Altenbergerstr.69, A4040 Linz, email: RichODL@mechatronik.uni-linz.ac.at December 20, 2000

Contents
1 Motivation 2 Mechanical Systems 2.1 Derivation of the Lagrange Equations . . . . . 2.1.1 Example I . . . . . . . . . . . . . . . . . 2.1.2 Example II . . . . . . . . . . . . . . . . 2.2 Variational Principle and Lagranges Equations 2.3 State Functions . . . . . . . . . . . . . . . . . . 2.3.1 Example I . . . . . . . . . . . . . . . . . 2.3.2 Example II . . . . . . . . . . . . . . . . 2.4 Energy of Mechanical Systems . . . . . . . . . 2.5 Legendre Transformations . . . . . . . . . . . . 2.6 Case Studies . . . . . . . . . . . . . . . . . . . 2.6.1 Atwoods Machine . . . . . . . . . . . . 2.6.2 Car and Beam . . . . . . . . . . . . . . 2.6.3 Double Pendulum . . . . . . . . . . . . 2.6.4 Bead and Hoop . . . . . . . . . . . . . . 2.6.5 Ball on a Wheel . . . . . . . . . . . . . 2.6.6 Two dimensional truck model . . . . . . 3 Electrical Systems 3.1 Energy and Coenergy of Simple Devices 3.1.1 Static terminals . . . . . . . . . . 3.1.2 Dynamic Devices . . . . . . . . . 3.1.3 Example . . . . . . . . . . . . . . 3.2 Equations of Motion of Simple Networks 3.2.1 Example I . . . . . . . . . . . . . 3.2.2 Example II, CukConverter . . . 3.3 The Energy of Electrical Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 5 5 8 9 9 11 13 15 15 16 17 17 18 18 18 18 18 19 20 20 22 24 24 27 28 33

CONTENTS

CONTENTS

4 Electromechanical Systems 4.1 Introduction . . . . . . . . . . . . 4.2 Electromechanical Coupling . . . 4.2.1 Magnetic Field Coupling . 4.2.2 Electrical Field Coupling 4.3 Equations of Motion . . . . . . . 4.4 Electrical Drives . . . . . . . . . 4.4.1 Elementary Machine . . . 4.4.2 DCdrive . . . . . . . . . 4.5 Case Studies . . . . . . . . . . . 4.5.1 WardLeonard drive . . . 4.5.2 Ball in a Magnetic Field . 4.5.3 Electromagnet . . . . . . 4.5.4 Relay Device . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

34 34 36 37 39 41 42 42 43 45 45 45 45 45

1 MOTIVATION

Motivation
Why should we study modeling?

For the answer we need some basic concepts and denitions. We start with a denition for the term system: A system is characterized by the fact that we can say what belongs to it and what does not. We can specify how it interacts with its environment Hence it can be observed and controlled. There are variables that are generated by the environment and that inuence the behavior of the system. These are called the inputs of the system. There are other variables that are determined by the system and that in turn inuence the behavior of its environment. These are called the outputs of the system. This leads to a possible denition for the term system [1]: A system is a potential source of data This leads to a denition for the term experiment: An experiment is the process of extracting data from a system by exerting it through its inputs To perform an experiment on the system means to apply a set of external conditions to the inputs and to observe the reaction of the system to these inputs by recording the behavior of the outputs. The major disadvantage with real systems is the fact that these systems are under the inuence of inaccessible inputs (so called disturbances) and a number of useful outputs are not accessible through measurements. The denition of a system and an experiment gives a way to dene the term model: A model M for a system S and an experiment E is anything to which E can be applied in order to answer questions about S In the present course, we concentrate ourselves to a subclass of models which are called mathematical models. This is a description of the relationship among the system variables in terms of mathematical equations. By performing experiments, we collect knowledge about a system. Modeling means the process of organizing knowledge about a given system In the beginning, this knowledge is unstructured. By understanding what are causes and what are eects, we organize the knowledge. In fact, we are engaged in a process of modeling. The major task in which a model is to be used has basic implications on the choice of the particular form of a model. In other words, a model can be considered as a specialized tool, developed for a particular application. Of course, such approach leads to dierent models for dierent uses of the same system. In particular a control engineer uses a model for the development of control algorithms. Thus models, for control reasons, should be as simple as possible. Often a model which takes into account the rst order eects is adequate. A well designed controller suppresses the second order eects. Another point is that the eort of the controller development is often related to the complexity of the model. Simulation engineers want to develop models which t the experimental data in an appropriate way. This approach may lead to more complex models. There are dierent types of mathematical models [1], [6]. We concentrate our investigations to lumped parameter models which are described by ordinary dierential equations of the form dx = f (x, u, t) . dt Moreover, we are interested especially in electromechanical systems.

1 MOTIVATION

How do we perform the process of modeling? 1. We start with a physical description of the dynamical systems. This includes a discussion of physical eects which should be taken into account. 2. Calculation of a single quantity which determines the time behavior of the dynamical system. Later, we will show that this quantity is called the extended Lagrangian. 3. Derivation of the equations of motion. A computer algebra program (MAPLE V) is used to derive the mathematical model in an automatic way. 4. Simulation code (MATLAB or DYNAST) for the model is generated automatically. 5. Simulation

Summary The derivation of the equations of motion (the mathematical model) can be obtained from variational principles applied to energy functions. There exists a well established common terminology for all type of systems, whether electrical, mechanical, magnetic, etc., by dening energy functions in terms of the generalized coordinates. Then by the use of a single fundamental postulate, e.g. Hamiltons principle, the equations of motion are determined. The variational approach is quite formal analytically and as a result insight into physical processes can be lost in the mathematical procedure. Nevertheless, if the method is properly understood, physical insight can be gained due to the generality of the method. There are a number of dierent energy functions (e.g. the Lagrangian, the total energy) which can be used as a energy function. In this course the modeling of purely mechanical systems is mainly based on the Lagrangian which is a function of the generalized coordinates and the associated velocities. If all forces are derivable from a potential, then the time behavior of the dynamical systems is completely determined. For simple mechanical systems, the Lagrangian is dened as the dierence of the kinetic energy and the potential energy. There exists a similar approach for electrical system. By means of the electrical coenergy and well dened power quantities, the equations of motions are uniquely dened. The currents of the inductors and the voltage drops across the capacitors play the role of the generalized coordinates. All constraints, for instance caused by the Kirchho laws, are eliminated from the considerations. In consequence, we have quantities (kinetic and potential energy, generalized forces) which determine the mechanical part and quantities (coenergy, powers) for the description of the electrical part. This oers a combination of the mechanical and electrical parts by means of an energy approach. As a result, we get an extended Lagrangian.

2 MECHANICAL SYSTEMS

2
2.1

Mechanical Systems
Derivation of the Lagrange Equations

In the case of systems of N particles we need, in general, 3N coordinates to specify the position of all particles. If there exist constraints, then the number of coordinates actually needed to describe the system is reduced. For instance, for the specication of a rigid body, we need six coordinates, three for the reference point and three for the orientation. In general, a certain minimum number n of coordinates, called the degrees of freedom, is required to specify the conguration. Usually, these coordinates are denoted by qi and are called generalized coordinates. The coordinate vector xi , i = 1, . . . , N (1)

of a specic particle and the generalized coordinates are related by equations of the form xi = xi (q1 , q2 , . . . , qn , t) . (2)

The time t appears explicitly in the case of moving constraints, such as a particle is constrained to move on a surface which itself is moving in a predened way. The choice of the generalized coordinates is usually somewhat arbitrary, but in general each individual energy storage element of the system have a set of generalized coordinates. For a dynamic system the generalized coordinates do not completely specify the system and an additional set of dynamic variables equal in number to the generalized coordinates must be used. These dynamic variables can be the rst time derivatives of the generalized coordinates, the velocities, or can be a second set variables (e.g. the generalized momenta). In order to nd the dierential equations of motion in terms of the generalized coordinates, we use the energy of the system. The kinetic energy T in terms of Cartesian coordinates is given by T =
N X 1 i=1 T mi vi vi ,

dxi = vi . dt

(3)

Remark 1 It is assumed that masses are not functions of the velocities or coordinates. From the relation (2), we obtain vi = and xi vi = , q j qj which gives d dt T xi d d T vi T xi T vi vi + vi = = vi vi q j dt qj dt qj qj (6) (5)
n X dqj xi xi , + dt qj t j =1

(4)

Next, we multiply by mi and make use of the relation

T after multiplication with vi and dierentiation with respect to t. This leads to T d xi 1 T 1 d T vi vi vi = v vi . + 2 dt q j dt qj 2 qj i

(7)

Fi = mi

d vi . dt

(8)

2 MECHANICAL SYSTEMS

2.1 Derivation of the Lagrange Equations

Hence, by summing over all i, we nd X T xi d T = FT + . i dt q j qj qj i=1 Here the kinetic energy T (q ) is assumed to be a function of q . The expression
N X i=1 N

(9)

FT i

xi = Qj , qj

j = 1, . . . , n

(10)

denes the generalized forces Qj . Hence, we obtain the result T d T = Qj + , dt q j qj j = 1, . . . , n. (11)

These are dierential equations of motion in the generalized coordinates qj . They are known as Lagrange equations of motion. If part of the generalized forces are conservative, then some Qj can be expressed as ! T N N X X V xi T xi Fi = (12) + Qe Qj = j q x q j i j i=1 i=1 and nally Qj = V + Qe j. qj (13)

V is called the potential energy function and the Qe j are generalized forces not derivable from a potential energy function V . Now, the equations can be written more compactly by dening the Lagrangian L=T V which leads to L d L = Qe j, dt q j qj j = 1, . . . , n. (15) (14)

The Lagrange equations have been derived from Newtons laws. In fact, they are a redenition of Newtons laws written out in terms of appropriate variables such that constraint forces are eliminated from considerations. The dynamical system is dened by a single function L, at least if all forces are conservative. The general procedure for nding the dierential equations of motion for a system is as follows: 1. Select a suitable set of coordinates to represent the conguration of the system. 2. Obtain the kinetic energy T as a function of these coordinates and their time derivatives. 3. If the system is conservative, nd the potential energy V as a function of the coordinates, or, if the system is not conservative, nd the generalized forces Qe j. 4. The dierential equations of motion are then given by equations (15). Remark 2 The application of the Lagrangian formulation is not restricted to mechanical systems. So, there exists Lagrangians which are not dened as the dierence between the kinetic and potential energy. Remark 3 The Lagrangian function determines the equations of motion uniquely, the converse of this fact is not true. 6

2 MECHANICAL SYSTEMS

2.1 Derivation of the Lagrange Equations

Remark 4 The Lagrange equations were derived without specifying a particular generalized coordinate system. Hence, they are also valid in other coordinate systems. Lagranges equations are coordinate independent. Remark 5 The Lagrangian function is a so called state function. Its value at a given instant of time is given by the state of the system at that time, and not on the history. Remark 6 The Lagrangian depends on the generalized coordinates q , the associated velocities q , and the time t. As mentioned above, external forces can be subdivided into two groups: The rst group consists of forces F which are given by a potential function F= The second group is formed by non potential forces. Suppose a non potential force which is a function of the velocity and that the force is directed opposite to the velocity of the particle, e.g. FD = gv, (17) V . xi (16)

with g > 0. Hence, the force does negative work and this leads to energy loss. Such forces are called dissipative. From the relations Qj = we get QD = FD v x = gv . q q (19)
N X i=1

FT i

xi qj

QD = FD

x , q

and

x v = q q

(18)

Let us dene the so called dissipative function or Rayleigh potential P R with QD = The combination of the relations (19) and (20) leads to v R P = gv , q q P
R

R P . q

(20)

Zv
0

gv dv = g

v2 . 2

(21)

The prime denotes the variable of integration. If g > 0 is positive, then P R is a positive function. The modied Lagrange equations now read L P R d L + = 0, dt q j qj q j j = 1, . . . , n. (22)

Of course there exist other dissipative forces not related to an equation like (17).

2 MECHANICAL SYSTEMS

2.1 Derivation of the Lagrange Equations

c1 l3 c3 m1

F1

l1 xa

c2 xb m2 F2

Figure 1: Mechanical example. 2.1.1 Example I

Consider the massspring system given in gure (1). In the equilibrium (zero forces F1 and F2 , the system is forced by the Earth gravitational force mg) the length of the springs are given with l1 , l2 , and l3 . Then, the coordinates xa and xb measure the deviation from the equilibrium. If xa and xb are specied, then the geometric conguration of the system is completely determined. So, we have found a set of generalized coordinates xa and xb and their associated velocities va and vb . Referring to equation (14) and equation (15), we start with the calculation of the kinetic energy and nd 1 1 2 2 + m2 vb . T = m1 va 2 2 Next, the potential energy is given as 1 1 1 2 2 V = c1 x2 1 + c2 x2 + c3 x3 m1 gx1 m2 gx2 2 2 2 with the lengths xi of the springs. With the geometric relations x1 = xa + l1 , the Lagrangian follows as L=T V = x2 = xb + l3 xa l1 = xb xa + l2 , x3 = xb + l3

1 1 2 2 m1 va + m2 vb . 2 2 1 1 1 c1 (xa + l1 )2 c2 (xb xa + l2 )2 c3 (xb + l3 )2 + m1 g (xa + l1 ) + m2 g (xb + l3 xa l1 ) 2 2 2 The generalized forces are Qe 1 = F1 , Qe 2 = F2 .

The application of the Lagrange Formalism leads to the equations of motion. xa : m1 dva + c1 (xa + l1 ) c2 (xb xa + l2 ) m1 g + m2 g = F1 dt dvb + c2 (xb xa + l2 ) + c3 (xb + l3 ) m2 g = F2 . dt 8

xb : m2

2 MECHANICAL SYSTEMS

2.2 Variational Principle and Lagranges Equations

T1
1 2

T2

J1

J2

Figure 2: Torsion drive. 2.1.2 Example II

Consider the motorshaftload system given in gure (2). The motor is represented by the rotating inertia J1 and the torque generated by the motor is a given function T1 . The load represented by the inertia J2 is coupled to the motor by means of an elastic shaft with stiness c. In addition, there is a load torque T2 . The coordinates 1 and 2 determines the geometric conguration of the system completely. Therefore the generalized coordinates are 1 and 2 and their associated angular velocities 1 and 2 . Referring to equation (14) and equation (15), we start with the calculation of the kinetic energy and nd 1 1 2 T = J1 2 1 + J2 2 . 2 2 Next, the potential energy is given as 1 V = c (1 2 )2 2 with the angular 1 2 of the torsion spring. The Lagrangian follows as 1 1 1 2 2 L = T V = J1 2 1 + J2 2 c (1 2 ) . 2 2 2 The generalized forces are Qe 1 = T1 , Qe 2 = T2 .

The application of the Lagrange Formalism leads to the equations of motion. 1 : J1 2 : J1 d 1 + c1 (1 2 ) = T1 dt

d 2 c1 (1 2 ) = T2 . dt

2.2

Variational Principle and Lagranges Equations

There exists an alternative way of deriving Lagranges equations which gives new insights. This method is based on Hamiltons variational principle: The motion of a system takes place in such a way that the integral Zt2 (T + W ) dt (23)

t1

is an extremum. The work W of the external forces is given by W =

N X i=1

FT i xi .

(24)

2 MECHANICAL SYSTEMS

2.2 Variational Principle and Lagranges Equations

In other words, Hamiltons principle says that out of all possible ways a system can change within a given nite time t2 t1 , that particular motion which will occur, for which the integral is either a maximum or a minimum. The statement can be expressed in mathematical terms as Zt2 (T + W ) dt = 0 (25)

t1

in which denotes a small variation. This variation results from taking dierent paths of integration by varying the generalized coordinates qj . Note, no variation takes place with respect to the time t. Caused by the variations qj we have virtual displacements xi of the coordinates xi . This leads to W = and W = A rst fact is that the product Qj qj (28)
n X N X j =1 i=1 N X i=1

FT i xi =

N X i=1

FT i

n X xi j =1

qj

qj

(26)

FT i

X xi qj = Qj qj . qj j =1

(27)

is the work done on the system by the external forces, when the coordinates qj change a virtual amount qj . The other generalized coordinates are remaining constant. For example, if the system is a rigid body, the work done by the external forces when the body turns through an angle about a given axis is M , (29)

where M is the torque about the axis. In this case, M is the generalized force associated with the angle . The combination of (25) and (27) gives Zt2 T dt + Zt2 X n Qj qj dt = 0. (30)

t1

t1 j =1

Let Qj be a generalized force which is derivable from a potential energy function V . In this case, we get by integration by parts Zt2 X n Qj qj dt = Zt2 X n V qj dt = qj Zt2 V dt . (31)

t1 j =1

t1 j =1

t1

The combination of (30) and (31) gives Zt2 (T V ) dt + Zt2 X

j

Qe j qj dt = 0 ,

(32)

t1

t1

where the summation goes over the generalized forces which are not derivable from a potential function. The Lagrangian L=T V 10 (33)

2 MECHANICAL SYSTEMS

2.3 State Functions

is a function of qj and q . We have and by integration by parts Zt2 Zt2 X n

j =1

t1

Zt2

Ldt =

t1

Zt2

Ldt =

t1 j =1

Zt2 X n

L L qj + q j dt qj q j

(34)

Ldt =

t1

t1

For xed values of the limits t1 and t2 , the variation qj = 0 at time t1 and t2 . Hence, we get Zt2 X n d L L e + Qj qj dt = 0. qj dt q j (36)

t2 Zt2 X n n X L L d L qj dt + qj qj dt. qj q j dt q j j =1 j =1
t1 t1

(35)

t1 j =1

If all the generalized coordinates are independent, then their variations are all independent, too. Therefore, each term in the bracket must vanish in order that the integral itself vanishes. Thus, L d L = Qe j, dt q j qj j = 1, . . . , n. (37)

Remark 7 Similar equations were rst derived by Euler for the general mathematical variational problem. Therefore, the equations (37) are called Euler Lagrange equations, too. Remark 8 The variational approach leads to the EulerLagrange equations even when relation (23) does not give a minimum. The minimum requirement is always satised if L = T V holds and V is independent of the velocities (or if V depends linearly on the velocity).

2.3

State Functions

As mentioned above the value of the Lagrangian at a given instant of time is a function given by the state of the system at that time, and not on the history. Such functions are called state functions. Examples are the total energy of the system and other closely related functions. The are of central importance in the characterization of physical systems. For example, let dW be a dierential change in energy produced by a dierential displacement dq in the variable q . Then we have dW = Qdq (38)

with the generalized force Q see equation (28) also. The product of the variables Q and q describes an energy relation, which is usually a state function. It contains much valuable information about the system. Unfortunately some physical eects (dissipation, hysteresis, inputs) must be excluded from systems if they are to be described by state function. So, we restrict our attention to conservative systems. This is not a serious drawback, because this formulation is mainly used for the coupling of electrical and mechanical part. Fortunately, these couplings are derivable from state functions. The state of the dynamical system can be described either by n generalized coordinates qi and its time derivatives q i or by the qi and the n generalized momenta pi . The associated 2n dimensional space is called the phase space. A pair qi and pi is called canonically conjugate variables. Associated with every set of independent i . So, for a mechanical system we have four dierent variable qi and pi is a set of dependent variables Qi and q kind of variables: q , the generalized mechanical coordinate, it is also called mechanical displacement, 11

2 MECHANICAL SYSTEMS

2.3 State Functions

q , the generalized mechanical velocity, it is also called mechanical velocity, Q, the generalized mechanical force, a mechanical force depends upon the position only Q = K (q ) q. (39)

p, the generalized mechanical momenta see equation (47), a mechanical momenta depends usually upon the velocity only p = M (q ) q. (40)

We have mentioned that there are Lagrangian which cannot be expressed as the dierence of the kinetic and potential energy. Nevertheless it is possible to decompose L as the dierence of two functions. To show this we express the dierential of the Lagrangian 1 , . . . , q n , t) L (q1 , . . . , qn , q as dL =
n n X X L L L dt. dqj + dq j + q q t j j j =1 j =1

(41)

(42)

From this expression L can be calculated by integration. Since L is a state function, a arbitrary path for the integration can be chosen. For example the q j are constant for integration with respect to qj and the qj are constant for integration with respect to q j . Furthermore, these integrations can be performed for a specic value of t. We get L=
q1 ,... Z ,qn q 1 ,... n Z ,q

0,... ,0 j =1

0 0 n L q , . . . , q , 0, . . . , 0, t X 1 n qj

dqj + (43)
0

0,... ,0 j =1

0 0 n L q , . . . , q , q n , t X 1 n 1 , . . . , q q j

dq j

and L is decomposed in two functions. The rst function is exactly the denition for the negative of the potential energy. Therefore a generalized force Qj associated to a potential is dened as 0 0 L q1 , . . . , qn , 0, . . . , 0, t (44) Qj = qj and the potential energy is dened as V =
q1 ,... Z ,qn n X

0,... ,0 j =1

0 0 0 Qj q1 , . . . , qn , t dqj .

(45)

This claries the introduction of the potential energy in equation (31). The second term 0 0 q 1 ,... n Z ,q n L q , . . . , q , q n , t X 1 n 1 , . . . , q 0 dq j q j j =1
0,... ,0

(46)

12

2 MECHANICAL SYSTEMS

2.3 State Functions

is a function of the nal values of qj and the velocities. This acts as a denition of the generalized momenta 0 0 1 , . . . , q n , t L q1 , . . . , qn , q (47) pj = q j and the so called kinetic coenergy T =
0

q 1 ,... n n Z ,q X 0,... ,0 j =1

0 0 0 1 , . . . , q pj q n , t dq j .

(48)

Remark 9 At this point the reason for this terminology seems to be articial. The analogous discussion for electrical systems shows that the introduction of the kinetic coenergy is a direct consequence of the denition of the magnetic coenergy. Remark 10 In this concept the denition of the kinetic energy has the general form Z T = q dp whereas the denition of the kinetic energy has the general form Z 0 T = pdq. As a consequence, the Lagrangian becomes simply L = T V.
0 0

(49)

(50)

(51)

If the masses of a mechanical system are constant, then the kinetic coenergy T and the kinetic energy T are equal. For example suppose a mass m with velocity q . The momenta is given as p = mq and we obtain the kinetic coenergy T = which is equal to the kinetic energy T . 2.3.1 Example I
0

q Z

0 0 mq 2 , = p q dq 2

(52)

The example should illustrate the calculation of the kinetic and potential energies. Consider the massspring system given in gure (3). In the equilibrium (zero force F ) the length of the springs are given with a and b. Then, the coordinates x1 and x2 measure the deviation from the equilibrium. If x1 and x2 are specied, then the geometric conguration of the system is completely determined. So, we have found the generalized coordinates x1 and x2 and their associated velocities v1 and v2 . To nd the potential energy the equation (45) is used, thus V = where 0 0 0 Q1 = c1 x1 + c2 x2 x1 , 13 0 0 Q2 = c2 x2 x1 .
x1 Z,x2 2 X

0,0 j =1

0 0 0 Qi x1 , x2 , t dxj

2 MECHANICAL SYSTEMS

2.3 State Functions

c1 b m1 c2 x2 m2 F

a x1 x

Figure 3: Example. The potential energy is V =

x1 Z,x2 0,0

0 0 0 0 0 0 0 c1 x1 + c2 x2 x1 dx1 c2 x2 x1 dx2 .
0 0 0

This integral is evaluated by holding x2 = 0 and displacing x1 from 0 to x1 , then holding x1 = x1 and displacing 0 x2 from 0 to x2 . This results in V = (c1 + c2 )
x1 Z 0

x1 dx1 +

x2 Z 0

0 0 c2 x2 x1 dx2

V = V =

(c1 + c2 ) 2 c2 2 x1 + x2 c2 x2 x1 2 2 c1 2 1 x + c2 (x2 x1 )2 . 2 1 2

The kinetic coenergy can now be derived using equation (48), which is T = The momenta are p1 = m1 x 1, and we get T = The line integral is T =
0 0 0

x 1 2 Z,x

0,0 j =1

2 X

0 0 0 1, x j. pj x 2 dx

p2 = m2 x 2

x Z 2 1 ,x 0,0

m1 x 1 dx 1 +

x 1 2 Z,x 0,0

m2 x 2 dx 2.

x Z 1 ,0 0,0

m1 x 1 dx 1 +

x 1 2 Z,x x 1 ,0

m2 x 2 dx 2 =

m1 2 m2 2 x + x . 2 1 2 2

14

2 MECHANICAL SYSTEMS

2.4 Energy of Mechanical Systems

2.3.2

Example II

The upper point of the ideal pendulum of length l is constrained to move at a constant angular velocity around a circle of radius r. At time t = 0 the upper point of the pendulum is located at the bottom of its

x
t

m y

Figure 4: Pendulum on a circle. circular path. We assume that there is no friction. If is specied, then the position of the pendulum is completely determined. So, is the generalized coordinate and the associated velocity. In terms of the Cartesiancoordinate system the kinetic coenergy of the mass m is given by T =
0

1 1 mx 2 + my 2. 2 2

The coordinates x and y are expressed in terms of the generalized coordinate, thus x = r sin (t) + l sin , Using the rst time derivatives x = r cos (t) + l cos , we have T = =
0

y = r cos (t) + l cos .

y = r sin ( t) l sin

1 2 m (r cos (t) 1 2m

The potential energy is associated to the gravitational force, thus

2 2 r + l2 2 + l r cos ( t ) . V = mgl (1 cos ) .

+ l cos )2 + 1 sin )2 2 m (r sin ( t) l

This denes the Lagrangian with 0 1 2 + l r cos ( t ) mgl (1 cos ) . L = T V = m r2 2 + l2 2 : g 1 2 r sin ( t ) + sin = 0. 2l l

The application of the Lagrange formalism leads to the equation of motion

2.4

Energy of Mechanical Systems

0

Let us assume that equality of the kinetic energy T and the kinetic coenergy T . Then, the dierence between the usual Lagrangian L=T V 15 (53)

2 MECHANICAL SYSTEMS

2.5 Legendre Transformations

and the energy E = T + V is just the sign of V . Is there some general way to calculate E from the knowledge of L? We start with a denition E=
n X i=1

q i

LL q i

(54)

and prove, whether E satises the conditions to be an energy (state function) or not. The candidate E meets the relations n d P d E= q i LL dt dt i=1 q i =
i=1 n P n P n P

q i

n n n P P P d d d i i L L+ L q q i L L q dt q i i dt i dt t i=1 q i=1 qi i=1 q

q i

i=1

d L L L dt q i qi t L t

(55)

i=1

q i Qe i

If no external generalized forces Qe i exist and L is time independent than the relation d E=0 dt (56)

will be met. In this case, E is a constant of motion like the energy. However we have not established, whether E is in fact the energy T + V . We state without proof that if T is a homogenous quadratic function of q i , then E will be the energy [4]. Note, a function f is called homogenous quadratic, i the condition f (az1 , . . . , azk ) = a2 f (z1 , . . . , zk ) is satised. The conditions for T to be homogenous quadratic in q i are: The potential V is independent of q i . The transformation from the Cartesian coordinates to the generalized coordinates is time independent. L = T V is time independent. (57)

2.5

Legendre Transformations

The Lagrangian can be used to formulate the equations of motion of dynamical systems. In this section we discuss alternate state functions. The Hamiltonian or total energy can be obtained from the Lagrangian by a transformation of the variables. The generalized velocity q i can be replaced by the associated variable pi to get the Hamiltonian H which is a function of qi and pi . Using a Legendre transformation to dene H gives H = L
n X i=1

pi q i .

(58)

Taking the total dierential of H gives either see equation (47) dH = X n n X L L L dqi dq i + q i dpi + pi dq i = dqi + q i dpi . qi q i qi i=1 i=1 16 (59)

2 MECHANICAL SYSTEMS

2.6 Case Studies

or dH = In case of no external forces we have d L L=0 dt q i qi and we can obtain dH =

n X i=1 n X H i=1

H dqi + dpi . qi pi

(60)

d pi = L dt qi

(61)

(p i dqi + q i dpi )

(62)

Finally, the comparison of (62) and (60) gives the Hamiltons equations of motion H = p i , qi H =q i . pi (63)

The Legendre transformation oers the denition of other state functions, e.g. L (p1 , . . . , pn , Q1 , . . . , Qn ) = H + or 1 , . . . , q n , Q1 , . . . , Qn ) = L H (q
0 0

n X i=1 n X i=1

qi Qi

(64)

qi Qi .

(65)

Usually these relations are not used in the modeling. Nevertheless the are of some theoretical interest. We have established that the Hamiltonian H and the total energy E are equal besides some less restrictive conditions 0 compare equation (54) and equation (58). Now, the quantity H is dened to be the total coenergy H =T +V
0 0 0

(66)
0

and it is called the coHamiltonian. Moreover, the quantity L is called the coLagrangian in dened as L =T V . Remark 11 In this concept the denition of the potential energy has the general form Z V = f dq whereas the denition of the potential coenergy has the general form Z 0 V = q df.
0 0

(67)

(68)

(69)

2.6
2.6.1

Case Studies
Atwoods Machine

Link to the HTMLpage ../../../Examples/Atwood/HTML/Atwood.html

17

2 MECHANICAL SYSTEMS

2.6 Case Studies

2.6.2

Car and Beam

Link to the HTMLpage ../../../Examples/CarAndBeam/HTML/CarAndBeam.html 2.6.3 Double Pendulum

frictionless: Link to the HTMLpage ../../../Examples/DoublePendulum/HTML/DoublePendulum.html with friction: Link to the HTMLpage ../../../Examples/DoublePendulumWithFriction/HTML/DoublePendulumWithFriction.html 2.6.4 Bead and Hoop

Link to the HTMLpage ../../../Examples/BeadAndHoop/HTML/BeadAndHoop.html 2.6.5 Ball on a Wheel

Link to the HTMLpage ../../../Examples/BallOnWheel/HTML/BallOnWheel.html 2.6.6 Two dimensional truck model

Link to the HTMLpage ../../../Examples/Truck/HTML/Truck.html

18

3 ELECTRICAL SYSTEMS

Electrical Systems

Section 2 has presented a modeling technique for mechanical systems by means of energy terms. For the coupling of electrical and mechanical systems, we have to extend this idea to electrical systems. We consider networks made up of resistors, capacitors, inductors, and sources. Resistors and sources are called the static terminals of the network. Capacitors and inductors are called the dynamic terminals. Later we discuss briey the nature of these objects, called the branches of the circuit. At present, it suces to consider them as devices with two terminals. The network is formed by connecting together various terminals. The connection points are called nodes. To nd a mathematical description of the network, we dene a graph which corresponds to the networks. This graph G = (N , B) consists of the following data: A nite set N of points called nodes. The number of nodes is a. A nite set B of lines called branches. The number of branches is b. A branch (or port) has exactly two end points which must be nodes. A current state of the network will be some vector iT = [i1 , . . . , ib ], where ik represents the current owing through the kbranch at a certain moment. Kirchho s current law states that the amount of current owing into a node at a given moment is equal to the amount owing out. For a node k, we get X dkl il = 0, k B . (70)
l

The sum is taken over all branches and dkl is dened as : dkl = 1: if node k and branch l are connected and the direction of il to node k is positive dkl = 1: if node k and branch l are connected and the direction of il to node k is negative dkl = 0: otherwise Next, a voltage state of the network is dened to be the vector uT = [u1 , . . . , ub ], where ul represents the voltage drop across the lth branch. Kirchho s voltage law states that there is a real function on the set of nodes, a voltage potential, so that vj djl + vk dkl = ul holds for each branch l. The power P of a network is a real function dened as P =
b X

(71)

ik uk

(72)

k=1

A current iT = [i1 , . . . , ib ] Rb , and a voltage uT = [u1 , . . . , ub ] Rb are said to be admissible, if they obey the current law (70) and the voltage law (71). Telegens theorem oers a very ecient way to characterize admissible currents and voltages. We state without proof [5]: Theorem 1 (Telegen) The relation
b X k=1

ik uk = 0

(73)

is met for any admissible current i and any admissible voltage u of a Kirchho network with graph G = (N , B). A direct consequence is that the power is zero for any admissible current i and any admissible voltage u of a Kirchho network. 19

3 ELECTRICAL SYSTEMS

3.1 Energy and Coenergy of Simple Devices

3.1

Energy and Coenergy of Simple Devices

Now, we describe in mathematical terms the three dierent types of devices in the network; static terminals, capacitors, and inductors. 3.1.1 Static terminals

Each static terminal S imposes a relation h(u, i) = 0 on the current i and the voltage u of its branch. Typical examples are the linear resistor u = Ri, the voltage source u = constant, i arbitrary and the current source i = constant, u arbitrary. Next, we introduce functions P u (u), P i (i) such that the relations u P = i, u i P = u, i P u + P i = ui (75) (74)

are met. Note, the physical dimension of these quantities is Watt. If u and i has the same direction, that means P u + P i is positive, then we will have dissipated power. Let us compare the denition Qj = R P q j (76)

of the Rayleigh potential P R of dissipative forces with the relations (75). For the dissipative forces of a mechanical system, we can dene powers P u and P i in the way u P = Qj , q j i P = q j , Qj Pu + Pi = P (77)

If q j and Qj are directed in opposite, then P u , P i , P will be positive functions. This is the usual convention of the direction of forces and velocities of dissipative forces. Taking the example QD = gq, then short calculations show u P = QD , q P i = q, QD and Pu + Pi = g Pu =
q R 0

(78)

gq dq =g
0

q 2 = PR 2 (79)

Pi =

Q RD 0

Q2 QD 0 dQD = D g 2g

q 2 g 2 q q 2 Q2 2 + D =g + = gq 2 = QD q. 2 2g 2 2g

(80)

Relations (75) and (77) look very similar. The dierence is just the sign. If u and i are directed in the same direction, then the power P u + P i > 0. This is a usual convention in electrical networks. For this similarity, we

20

3 ELECTRICAL SYSTEMS

3.1 Energy and Coenergy of Simple Devices

call the powers of the electrical network generalized potentials, too. For example, a linear resistance satises u = Ri, which leads to Pu =
u 0 u u0 R R u2 0 0 du = i u du = 2R 0 0 R

The resistor is a dissipative element. Hence i and u of a resistor have the same direction. The relations (81)

Ri 0 0 Ri 0 0 Ri2 . P i = u i di = Ri di = 2 0 0

(81)

Figure 5: Resistor. allow a simple interpretation of the quantities P u (u), P i (i). Remark 12 Figure (6) shows a nonlinear resistor law. We can deduce that P i and P u can be interpreted as areas above and beyond the curve.
u R (i ) p
u

pi

Figure 6: Interpretation A voltage source satises u = u0 , which leads to P =

i

Zi
0

0 0 u i di = u0 i,

P u = 0.

(82)

Note, u and i dont have the same direction. Hence P i < 0. Analogous, a current source satises i = i0 , which

i u

Figure 7: Voltage source.

21

3 ELECTRICAL SYSTEMS

3.1 Energy and Coenergy of Simple Devices

leads to Zu 0 0 P = i u du = i0 u,
u 0

P i = 0.

(83)

i u

Figure 8: Current source.

3.1.2

Dynamic Devices

An inductor or capacitor does not impose conditions directly on the state, but denes how the state in that branch changes in time. In particular a capacitor and an inductor link the charge q with the voltage u and the ux with the current i by the nonlinear relations q = f (u) , Their dynamics is given by the relations d q = i, dt d =u dt (85) = g (i) . (84)

for a capacitor and an inductor, respectively. The combination of (84) and (85) immediately leads to d f u = i, u dt d g i=u. i dt (86)

Remark 13 A linear capacitor is given by q = f (u) = Cu, which leads to the relation d d q = C u = i . dt dt Remark 14 A linear inductor is given by = g (i) = Li, which leads to the relation d d =L i=u . dt dt (88) (87)

Now, let us introduce the energy W q of the capacitor and the energy W of the inductor. We assume that q there exist functions W , W such that

W q = u, q or equivalently
Zq 0 0 q, d W = u q q 0

W = i

(89)

Z 0 0 W = i d
0

(90)

22

3 ELECTRICAL SYSTEMS

3.1 Energy and Coenergy of Simple Devices

is met. The prime denotes the variable of integration. One gets q d q d W = W q = ui dt q dt and d d W = W = iu. dt dt Remark 15 Here, the coordinates q and are considered as independent variables. Remark 16 The relations d d q = i, =u (93) dt dt motivate the interpretation of the charge q and the ux linkage as position coordinates. The associated velocities are the currents and the voltages. Such an interpretation is sometimes used in modeling, see [?]. In a similar way, we can dene dual objects W u , W i with i Wu = q W = , , u i which are called electrical coenergies. They meet the relations d d d q d q d d W + Wu = W q + Wu u = u q + q u= (uq ) dt q dt u dt dt dt dt and d d d d d d W + Wi = W + Wi i = i + i = (i) dt dt i dt dt dt dt d dt Wu u = i, d dt i W i = u. (94) (92) (91)

(95)

(96)

as well as

(97)

Remark 17 Here, the coordinates u and i are considered as independent variables. For example, a linear capacitor satises q = Cu, which leads to the energy
0 Zq 0 0 Zq q 2 q 0 d q = q = d W = u q C 2C q 0 0

(98)

and the coenergy Zu Zu 0 0 0 0 Cu2 W = q . u du = Cu du = 2

u 0 0

(99)

For linear capacitors, the energy and the coenergy are equal. This fact does not hold for nonlinear devices. Analogous, a linear inductor satises = Li, which leads to the energy Z Z 0 0 0 0 2 d = W = i d = L 2L
0 0

(100)

and the coenergy W =

i

Zi
0

0 0 Zi 0 0 Li2 . i di = Li di = 2
0

(101)

23

3 ELECTRICAL SYSTEMS

3.2 Equations of Motion of Simple Networks

Figure 9: Capacitor.

Figure 10: Inductor. 3.1.3 Example

Simulation models of MOSFET devices take into account that the charge storage can be described by voltage depending capacitors q = C (u) u. If the voltage u drop across the bulkdrain branch is positive, then we have the relation for the capacitor u C = C0 1 + C0 = 2.4 1014 [F] , = 0.87 [V] . Then, the energy W q and the coenergy W u are given as: W =
q u Z 0

q (z ) dz = C0 u

1 1u + 2 3

Wu =

p p 1 C0 + C0 (C0 + 4q ) 8q C0 + C0 (C0 + 4 q ) /C0 24

3.2

Equations of Motion of Simple Networks

As mentioned above, a simple network is built up by three types of terminals. Therefore, the set B can be subdivided into 3 disjoint sets L, C and S of subsets of B in such a way, that the sets L, C , S contain the inductors, the capacitors and the static terminals, respectively, and the relation B = L C S is met. Remark 18 For our proposed type of electrical networks, the current in the inductors and the voltage drops along the capacitors, via Kirchho s laws and the laws of the static terminals, determine the currents and the voltages in all the branches. We call such networks simple. If capacitors are connected in parallel or inductances are connected in series, then such devices should be combined to one single device. All voltages (uC , uL , uS ) are expressed as functions of the voltages uT C = [uC,1 , . . . , uC,nC ] of the capacitors = [ i , . . . , i ] of the inductors. In a similar way, we express all currents as functions of and the currents iT L,1 L,nL L uC and iL . This gives two maps : (uC , iL ) (uC , uL , uS ) , 24 : (uC , iL ) (iC , iL , iS ) (102)

3 ELECTRICAL SYSTEMS

3.2 Equations of Motion of Simple Networks

Without saying, (uC , uL , uS ), and (iC , iL , iS ) meet Telegens theorem. The total coenergy of the network is dened by X X W u (uC ) + W i (iL ) . (103) W el (uC , iL ) =
C L

We will follow the convention that sums over C or L means summation over all capacitor or inductor branches. Further, we get ! !! X X d d d el u i k = ik W = W (uC ) + W (iL ) = q (104) dt uC,k dt uC,k uC,k dt C L and d d W el = dt iL,l dt iL,l X
C

W (uC )

+ iL,l

X
L

W (iL )

!!

d = ul dt l

(105)

with k C , l L. We dene the quantities P C (uC , iL ) = and P u (uC , iL ) = X

S

X
C

ui,

P L (uC , iL ) =

X
L

ui

(106)

P u (u) ,

P i (uC , iL ) =

X
S

P i (i)

(107)

which are powers. They are needed for the derivation of the equations of motion. Telegens theorem can be rewritten as
b X

ik uk =

k=1

X
C

ui +

X
L

ui +

X
S

ui = P C + P L + P u + P i = 0.

(108)

Please remind that sums over C or L means summation over all capacitor or inductor branches. At a given time t0 the circuit is in a particular current [i1 , . . . , ib ] Rb voltage [u1 , . . . , ub ] Rb state. In this way, a curve is obtained, depending on the initial state of the circuit. The components ik , uk , k B of this curve must satisfy the conditions imposed by Kirchho s laws and the static terminal laws. In addition, at a given time the components duk /dt and dik /dt of the tangent vectors of the curve must satisfy the relations (104) and (105). A curve satisfying these conditions is called a trajectory. The coordinates ik , uk , k B has a property in common with the coordinates xi of a specic particle of a mechanical system. They describe the system completely but in general they are not independent from each other due to the restrictions. For this reason, we have introduced the generalized coordinates qj . In the present case, the currents through the inductors and the voltage drops along the capacitors play the analogous role. As supposed in remark (18), they determine the system completely. Moreover, they are independent from each other. Next, we are going to state this set of equations of motion for simple electrical networks. Let u be any admissible voltage. Then, the time derivative du/dt fullls the Kirchho voltage law and is an admissible voltage, too. Hence, Telegens theorem tells us
b X k=1

ik

duk =0 dt

(109)

for any admissible current i. We rewrite this as X du X du X du + + = 0. i i i dt dt dt

C L S

(110)

25

3 ELECTRICAL SYSTEMS

3.2 Equations of Motion of Simple Networks

From the Leibnitz rule we get X du d = i dt dt

L

X
L

iu

X du X du X di u = i i . dt dt dt
L C S

(111)

This leads to d dt X
L

iu

X di X du X du + =0 u + i i dt dt dt L C S

(112)

and see equations (106) and (107) X di X du d d L P + Pu =0. u + i dt dt dt dt

L C

(113)

P L and P u are functions of the currents through the inductors and the voltage drops along the capacitors. Hence, the chain rule gives X du X L di X di X du PL + Pu + P + Pu =0 u + i u dt i dt dt dt
C L L C

(114)

and X du X L di L u u P +P +i + P +P u = 0. u dt i dt C L Since duk /dt and dik /dt can take any value, L P + P u + ik = 0, uk d W el = iC,k , dt uC,k we get L d P + P u = 0, W el + dt uC,k uC,k L d P + Pu = 0 W el dt iL,k iL,k C d P + P i = 0, W el + dt iL,k iL,k kC (118) L P + P u uk = 0. ik d W el = uL,k dt iL,k (116) (115)

Taking into account

(117)

and

k L.

(119)

Using the equation (108) gives

k L.

(120)

The equations (118) and (120) state the set of equations of motions for simple electrical networks see [5]. Remark 19 The right hand sides of the dierential equations (118) and (120) are functions of all uC,k , iL,k . This fact coincides with remark (18). Remark 20 The set of the dierential equations (118) and (120) is a set of rstorder dierential equations. 26

3 ELECTRICAL SYSTEMS

3.2 Equations of Motion of Simple Networks

iC,1 uC,1 u0 iL

iC,2 uC,2 uL iR uR

Figure 11: A simple electrical network. 3.2.1 Example I

Suppose we have the simple electrical network given in gure 11. First we will calculate the maps and as introduced in (106) and (107). This means nothing that all voltages and currents must be expressed as functions of the voltages across the capacitors and currents through the inductances. From the voltage and current laws it follows u0 = uC,1 + uL uL = uC,2 + uR iC,1 = iC,2 + iL iC,2 = iR and uL = u0 uC,1 uR = uL uC,2 = u0 uC,1 uC,2 iR = u0 uC,1 uC,2 uR = R R u0 uC,1 uC,2 + iL = iC,2 + iL = R u0 uC,1 uC,2 . iC,2 = iR = R

iC,1

Now, the power P C (uC , iL ) follows with P C (uC , iL ) = uC,1 iC,1 + uC,2 iC,2 u0 uC,1 uC,2 u0 uC,1 uC,2 + iL + uC,2 = uC,1 R R and P u (uC , iL ) is given by P u (uC , iL ) = (u0 uC,1 uC,2 )2 u2 R +0= . 2R 2R X
L

The calculation of the total coenergy of the network leads to W el (uC , iL ) = Moreover we have P L (uC , iL ) = (u0 uC,1 ) iL 27 X
C

W u (uC ) +

W i (iL ) =

C2 u2 C1 u2 Li2 C,1 C,2 + + L. 2 2 2

3 ELECTRICAL SYSTEMS

3.2 Equations of Motion of Simple Networks

and R P (uC , iL ) = 2
i

u0 uC,1 uC,2 R

u0

u0 uC,1 uC,2 + iL . R

The evaluation of L d P + P u = 0, W el + dt uC,k uC,k kC

leads to

and

L uC,1 + uC,2 u0 d d iL = 0 P + P u = C1 uC,1 + W el + dt uC,1 uC,1 dt R R L uC,1 + uC,2 u0 d d = 0. p + pu = C2 uC,2 + W el + dt uC,2 uC,2 dt R R C d P + P i = 0, W el + dt iL,k iL,k kL

The evaluation of

leads to

Link to HTMLpage ../../../Examples/ElectricalNetwork/HTML/ElectricalNetwork.html 3.2.2 Example II, CukConverter

d C d P + P i = L iL + uC,1 u0 = 0. W el + dt iL iL dt

The CukConverter is a special case of a dcdc converter, which is widely used in switchmode dc power supplies and dc motor drive applications. As shown in gure (12), often the input of such converters is an unregulated dc voltage Ue . Switchmode dcdc converters are used to convert the unregulated dc input into a controlled dc output uC,2 at a desired voltage level. The output voltage uC,2 may be higher or lower than the input voltage. For the analysis, the switch is treated as being ideal and the capacitive elements C1 , C2 have no losses. The
u C, 1 i L,1 L 1, R 1 1 Ue S C1 0 L 2, R 2 C2 u C, 2 R i L,2

Figure 12: CukConverter. losses in the inductances L1 , L2 are modelled by resistors R1 , R2 . The dc input voltage Ue to the converter is assumed to have zero internal impedance. The output is assumed to supply a load that can be represented by 28

3 ELECTRICAL SYSTEMS

3.2 Equations of Motion of Simple Networks

u C, 1 i L,1 L 1, R 1 Ue C1 0 S S 0 L 2, R 2 C2 u C, 2 R i L,2

Figure 13: The switch S is in position 0. an equivalent resistor R. The control input u is called the duty ratio u, 0 u 1 and this quantity species the ratio of the duration of the switch S in position 1 to the xed modulation period T . If the switch S is in position 0, then the circuit can be divided in two parts as shown in gure (13). First, we will calculate the maps and as introduced in (106) and (107). From the voltage and current laws it follows Ue = uC,1 + uL,1 + uR,1 0 = uC,2 + uL,2 + uR,2 iL,2 = iR,2 = iR + iC,2 and uL,1 = Ue uC,1 iL,1 R1 uL,2 = uC,2 iL,2 R2 uC,2 + iL,2 . iC,2 = R Trivial equations are not displayed. The proposed modeling technique requires the quantities u X C,2 + iL,2 ui = uC,1 iL,1 + uC,2 P C (uC , iL ) = R C X P L (uC , iL ) = ui = (Ue uC,1 iL,1 R1 ) iL,1 + (uC,2 iL,2 R2 ) iL,2 P u (uC , iL ) = P i (uC , iL ) = X
S S L

P u (u) =

i2 u2 i2 L,1 R1 L,2 R2 C,2 + + 2 2 2R

P i (u) =

i2 u2 i2 L,1 R1 L,2 R1 C,2 + + Ue iL,1 2 2 2R

and the total coenergy of the network W el (uC , iL ) = The evaluation of d L P + P u = 0, W el + dt uC,k uC,k 29 kC X
C

W u (uC ) +

X
L

W i (iL ) =

C2 u2 L1 i2 L2 i2 C1 u2 C,1 C,2 L,1 L,2 + + + . 2 2 2 2

3 ELECTRICAL SYSTEMS

3.2 Equations of Motion of Simple Networks

leads to L d d P + P u = C1 uC,1 iL,1 = 0 W el + dt uC,1 uC,1 dt L uC,2 d d =0. P + P u = C2 uC,2 iL,2 + W el + dt uC,2 uC,2 dt R C d P + P i = 0, W el + dt iL,k iL,k kL

and

The evaluation of

leads to

and

C d d P + P i = L1 iL,1 + uC,1 + iL,1 R1 Ue = 0 W el + dt iL,1 iL,1 dt C d d P + P i = L2 iL,2 + uC,2 + iL,2 R2 = 0 . W el + dt iL,2 iL,2 dt
u C, 1 iL,1 L 1, R 1 1 Ue S S C1 L 2, R 2 C2 u C, 2 R i L,2

If the switch S is in position 1, than the circuit can be divided in two parts as shown in gure (13). First we

Figure 14: The switch is in position 1. will calculate the maps and as introduced in (106) and (107), again. From the voltage and current laws it follows Ue = uL,1 + uR,1 0 = uC,1 + uC,2 + uL,2 + uR,2 iL,2 = iR,2 = iC,2 = iR + iC,2 and uL,1 = Ue iL,1 R1 uL,2 = uC,1 uC,2 iL,2 R2 uC,2 + iL,2 . iC,2 = R 30

3 ELECTRICAL SYSTEMS

3.2 Equations of Motion of Simple Networks

Trivial equations are not displayed. The proposed modeling technique requires the quantities u X C,2 + iL,2 ui = uC,1 iL,2 uC,2 P C (uC , iL ) = R C X P L (uC , iL ) = ui = (Ue iL,1 R1 ) iL,1 + (uC,1 uC,2 iL,2 R2 ) iL,2 P u (uC , iL ) = P i (uC , iL ) = X
S S L

P u (u) =

i2 u2 i2 L,1 R1 L,2 R1 C,2 + + 2 2 2R

P i (u) =

i2 u2 i2 L,1 R1 L,2 R1 C,2 + + Ue iL,1 2 2 2R

and the total coenergy of the network W el (uC , iL ) = The evaluation of d L P + P u = 0, W el + dt uC,k uC,k kC X
C

W u (uC ) +

X
L

W i (iL ) =

C2 u2 L1 i2 L2 i2 C1 u2 C,1 C,2 L,1 L,2 + + + . 2 2 2 2

leads to

and

L d d P + P u = C1 uC,1 iL,2 = 0 W el + dt uC,1 uC,1 dt L uC,2 d d =0. P + P u = C2 uC,2 iL,2 + W el + dt uC,2 uC,2 dt R d C P + P i = 0, W el + dt iL,k iL,k kL

The evaluation of

leads to

and

C d d P + P i = L1 iL,1 + iL,1 R1 Ue = 0 W el + dt iL,1 iL,1 dt C d d P + P i = L2 iL,2 + uC,1 + uC,2 + iL,2 R2 = 0 . W el + dt iL,2 iL,2 dt

converter is described by two systems of dierential equations Generally a PWM controlled converter like the Cuk of the form x 1 x 2 = a1 (x1 ) = a2 (x2 ) t (iT, (i + u)T ] t ((i + u)T, (i + 1)T ] S in 1 S in 0 (121)

31

3 ELECTRICAL SYSTEMS

3.2 Equations of Motion of Simple Networks

converter we for i = 0, 1, . . . with the smooth vector elds a1 , a2 and the duty ratio u, 0 u 1. For the Cuk have Ue iL,1 R1 Ue iL,1 R1 uC,1 L1 L1 u +u +i R iL,2 R2 + uC,2 C,2 C,1 L,2 2 L2 L2 and a ( x ) = a1 (x) = 2 i i L,2 L,1 C1 C1 uC,2 uC,2 iL,2 iL,2 R R C2 C2

with the state xT = [iL,1 , iL,2 , uC,1 , uC,2 ]. The duty ratio thus species the ratio of the duration of the switch S in position 1 to the xed modulation period T (see Fig. (15)). From the theory of dierential equations it
S in 1 x1 iT x2 (i+u )T (i+ 1)T S in 0

Figure 15: Duty ratio of a PWM controlled system. is a well known fact that the state variables of a system x = f (x, u) with piecewise continuous inputs u are continuous [5]. Therefore, the two systems of (121) are connected by the conditions x1 (iT ) = x2 (iT )

x1 ((i + u)T ) = x2 ((i + u)T ) . Under the assumption that the switching frequency is much higher than the natural frequencies of the converter system and the switches are realized with common power semiconductor devices, we can derive the so called average model for the PWM controlled converter (121) in the form x = a2 (x) + (a1 (x) a2 (x)) u (122)

with the state xT = [iL,1 , iL,2 , uC,1 , uC,2 ]. Link to HTMLpage ../../../Examples/CukConverter/HTML/CukConverter.html 32

converter with the average state vector x and the duty ratio u see [4]. Hence, the average model for the Cuk reads as U x R x e 1 1 3 x3 L1 L1 x2 R2 + x4 x 3 L2 L2 + x = u x1 x2 x1 C1 C1 x4 x2 0 R C2

3 ELECTRICAL SYSTEMS

3.3 The Energy of Electrical Systems

3.3

The Energy of Electrical Systems

The energy of all capacitors and inductors of an electrical network is given by X X W el (uC , iL ) = W u (uC ) + W i (iL ) .
C L

(123)

Using Telegens theorem we get X

C

ui +

X
L

ui +

X
S

ui = P C + P L + P u + P i !

(124)

X
C

ui +

X
L

ui +

X
S

d ui = dt

X
C

W +

X
L

+ Pu + Pi = 0

(125)

and d dt X
C

Wu +

X
L

Wi

= P u P i = P S .

(126)

These relations may be interpreted in the way that in a circuit the energy in the inductors and capacitors varies according to the power dissipated in the resistors and supplied by the sources.

33

4 ELECTROMECHANICAL SYSTEMS

4
4.1

Electromechanical Systems
Introduction

The overall idea of the coupling of electrical and mechanical systems will be described by the following example. The system consists of a capacitor realized by two plates, one is xed and the other is movable but attached to a spring see gure (16). Further, we know that the capacitance can be described by C (x) = k/x. If we
i u R uC x capacitor

spring

Figure 16: Example of an electromechanical system. change the position x of the moveable plate, then a force F will take place. This force can be calculated from the equation d d q W ( q , x) = iC uC F x. dt dt (127)

In other words, the rate of energy of the capacitor is equal to the dierence of the mechanical power supplied by the force and the electrical power iC uC supplied by the electrical part. This energy relation is based on the choice q and x for the independent variables. For electrical engineers it is more common to use the voltage. Therefore, using the relation d d q W ( (uC q q , x) + W u (uC , x) = ) , dt dt d d d (uC q ) W u (uC , x) = iC uC F x dt dt dt W u d W u d d d uC q + q uC x = iC uC F x dt uC dt x dt dt uC iC and nally W u = F. x (130) W u d d x = iC uC F x x dt dt (128)

we get

(129)

34

4 ELECTROMECHANICAL SYSTEMS

4.1 Introduction

Hence, the force F of the mechanical part is coupled to the electrical part by the coenergy. For the present example, it follows
uC uC Z Z 0 0 0 0 C (x) u2 C uC duC = C (x) uC duC = W (uC , x) = q 2 u 0 0

(131)

and F = ku2 W u (uC , x) = C . x 2x2 (132)

In principle, we can apply Newton second law for the moveable plate with mass m m ku2 d v= C cx, dt 2x2 (133)

which takes into account the law of the linear spring. On the other hand, we know that F is related to the coenergy by equation (130). This fact is a motivation for the introduction of an extended Lagrangian Lex (uC , x, v ) = L (x, v ) + W u (uC , x) with Lagrangian L = T V of the mechanical part. Remark 21 For simplicity, we have assumed that the kinetic energy is equal to the kinetic coenergy. For our example the equations of motion of the mechanical part are given by ex d ex L (uC , x, v ) L (uC , x, v) = 0 dt v x with 1 1 k u2 C . Lex (uC , x, v) = mv2 cx2 + 2 2 2 x We get m k u2 d C v = cx . dt 2 x2 (137) (136) (135) (134)

The equation for the electrical part (a capacitor in series with a resistor) is given by L d P + Pu = 0 W u (uC , x) + dt uC uC W u (uC , x) = We get k k d kuC d d uC (u uC ) R = uC 2 x (u uC ) R = 0. dt x x dt x dt (140) ku2 C , 2x P L = 0, Pu = (u uC )2 R . 2 (138)

with

(139)

35

4 ELECTROMECHANICAL SYSTEMS

4.2 Electromechanical Coupling

4.2

Mechanical Forces of Electromechanical Coupling

The rst step in analyzing a complicated electromechanical system by a conservation of energy approach is to reduce the system containing electromechanical coupling terms to a minimum. To do this, separate out all purely electrical parts and all purely mechanical parts of the system including losses. This separation procedure is carried out to the extent that each electrical terminal pair is coupled to one energy store, either magnetic or electrical. Any internal interconnections between circuits that are coupled to dierent energy storages are included in the external electrical network. The mechanical variables represented by the mechanical terminal pairs are those which aect energy storage in the electric and magnetic elds. The separation procedure results in the general conservative electromechanical coupling network in gure 17 in which there are n electrical terminals and m mechanical terminals pairs. Each electrical terminal pair will be coupled to either a magnetic eld energy storage or an electric energy eld storage. The total stored energy W in the coupling network is
ii

electrical energy

Electrical Network

ui

Qei

Coupling Network

mechanical energy

Mechanical Network

qi

Figure 17: Simplication of electromechanical systems. given by

W = Wq + W ,

(141)

where W q is energy stored in electric elds and W is energy stored in magnetic elds. We assume that W is state function and given by the instantaneous conguration of the system.

Remark 22 Hysteresis can not be taken into account. Otherwise the assumption that W is a state function would be violated. Consider an electrical terminal pair coupled to the electrical eld storage. When the q i and qi are specied qi /dt and the voltage ui at the ith terminal is given by independently, the current in the ith terminal is ii = d the internal constraints. Next, consider an electrical terminal pair that are coupled to magnetic eld storage. When the i and xi are specied independently, the voltage in the ith terminal is ui = di /dt and the current ii at the ith terminal is given by the internal constraints. It should be mentioned that instead of specifying the q i and i the voltages ui and the currents ii could have been considered as independent. This is in accordance to the results obtained in section ??. The next problem is to nd the generalized force due to the electromechanical coupling. Since the m mechanical terminal pairs are characterized by m independent variables, it is possible to consider each mechanical terminal pair individually to nd the force. Let us dene the generalized force Qe k see gure 17 as the force applied to the kth mechanical coordinate by the coupling network. Qe k can be found by considering that an arbitrary placement dqk of the kth mechanical coordinate during the time dt takes place. All other mechanical coordinates are xed and the electrical variables may change in accordance to the internal constraints due to the electrical network. This means that only one electrical variable at each electrical terminal can be changed arbitrary. During the displacement the conservation of energy must hold. The various energies involved in the arbitrary displacement are 36

4 ELECTROMECHANICAL SYSTEMS

4.2 Electromechanical Coupling

energy supplied at electrical terminals:

n X i=1

ui ii dt

(142)

energy supplied at the kth mechanical terminal: k dt = Qe Qe kq k dqk change in stored electrical and magnetic energy of coupling eld: dW All lossy elements are either part of the purely electrical network or of the purely mechanical network. Hence, the conservation of energy requires that the sum of the input energy must be equal to the change in stored energy
n X i=1

(143)

ui ii dt Qe k dqk = dW.

(144)

Then, the generalized force applied to the kth terminal is

n X i=1

Qe k

ui ii dt dW dqk

(145)

We assume the all electrical energy storage will be in capacitances and all magnetic eld storage will be in inductances. Thus, the problems of electrical and magnetic eld coupling can be treated separately. 4.2.1 Mechanical Forces Due the Magnetic Field Coupling

In such cases the coupling network consists of n coils see Figure 18 for an example with one coil. The
i

Electrical Network
Q

Mechanical Network

Figure 18: The coupling network is represented by a coil. conservation of energy establishes that the energy input from all sources is stored as magnetic eld energy W = W (1 . . . , n ; q1 , . . . , qm ) = input electrical energy + input mechanical energy or W

(146)

1 ,... Z ,n 0,... ,0

n X i=1

ii di +

q1 ,... Z ,qm 0,... ,0

m X i=1

Qi dqi .

(147)

37

4 ELECTROMECHANICAL SYSTEMS

4.2 Electromechanical Coupling

which is in accordance to equation (90). The energy stored in the magnetic eld coupling can be determined by bringing all system variables to their nal values in an arbitrary manner. For example all ux linkages are hold at zero (W = 0) and the mechanical coordinates are assembled, then establish the ux linkages with the mechanical coordinates held at their nal positions. For this case, we get
1 ,... Z ,n 0,... ,0 n 0 0 X 0 ii 1 . . . , n ; q1 , . . . , qm di , i=1

W =

(148)

where W is evaluated as the integral of id for any xed qi . Now that the stored magnetic energy has been determined, the mechanical forces due to the magnetic eld coupling can be calculated. Using the relation ui = di /dt and the equation (144) we get Qe k dqk =
n X i=1

ii di dW .

(149)

Since the i , i = 1, . . . , n and the qk are independent variables, the dierentiation of dW yield dW =
n X W W dqk + di . qk i i=1

(150)

All other qi are hold constant. Taking into account equation (148) we get Qe k dqk = and nally Qe k = W . qk (152)
n X i=1

ii di dW =

n X i=1

ii d i

n X W W dqk + ii di = dqk qk qk i=1

(151)

This relation states how the generalized force applied to the kth terminal depends on the magnetic energy in terms of the ux linkages at a certain choice for the mechanical coordinates. In most cases it is preferred to express this relation in terms of the currents through the magnetic coils. Currents are usually used in the description of the electrical network. This requires the introduction of the coenergy. We start with the equation (151) and integrate it by parts
n X i=1 i1 ,... Z ,in n X

W =

ii i

0,... ,0 i=1

0 0 0 i i1 . . . , in ; q1 , . . . , qm dii ,

(153)

where the second term is called the magnetic coenergy

i i1 ,... Z ,in n X

W =

0,... ,0 i=1

0 0 0 i i1 . . . , in ; q1 , . . . , qm dii

(154)

see equation (94). Substitution of equation (154) into (152) leads to the desired expression. Qe k = W i . qk (155)

38

4 ELECTROMECHANICAL SYSTEMS

4.2 Electromechanical Coupling

The several forms of the generalized electromechanical coupling force Qe k applied to the k th terminal by a magnetic eld as found by an arbitrary displacement of the kth mechanical coordinate qk are summarized in Table (156). Stored magnetic energy W

1 ,... Z ,n 0,... ,0

n 0 0 X 0 ii 1 . . . , n ; q1 , . . . , qm di i=1

Magnetic coenergy

W =

i1 ,... Z ,in X n 0,... ,0 i=1 n X i=1

0 0 0 i i1 . . . , in ; q1 , . . . , qm dii (156) ii i

Relation

W + Wi =

Independent coordinates qi , ii

Force: Qe k =

W i qk W qk

Independent coordinates qi , i

Force: Qe k =

i Remark 23 Suppose the force Qe k = W / qk . This force is independent of the changes in ii and i which take place during the arbitrary displacement. Consequently this expression is valid regardless of how ii and i vary, if the variation is compatible with the internal constraints given by the electrical network.

Remark 24 From a mathematical point of view the partial derivative is taken with respect to qk holding all other qi and the ii constant. The holding of ii constant has nothing to do with electrical terminal constraints. Remark 25 Suppose a electrical linear system. That means the uxes are related to the currents via i = Li ii . Then we have W = and W =
n X i=1 i i1 ,... Z ,in X n 0,... ,0 i=1

Li ii dii =

n X Li i2 i i=1

(157)

ii i W i =

Thus for electrical linear systems the stored magnetic energy is equal to the magnetic coenergy. 4.2.2 Mechanical Forces Due the Electrical Field Coupling

n X i=1

i2 iL

n X Li i2 i i=1

n X Li i2 i i=1

= W i.

(158)

We have determined the mechanical forces produced by the magnetic eld coupling. A similar development can be made for nding the mechanical forces due to the electrical coupling. In such cases the coupling network consists of l capacitances. The conservation of energy establishes that the energy input from all sources is stored as electric eld energy
= Wq ( q1 . . . , q l ; q1 , . . . , qm ) = input electrical energy + input mechanical energy Wq

(159)

or W =
q q 1 ,... l Z ,q l X

ui d qi +

0,... ,0 i=1

q1 ,... Z ,qm 0,... ,0

m X i=1

Qi dqi .

(160)

39

4 ELECTROMECHANICAL SYSTEMS

4.2 Electromechanical Coupling

which is in accordance to equation (90). The energy stored in the electrical eld coupling can be determined as W =
q q 1 ,... l Z ,q l X

0,... ,0 i=1

0 0 0 1 . . . , q qi , ui q l ; q1 , . . . , qm d

(161)

where W q is evaluated as the integral of ud q for any xed qi . For the electrical eld case, just as it was for the magnetic eld case, it is the interchange of energy among electrical and mechanical sources and the stored electrical energy that is a manifestation of energy conversion. This, and the fact that the stored energy is a state function which is determined by the instantaneous values of the variables, allows the use of the stored electrical energy to nd the mechanical forces. Using the relation q i = dii /dt and the equation (144) we get l X i=1

Qe k dqk

ui d qi dW q .

(162)

Since the q i , i = 1, . . . , l and the qk are independent variables the dierentiation of dW q yield l X W q W q dqk + d qi . qk q i i=1

dW q =

(163)

Taking into account equation (161) we get Qe k dqk = and nally Qe k =

W q . qk l X i=1 ui d qi dW q = l X i=1

ui d qi

l q X W W q dqk + ui d qi = dqk qk qk i=1

(164)

(165)

This relation states how the generalized force applied to the kth terminal depends on the electrical energy in terms of the charges at a certain choice for the mechanical coordinates. In most cases it is preferred to express this relation in terms of the voltages along the capacitors. Voltages are usually used in the description of the electrical network. Like for the magnetic eld case, this requires the introduction of the coenergy. We start with the equation (160) and integrate it by parts W =
q l X i=1

q i ui

u1 ,... Z ,ul

0,... ,0 i=1

l X

0 0 0 q i u1 . . . , ul ; q1 , . . . , qm dui ,

(166)

where the second term is called the electrical coenergy W =

u u1 ,... Z ,ul l X

0,... ,0 i=1

0 0 0 q i u1 . . . , ul ; q1 , . . . , qm dui

(167)

see equation (94). Substitution of equation (167) into (165) leads to the desired expression. Qe k = W u . qk (168)

The several forms of the generalized electromechanical coupling force Qe k applied to the k th terminal by a electrical eld as found by an arbitrary displacement of the kth mechanical coordinate qk are summarized in 40

4 ELECTROMECHANICAL SYSTEMS

4.3 Equations of Motion

Table (169). Stored electrical energy W =

q q 1 ,... l Z ,q l X

0,... ,0 i=1

0 0 0 1 . . . , q qi ui q l ; q1 , . . . , qm d 0 0 0 q i u1 . . . , ul ; q1 , . . . , qm dui (169) q i ui

Electrical coenergy

W =

u1 ,... Z ,ul

0,... ,0 i=1 l X i=1

l X

Relation

W +W =

Independent coordinates qi , ui Independent coordinates qi , q i

Force: Qe k =

W u qk
W q qk

Force: Qe k =

4.3

Equations of Motion

In the previous sections we have dened the generalized coordinates and state functions for dierent kinds of physical domains separately. Now, the way of modeling is given as follows 1) mechanical part a) Select a suitable set of coordinates qT = [q1 , . . . , qm ] to represent the mechanical conguration of the system. b) Obtain the kinetic coenergy T and the Rayleigh function P R as a function of the time derivatives. c) If the system is conservative, nd the potential energy V as a function of the coordinates, or, if the system is not conservative, nd the generalized forces Qe j. 2) electrical part: a) The generalized coordinates are chosen as the currents iT L = [iL,1 , . . . , iL,nL ] through the inductances and the voltages uT C = [uC,1 , . . . , uC,nC ] along the capacitors. b) Obtain the total electric coenergy W el = W u (uC , u0 , i0 ; q) + W i (uC , u0 , i0 ; q) (170)
0

as a function of the mechanical coordinates and the electrical coordinates and the sources u0 , i0 . c) Calculate the power quantities P L (uC , u0 , i0 , iL ) , P u (uC , u0 , i0 , iL ) , P C (uC , u0 , i0 , iL ) , P i (uC , u0 , i0 , iL ) 3) Dene the extended Lagrangian Lex = T + W el V. Then, the equations of motion of the mechanical part are given by d qi = q i , dt d ex ex R L L + P = Qe i, dt q i qi q i 41 i = 1, . . . , m (173)
0

(171)

(172)

4 ELECTROMECHANICAL SYSTEMS

4.4 Electrical Drives

and the equations for the electrical part are L d P + P u = 0, W el + dt uC,k uC,k d C P + P i = 0, W el + dt iL,k iL,k kC (174) i L.

4.4
4.4.1

Electrical Drives
Elementary Machine

The purpose of this section is to derive the magnetic coenergy of an elementary machine. This will act as starting point for the considerations on DCdrives. Figure 19 presents an elementary two pole machine with one winding on the stator and one on the rotor. These windings are distributed over a number of slots so
s

+ + + + + + + + + + + +

magnetic axes of s

magnetic axes of r

Figure 19: Elementary machine. that their magnetomotive force can be approximated by space sinusoids. The stator and rotor are concentric cylinders, and slot openings are neglected. On these assumptions the stator and rotor selfinductances Lss and Lrr are constant. The statorrotor mutual inductance depends on the angle between the magnetic axes of the stator and rotor windings. The mutual inductance is a positive maximum when = 0 (2), is zero when = /2, and is a negative maximum when = . On the assumptions of sinusoidal waves and a uniform air gap, the spacewise distribution of the airgap ux in sinusoidal, and the mutual inductance is L () = Lsr cos () . (175)

Lsr is the value of L when the magnetic axes of the rotor and the stator are aligned. In a linear system, the relationship between the uxes and currents are given with s = Lss is + L () ir , r = Lrr ir + L () is . (176)

The coenergy in the magnetic eld in the air gap is given by dW i (is , ir , ) = s dis + r dir + M d 42 (177)

4 ELECTROMECHANICAL SYSTEMS

4.4 Electrical Drives

and we have Z dW (is , ir , ) =

i ir Z 0 ir Z 0 is 0 Z 0 0 0 r is = 0, ir , dir + s ir = constant, is , dis 0

Lrr ir dir +

is Z 0

(Lss is + L ( ) ir ) dis

(178)

= = 4.4.2 DCdrive

Lss i2 Lrr i2 r s + + L () ir is 2 2 Lss i2 Lrr i2 r s + + Lsr cos () ir is 2 2

For our purpose it is suces to say that a DCdrive is an elementary machine with commutator. The task of the rotating commutator is to convert the AC voltage generated in each rotating armature coil to DC in the external armature terminals by means of the stationary brushes to which the armature leads are connected. Moreover, the magnetic axes of the armature winding is perpendicular to the magnetic axes of the eld winding. See [2] for a detailed introduction in the theory and application of DCdrives. For convenience we assume a sinusoidal ux density wave in the air gap. Then, we apply the coenergy equation (178) W i (iE , iA , ) = LE i2 LA i2 A E + c cos ( ) iA iE , 2 2 (179)

with the external armature current iA , the exciting current iE , the angle between the magnetic axes of rotor and stator, the inductances LA and LE , and the coupling constant c. For DCgenerators the plus sign has to be taken, the minus sign is related to a DCmotor. The losses in the windings are taken into account using the resistors RA and RE . It follows the torque with M= W i (iE , iA , ) = c sin () iE iA . (180)

Next the process of commutation has to be taken into account. The commutator ensures that the angle between the airgap ux and the armature magnetomotive force is 90 electrical degrees. Hence, the commutation leads to M = ciE iA . (181)

The Figure 20 shows a schematic representation of the separate excited DCdrive. The derivation of the equations of motion follows the following procedure using the coenergy of the magnetic eld of a DCdrive before commutation see equation (179): Mechanical part and armature current part: The derivatives with respect to the coordinates , iE and the time have to be carried out. Finally, the commutator condition requires sin () = 1. Exciting current part: First, the commutator condition requires sin () = 1. Then, the derivatives with respect to the coordinates , iE and the time have to be carried out. Separate excited DCdrive: Next, this procedure is applied to a simple DCdrive with load. The extended Lagrangian is the sum of the mechanical Lagrangian and the magnetic coenergy (coordinates iA , iE ) Lex (iA , iE , , ) = L (, ) + W i (iA , iE , ) . 43 (182)

4 ELECTROMECHANICAL SYSTEMS

4.4 Electrical Drives

RA

LA

iA

RE uA uE

LE

iE

Figure 20: Separate excited DC-drive. At this state the mechanical rotor angle and the angle between the magnetic axes of rotor and stator are the same. This leads for a DCmotor Lex (iA , iE , , ) = LE i2 J 2 LA i2 A E + + c cos ( ) iA iE 2 2 2 (183)

with the inertia J . Next, we apply the relations (173) and get J c sin () iA iE = ML with the external load ML . The commutation requires sin ( ) = 1 which gives d = , dt J = ciA iE ML . (185) (184)

Here, the rotor angle is the generalized coordinate of the movement and is dierent from . We assume constant excitement that means iE = constant for the derivation of the equations for the electrical part. Hence, we have C d P + Pi = 0 W el + dt iL iL P C = 0, Finally, we get LA d d iA c cos ( ) iE + ciE sin ( ) + iA RA uA = 0 dt dt d iA + ciE + iA R uA = 0 dt (188) Pi = i2 RE i2 A RA + E uA iA uE iE . 2 2 (186)

with

(187)

and the commutator conditions = /2 leads to LA (189)

Shunt DCdrive and Series DCdrive: Other DCdrives of interest are the shunt DCdrive with uA = uE . This has no inuence on the magnetic coenergy, thus equation (179) is still in force. A schematic gure of the series DCdrive is shown in gure 22. The condition iA = iE is in force, which gives the magnetic coenergy see equation (179) LE 2 LA i 2 + W = c cos () iA + iA . 2 2

44

4 ELECTROMECHANICAL SYSTEMS

4.5 Case Studies

RA

LA

iA

RE uA

LE

iE

Figure 21: Shunt DC-drive.

RA LA iA

uA

RE

LE

iE

Figure 22: Series DCdrive.

4.5
4.5.1

Case Studies
WardLeonard drive

Link to HTMLpage ../../../Examples/WardLeonard/HTML/WardLeonard.html 4.5.2 Ball in a Magnetic Field

Link to HTMLpage ../../../Examples/BallInMagnet/HTML/BallInMagnet.html 4.5.3 Electromagnet

Link to HTMLpage ../../../Examples/Electromagnet/HTML/Electromagnet.html 4.5.4 Relay Device

The relay shown in gure 23 is made from innitely permeable magnetic material with a moveable plunger, also of innitely permeable material. The height of the plunger is much greater than the airgap length (h g). The magnetic coenergy is dened by W (i, x) =
i

Zi

(i , x) di =

i Z

L (x) i0 di = L (x)

i2 . 2

Thus, the calculation of L (x) is required. Because of the high permeability the ux is conned almost entirely in the core. The relationship between the magnetomotive force iN (N is the number of turns) and the magnetic 45

4 ELECTROMECHANICAL SYSTEMS

4.5 Case Studies

magnetic core
i g

moveable plunger
x h

N turn coil

Figure 23: Relay with moveable plunger. eld intensity in the core Hc , in the plunger Hp , and the gap Hg is given by iN = Hc lc + Hp h + Hg 2g with the mean core length lc . Further, the general relationship between B and H is given by B = H. This leads to iN = Bc Bp Bg lc + h+ 2g c c 0

with the ux density Bc in the core (uniform distributed over a cross section area Ac = ld of the core), the ux density Bg in the air gap (uniform distributed over a cross section area Ag = l (d x) of the air gap), and the ux density Bp in the plunger. The eld follows the path dened by the core, thus the relation = Ac Bc holds. In the air gap and in the plunger the ux is almost the same as in the core. Hence, = Ag Bg = Ap Bp . This gives iN = and L (x) = The air gap cross section area is N = = i i N2
lc c Ac h c Ag 2g 0 Ag

lc h 2g + + c Ac c Ag 0 Ag

0 N 2

Ag . g

Ag = l (d x) which gives L (x) = 0 N 2 x ld 1 . 2g d

46

4 ELECTROMECHANICAL SYSTEMS

4.5 Case Studies

The magnetic energy follows with W i (i, x) = 0 N 2 and the related force with F = W i (i, x) l = 0 N 2 i2 . x 4g x 2 ld 1 i 4g d

47

REFERENCES

REFERENCES

References
[1] F. Cellier: Continuous System Modeling, Springer Verlag, 1991. [2] A. Fitzgerald, C. Kingsley, S. Umans, Electric Machinery, McGrawHill, 1990. [3] G. Folwes: Analytical Mechanics, Saunders, 1982. [4] H. Goldstein: Klassische Mechanik, AULA Verlag, 1991. [5] M. Hirsch, S. Smale: Dierential Equations, Dynamical Systems, and Linear Algebra, Academic Press, 1974. [4] A. Kugi, K. Schlacher: Nonlinear H Controller Design for a DC-to-DC Power Converter, IEEE Transactions on Control Systems Technology , in press. [6] J. Shearer, B. Kulakowski: Dynamic Modeling and Control of Engineering Systems, Mc Millan, 1990. [7] K. Schlacher, A. Kugi, R. Scheidl: Tensor Analysis Based Symbolic Computation for Mechatronic Systems, Mathematics and Computers in Simulation, Vol. 46, 517-525, 1998. [8] A. Zemanian: Transnitness for Graphs, Electrical Networks, and Random Walks, Birkhuser, 1996.

48