Escolar Documentos
Profissional Documentos
Cultura Documentos
T LINZ
UNIVERSITA
N e t z w e r k f u r F o r s c h u n g , L e h r e u n d P r a x i s
Eingereicht von:
Betreuung:
for Susanne,
for her love, inspiration and patience.
Preface
This thesis aims at reflecting selected issues of my research activities done during the last
years as a research assistant at the Christian Doppler Laboratory for Automatic Control
of Mechatronic Systems in Steel Industries, installed at the Institute of Automatic Control
and Control Systems Technology at the Johannes Kepler University of Linz. Accompanied
and guided by the scientific advice of Professor Kurt Schlacher I had the pleasure to find
a very stimulating scientific culture. For his advice, support and for always finding time
for me, even if he was covered up with work, I wish to express my gratitude.
For having a significant contribution on arousing my interests on control theory, when
I was an undergraduate student, via his incomparable style of transporting the key ideas
in his lectures, for his advice, his help, and for preparing an experts report for this thesis,
I wish to address my thanks to Professor Andreas Kugi, chair of System Theory and
Automatic Control, University of Saarbr
ucken, Germany.
Special thanks I wish to express to our industrial partner Voest-Alpine Industrieanlagenbau GmbH Linz for the cooperativeness, the financial support, and for providing space
for research. In particular, I wish to acknowledge the contributions of Georg Keintzel,
Bruno Lindorfer and Karl Aistleitner.
To my friends and colleagues Werner Haas, Rainer Novak, Gernot Grabmair, Johann Holl, Reinhard Gahleitner, Hannes Seyrkammer, Kurt Zehetleitner, Bernhard Roider, Johannes Schrock, Markus Schoberl, Richard Stadlmayr, Martin Staudecker, Helmut
Ennsbrunner, Brigitta Peitl and Harald Pachler I wish to convey my thanks for many
non-scientific and scientific discussions, the good atmosphere, and stimulating thought
experiments from most various fields of concern.
Besides my occupation as a lecturer, I particularly enjoyed the time I had the opportunity and pleasure to spend with my friends and diploma students Klaus Straka, Gunnar
Grabmair, Marc Polzer, Thomas Rittenschober and Achim Berger. The endeavour of
conjoint research and all the conversations beyond this scope were very valuable for me.
However, the center of all my thoughts and the key reason for enjoying my life like this is
my beautiful, bright and charming wife Susanne. I love you, Susanne!
Last but not least I wish to express my deepest gratitude to my parents Maria Anna and
Franz Fuchshumer, who shaped my senses from the very beginning of my especially beautiful childhood, and made me admire nature and literature. Liebe Eltern, ich danke Euch
f
ur all Euer Engagement und daf
ur, dass Ihr mir eine unbeschwerte Zeit des Studiums
geschenkt habt!
Stefan Fuchshumer
Kurzfassung
Befasst mit der Anwendung moderner linearer und nichtlinearer Regelungstheorie adressiert diese Arbeit ausgewahlte Themen und Anwendungsbeispiele aus den Schwerpunkten
der Identifikation linearer zeitdiskreter zeitinvarianter dynamischer Systeme, einem algebraischen Zugang folgend, der Modellbildung basierend auf physikalischen Betrachtungen
sowie der modellbasierten nichtlinearen Reglersynthese. Der Beitrag zur Identifikation
Abstract
This thesis, dealing with applications of modern linear and non-linear control theory, addresses selected issues and application examples from the fields of identification of linear
time-invariant discrete-time systems following an algebraic approach, mathematical modelling based on physical considerations, and model-based non-linear control synthesis. The
contribution on identification of discrete-time linear systems is essentially stimulated by
an algebraic approach to parameter identification of continuous-time linear systems proposed in the recent literature. The discussions are given for n-th order systems resorting
to the operational domain. On the basis of a 3-mass drive-train model being available
as a laboratory setup the numerical conditions of two different parametrizations of the
identifier, which are linked by means of the bilinear Tustin transform, are investigated
and found to significantly differ with decreasing sampling times.
The second subject area of this thesis is concerned with two industrial applications
stemming from different engineering disciplines, but connected via their common differential flatness property: a cold rolling mill application from steel industries and a vehicle
dynamics control application, regarding the steering angle and the longitudinal tire forces
as control inputs, from automotive industries. In both examples the control design, based
on physically motivated models, is performed invoking the flatness-based methodologies.
It is worth mentioning that in both cases the flat output is attached with a clear physical
meaning which is valuable for control synthesis and trajectory planning as well. The focal
points of these two applications are very different, however: While in the vehicle dynamics
application the mathematical model in the shape of the well-known planar (holonomic)
bicycle model is taken from the literature and the central new contribution is the observation of the flatness property and the determination of a representative of the flat output,
the focus of the rolling mill example is laid on the development of a model being suitable
for the simulation and control purpose. This effort on modelling arises from the objective
to rendering also those roll gap scenarios adequately which are characterized by significant
elastic work roll deformations, and, thus, a non-circular shape of the roll gap. In order to
develop a model being suitable for the control purpose due to manageable computational
effort, an approximation of the elastio-static work roll deformation problem invoking the
Rayleigh-Ritz method is proposed, with particular emphasis laid on the choice of a set of
appropriate shape functions for the displacement fields. This approach, combined with
a mathematical description of the deformed strip taken from the literature, finally gives
the non-circular arc roll gap model. With this model as a central point, a flatness-based
approach to multivariable control of a cold rolling mill is proposed.
CONTENTS
1 Introduction
5
5
7
8
14
17
18
19
21
34
35
36
37
37
38
39
41
44
47
68
70
72
79
82
83
84
85
88
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
90
90
90
92
93
A Flatness-based Approach
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
96
97
99
100
102
103
109
116
119
. . . .
. . . .
Rolls
. . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
122
II
chapter
ONE
Introduction
ddressing selected issues of the field of modern control theory, this thesis spans from
an algebraic approach to discrete-time linear systems identification, mathematical
modelling of complex industrial systems based on physical considerations, to the application of non-linear model-based control theory, in particular flatness-based methods. Both
main examples of this thesis, namely a cold rolling mill example from steel industries to
be dealt with in the Chapters 3 and 4, and an example from the automotive industry
on vehicle dynamics control (to be addressed in Chapter 5) are linked by their common
property of differential flatness, with the flat output attached with a clear physical meaning. Differential flatness, introduced by Michel Fliess, Jean Levine, Philippe Martin and
Pierre Rouchon in [FLMR92], [FLMR95], is a structural property that allows for a complete, finite and free differential parametrization of a dynamic system by means of the flat
output. Thus, once chosen the trajectories of the flat output, the associated trajectories
of the system variables, in particular those of the control inputs, can be derived from
this flat outputs trajectory and its time derivatives, without the need for integrating the
systems differential equations. Moreover, given a flat output there exists a systematic
approach to the control synthesis.
Additionally to their common structural property regarding flatness, both applications
have in common that their mathematical models are evolved on the basis of physical
considerations thus allowing for a parametrization in terms of the geometry data and
the material parameters. This firm reference to physics exploits significant advantages
for the system analysis and the control synthesis as it provides additional insight into
the system, and, not at least, for the re-usability of the models (and the controllers) for
different system configurations.
Finally, the third common ground relating these applications is that in both cases
the trajectory tracking problem is central, an objective well complying with the flatness
based control approach. Clearly, in the field of rolling mills, for productivity reasons the
time made available to the acceleration and the deceleration of the mill is required to be
kept as short as possible, with simultaneously meeting the tight tolerances imposed on
the control variables. However, apart from demands specified on-line by the plant opera-
1. Introduction
tor, the trajectories of the flat output of the rolling mill, which is found to coincide with
the control variables, might be designed off-line. Contrary, the trajectory planning task
for the vehicle, when addressing vehicle dynamics control to improve the safety, clearly
requires a real-time trajectory generation permanently accounting for the inputs supplied
by the driver, i.e. the current angle of the steering wheel and the current position of
the throttle/brake pedal. Besides the objective of supporting the driver in case of an
emergency situation via conjoint actuation of the steering angle and the engine/braking
torque, i.e. vehicle dynamics control as primarily addressed in Chapter 5, there is a possible second field of application of the vehicles (configuration) flatness property, namely
the task of tracking a vehicle along a pre-specified path on a test or race track. This
task, typically left to test pilots, is arranged in order to investigate e.g. the consequences
of different chassis adjustments. Automatic tracking control for test vehicles (equipped
with global-positioning systems) on test tracks might help in order to produce particular
traceable runs.
While in the vehicle dynamics control example the mathematical model the control design
is based upon is taken from the literature and the main new contribution is the observation
of the differential flatness property, the focus of the rolling mill example is laid on the
derivation of a mathematical model, being suitable for the simulation and the control
purpose. The impulse for setting up a new rolling mill model, particularly concerning
the modelling of the roll gap, crucially stems from the following observations: In order
to being applicable to cold, thin strip and temper rolling, the roll gap model is required
to include a detailed description of the elastic work roll deformations which occur under
the action of the rolling load. In the case of cold rolling, typically the assumption of
a circular roll/strip contact shape, though with a larger so-called equivalent roll radius,
qualifies to be appropriate, cf, e.g., [BF52]. However, with decreasing strip thickness and
in the temper rolling case as well where the thickness reduction of the strip is very small,
this simple assumption is to be dropped. There is a huge literature on roll gap models
including a detailed description of the work roll deformations, typically referred to as noncircular arc roll gap models, see, e.g., [JOZ60], [FJ87], [FJMZ92], [DET94], [LS01]. They
are usually either based on the so-called elasto-static half-space solution or on Jortners
solution respectively, i.e., the solution to the problem of an elastic half-space/cylinder
loaded with a constant normal stress. Due to the computational effort involved, these
models are primarily intended for off-line calculations. So, Chapter 3 aims at introducing
a novel non-circular arc roll gap model exhibiting reduced computational effort and being
thus applicable for the control purpose. To this end, the elasto-static work roll deformation
problem is addressed in the sense of the Rayleigh-Ritz method, with special emphasis laid
on an appropriate choice of shape functions. For the case study discussed in Chapter 3, a
total number of 56 generalized coordinates (i.e. degrees of freedom) of the proposed Ritz
ansatz suffices to obtain a very suitable result, compared with a model incorporating a
detailed description of the roll deformations. The approximation of an infinite-dimensional
system via a suitable Ritz approximation in order to obtain a model the control design
can be based upon is an approach very often encountered in control applications, as, e.g.,
1. Introduction
in structural control, for systems including beams and plates, etc. Typically, as also seen
in the roll gap modelling, the key problem is to find an appropriate set of shape functions.
As sketched above, the key contribution concerning the vehicle dynamics control application of Chapter 5 is the observation of the differential flatness property of the planar
holonomic bicycle model and the proposition of a flatness-based approach to vehicle dynamics control regarding the (front) steering angle and the longitudinal forces of the
tires as control inputs. Vehicle dynamics control systems as e.g. ESP (electronic stability program) acting on the brakes and the traction control system are implemtented
in series-production vehicles. These systems support the driver in emergency situations
by producing a (counter-) yaw torque due to individually controlled braking of all four
wheels in the case of exceedance of a certain yaw rate. Accordingly, the potential of involving the steering system to handle emergency situations is very rich. The automotive
industry aims at utilizing these possibilities e.g. by means of supporting the driver for
the difficult task of counter-steering by application of a respective artificial torque to the
steering wheel, i.e., as a haptic recommendation for the driver. As a very recent advance,
active front steering (AFS) [KB02] has been implemented in series-production vehicles.
The basic function of AFS is to mechanically add an additional steering angle (adjusted
e.g. by an electric drive) to the steering angle given by the driver. Besides the objective
to enable a velocity-dependent gain of the steering mechanism, the AFS system can be
used for vehicle dynamics control, too.
The bicycle model, proposed in [RS40], evolves from the four-wheel car by gluing
together the front and the rear wheels to a single (mass-less) front and rear wheel, located
at the longitudinal axis of the car. The contact between the tires and the road is modelled
in terms of contact forces, which implies that the tires are enabled to slip and slide on
the road. This planar model, known as a well-established basis for the design of vehicle
dynamics control systems, see, e.g., [ABO99], [B
un98], [Rit04], is capable of rendering the
longitudinal, lateral and yaw dynamics of the vehicle. The pitch and roll dynamics of a
vehicle are clearly not involved in the scope of this model.
Within the scope of vehicle dynamics control as an assistance for the driver to cope
with emergency situations, the components of a flat output of the bicycle dynamics,
with the subsystem related to the global position dropped as not being involved in the
control objective, are revealed as the lateral and the longitudinal velocity component of
a distinguished point located on the longitudinal axis of the vehicle. This property is
shown for the front-, rear- and all-wheel driven vehicle, without referring to particular
representatives of the functions modelling the lateral tire forces. In case the global position
of the vehicle is involved in the control objective (e.g., tracking of a prescribed path), the
configuration flatness property of the bicycle dynamics, with the position of the point
qualifying as a flat output, can be exploited.
In order to leading over directly to the matter of this thesis, we will proceed by introducing
a contribution on discrete-time linear systems parametrical identification in Chapter 2, essentially stimulated by the algebraic approach [FSR03] to continuous-time linear systems
1. Introduction
identification introduced by Michel Fliess and Hebertt Sira-Ramrez. By now, this algebraic approach [FSR03] has led to a variety of related research, as e.g. on fault diagnosis
[FJ03], on signal processing and signal compression [FMMSR03], and on the estimation
of the derivatives of signals [RSRF05].
In accordance to the continuous-time framework of [FSR03], the operational representation of the discrete-time constant linear system (in the z-domain) is considered. Initial
conditions are allowed to being ignored by taking derivatives with respect to the operational operator z. To determine the unknown system parameters, subsequent iterated
summations of the discrete-time counterpart of the resulting operational equation are
carried out to set up a system of linear equations. To cope with measurement noise,
besides the possibility of a straight-forward incorporation of linear filters, the setup of an
overdetermined system of linear equations by means of additional iterated summations
qualifies to be suitable. Starting with introductory examples to illustrating the proposed
approach, the discussion then proceeds with the elaboration of a linear identifier for a
general n-th order discrete-time constant linear dynamics.
On the basis of a fifth order model of a drive-train, which is available as a laboratory
experiment, the problem of inaccurate estimation of those system zeros, which have only
minor effect on the system response, and, hence, are difficult to estimate in presence of
noise, is illustrated. In order to overcome this problem, the idea to discarding (or presetting) those non-essential zeros is proposed, first commencing with the application of
the bilinear Tustin transform, also referred to as q-transform for short, to the z-domain
representation. The motivation for discussing this idea for the q-domain case first simply
is that, by virtue of the close similarity to the continuous-time domain, things might be
particularly apparent from the control engineers point of view. Clearly, discarding certain
zeros of the q-domain transfer function, i.e. shifting those zeros to infinity, corresponds
to placing the associated zeros of the z-domain transfer function at z = 1. This idea of
pre-setting those non-essential zeros to obtain a reduced-order linear identifier is shown
to providing particular attractiveness.
Again referring to the drive-train example, it is found that the numerical condition of
the linear identifier given in terms of the q-domain parameters shows significant advantages compared to the z-domain parametrization, becoming particularly apparent with
decreasing sampling times. This effect is observed independently from whether or not the
idea of pre-setting certain zeros is applied. The increasingly poor numerical condition of
the z-domain approach emerging with decreasing sampling times might become apparent
by reflecting the relation zi = exp (si Ta ) between the poles si of the continuous-time system and the poles zi of the according discrete-time representation. Hence, with decreasing
sampling time Ta , the poles zi approach the point z = 1.
To resume, the algebraic theory on linear identification of [FSR03], which, in turn,
has led to the research to be introduced in Chapter 2, provides easy-to-implement on-line
identifiers, which have shown to perform well in applications.
chapter
TWO
An Algebraic Approach to Discrete-Time Linear Systems
Identification
he following investigations aim at introducing an approach to discrete-time linear systems parametrical identification, essentially stimulated by the theory of M.Fliess and
H.Sira-Ramrez developed for continuous-time linear systems [FSR03]. As a prerequisite
for addressing the discrete-time framework, we will briefly revisit some key ideas of the
continuous-time identification approach on the basis of a simple example, following the
lines of [FSR03]. The underlying theoretical framework utilizing, in particular, module
theory and Mikusi
nskis operational calculus, can also be found in [FSR03].
y (0) = y0 ,
y (0) = y 0
(2.1)
s2 + a1 s + a0 y y 0 (s + a1 ) y0 = b
u+ .
(2.2)
s
Now, multiply by s and then take derivatives, three times, with respect to s in order to allow
for ignoring the initial conditions y0 , y 0 and the constant load perturbation. To avoid, in
the according time-domain representation, derivatives with respect to time, divide both
sides of the resulting equation by s3 and re-sort w.r.t. the parameters a0 , a1 and b,
3
3 d2 y
6 d2 y
6 d
y
3 d2 u
1 d y
1 d3 u
1 d3 y
a0 +
b=
+
+
+
a1
+
s2 ds3 s3 ds2
s ds3 s2 ds2 s3 ds
s2 ds3 s3 ds2
d3 y 9 d2 y 18 d
y
6
= 3
2
3 y. (2.3)
2
ds
s ds
s ds s
5
2. Identification. . .
(2.4)
(2.5)
with1
Z
(2)
(3)
(2)
(3)
ty,
t y6
t y+6
(2.6)
Z
Z (2)
Z (3)
3
2
Q1 (t) = t y 9 t y + 18
ty 6
y.
t y,
t y+3
P11 (t) =
Z (2)
Z (3)
3
P13 (t) =
t u3
t2 u,
P12 (t) =
(i1)
P1j (t) ,
Qi (t) =
(i1)
Q1 (t) ,
i = 2, 3,
j = 1, 2, 3,
(2.7)
T =
a0 a1 b
(2.8)
Remark 2.1 There are no derivatives with respect to time involved, but only integrations. This is particularly valuable in the presence of high frequency perturbations or
measurements corrupted by noise.
Remark 2.2 The incorporation of a filter yf = F (s) y is straight-forward. Clearly, the
order of derivatives w.r.t. s increases by the order of F (s).
Remark 2.3 (Implementation issues). Instead of taking dim () 1 iterated integrals for
setting up a system of dim () linear equations as done in the introductory example, it is
useful also to consider subsequent time integrals thus yielding an over-determined set of
equations, to be solved e.g. by means of the least-squares method. Additionally, dropping
the first few equations of the (over-determined) system P = Q, which are clearly more
articulately affected by noise than the subsequent time integrals, has turned out to improve
the performance of this parameter estimation method significantly.
1
The notation
is arranged.
R ()
(t) =
R t R 1
0
R 1
0
( ) d d2 d1 ,
(t) =
R (1)
(t) =
Rt
0
(1 ) d1
2. Identification. . .
t0
namely = t, we have
Z
Z t
f ( ) d =
0
f (
t) td
=t
+k+1
1
0
(1 ) k x (
t) d
.
(2.10)
Example 2.2 (Example 2.1 contd) With (2.9) and (2.10) we obtain
Z 1
ti+4
Pi1 (t) =
((i + 4) 3) (1 )i 2 y (
t) d
(i + 1)! 0
Z 1
ti+3
Pi2 (t) =
(i + 4) (i + 3) 2 6 (i + 3) + 6 (1 )i1 y (
t) d
(i + 1)! 0
Z 1
ti+4
((i + 4) 3) (1 )i 2 u (
t) d
,
Pi3 (t) =
(i + 1)! 0
for i 1, and
Q1 (t) = t
y (t) 3
Z 1
i+2
1
2
10
8
+ 1 y (
t) d
t
24 (i) 3 923 (i) 2 + 1822 (i) 6 (1 )i2 y (
t) d
,
(i + 1)! 0
Q
i 2, with (i) = j= (i + j). Hence, for numerical reasons it is suitable to factor
(t), with (t) = diag (ti+2 / (i + 1)!),
P (t) and Q (t) as P (t) = (t) P (t), Q (t) = (t) Q
i = 1, . . . , dim Q.
Qi (t) =
(2.11)
i=0
2. Identification. . .
X
fi z i
(2.12)
fz (z) =
i=0
is absolutely convergent for all z with |z| > A. The holomorphic function f z (z) is called
the z-transform of (fk ), indicated as fz (z) (fk ) for short. Notice that
!
n1
X
(2.13)
fj z j .
(fk+n ) z n fz (z)
j=0
Y
X
X
dj
dj
(i + s) fi z i ,
fz (z) =
fi j z i = (1)j z j
dz j
dz
s=0
i=0
i=0
thus, we have
z j f (j) (z) (1)j
j1
Y
(k + s) fk
s=0
(2.14)
arranging the notation ()(j) = (z )j . To start with, let us first illustrate the basic ideas
of the identification method to be discussed with a simple introductory example.
Example 2.3 Consider the system
yk+1 + ayk = buk ,
(2.15)
(2.16)
Taking derivatives, twice, with respect to z in order to allow for ignoring the initial condition yields
2yz(1) + (z + a) yz(2) = bu(2)
(2.17)
z .
Notice that
zf (1) (z) (kfk ) ,
z 2 f (2) (z) (k (k + 1) fk ) ,
(2.18)
2. Identification. . .
(2.19)
Following the lines of the continuous-time approach, e.g. calculate N 1 iterated sums of
the right-hand side of (2.19), i.e., divide (2.18) N times by (z 1), to set up the system
of (N + 1) linear equations for determining the parameters a and b,2
k (k 1) yk
k (k 1) yk1
k (k 1) uk1
k1
k1
k1
P
P i (i 1) yi
P i (i 1) yi1
i
(i
1)
u
i1
i=1
i=1
,
i=0
+
e
=
..
..
..
.
.
.
(N )
(N
(N
P)
P)
P
iN (iN 1) yiN
iN (iN 1) uiN 1
iN (iN 1) yiN 1
eT = e0k e1k eN
. The least-squares solution to this (under-determined) problem,
k
1 T
2
i.e., min kek2 , T = [a, b], P + e = Q, reads = P T P
P Q.
Though aiming at discarding the initial conditions by taking derivatives w.r.t. z, the
number of derivatives for achieving that goal is clearly not unique. In Example 2.3, the
initial condition y0 was discarded by differentiating twice, i.e., (n + 1) times, w.r.t. z.
However, as an alternative way, one might think of first dividing (2.16) by z, followed
by an n times differentiation in order to allow for ignoring the initial conditions. This
approach is illustrated via the following example, and further pursued in Remark 2.5.
Example 2.4 Consider again (2.15) and (2.16). Now, divide (2.16) by z, then differentiate once w.r.t. z to obtain
2
1 + az 1 yz(1) az 2 yz = b z 1 u(1)
z z uz .
Collecting for the parameters a, b,
1
1
1
1
(1)
(1)
yz yz a
uz uz b = yz(1) ,
z
z
z
z
Compare this expression with (2.19). The iterated summation and the calculation of the
parameters proceeds as discussed in Example 2.3.
2
The notation
()
P
fi =
1
k1
P i1P
i1 =0 i2 =0
i1
P1
i =0
fi is arranged.
2. Identification. . .
10
j=0
i=0
j=0
hence,
A z yz
n
X
ai z i
i=0
i1
X
j=0
yj z j = Bz uz
m
X
an = 1,
bi z i
i1
X
uj z j .
(2.21)
(2.22)
j=0
i=0
nj)
(j) (
nj)
Bz(j) u(
,
(2.23)
A z yz
=
z
j
j
j=0
j=0
(j)
(j)
noticing that Az (z) = 0, j > n, and Bz (z) = 0, j > m. With the j-th order derivatives
of the polynomials Az and Bz reading
A(j)
z
n
X
s=j
s!
as z sj ,
(s j)!
Bz(j)
m
X
s=j
s!
bs z sj ,
(s j)!
j)!
(
n
j)!
j
j
s=j
s=j
j=0
j=0
(2.24)
To further reveal the structure of (2.24), and, in particular, to facilitate the transformation
back to the discrete-time domain, see also (2.14), let
(j)
. z j yz
Yz(j) =
,
j!
(j)
. z j uz
Uz(j) =
,
j!
(2.25)
2. Identification. . .
11
Yz
Uz
n
!
as z
=n
!
bs z sn .
(2.26)
j
j
s=j
s=j
j=0
j=0
Y
,
Uz
bi =
n!
z
Yz
ai
n
!
z
n
!
j z
j
j
j=0
i=0
j=0
i=0
j=0
(2.27)
noticing that an = 1 by (2.11). Thus, each parameter ai (or bi ) is associated with an
(
nj)
(
nj)
, respectively), 0 j i, which is to be shifted
(or Uz
expression involving Yz
(n i) times.
Now, let us determine the discrete-time representation of (2.27). First of all, from
(2.14) and (2.25) we obtain
!
j1
j
o
n
Y
(1)
Fz(j)
(2.28)
Yz , (yk ) , Uz , (uk ) .
Fz , (fk )
(k + s) fk ,
j!
s=0
i
X
i
j=0
= (1)n
Fz(nj) (1)n
n
1j
i
X
i (1)j n
! Y
j=0
(
n j)!
s=0
(k n + i + s) fkn+i =
! n 1
!
ij1
Y
Y
i
n
!
(1)j
(k n + n
+ s)
(k n + s) fkn+i ,
j
(
n
j)!
s=0
j=0
s=i
i
X
j)!
j
s=0
s=0
j=0
we finally find the relation
!
i
X
i
Fz(nj) (1)n
!
z in n
j
j=0
n
1
Y
s=0
(k n + s) fkn+i
(2.29)
Thus, with (e0k ) denoting the equation error sequence of (2.27), we obtain the discrete-time
representation
!(
)
n
1
n1
m
Y
X
X
e0k = (1)n
(k n + s)
yk +
ykn+i ai
ukn+i bi =
= (1)n
s=0
n
1
Y
s=0
i=0
(k n + s)
yk
hTa,k hTb,k
i=0
(2.30)
2. Identification. . .
hTb,k =
ukn ukn+m
12
(2.31)
referred to as the data vectors as containing the measurements. The (n + m + 1) parameters are arranged to form the vector ,
(2.32)
T = a0 an1 b0 bm .
Remark 2.5 (following the idea as sketched in Example 2.4) Take (2.22), divide by z,
and subsequently differentiate n times w.r.t. z in order to allow for ignoring the initial
(n)
(n)
conditions. Then, (z 1 Az yz ) = (z 1 Bz uz ) . By introducing fz = z 1 fz , fz {yz , uz },
corresponding to an index shift by 1 in the discrete time domain, we have
(Az yz )(n) = (Bz uz )(n) .
This problem can be immediately traced back to the case discussed above (with n
= n),
thus yielding
!
n1
Y
e0k = (1)n
(k n + s)
yk1 hTa,k1 hTb,k1 ,
s=0
see (2.30)(2.32). The application of an index shift by +1 on the right-hand side finally
gives the result
!
n
Y
n
(k n + s)
yk hTa,k hTb,k .
(2.33)
e0k = (1)
s=1
(2.34)
For calculating the parameters , the equation error sequence (2.30) or (2.33), respectively, is subject to N n + m fold iterated summation to assemble the set of (N + 1)
linear equations
P + e = Q,
(2.35)
cf. Example 2.3.
Implementation Issues
Referring to Figure 2.1, which depicts the structure of the proposed z-domain online
identification approach, the subsequent notes aim at drawing the link to a computeralgebra implementation available at http://regpro.mechatronik.uni-linz.ac.at.
..
n2.Identification.
m
(yk )
z 1
..
.
n
shifts
n1
...
z 1
..
.
z 1
...
2n
n
ykn
ykn+1
yk1
multiply by (1)n
P1,1,k
P1,2,k
1
z1
P1,n,k
...
..
.
..
.
1
z1
1
z1
a (0)+N
Qn 1
s=0
...
P1,n+1,k
1
z1
..
.
1
z1
1
z1
an1
yk
b (0)+N
b0
Q1,k
P1,n+m+1,k
...
1
z1
b (0)+1
..
.
a (n1)+N
ukn+m
a (n1)+1
a (1)+N
a1
ukn
1
z1
a (1)+1
a (0)+1
a0
n+1
n
shifts
2nm
1
z1
(yk )
z 1
.. 1
.
..
.
13
1
z1
b (m)+1
...
+1
..
.
..
.
1
z1
1
z1
b (m)+N
+N
bm
Figure 2.1: Scheme of the online z-domain identification approach emanating from (2.30)
or (2.33), respectively. Arrangement of the shifter and integrator states , and link to
the composition of (2.37).
The focal point of Figure 2.1, introducing a possible scheme of a state-space realization
(with the state ), is the representation of (2.30) or (2.33), respectively, processing the
measurements ykn+i , i = 0, . . . , n, and ukn+i , i = 0, . . . , m, made available via the shifter
chains. The expressions of (2.30), (2.33), decomposed with respect to the parameters
{ai , bi }, are further processed by means of N -fold iterated summation each, also referred
to as (discrete-time) integrator chains. With
a (s) = 2n + sN,
s = 0, . . . , n 1
b (s) = 2n + (n + s) N,
s = 0, . . . , m,
(2.36)
2. Identification. . .
P1,1
...
P1,n
P1,n+1 . . . P1,n+m+1
a (0)+1
a (n1)+1
b (0)+1
b (m)+1
..
..
..
..
.
.
.
.
a (0)+N
a (n1)+N
b (0)+N
b (m)+N
...
...
e =
Q1
+1
..
.
+N
14
(2.37)
Q1,k = (k) yk ,
i = 0, . . . , n 1, j = 0, . . . , m, with
(k) = (1)
n
1
Y
s=0
(k n + s)
or
(k) = (1)
n
Y
s=1
(k n + s) ,
(2.38)
1 + q/0
,
1 q/0
q = 0
z1
,
z+1
0 =
2
,
Ta
(2.39)
referred to as Tustin transform or q-transform for short, let us state a preliminary remark
on subtleties associated to the parametrization of the linear identifier in terms of the zdomain parameters as introduced above. These issues will be illustrated in Section 2.2.6
on the basis of simulation and measurement results from a 5th order drive-train example.
Remark 2.6 Given a continuous-time system G (s) with poles si , then the poles of the according discrete-time system G (z) are located at zi = exp (si Ta ). Thus, with the sampling
time Ta decreasing, the poles zi approach the point 1.
The q-domain transfer function, outlined in terms of the z-domain parameters a i ,
i = 0, . . . , n 1, and bi , i = 0, . . . , m, is
nm
m
m
bm 1 + q0
+ . . . + b0 1 q0
1 q0
# (q) = G (z)| 1+ q =
G
n
n1
n ,
0
z=
q
q
q
q
q
1
1 + 0 + an1 1 0
+ . . . + a 0 1 0
1 + 0
0
# (q) arranged to indicate the representation in terms of {ai , bi }.
with the prime on G
Clearly, an (n m)-fold zero at q = 0 occurs. Next, arrange a change of the (n + m + 1)
parameters,
(a0 , . . . , an1 , b0 , . . . , bm ) 7 (A1 , . . . , An , B0 , . . . , Bm ) ,
(2.40)
2. Identification. . .
0
0
i=0
=
G# (q) =
.
n
# (q)
P
A
i
1+
Ai q
15
(2.41)
i=1
Notice that, in general, the map (2.40) is non-linear due to the scaling chosen for the denominator, A# (0) = 1. The following (closely related) remarks state some facts on G# (q)
to motivate the ideas leading to the q-domain identification method to be introduced.
Remark 2.7 Given the continuous-time system G (s), the approximation
Ta
2
#
tan
,
|Ta | 1
G (j) G (j) ,
=
Ta
2
holds. In particular, this fact gives rise to discard, except for the (n m)-fold zeros at
0 , those zeros of G# (q) located significantly beyond the locus of the system poles, as they
affect the system response inessentially.
Remark 2.8 Let s = a be a pole of G (s), then the according pole of G# (q) is q =
0 tanh (a/0 ). According relations hold for complex conjugate poles. This observation
draws a close link between the q-domain and the s-domain identification.
Now we are ready to state the point of departure for the q-domain parametrization
of the identification approach. Following Remark 2.7, we propose to (optionally) discard
those zeros located significantly beyond the domain of the poles of G# (q) to obtain
nm m
nm
P i
q
q
# (q)
1
Bi q
1
B
0
0
i=0
# (q) =
G
=
,
m
m,
(2.42)
n
# (q)
P
A
i
1+
Ai q
i=1
with the tilde indicating the approximation of G# (q). Next, we will trace back to the
z-domain solution, and re-cast those results ((2.30) and (2.33), respectively) in terms of
the q-domain parameters ,
0 B
m .
(2.43)
T = A1 A n B
To this end, let us first determine the parameter transformation. The z-domain transfer
function (with an 6= 1 in general) according to the specified q-domain representation
# (q) reads
G
2nm
G (z) =
P
i i0 (z 1)i (z + 1)mi
B
i=0
n
(z + 1) +
n
P
i=1
Ai i0
(z 1) (z + 1)
ni
m
P
= s=0
n
P
s=0
bs z s
.
as
zs
(2.44)
2. Identification. . .
x, y C,
n N0+ ,
16
(2.45)
we have
(z 1) (z + 1)
ni
i
ni X
X
i n i j+l
z =
(1)
= (1)
j
l
j=0 l=0
!
s
n
X
X
ni
i
zs,
(1)s
= (1)i
(1)j
j
sj
s=0
j=0
i
and, finally,
m
m
X
X
s
X
i
) !
i
mi
i z s
(1)s+j
2nm
(0 )
B
sj
j
s=0
j=0
i=0
(
G (z) = n
) ! .
n
s
X
X
X
n
i
ni
i
s+j
+
(0 )
(1)
Ai z s
s
s
j
j
s=0
i=1
j=0
(
(2.46)
X
i ,
bs,i B
s = 0, . . . , m
(2.47)
bs =
i=0
and
X
n
n
+
as,i Ai ,
as =
s
i=1
s = 0, . . . , n,
(2.48)
with the (m+ 1) (m
+ 1) dimensional matrix bs,i and the (n + 1) n dimensional
matrix as,i ,
s
X
i
mi
i
s+j
b
nm
s,i = 2
(0 )
(1)
,
sj
j
j=0
(2.49)
s
X
i
ni
i
s+j
a
s,i =
(0 )
(1)
.
sj
j
j=0
Plugging the transformation (2.47)(2.49) into the z-domain solution (2.30) or (2.33)
(notice that these equation have to be slightly adapted so as to allow for arbitrary a n ),
we finally find
! )
!
( n
m
m
n
n
X
X
X
X
X n
i
ukn+r br,i B
ykn+r ar,i Ai
e0k = (k)
ykn+r +
r
r=0
r=0
r=0
i=0
i=1
2. Identification. . .
n
a T b
0.
0
T
T
ha,k yk ,
ha,k yk ..
hb,k ,
ek = (k)
17
(2.50)
Remark 2.9 The procedure of setting up the set of linear equations by means of iterated
summation of (2.50), and the solution for as well, proceeds as discussed above.
zero at z = 1).
Now, introduce the counterpart of (2.42),
(z + 1)mm
(z) =
G
P
bi z i
i=0
n
P
ai
zi
i=0
m
m
Pm P
bi
s=0 i=0
n
P
mm
ai
zi
m
P
z s+i
i=0
= P
n
i=0
bi z i
,
ai
(2.51)
zi
i=0
with the tilde indicating the approximation of G (z). There should be no confusion
by labeling the coefficients of the numerator again as bi , i = 0, . . . m, though now
(z) instead of G (z). The injective linear map relating the parameters
referring
to G
bT = b0 . . . bm to the numerator coefficients bT = b0 . . . bm is represented as
b = bb. Let
T = a0 an1 b0 bm ,
(2.52)
then
with
E nn 0
0
b
reads
and the linear identifier (2.37), re-casted in terms of the parameters ,
+ e = Q,
P
P = P .
(2.53)
2. Identification. . .
18
n1
X
ai ykn+i =
i=0
ek = yk
ykn+i ai +
i=0
bi ukn+i ,
m n,
i=0
m
X
m
X
i=0
ukn+i bi
= yk
hTa,k hTb,k
(2.54)
see (2.31) and (2.32) for the definitions of ha,k , hb,k and . Compare this definition of (ek )
to (2.30) and (2.33).
Referring to the z-domain representation, with G (z) due to (2.21), this error ez (z),
also referred to as the generalized error as linearly depending on the parameters ,
reads
Pm
bi z i
zn
1
e z = M 2 y z M 1 uz ,
M1 (z) = i=0n
,
M2 (z) =
.
P
i
z
z n + n1
i=0 ai z
Taking N + 1 measurements, ek = yk hTa,k hTb,k , k = 0, . . . , N , we end up with
the (under-determined) system of N + 1 linear equations
hTa,0 hTb,0
y0
e0
.. ..
.. ,
..
(2.55)
. = .
.
.
T
T
ha,N hb,N
yN
eN
1 T
or e = Q P for short. The least-squares solution to (2.55), = P T P
P Q, finally
2
gives the estimate for in the sense min kek2 .
This is a standard procedure for discrete-time linear systems identification. Analogously to the previous discussion, we will recast this method in terms of the according
q-domain parameters , see (2.43). To this end, commence from (2.54), adapted to allow
for an arbitrary, and apply the transformation (2.47), (2.48) to obtain
!
!
m
m
n
n
n
X
X
X
X
X
n
i .
ukn+r br,i B
(2.56)
ykn+r ar,i Ai
ykn+r +
ek =
r
r=0
r=0
r=0
i=0
i=1
In a more compact notation, the resulting set of (N + 1) linear equations evolving from
(2.56) reads
n
T
T
h
hTa,0 y0
h
y
e0
0
a,0
b,0
0
.. ..
.. ..
..
.. a , .. b . (2.57)
=
. .
.
. .
.
.
n
T
T
T
h
y
h
y
eN
hb,N
N
N
a,N
a,N
n
Compare this result to (2.50).
2. Identification. . .
19
Remark 2.10 Each equation of (2.55) or (2.57), evaluated at k, i.e., (2.54) or (2.56),
involves the measurements yks , uks , s = 0, . . . , n, but, in contrast to the approach
introduced in the Sections 2.2.1 and 2.2.2, clearly not the prior history.
X
1
+
1
+
z1
.
0
0
fq (q) = fz
fi
,
(2.58)
=
,
fz (z) = fq 0
q
1 q0
1
z
+
1
0
i=0
indicated as fq (q) fz (z) (fk ) for short. For notational convenience the abbreviation z = (q) is arranged for (2.39), thus fq = fz . The identification procedure to be
outlined will involve the repeated derivatives of signals fq w.r.t. q,
2/0
(1)
2 fz ,
(1 q/0 )
2
2/0
q
2
(2)
(1)
(q ) fq (q) =
fz + 1
fz
.
0
(1 q/0 )2
q fq (q) =
(i)
(2.59)
(i)
fq(1) (kfk )
2
0
and
0
2
q
0
2 !!2
cf. (2.14), with ? as the convolution operator. However, for convenience, within the
subsequent discussions we will not carry out the transformations up to the discrete time
domain, but evolve from the representations
2/0
(1)
(1)
fq =
F
(1 q/0 ) (1 + q/0 ) z
(2.60)
2
q
2/0
(2)
(1)
(2)
Fz
2F z + 1 +
fq =
(1 q/0 ) (1 + q/0 )
0
(j)
see (2.25) for the definition of Fz .
2. Identification. . .
20
Example 2.5 (Example 2.3 contd) Application of (2.39) to (2.16) yields the q-domain
representation of (2.15),
q
q
q
yq 1 +
y0 = 1
buq .
(2.61)
1 + a + (1 a)
0
0
0
The transfer function, obtained via the re-parametrization a = (1 A0 ) / (1 + A0 ),
b = 2B/ (1 + A0 ), is
(1 q/0 ) B
G# (q) =
,
1 + Aq
see (2.40) and (2.41). Then, (2.61) takes the form
q
1
q
y0 = 1
Buq ,
(1 + Aq) yq (1 + A0 ) 1 +
2
0
0
establishing the point of departure. Taking derivatives, twice, with respect to q in order to
allow for ignoring the initial condition yields
q
2 (1)
(2)
(1)
(2)
(1 + Aq) yq + 2Ayq = B
1
u
.
uq
0
0 q
2
2
1 + q0 , and re-sort with respect to the
Next, apply (2.60), multiply by 20 1 q0
parameters {A, B},
!
!
q
(2)
(2)
2Yz q
2
Y
z
0 (2)
2
+ 0 Yz(1) A 4
U B = 2
+ Yz(1) .
(2.62)
1 + q0
1 + q0 z
1 + q0
(j)
(j)
(For notational brevity, the composition of Yz , Uz with has been dropped). It is easy
to verify that the discrete-time representation of (2.62) coincides with (2.50).
(2.11), outlined in terms of the parameters Ai , Bi , see (2.42). Taking derivatives w.r.t.
q, n
= n + 1 times, to discard the initial conditions, we obtain
!n+1
nm
q
n+1
# uq
B
,
A # yq
=
1
0
and, by virtue of Leibniz rule,
n
X
n + 1 #(j) (n+1j)
A
yq
j
j=0
!(j)
nm
q
#
1
B
uq(n+1j) = 0.
0
(2.63)
2. Identification. . .
21
(i)
i 0, with
ci,j = 2
ji
i! i 1
.
j! j 1
(2.65)
(j)
0
j=1
(i)
Clearly, by virtue of this equation, the relation of fq to the discrete-time domain is easily
established.
n+1
q
q
n+1
1
Again, the coincidence of (2.63), multiplied by 2 1 + 0 0
, with
1 0
(2.50) can be verified by straight-forward, though tedious, computations.
=
u
23.7
s 2 ,
s
s
s
s 2
1 + 2 0.1
1 + 2 0.0083
1+
+
+
8.1
12.4
12.4
27.7
27.7
PSfrag replacements
2. Identification. . .
permanent-magnet
dc motor
2
c1
22
c2
2. Identification. . .
23
the system response is seen to be appropriate, see in particular the top right subplot
of Figure 2.5. The estimation of the poles is again accurate. Additionally, due to the
reduced number of parameters, this approach is associated with the advantage of having better numerical condition compared to the case m
= m of the previous example,
illustrated by the following condition numbers: (10ms, 1.8e7, 2.3e7), (5ms, 6.0e8, 2.7e7),
(2ms, 3.5e10, 3.0e7), (1ms, 6.4e11, 3.1e7), (0.5ms, 1.1e13, 3.2e7). Again, the numerical
condition of the z-domain approach suffers with decreasing sampling times, whereas the
parametrization in terms of does not experience these problems. The according condition
numbers obtained by noise-free simulations are: (10ms, 1.9e7, 2.3e7), (5ms, 5.8e8, 2.7e7),
(2ms, 3.5e10, 3.0e7), (1ms, 6.4e11, 3.1e7), (0.5ms, 1.1e13, 3.2e7). To close this example,
Figure 2.7 finally depicts the Bode diagrams of the z- and q-domain identification results,
obtained with Ta = 1ms.
Example 2.8 (measurement results) Let m
= 0 and Ta = 10ms. The Figures (2.8, 2.9)
depict the identification results obtained from measurements of the lab model drive-train.
See again the figure captions for comments.
Remark 2.11 Regarding the performance of the identifiers (and the numerical condition
as well), investigated on the basis of the drive-train example, no clear preference was
found between the different approaches (2.30), and (2.33) to setting up the equation error
sequences.
Remark 2.12 Though this algebraic approach, following the light of [FSR03], provides
promising results, it is worth mentioning that, clearly, the linear identifier cannot incorporate a-priori knowledge on stability. More concretely, for the drive-train example, the
pole-pair related to the fast eigenfrequency is located very close to the stability margin, hence, in presence of noisy signals, this pole-pair might be found to shift beyond the
stability margin.
To resume, let us finally accent some characteristics of the proposed algebraic z- and
q-domain identification approach. As illustrated in the Figures 2.32.9, the idea to discarding those zeros of G (z) and G# (q) having negligible effect on the system dynamics
(regarding the interesting frequency domain), qualifies as attractive. Additionally, referring to the discussed example, the numerical condition of the linear identifier parametrized
respectively) is found to suffer with decreasing sampling times, which
in terms of (or ,
is not the case for the q-domain parametrization.
The problem of determining the system order n and to a-priori deciding whether or
not to discard certain zeros to find a suitable setting for 0 m
m is preferably tackled
by deriving a mathematical model based on physical considerations.
Finally, this chapter closes with examples on invoking the standard least-squares (LS)
identification method of Section 2.2.4 to the drive-train example, with the z- and q-domain
representation entitled as LS-z and LS-q for short. These following Examples 2.9,
2.10 represent the counterparts of the Examples 2.6, 2.7.
2. Identification. . .
24
2. Identification. . .
25
80
2
40
[rad/s]
u [V]
1
0
1
0
40
2
0.2 0.4 0.6 0.8
t [s]
1.2 1.4
120
30
1.2 1.4
q-ident. (detail)
20
10
Im
40
30
q-ident. (red)
20
10
0
10
20
30
40
600 400 200 0
Re
1
Im
10
20
200 400 600
30
10
z-ident. (red)
Re
z-ident. (detail)
0.5
Im
0
0.5
1
25 20 15 10 5
Re
0.5
Im
sim.
q-ident.
z-ident.
80
0
0.5
1
0.5
0.5
Re
1.5
2. Identification. . .
40
#
G (j) [dB]
q-ident.
z-ident.
nom.
40
80
120
0.1
10
100
1000
10000
q-ident.
z-ident.
nom.
90
180
270
360
450
540
0.1
10
100
= 0 tan (/0 ) [rad/s]
1000
10000
26
2. Identification. . .
27
80
2
40
[rad/s]
u [V]
1
0
1
0
40
2
0
30
20
1.2 1.4
120
30
q-ident. (red)
20
Im
1.2 1.4
q-ident. (detail)
10
10
20
20
30
600 400 200
0
Re
30
10
Re
1
z-ident. (red)
z-ident. (detail)
0.5
Im
0.5
Im
10
10
Im
sim.
q-ident.
z-ident.
80
0
0.5
1
25 20 15 10 5
Re
0
0.5
0.5
0
Re
0.5
2. Identification. . .
0
90
180
270
360
450
540
0.1
28
q-ident.
z-ident.
nom.
#
G (j) [dB]
40
0
40
80
120
160
200
240
0.1
10
100
1000
10000
q-ident.
z-ident.
nom.
10
100
= 0 tan (/0 ) [rad/s]
1000
10000
#
G (j) [dB]
40
0
40
80
120
160
200
240
0.1
0
90
180
270
360
450
540
0.1
10
100
1000
10000
q-ident.
z-ident.
nom.
10
100
= 0 tan (/0 ) [rad/s]
1000
10000
2. Identification. . .
29
60
2
40
[rad/s]
u [V]
1
0
1
20
0
20
2
0
30
20
1.2 1.4
60
20
1.2 1.4
q-ident. (detail)
10
Im
Im
30
q-ident. (red)
10
0
10
10
20
20
30
600 400 200
0
Re
30
10
Re
0.3
1
z-ident. (red)
0.2
0.5
0.1
Im
Im
meas.
q-ident.
z-ident.
40
0.1
0.5
1
25 20 15 10 5
Re
0.2
0
0.3
0.92
0.94
0.96
Re
0.98
2. Identification. . .
q-ident.
z-ident.
nom.
#
G (j) [dB]
40
0
40
80
120
160
200
240
0.1
10
100
1000
10000
q-ident.
z-ident.
nom.
90
180
270
360
450
540
0.1
10
100
= 0 tan (/0 ) [rad/s]
1000
10000
Figure 2.9: (Figure 2.8 contd; Ta = 10ms) Bode diagrams (measurement results).
30
2. Identification. . .
31
80
2
40
[rad/s]
u [V]
1
0
1
0
40
2
0
300
1.2 1.4
120
Im
Im
1.2 1.4
LS-q (detail)
10
100
10
200
20
300
-3e5
-2e5
-1e5
Re
0e0
1e5
30
10
Re
LS-z (detail)
0.5
Im
0
0.5
2
3
25 20 15 10 5
Re
LS-z (red)
1
Im
20
100
30
LS-q (red)
200
sim.
LS-q
LS-z
80
1
0.5
0.5
Re
1.5
2. Identification. . .
32
80
2
40
[rad/s]
u [V]
1
0
1
0
40
2
0
300
1.2 1.4
120
30
LS-q (red)
200
20
100
1.2 1.4
LS-q (detail)
100
10
200
20
300
-9e3
-6e3
-3e3
Re
0e0
3e3
30
10
Re
1
LS-z (red)
LS-z (detail)
0.5
Im
0.5
Im
10
Im
Im
sim.
LS-q
LS-z
80
0
0.5
1
25 20 15 10 5
Re
0
0.5
0.5
0
Re
0.5
2. Identification. . .
40
33
#
G (j) [dB]
LS-q
LS-z
nom.
40
80
120
0.1
0
90
180
270
360
450
540
630
720
810
0.1
10
100
1000
10000
LS-q
LS-z
nom.
10
100
= 0 tan (/0 ) [rad/s]
1000
10000
LS-q
LS-z
nom.
#
G (j) [dB]
40
80
120
160
0.1
0
90
180
270
360
450
540
0.1
10
100
1000
10000
LS-q
LS-z
nom.
10
100
= 0 tan (/0 ) [rad/s]
1000
10000
chapter
THREE
A Mathematical Model of a Rolling Mill
PSfrag replacements
ased on physical considerations, a mathematical model of the rolling mill schematically
B depicted in Figure 3.1 is to be evolved in the scope of this chapter, intended as the
basis for the controller design to be addressed in Chapter 4.
hydraulic
actuator
upper
work roll
entry
bridle rolls
en
vebr
Feco
upper
backup roll
exit
bridle rolls
ex
vxbr
Mebr,1
Mxbr,1
Mebr,2
Mxbr,2
Fxco
35
in the sense of the RayleighRitz method. The implicit algebraic equations of the roll gap
model together with the mill stand and bridle roll dynamics finally yield a DAE system
with replacements
index 1, with the systems motion constrained to the manifold given by the roll gap
PSfrag
model equations.
m1
Q
A1
A2
exit bridle rolls cg
roll gap
ex
Fh
xp
x1
m0
d1
d0
Fr
h0
Fr
h0
base-line
m1 v 1 = Fh cg (x1 l1 ) d1 v1 m1 g
m0 v 0 = Fr Fh d0 v0 m0 g,
(3.1)
where g denotes the constant of gravity, Fr the roll force, d0 and d1 coefficients of viscous
friction, and l1 the length of the unloaded mill spring cg . The piston position is xp =
x1 h0 l2 with some offset length l2 .
36
forward chamber
A1
A2
xp
return chamber
(3.2)
from the continuity equation. Using the definition of the isothermal bulk modulus E oil of
the oil,
oil
1
1
=
Eoil
oil p1 T =const.
the mass balance equation (3.2) can be rewritten as
p1 =
Eoil
(A1 vp + Q Qleak ) ,
V 0 + A 1 xp
vp = x p .
Eoil A1
(A1 vp + Q Qleak ) .
V 0 + A 1 xp
(3.3)
The dynamics of the servo valve are neglected, because they are very fast compared to the
dynamics of the hydraulic actuator. Thus, the displacement xv of the servo valve piston
is regarded as the control input. In order to describe the (quasi-static) valve behavior,
37
the cases xv 0 with the supply pressure pS being connected to the forward chamber,
and xv < 0 with the forward chamber connected to the tank (with pressure pT ) have to
be distinguished. With Kd denoting the valve coefficient, the behavior of the servo valve
is modelled as
xv 0
Q = K d xv p S p 1 ,
(3.4)
Q = K d xv p 1 p T ,
xv < 0.
Eoil A1
Eoil A1
.
A 1 vp =
Q = Eoil u
V 0 + A 1 xp
V 0 + A 1 xp
V 0 + A 1 xp
u.
A1
(3.6)
The controllers for the position control mode (PCM) as well as for the force control mode
(FCM) to be discussed subsequently are based on (3.6).
V0 + A1 xdp
V 0 + A 1 xp
(3.7)
38
and, thus,
Q = pcm (V0 + A1 xp ) ln
V0 + A1 xdp
V 0 + A 1 xp
(3.8)
is proposed in [KSN01], [Kug01], where xdp is the desired piston position and pcm serves
as a parameter to adjust the desired closed loop dynamics. From (3.3) and (3.8) the
closed-loop dynamics
!
!
V0 + A1 xdp
A
v
+
Q
1
p
leak
F h = Eoil A1 pcm ln
(3.9)
V 0 + A 1 xp
V 0 + A 1 xp
are obtained. By including the compensation of the servo valve function of (3.4), we
finally get the non-linear state feedback control law
u>0:
u0:
V 0 + A 1 xp
u
A 1 Kd p S p 1
V 0 + A 1 xp
xv =
u
A 1 Kd p 1 p T
xv =
(3.10)
with u from (3.7). The proof of the asymptotic stability for the parameter 0 < pcm <
max with a suitable max is based on the Popov criterion and can be found in [KSN01],
[Kug01], where also the more general case of a double-acting double-ended piston configuration is discussed.
(3.11)
with the parameter f cm is applied. Thus, with (3.6) the control law for the force control
mode reads
F d Fh
.
(3.12)
Q = f cm (V0 + A1 xp ) h
A1
From (3.3) and (3.12) the closed-loop dynamics
d
A
v
+
Q
F
F
1
p
leak
h
h
F h = Eoil A1 f cm
A1
V 0 + A 1 xp
(3.13)
for the force-controlled actuator are obtained. The compensation of the servo valve function proceeds as given above.
39
Fr
xx xx
vex
x
ex
hex
hen
Fr
r
lower work roll
Figure 3.4: Scheme of the roll gap (the indicated directions of the roll/strip contact load
{
, } are related to the strip).
As the dynamics of the processes taking place in the roll gap are considerably faster
than the dynamics of the mill stand and the hydraulic actuator, it is appropriate to follow
a quasi-static point of view.
Besides the forming of the strip, the normal and frictional stresses
, acting in
the roll/strip interface also invoke elastic deformations of the work rolls, see again the
Figures 3.4 and 3.5. In the case of cold rolling, the assumption of a circular roll gap
shape, though with a larger so-called equivalent radius, qualifies as appropriate, see e.g.,
[BF52]. However, for the case of temper and thin strip rolling, this approximation is no
longer valid. Thus, the elastic work roll deformations have to be accounted for in detail.
Manifold research results on the temper and thin strip rolling case have been reported
in the metal forming literature, see, e.g., [JOZ60], [FJ87], [FJMZ92], [DET94], [LS01],
commonly referred to as non-circular arc roll gap models. They all have in common
that a detailed description of the work roll deformations is incorporated in the model,
either based on the elastic half-space solution (see, e.g., [Joh85]) or on Jortners solution
[JOZ60] for the radial displacements of an elastic cylinder caused by a diametrically
applied constant normal load (to be briefly revisited in Remark 3.3). As these models
typically involve considerable computational effort, they are primarily used to calculate
40
h [mm]
0.996
0.994
0.992
0.990
0.988
0.986
1200
, , xx [N/mm2 ]
1000 elastic
800
600
400
compression
zone A
plastic reduction
zone B
(backward slip)
plastic reduction
zone C
(forward slip)
elastic
recovery
zone D
200
0
-200
-400
xx
2.5
Figure 3.5: A non-circular arc roll gap with the strip thickness h and the stresses
,
11
2
and xx . R = 0.22m, E = 2.1 10 N/m , = 0.3, hen = 1mm, hex = 0.99mm,
en = 70N/mm2 , ex = 70N/mm2 , yield stress kf = 750N/mm2 , and Coulomb friction
coeff. = 0.1.
Contrary to these approaches, the proposed non-circular arc roll gap model addresses
the work roll deformation problem in the sense of the RayleighRitz method. This method
amounts to defining a finite number of shape functions for the radial and circumferential
displacement fields u and v, which are to be associated with generalized coordinates q, in
order to approximate the exact solutions, see, e.g., [Zie92]. Thus, the elasto-static work
roll deformation problem is approximated via a finite-dimensional one. The motivation
for this approach emanates from a control point of view: The objective is to design a
mathematical model which contains the main physical effects and at the same time involves a manageable computational effort that makes it applicable for simulation and
41
control. Clearly, besides the number of shape functions, the accuracy of this approximation intrinsically depends on the choice of such functions. This key issue of defining a
suitable set of ansatz functions for the displacement fields of the roll is the focal point of
Section 3.4.3, closing with a comparison of the displacements obtained with the proposed
Ritz approach and the numerical solution to this elasto-static problem.
The stresses building up in the formed steel strip are calculated following the classical
stripe model of metal forming, see, e.g., [PP00] or the literature cited above. This model,
briefly revisited in Section 3.4.5, is essentially based on the assumption that plane sections of the strip remain plane throughout the roll gap. This yields the spatial ordinary
differential equations (ODEs) of the strip model for the elastic and the plastic regions of
deformation. The roll gap model introduced in Section 3.4.6 is finally given as a set of
implicit algebraic equations, whose evaluation involves the numerical integration of the
strip ODEs. The feasibility of the proposed approach is discussed on the basis of a comparison with numerical results obtained from a roll gap model based on Jortners solution
to the roll deformation problem. Conclusions finally close the discussions on the roll gap
model.
Besides the advantage of involving significantly reduced computational effort compared
to the roll gap models based on the elastic half-space concept or on Jortners solution,
respectively, the proposed approach offers additional flexibility to adjust to different rolling
conditions. With increasing indentations of the work rolls, it is advisable to refine the
Ritz ansatz by means of introducing additional degrees of freedom. Section 3.4.3 provides
a discussion on suitable possibilities for such a refinement. For rolling conditions near the
classical cold rolling case, however, only a few degrees of freedom will suffice to obtain
a reasonable accuracy. Thus, given a desired choice of the shape functions, the setup
and the processing of the Ritz ansatz, and in particular the code-generation as well, is
preferably left to a computer algebra program.
PSfrag replacements
42
r0
B
deformed
shape
2r0
r
en
R
strip
ex
Figure 3.6: The plane strain elasto-static work roll deformation problem (left) with the
accentuation of a roll gap vicinity G to be defined.
in mind that the focus of interest is at last laid on the roll deformations evolving in a
suitable roll gap domain, say G.
Due to the finite width of the strip (and the rolls), the rolling load is not exactly
constant along the axial direction of the roll entailing a deformation mode known as work
roll bending, and, thus, non-flat products, see Figure 3.6 (right). In order to reduce
this undesired bending of the rolls, additional hydraulic actuators (so-called bending
cylinders) are usually installed acting on the bearings of the rolls. As this thesis is
not concerned with the issue of work roll bending control, the presence of a bending
mode attenuation control is assumed. Thus, the assumption of a plane strain work roll
deformation problem in the scope of the roll gap modelling qualifies as a reasonable
approximation.
Of course, the load acting on the work rolls is not exactly point-symmetric with respect
to the roll center in the case of rolling mills, as the shape of the loading evolving in the
roll gap clearly differs from the work roll / backup roll contact stress. In particular,
this applies to the tangential load . The following argumentation essentially utilizes
St.Venants principle, which states that statically equivalent forces, when applied to a
small part of the body, render approximately the same stresses and deformations at a
sufficiently large distance from this loaded region, [Zie92].
Regarding the normal load
, the point-symmetry assumption is seen as appropriate in view of St.Venants principle: The normal forces being statically equivalent to
the actual rolling stress and work roll / backup roll contact stress, respectively, imply
static equilibrium, neglecting the body force due to gravity and the (slight) deflection
between the respective lines of action. The applicability of St.Venants principle to the
problem under consideration is to be revisited and illustrated in the following paragraphs.
43
1
( r + rr ) = 0,
r
1
2r
+ r r +
=0
r
r
1
1
2
(r ) + r + 2 ( ) (rr + ) = 0
r
r
2
(3.15)
(3.16)
E
E
1
1
E
E
= (1 + ) r =
r ,
G=
=
,
E
2G
2 (1 + )
2 (1 + )
rr =
r
(3.18)
with G denoting the shear modulus. The compatibility equation (3.16) stems from the
integrability requirement of (3.17).
Remark 3.1 An alternative formulation of the linear elasticity problem (3.15), (3.16) is
obtained by introducing the Airy stress function (r, ), see e.g., [Joh85],
1
1
1
2
2
= (r ) ,
r = r
rr = r + 2 ( ) ,
(3.19)
r
r
r
44
in accordance with the PDEs (3.15). Then, the elasticity problem amounts to solving for
a solution of the PDE
1
1
1
1
2
2
2
2
(r ) + r + 2 ( )
(r ) + r + 2 ( ) = 0,
(3.20)
r
r
r
r
which stems from the compatibility equation (3.16).
As a prerequisite for addressing the elasto-static problem (3.15)(3.18), with the pointsymmetric loading {
, }, and the balancing tangential load 0 applied at the inner boundary due to (3.14), cf. Figure 3.6 (left), let us first focus on two sub-problems which will
qualify to play a vital role for the setup of a suitable Ritz ansatz.
PSfrag replacements
P
0
0 =
PR
r02
F
problem J
2r0
problem M
Figure 3.7: Two sub-problems of Figure 3.6 (left) attached with particular meaning for
the proposed Ritz approach.
Let q 1 = F/E . The displacement fields uJ (q 1 , r, ), v J (q 1 , r, ) : R [0, R] S 1 R
of the problem depicted in Figure 3.7 (left) are given as
uJ =
q1 J
(r, ) ,
vJ =
q1 J
(r, )
(3.21)
45
with
r
J (r, ) = 1 cos + 23 R2 + r2 sin2 + (1 )
2 sin ,
R
J
2
2
(r, ) = 1 sin (1 ) 2 cos + 3 R r sin (2) .
(3.22)
(3.23)
rr =
1 (r R cos ) +
2 (r + R cos ) (
1 +
2 ) (R sin )
2R
1
2q
1
2
2
2
J
=
1 (r R cos ) +
2 (r + R cos ) (
1 +
2 ) (R sin ) +
2R
2q 1
Jr =
(1 + ) (
1 (r R cos )
2 (r + R cos )) R sin
and
J
rr
J
J
r
1
2q 1 E
2
2
1 (r R cos ) +
2 (r + R cos )
=
2R
1
1
2q E
=
(
1 +
2 ) (R sin )2
2R
1
2q E
=
(
1 (r R cos )
2 (r + R cos )) R sin ,
1 (r, ) =
R r cos
,
(R2 + r2 2Rr cos )2
2 (r, ) =
R + r cos
.
(R2 + r2 + 2Rr cos )2
In contrast to (3.21)(3.23), the solution to the problem of calculating the displacements evolving in the (hollow) cylinder loaded by tangentially acting forces P , see Figure 3.7 (right) is a more recent result and was given by [Mei01]. All quantities related to
this solution will be augmented with the superscript M. In order to enforce the static
equilibrium, the circumferential loading
P
R/ (r02 ) is applied at the inner boundary
M0 =
of the hollow cylinder. The solution u , v M of Meindl to be given below is an approximation of the exact solution to this problem, however, for r0 R, this approximation is
46
very accurate, see [Mei01] for a detailed investigation. With D = [r0 , R] S 1 , q 2 = P/E ,
the (approximate) solution uM (q 2 , r, ), v M (q 2 , r, ) : RD R to the problem depicted
in Figure 3.7 (right) is
uM =
q2 M
(r, ) ,
vM =
q2 M
(r, )
(3.24)
with
M = 1 sin (1 ) 2 cos 3 R2 r2 sin (2) ,
R
r (3.25)
M = 1 cos + 23 R2 + r2 sin2 (1 + ) + (1 ) 2 sin
.
r
2R
2q 2
(1 + )
32
42 R2 r sin3 (
3
4 ) r sin
2q 2
(1 + )
32
42 R2 r sin3 + (
3
4 ) r sin
=
2q 2
1
2
=
(1 + ) R (R (
1 +
2 ) (
3 +
4 )) sin + 2
2r
M
rr =
M
M
r
with
3 (r, ) =
1
,
R2 + r2 2Rr cos
4 (r, ) =
1
.
R2 + r2 + 2Rr cos
Remark 3.2 The coordinates q 1 and q 2 are to be attached with the meaning of generalized
coordinates of the elasticity problem of Figure 3.6 (left) outside the roll gap vicinity G in
the scope of the proposed Ritz ansatz.
1
Example 3.1 Let R = 0.22m, = 0.3 and
q 2 = 1.
Figures
3.9 and 3.10 illus
The
Jq =
M M
J
, respectively, evaluated
and u , v
trate the shape of the displacement fields u , v
at five equidistant slices (w.r.t. r) of the sub-domain [R/10, R] [/2, /2].
E en
(3.26)
Z ex
R
v (r, ) =
M (r, ) () J (r, )
() d.
E en
Example 3.2 Let R = 0.22m, E = 2.11011 N/m2 , = 0.3 and suppose that the loading
{
, } as given in Figure 3.5 is applied to the cylinder. Figure 3.11 illustrates the resulting
radial and circumferential displacements {u, v} due to (3.26) (solid lines), as well as the
displacements due to the equivalent forces (dashed) using (3.21) and (3.24).
47
Figure 3.11 also clearly illustrates the fact that the influence of the particular shape
of the distributed loading on the shape of the displacement fields rapidly decreases with
increasing distance from the loaded region (St.Venants principle). On the one hand, this
observation justifies the assumption of a point-symmetric loading with respect to the roll
center. On the other hand, it serves as the key motivation for using (3.21) and (3.24) as
ansatz functions for the work roll displacement fields outside a suitable roll gap vicinity
G to be specified in the subsequent section.
Remark 3.3 (Jortners influence functions) Let
() =
0 , [, ] + k, k Z,
zero elsewhere, and = 0. To calculate the radial displacements of the roll surface, first
notice that
J
2
(R, ) = 2 + cos ln tan
(1 ) sin sign (sin ) .
2
2
From (3.26) we have
0 R
u (R, ) =
E
J (R, ) d =
0 R
(, ) ,
E
, (3.27)
+ (1 ) cos () sign (sin ())
o (, ) = sin () ln tan2
2
2
=
= o (, ) (1 ) .
(R, ) d
=
(3.28)
In the metal forming literature, (3.27) and (3.28) are typically referred to as Jortners
influence functions. They are numerously applied to cope with the work roll deformation
problem, associated with the very common approximations v (R, ) 0 and = 0, see,
e.g., [JOZ60], [DET94].
48
(3.29)
max (|en | , |ex |) < 1 < /2, parametrized in terms of r1 ]r0 , R[ and 1 to be chosen
appropriately. Note that the domain [1 , 1 ] of G is not required to coincide with the
actual extent of the roll/strip contact arc [en , ex ].
Following the idea
above,
of the
propose to arrange the superposition
M we
Jsketched
1
M
J
, associated with the coordinates q and q 2 ,
and u , v
displacement fields u , v
respectively, as approximations for u and v on the domain F = B\G.
Remark 3.4 Due to the use of uJ , v J and uM , v M as ansatz functions on the
domain F, the boundary conditions on F\G are naturally fulfilled.
As a prerequisite for the design of the shape functions uG , v G to be defined on G,
which are required to be (at least) C 1 coupled to the functions uJ , uM and v J , v M on
the boundary F G, let us first introduce the decomposition
uG = uJ + uM + u,
v G = vJ + vM + v.
(3.30)
21
[ (r, 1 ) (r1 , 1 ) r (r1 , 1 ) (r r1 )]
cos
+ [r (r1 , ) r (r1 , 1 )] (r r1 )
1 2
(3.31)
(r1 , )
(r1 , )
g () (r, ) =
(r, 1 ) + r (r1 , )
r (r1 , 1 ) (r r1 ) +
(r1 , 1 )
(r1 , 1 )
1
(r1 , 1 )
[ (r, 1 ) r (r1 , 1 ) (r r1 )] sin
+
+
(r1 , 1 )
1
1
(3.32)
and
49
and
q2
q2
uM = g M ,
vM = f M
(3.34)
(, ) {(, J ) , (, M)} .
c1 = r (r1 , 1 ) ,
c4 = ( )1 (r1 , 1 ) ,
c2 = c (r1 , 1 ) c3 r1 ,
c5 = r (r1 , 1 ) ,
c6 = (r1 , 1 ) ,
c3 = cr (r1 , 1 ) ,
c7 = r (r1 , 1 ) ,
b0 () = (r1 , ) ,
b2 () = (r1 , ) ,
Then, we have
f ( ) =
a0
(r) + b0
() (a1
(r)
c3 r
c2 ) cos
+ (b1 () c1 ) (r r1 ) c0 (3.35)
c
and
g ( ) = c4 a2 (r) b2 () + (b3 () c4 c5 b2 ()) (r r1 ) +
+
(c4 c6 a2
(r) +
c8
(r r1 )
a3
c
(r)) sin
2
2
c
, (3.36)
50
Remark 3.6 (on the notation) In the following, variables of the work roll deformation
problem referring to the roll surface r = R are indicated by means of a superscript bar,
.
e.g., u (q, ) = u (q, R, ).
Remark 3.7 (Remark 3.5 contd; evaluation of f ( ), g ( ) at the roll surface). By
introducing the constants
C0 = a0 (R) c0 c1 C2 ,
C3 = c4 (a2 (R) C2 c5 ) ,
C1 = a1 (R) c3 R c2 ,
c
C4 = (c4 c6 a2 (R) + c8 C2 a3 (R)) ,
2
and C2 = R r1 , we have
f ( ) =
+ () +
()
cos
,
c
2
g ( ) = C3 b2 () + C2 b3 () + C4 sin
.
c
C0
Thus, uJ (q 1 , ) = uJ (q 1 , R, ) =
2
vM = q f (M ), cf. (3.33), (3.34).
b0
q1
f
C2 b1
C1
(J ) (), vJ =
q1
g
(J ), and uM =
(3.37)
q2
g
(M ),
In order to achieve the required accuracy of the solution obtained with the Ritz
approach
to further decompose the domain G into subdomains Gk =
k k it kis appropriate
k
ra , r b a , b ,
[
G=
Gk ,
Gk Gl = {} , k 6= l,
(3.38)
k
the set of generalized coordinates, collected to the nu and nv -dimensional vectors qu and
qv is obtained. To this end, compute bases of ker Ru , ker Rv , i.e.,
ker Ru = span {wu,1 , . . . , wu,nu } ,
(3.40)
(3.41)
with
nu = dim (cu ) rank (Ru ) ,
s=1
(3.42)
51
Finally, the vector q representing the generalized coordinates of the displacement fields
on B is
(3.43)
q T = q 1 q 2 quT qvT .
Thus, the linear elasticity problem depicted in Figure 3.6 (left) is approximated via a
finite-dimensional one of determining 2+nu +nv generalized coordinates q. To summarize
the composition of the Ritz ansatz {u , v }, we have
( J 1
u (q , r, ) + uM (q 2 , r, ) + uk (qu , r, ) ,
(r, ) Gk
u (q, r, ) =
(3.44)
uJ (q 1 , r, ) + uM (q 2 , r, ) ,
(r, ) J
for the radial displacement field on B, and, accordingly,
( J 1
v (q , r, ) + vM (q 2 , r, ) + vk (qv , r, )
v (q, r, ) =
v J (q 1 , r, ) + v M (q 2 , r, )
(r, ) Gk
(r, ) J
(3.45)
for the circumferential displacement field. Note that, by construction, the functions
{u , v } are linear with respect to the coordinates q.
The Elastic Potential of the Work Roll
Generally, in the 3-dimensional linear-elastic strain state the potential energy U per unit
volume, outlined in polar coordinates (r, , z), reads
E
1 2
2
2
2
U=
e 2 (rr + zz + zz rr ) + 2 r + z + rz , (3.46)
2 (1 + ) 1 2
with the first invariant e = rr + + zz of the strain tensor, see, e.g., [Par88]. In the
case of a plane strain problem, i.e., zz = rz = z = 0, we obtain
1 2
2
E
2
2
+ +
rr + 2r =
U =
2 (1 + ) 1 2 rr
1 2
E
2rr + 2 + 2 rr + 2 (1 ) 2r .
(3.47)
=
2
2 (1 )
Thus, the potential energy, also referred to as the elastic potential, stored in the deformed
body with domain B is given as
V =
U rdrd =
r0
U rdrd.
(3.48)
By virtue of the linearity of {u , v } w.r.t. q, the elastic potential V (q) can be rewritten
as
E
1
A > 0,
E =
(3.49)
V = E q T Aq,
2
2 (1 2 )
3. APSfrag
Rolling
Mill Model. . .
replacements
52
and i = /q i . In the following, we will focus on some issues regarding the computation
of the matrix A.
F
coords. q 1 , q 2
r0
Fa
en
Fb
ex
r1
Figure 3.8: On the decomposition of the domain B, see also Figure 3.6 (left).
G
Remark 3.8 (computation issues; domain F = B\G) Let Aij = AF
ij + Aij , i, j {1, 2},
with the superscripts {F, G} referring to the respective domains of integration, see also
1 2
2
Figure 3.8. Let F
kl (q , q , r, ) : R F R, k, l {r, }, denote the strain fields
obtained with (3.17), (3.21), (3.24) and (3.44), (3.45), and introduce the abbreviations
.
J
J
1 J
M
1 J
2 M
M
F
F
2 M
F
kl = kl + kl = q ekl + q ekl , ekl , ekl : F R, kl,i = i kl = i ekl + i ekl with the
n
Kronecker symbols m
. Then,
Z
F 2
F 2
F F
F 2
+
2
+
2
(1
rdrd =
AF
=
i
j
rr
rr
r
ij
F
Z
F
F
F
F
F
F
F
F
= 2
F
F
rr,i rr,j + ,i ,j +
rr,i ,j + rr,j ,i + 2 (1 ) r,i r,j rdrd,
F
hence, e.g.,
AF
11
AF
12
Notice that
= 0. Let Fa = [r0 , r1 ] [0, /2] and Fb = [r1 , R] [1 , /2], see FigFb
Fa
ure 3.8 again, then AF
ii = 2 Aii + Aii , i = 1, 2, due to symmetry reasons on F. The
integration over the domain Fa yields the result
a
=
AF
11
(1 2 ) (r12 r02 ) R2
(1 + 3 ) R4 2 (3 ) r02 r12
+
r r 2
0 1
2
2
4
4
2
(1 )
r0 + r 1
+ r0 + r1 + 2 R +
R
R
2
R + r02 R2 r12
2 2
ln
.
1
R2 r02 R2 + r12
(R2
2
r02 )
2
+ r12 )
r0 r1 4
(R2
53
a
Accordingly, a symbolic expression for AF
22 can be derived. However, symbolic results for
Fb
b
the terms AF
11 and A22 could not be established, thus, these terms are left to numeric
integration.
= r u
(qu , qv , r, ) = (
u + v) /r
2
2
with Gkl,i Gkl,j = eJkl i1 + eM
kl eJkl j1 + eM
kl , k, l {r, }, see (3.51).
kl i + i
kl j + j
Hence,
J
J 1
2
2
i, j {1, 2}
ekl j + eM
ekl i1 + eM
kl j
kl i
eJ 1 + eM 2 2
i = j {1, 2}
kl i
kl i
G
G
kl,i kl,j =
2
kl
i {1, 2} , j > 2
eJkl i1 + eM
kl i j
i kl j kl
i, j > 2
2
2
Accordingly, the expression Grr,i G,j = eJrr i1 + eM
rr eJ j1 + eM
is
rr i + i
j + j
processed.
G
Remark 3.10 Notice that AG12 = 0, thus A12 = AF
12 + A12 = 0, see also Remark 3.8.
54
2
,i ,j r =
p R [r, ]
pr R [r, ]
(3.52)
(3.53)
ex
en
(
i u + i v ) Rd 0
i v (r0 , ) r0 d,
J (r0 , ) d = 0.
M (r0 , ) d,
(3.54)
55
we end up with
Q1 = R
Q2 = R
Qu,i = R
Qv,j = R
ex
en
ex
en
Z ex
en
Z ex
en
1 uJ + 1 vJ d
2 uM + 2 vM K
qui ud,
i = 1, . . . , nu
qvj vd,
j = 1, . . . , nv .
(3.55)
Given the generalized forces QT = Q1 , Q2 , QTu , QTv , the deformed state of the roll reflected by the coordinates q is calculated by solving for the solution of the set of linear
equations
(q V )T = E Aq = Q.
(3.56)
Outlined in terms of the coordinates q = LT q, equation (3.56) reads L
q = Q/E .
Example 3.4 Let R, E, as given in Example 3.2, r0 = R/100, and adjust the domain
boundaries of G as r1 = 0.96R, 1 = 0.01/R, see also Figure 3.11. Let Gk = [r1 , R] Ik ,
Ik = ((2k 5) /3 + [0, 2/3]) 1 , k = 1, 2, 3, and define
!
2
2
3 X
2
X
R
u
i+2 j
1
c1,ij r
u1 =
1 +
r1
i=1 j=0
!
2
6
3 X
X
R
u
i+2 j
c2,ij r
u2 =
1
r1
i=1 j=0
!
2
3 X
2
2
X
R
u
i+2 j
u3 =
c3,ij r
1
1
r1
i=1 j=0
with r = (r r1 ) / (R r1 ) and = /1 . The polynomials vk (with the coefficients cvk,ij )
are chosen analogously with the same orders as for u k . Thus, dim (cu ) = dim (cv ) = 39,
rank (Ru ) = rank (Rv ) = 12, and, hence, nu = nv = 27. The Figures 3.14, 3.15 and 3.16
provide a comparison of the displacements obtained by means of this Ritz approximation
( red) and the solutions due to (3.26) ( black) applying the loading {
, } as given in
Figure 3.5.
Example 3.5 Given the Ritz ansatz as specified in Example 3.4 and the loading
due
Figure 3.5, = 0, the Figures 3.19 and 3.20 depict the displacement fields obtained via the
Ritz approximation, as well as the numerical solution obtained by invoking Femlab 3.1
[Mul]. The associated mesh of the finite element approximation of Femlab is illustrated
in Figure 3.17, and Figure 3.18 depicts the von Mises effective stress [Zie92]. Particularly,
Figure 3.18 provides a vivid impression of the extent of the roll gap domain compared to
the dimensions of the work roll.
56
A11 0 A1n
Q1
q1
2
A22
q Q2
,
E
=
.
.
..
..
qu Qu
qv
0
Ann
or, more compactly,
E
A11 A12
AT12 A22
q
qv
Q
0
q1
q = q 2 ,
qu
Q1
= Q2 .
Q
Qu
(3.57)
q =
Hence, in this particular case the problem reduces to a (2 + nu )-dimensional one, E A
1
T
with A = A11 A12 A22 A12 > 0.
Q,
57
radial displacement uJ (q 1 = 1)
0.5
0
0.5
1
1.5
2
2.5
3
2
1.5
0.5
0.5
1.5
1.5
0.5
0
[rad]
0.5
1.5
circumferential displacement v J (q 1 = 1)
0.4
0.3
0.2
0.1
0
0.1
0.2
0.3
0.4
2
J J
1
due to (3.21).
Figure
3.9:
(cf.
Example
3.1)
Let
q
=
1.
Shape
of
the
functions
u ,v
R
2 , 2 , r = 10
1 + i 49 , i = 0, . . . , 4. The displacements of the roll surface are
indicated with dashed lines.
58
radial displacement uM (q 2 = 1)
0.3
0.2
0.1
0
0.1
0.2
circumferential displacement v M (q 2 = 1)
0.3
2
1.5
0.5
0.5
1.5
1.5
4
3
2
1
0
1
2
3
4
5
2
1.5
0.5
0
[rad]
0.5
M M
2
due to
Figure 3.10: (cf. Example 3.1) Let
q
=
1.
Shape
of
the
functions
u ,v
R
(3.24). 2 , 2 , r = 10
1 + i 94 , i = 0, . . . , 4. The displacements of the roll surface
are indicated with dashed lines.
59
20
30
40
50
60
70
80
0.1 0.08 0.06 0.04 0.02
0.02
0.04
0.06
0.08
0.1
0
0.02
[rad]
0.04
0.06
0.08
0.1
4
3
2
1
0
1
2
3
4
5
0.1 0.08 0.06 0.04 0.02
Figure 3.11: (cf. Example 3.2) Displacement fields of the roll ( [21 , 21 ], r =
r1 + i (R r1 ) /4, i = 0, . . . , 4, 1 = 0.01/R, r1 = 0.96R) caused by the distributed load
{
, } due to Figure 3.5 (solid), and the equivalent forces (blue dashed).
60
uJ , uJ , (q 1 = 1)
1.5
2
2.5
3
3.5
4
4.5
-0.1
uJ (domain G)
uJ
-0.08 -0.06 -0.04 -0.02
0.02
0.04
0.06
0.08
0.1
0
0.02
[rad]
0.04
0.06
0.08
0.1
0.3
v J , vJ , (q 1 = 1)
0.2
0.1
0
0.1
0.2
0.3
-0.1
J J
Figure 3.12: (cf. Example 3.3) Let q 1 = 1. Shape
functions u , v (red) defined
J of Jthe
(blue dashed) for [21 , 21 ],
on G, 1 = 0.01/R, r1 = 0.96R. Additionally, u , v
r = r1 + i (R r1 ) /4, i = 0, . . . , 4.
0.4
61
uM (domain G)
uM
0.3
uM , uM , (q 2 = 1)
0.2
0.1
0
0.1
0.2
0.3
0.4
-0.1
0.02
0.04
0.06
0.08
0.1
0
0.02
[rad]
0.04
0.06
0.08
0.1
v M , vM , (q 2 = 1)
4.5
4
3.5
3
2.5
2
1.5
-0.1
Figure 3.13: (cf. Example 3.3) Let q 2 = 1. Shape
functions uM , vM (red) defined
Mof the
on G, 1 = 0.01/R, r1 = 0.96R. Additionally, u , v M (blue dashed) for [21 , 21 ],
r = r1 + i (R r1 ) /4, i = 0, . . . , 4.
62
20
25
30
35
40
45
50
55
60
-0.1
0.02
0.04
0.06
0.08
0.1
0
0.02
[rad]
0.04
0.06
0.08
0.1
4
3
2
1
0
1
2
3
4
-0.1
Figure 3.14: (cf. Example 3.4) Ritz approx. (red) vs. solution due to (3.26) (blue
dashed): Displacements ( [21 , 21 ], r = r1 + i (R r1 ) /4, i = 0, . . . , 4) caused by
the load {
, } due to Figure 3.5.
63
25
30
35
40
45
50
55
60
-0.05 -0.04 -0.03 -0.02 -0.01
0.01
0.02
0.03
0.04
0.05
0
0.01
[rad]
0.02
0.03
0.04
0.05
4
3
2
1
0
1
2
3
4
-0.05 -0.04 -0.03 -0.02 -0.01
Figure 3.15:
i = 0, . . . , 4.
64
25
30
35
40
45
50
55
60
0.211
0.212
0.213
0.214
0.215
0.216
0.217
0.218
0.219
0.22
0.212
0.213
0.214
0.215 0.216
r [m]
0.217
0.218
0.219
0.22
3
2
1
0
1
2
3
4
0.211
Figure 3.16: (cf. Example 3.4) Ritz approx. (red) vs. solution due to (3.26) (blue
dashed): Displacements (evaluated along radial slices of G: r [r1 , R], = i1 /4,
i = 0, . . . , 4) caused by the load {
, } due to Figure 3.5.
65
0.25
0.2
0.15
0.1
0.05
0
0.05
0.1
0.15
0.2
0.25
0.25
0.2
0.15
0.1
0.05
0.05
0.1
0.15
0.2
0.25
Figure 3.17: (cf. Example 3.5) The mesh of the finite element approximation via Femlab 3.1 [Mul].
Figure 3.18: (cf. Example 3.5) Illustration of the finite element solution: The von Mises
effective stress.
66
10
0
10
20
30
40
50
60
2
1.5
0.5
0.5
1.5
1.5
0.5
0
[rad]
0.5
1.5
8
6
4
2
0
2
4
6
8
2
Figure
3.19:
(cf.R Example
3.5) Ritz approx. (red) vs. FEM solution (blue dashed):
2 , 2 , r = 10
1 + i 94 , i = 0, . . . , 4.
67
25
30
35
40
45
50
55
60
0.05 0.04 0.03 0.02 0.01
0.01
0.02
0.03
0.04
0.05
0
0.01
[rad]
0.02
0.03
0.04
0.05
4
3
2
1
0
1
2
3
4
0.05 0.04 0.03 0.02 0.01
Figure 3.20: (cf. Example 3.5) Ritz approx. (red) vs. FEM solution (blue dashed):
[1 , 1 ], r = r1 + i (R r1 ) /4, i = 0, . . . , 4 (domain G).
68
PSfrag replacements
Assume we are given the generalized coordinates q reflecting the state of the deformed
roll in the sense of the preceding section. Then, with the surface displacements u, v of
the work roll, see Remark 3.6 on the notation, the strip thickness h reads
h (; , q, h0 ) = h0 + 2 (R (R + u (q, )) cos ( ) + v (q, ) sin ( )) ,
(3.58)
see Figure 3.21. Here, denotes the deflection of the polar coordinate system (r, ) of
the upper work roll with respect to the y-axis coinciding with the work rolls center-line,
and h0 is the so-called no-load gap, i.e., the roll gap in absence of a rolling load.
R
xx
xx
u (q, )
strip
centerline z
v (q, )
1
h (; , q, h0 )
2
surface of the
deformed roll
h0
2
x = (; , q)
Figure 3.21: On the shape of the roll/strip contact arc (right), and the notion of the
classical stripe model of metal forming (left).
As a prerequisite for the combination of the strip equations (to be discussed in the
subsequent section) with the roll deformation problem, the diffeomorphic change of the
spatial coordinates and x,
x = (; , q) = (R + u (q, )) sin ( ) + v (q, ) cos ( )
(3.59)
is arranged, see again Figure 3.21. Then, the strip thickness in terms of x, which is
= h 1 . Differentiation of the identity 1 = id,
indicated by an overset bar, is h
i.e., (1 (x; , q) ; , q) = x, with respect to x yields
x 1 (x; , q) = ( )1 1 (x; , q)
and, thus,
(x; , q, h0 ) = h 1 (x; , q)
x h
and
= 1
(x ) h
( )2
2
( )2
( ) h
h
2
1 .
(3.60)
(3.61)
69
i
1 (x; , q)
(3.62)
and, hence,
i
+ i h 1
( h)
1
i
i
h
2
2
x i h =
1 .
( ) h i h + i
( )
=
i h
(3.63)
0 + t
and a subsequent elastic deformation = elast (s, , q),
(r + u (q, r, )) sin ( ) + v (q, r, ) cos ( )
x
,
=
= h0
y
+ R (r + u (q, r, )) cos ( ) + v (q, r, ) sin ( )
2
i.e., = elast rot . Clearly, the material velocity of the point X is V (t, X) = dt (t, X).
However, for the purpose of the roll gap model, the Eulerian point of view is more
suitable, yielding the spatial velocity
v (t, ) = V 1 .
In particular, the velocities of surface points are of interest.
!
(R
+
u
(q,
))
sin
(
)
+
v
(q,
)
cos
(
)
V (t, 0 ) = dt
=
h0
+
R
(R
+
u
(q,
))
cos
(
)
+
v
(q,
)
sin
(
)
2
=0 +t
( u v) sin ( ) + (R + u + v) cos ( )
=
(R + u + v) sin ( ) + (
v u) cos ( ) =0 +t
Let x (R, R). Then, = 1 (x; , q), see (3.59), and 0 = t = 1 (x; , q) t.
Thus,
( u v) sin ( ) + (R + u + v) cos ( )
v (t, x) =
(3.64)
(R + u + v) sin ( ) + (
v u) cos ( ) =1 (x;,q)
70
(3.65)
are obtained. These equations are to be associated with a suitable friction model for the
roll/strip interface and a constitutive law of the strip.
Remark 3.12 In view of the geometry of the roll/strip contact arc typically the approximation yy
is encounterd in the rolling literature, cf., e.g., [JOZ60], [FJ87],
[FJMZ92], [DET94], [LS01]. In this case, (3.65) reads
1
d
+ 2
,
xx = (yy xx ) x h
dx
h
= yy .
(3.66)
Remark 3.13 (friction law for the roll/strip interface) An approach usually encountered
in the analysis of rolling is the assumption of the Coulomb friction law in the roll/strip
interface. We will also make use of this assumption. Thus, the friction law is given as
=
, > 0, with = +1 in the regions with backward slip, and = 1 otherwise,
see also Figure 3.5.
With the Coulomb friction law of Remark 3.13, the equilibrium conditions (3.65) read
!
!
d
1
yy
yy xx x h
+2
x h
1+
xx =
1
dx
h
h 1 12 x h
2 1 2 x h
(3.67)
1
1
yy .
= 1 x h
2
The subsequent paragraphs provide the strip equations for the elastic and the plastic
deformation zones, obtained by plugging the respective constitutive laws into (3.67).
71
yy = ES yy + 1 S
x h yy + ES yy x h+
dx
dx
h
2 1 21 x h
2S
yy
h
1 12 x h
1
1
= 1 x h
yy .
2
(3.68)
d
1
d
+ 2S yy ,
yy = ES yy + ( (1 S ) yy + ES yy ) x h
dx
dx
h
h
Furthermore, the approximation
d
2S
d
yy = ES yy + yy ,
dx
dx
h
is very frequently used in the literature, see e.g. [LS01].
= yy .
= yy
(3.69)
As the strip processed in the roll gap experiences considerable displacements, the
logarithmic strain measure (see, e.g., [Zie92]),
h (x)
(3.70)
yy (x) = ln
h0
(x)
=
ln
hen = hen exp
en
(3.71)
en
yy
ES
hen
ES
by means of (3.70). Analogously, for the elastic recovery zone D the equation
h (x)
D
yy (x) = ln
S ex
hex
ES
is obtained.
(3.72)
3. A Rolling Mill Model. . . 3.4.6. The Implicit Non-Circular Arc Roll Gap Model
72
Remark 3.15 (Remark 3.14 contd). Substitution of (3.71) or (3.72), respectively, into
(3.69) yields
d
1
+ 2 yy ,
yy = ES x h
= yy ,
(3.73)
S
dx
h
cf. again [LS01].
Plastic Reduction Zones
In the plastic reduction zones B and C the yield criterion of Tresca (see, e.g., [Zie92],
[PP00]) is applied, xx yy = kf , with the yield stress kf = const.
Remark 3.16 (Remark 3.12 contd.) Substitution of Trescas law into (3.66) and the
application of the Coulomb friction law yields the equation
1
d
+ 2
,
yy = kf x h
dx
h
= yy ,
(3.74)
= / ,
{xx , yy ,
, } ,
R.
(3.75)
yy = ES x h + 2S
yy ,
dx
h
yy ,
E
ES = S
(3.76)
and
d
1
kf
yy ,
yy = kf x h + 2
kf =
.
(3.77)
dx
= f i x,
dx
0i ,
i xi0 =
i {A, B, C, D}
(3.78)
3. A Rolling Mill Model. . . 3.4.6. The Implicit Non-Circular Arc Roll Gap Model
73
given in the previous Section 3.4.5. With the respective domain of integration indicated
by a superscript, these solutions are referred to as
A(
B (
C (
D(
x,
x,
x,
x,
, q, h0 , xen (, q, h0 ) , 0
, q, h0 ,
xAB ,
AB
, q, h0 , xCD (, q) , CD
, q, h0 ,
xex ,
0
),
),
),
).
(3.79)
0
0
1
C
=
(x
,
).
The
coordinate
(x
;
,
q)
of
the
neutral
point, i.e.,
and xC
,
CD CD
N
N
0
0
the boundary between B and C, is fixed by means of the condition
B (xN , , q, h0 , xAB , AB ) =
C (xN , , q, h0 , xCD (, q) , CD ) .
(3.80)
The implicit algebraic roll gap model is outlined in terms of the independent coordinates
T = q T xAB AB CD xex h0 .
(3.81)
The entry and exit strip thicknesses hen , hex , the backward and forward tensions en , ex ,
the yield stress kf , the Coulomb friction coefficient , the material parameters of the roll
(index R) and the strip (index S) and the parameter are collected to the parameter
vector ,
T = hen hex en ex kf ER R ES S .
(3.82)
In order to keep the following formulas short, the explicit notation of the dependence
on the parameters will be supressed whenever appropriate. The generalized forces
associated with the loading {
i , i }, i {A, B, C, D}, read
Q1 = R
+R
+R
+R
1 (xAB ,,q)
en (,q,h0 )
A 1 uJ + A 1 vJ d
1 (xAB ,,q)
CD (,q)
N (,q,h0 ,xAB ,AB ,CD )
1 (xex ,,q)
CD (,q)
B 1 uJ + B 1 vJ d
C 1 uJ + C 1 vJ d
D 1 uJ + D 1 vJ d
(3.83)
3. A Rolling Mill Model. . . 3.4.6. The Implicit Non-Circular Arc Roll Gap Model
74
1 (xex ,,q)
en (,q,h0 )
Qu,i = R
1 (xex ,,q)
en (,q,h0 )
Qv,j = R
(
) 2 uM + (
) 2 vM K
(
) qui ud ,
i = 1, . . . , nu
(
) qvj vd ,
j = 1, . . . , nv ,
d
(3.84)
1 (xex ,,q)
en (,q,h0 )
see (3.55).
As the central result of this section, we are now ready to state the implicit roll gap
model (, ) = 0, : Rdim() Rdim() Rdim() ,
Aq Q ()
(xex ) hex
h
h (xAB )
(1 S ) yy (xAB ) ES ln
+ S (en kf )
hen
(, ) =
A
yy
(xAB ) AB
(xCD )
(1 ) D (x ) E ln h
CD
S
yy
S
(xex ) + S (ex kf )
h
D
(xCD ) CD
yy
=0
(3.85)
The first equation represents the work roll deformation problem (see also Remark 3.17)
and the second equation is due to the demand that the deformed strip leaves the roll
gap with the desired exit thickness hex . The remaining four equations fix the transitions
between the elastic and plastic zones AB and CD.
Given the parameters , the solution of the non-circular arc roll gap model (3.85)
amounts to invoking the implicit function theorem. The function is C 1 w.r.t. , on
an open set S Rdim() Rdim() . Let (0 , 0 ) S, (0 , 0 ) = 0, (0 , 0 ) regular,
then there exist neighborhoods U Rdim() of 0 and V Rdim() of 0 such that for each
V the equation (, ) = 0 has a unique solution U. This solution is denoted as
= () with the C 1 function : V U.
The roll force Fr , the roll torque Mr and the forward and backward slip en = ven /vR ,
ex = vex /vR , vR = R R, cf. Fig. 3.4, are functions of and .
On the Calculation of the Jacobian of
Clearly, the calculation of the Jacobian of (3.85) at the point (, ) involves the knowledge
of the sensitivites of the interface stresses {
, } with respect to the parameters q, h 0 ,
and the initial values of the strip ODEs. From Remark 3.13, we have i =
i, i
i
i
{A, B}, and =
, i {C, D}, > 0. Let us arrange the abbreviation N () =
3. A Rolling Mill Model. . . 3.4.6. The Implicit Non-Circular Arc Roll Gap Model
75
Z
Z
N ()
1 (xAB ,,q)
CD (,q)
vi ) d+
f B j + SjB (i u +
f C j + SjC + T C j xCD (i u
vi ) d+
N ()
1 (xex ,,q)
SjD
(3.86)
(i u
vi ) d
j
A
B
+
(i u +
vi )
(xAB )
(xAB )
=1 (xAB ,,q)
+ 2 (j N ())
C vi = () +
N
C
D
vi ) = (,q)
+ (j CD (, q))
(i u
+
f j +
CD (,q)
CD
with
vi () = i v () i2 K.
(3.87)
Note that (3.62), (3.80) and the conditions
A (en (, q, h0 )) = 0,
D (xex ) = 0, see
(3.79), have been used. By differentiating the identities
(xen (, q, h0 ) ; , q, h0 ) = hen ,
h
(xCD (, q) ; , q, h0 ) = 0
x h
j h
(xen (, q, h0 ) ; , q, h0 ) ,
x h
j x h
j xCD (, q) =
2 (xCD (, q) ; , q, h0 ) .
(x ) h
j xen (, q, h0 ) =
and
j xen (, q, h0 ) j
j h
=
j en (, q, h0 ) =
h =en (,q,h0 )
=en (,q,h0 )
j xCD (, q) j
j CD (, q) =
=CD (,q)
are determined. Finally, differentiation of (3.80), i.e.,
B
C (N () ; , q) = 0,
w.r.t. q j yields
!
SjC SjB + T C j xCD (, q)
1
j N () =
.
j
fC fB
=N ()
3. A Rolling Mill Model. . . 3.4.6. The Implicit Non-Circular Arc Roll Gap Model
76
Let Sh0 = h0
. Then, taking the derivative of Q with respect to the no-load gap h0
yields
Z 1 (xAB ,,q)
h0 Q i = R
ShA0 + T A h0 xen (i u +
vi ) d+
en (,q,h0 )
+
+
N ()
1 (x
ShB0
AB ,,q)
1 (xex ,,q)
CD (,q)
with
CD (,q)
N ()
ShC0 (i u
vi ) d+
ShD0 (i u
vi ) d + 2 (h0 N ())
C vi =
h0 xen (, q, h0 ) = x h
and
1
h0 N () =
Accordingly, we have
Z N ()
xAB Qi = R
1 (xAB ,,q)
with
(i u +
vi ) d +
1
N ()
(xen (, q, h0 ) ; , q, h0 )
ShC0 ShB0
.
fC fB
=N ()
T B (i u +
vi ) d + 2 (xAB N ())
C vi =
+
A (xAB )
B (xAB ) ( )1 (i u +
vi )=1 (x
1
xAB N () =
AB ,,q)
N ()
TB
.
fC fB
=N ()
The sensitivities of Qi with respect to the initial values AB , CD for the strip ODEs of
the domains B and C read
!
Z N ()
B
C
AB Qi = R
X (i u +
vi ) d + 2 (AB N ())
vi = () ,
N
1 (xAB ,,q)
CD Qi = R
CD (,q)
N ()
with
1
AB N () =
X (i u
vi ) d + 2 (CD N ())
vi =
XB
,
fC fB
=N ()
1
CD N () =
N ()
XC
fC fB
=N ()
Finally, the derivative of Qi with respect to the roll gap exit point xex is
Z 1 (xex ,,q)
T D (i u
vi ) d.
xex Qi = R
CD (,q)
The calculation of the sensitivity functions Sj , Sh0 , T and X amounts to solving for
the solution of the sensitivity ODEs (cf., e.g., [Kha96]) associated to (3.78). See also the
Appendix for a geometric point of view on this issue.
3. A Rolling Mill Model. . . 3.4.6. The Implicit Non-Circular Arc Roll Gap Model
77
12S
ES
12S
ES
1+
1+
1S
ES
1S
ES
see [PP00], with the second identity referring to the plane strain case under consideration.
By virtue of the mass balance law hv = const applied to the elastic compression zone A,
the roll gap entry speed ven is related to the strip velocity at xAB by
ven =
(xAB ) (xAB )
h
v (xAB ) .
hen (xen )
At the point xAB , the yield criterion of Tresca, cf. Section 3.4.5, requires xx (xAB ) =
kf + AB , therefore, we obtain
1 + (1 S ) en /ES
(xAB )
=
,
(xen )
1 + (1 S ) (kf + 2AB ) /ES
see also (3.79). In the plastic reduction zones the deformation is assumed to be isochoric,
an assumption well verified by experiments on plastic forming, see, e.g., [PP00], [Zie92].
(xN ) vR /h
(xAB ). By combining these equations,
Hence, for zone B, we have v (xAB ) = h
the entry slip en follows as
en =
(xN )
h
1 + (1 S ) en /ES
.
hen 1 + (1 S ) (kf + 2AB ) /ES
(3.88)
vex
1 + E S ex
(xCD ) (xCD )
(xN ) (xCD )
(xN )
h
h
h
S
vR ,
=
v (xCD ) =
vR =
hex (xex )
hex (xex )
hex 1 + 1S (kf + 2CD )
E
S
(xN )
h
1 + (1 S ) ex /ES
.
hex 1 + (1 S ) (kf + 2CD ) /ES
(3.89)
1S
ex
ES
1S
en
ES
1+
1+
1S
ES
1S
ES
(kf + 2AB )
(kf + 2CD )
1.
(3.90)
3. A Rolling Mill Model. . . 3.4.6. The Implicit Non-Circular Arc Roll Gap Model
78
79
shows a comparison between the results, both evolving from the same Ritz ansatz, abtained
with the approximations due to Example 3.6 (x R, effect of on the shape of the
contact arc neglected) on the one hand, and the detailed model of the contact arc due to
(3.58), (3.59) on the other hand.
1.000
80
0.998
[mm]
h
0.996
0.994
0.992
0.990
0.988
0.986
-3
-2.5
-2
-1.5
-1
-0.5
0.5
1.5
2.5
1200
1000
elastic
compression
zone A
, , xx [N/mm2 ]
800
B
plastic reduction
zone B
(backward slip)
600
400
200
elastic
recovery
zone D
C
plastic reduction
zone C
(forward slip)
0
xx
-200
-400
-3
-2.5
-2
-1.5 -1 -0.5
0
0.5
1
1.5
distance from the roll centerline, x [mm]
2.5
1.000
81
0.998
[mm]
h
0.996
0.994
0.992
0.990
0.988
0.986
-3
-2.5
-2
-3
-2.5
-2
-1.5
-1
-0.5
0.5
1.5
2.5
-1.5 -1 -0.5
0
0.5
1
1.5
distance from the roll centerline, x [mm]
2.5
1200
, , xx [N/mm2 ]
1000
800
600
400
200
0
-200
Figure 3.23: (cf. Example 3.7) Comparison of the results obtained by using an approximate model of the contact arc (red) (x R, effect of on the shape of the contact arc
neglected; cf. Example 3.6 and Figure 3.22), and, displayed in black color, by using the
detailed description of the contact arc due to (3.58) and (3.59).
82
The bridle rolls in the entry and exit section of the mill, cf. Figures 3.1 and 3.24, are
intended to assigning, by application of the bridle torques, a certain difference of the strip
forces between the periphery (i.e., adjacent winders or other strip processing lines) and
the roll gap.
v1
slip arc
v > v1
Fext,1
v1 = 1 R
Fext,1
stick arc
v = v1 = 1 R
q
1
1 , M1
1 , M1
stick arc
v = v1
v1 = 1 R
pulley 1
pulley 1
D
v2
stick arc
v = v2 = 2 R
R
F
v2
pulley 2
v1
pulley 2
slip arc
v > v2
2 2 , M2
2 , M2
q
Fext,2
v2 = 2 R
stick arc
v = v2 = 2 R
slip arc
v < v1
slip arc
v < v2
Fext,2
v2
Figure 3.24: Configuration of the entry and exit bridle device (including the locations of
the slip arcs; operating range: Fext,1 > F > Fext,2 ).
The notation introduced in Figure 3.24 is arranged with the intention to commonly
treat the main characteristics of both bridle devices. When there is danger of confusion,
the variables are being attached with the subscript br {ebr, xbr} to distinguish between
the entry and the exit bridle rolls. Variables carrying the index 1 are associated with the
top pulley of the bridle device, the index 2 identifies variables associated with the bottom
pulley. The operating range of both bridle devices is defined as Fext,1 > F > Fext,2 . All
pulleys have the same radius R, and the vertical distance between the pulley centers is
denoted by D. The following assumptions are arranged:
The strip processed by the bridle rolls is assumed to be massless, with negligible
bending rigidity, and to exhibit a constant cross-sectional area A, i.e., A ebr = Bhnom
en
83
nom
and Axbr = Bhnom
(hnom
ex , with hen
ex ) denoting the nominal strip entry (exit) thickness and B the strip width;
in the slip arcs (to be identified), Coulomb friction is assumed, with the friction
coefficient br = const.;
the operating range of the bridle system is chosen such that the slip arcs do not
exceed the maximum wrap angle of the pulleys. This means that the strip does not
slide on the pulleys. Note that the term sliding must not be confused with the
notion of creep occurring in the slip arcs.
The maximum wrap angle max not to be exceeded by the slip arc is given as max =
arccos (2R/D). The moments of inertia, i.e., 1 for the top roll and 2 for the
bottom system, summarize the moments of inertia of the pulley, the associated drivetrain and the motor, related to the pulley shaft. The torques M1 and M2 are regarded
as control inputs. The friction torques occurring in the drive-train of the pulleys are
modelled as di i , i = 1, 2, di = const.
slip arc
v > R
Fa
, M
stick arc
v = vb = R
R
vb
Fb
Figure 3.25: On the discussion of the slip arc and the creep (Fa > Fb ).
Referring to Figure 3.25, let Fa and Fb < Fa be the forces at the tight and the slack
side of the system. The associated strains in the strip segments given by Hookes law are
a = Fa , b = Fb , with = 1/ (ES A) and ES denoting Youngs modulus of the strip.
With ref denoting the mass density of the unloaded strip, the mass density of the
84
1 + b
1 + Fb
va =
va < v a .
1 + a
1 + Fa
(3.91)
Hence, a material point of the strip experiences an acceleration while passing from the
slack side to the tight side. Clearly, the torque to be applied in order to establish the
difference between Fa and Fb is of negative sign (referring to the sign convention of Figure 3.25). Therefore, the Coulomb frictional traction q acting on the strip is oriented as
indicated. The calculation of the strip force F () evolving in the creep domain is done
via Eulers equation of cable friction (see, e.g., [Zie92]), dF/d + F = 0, which admits
the solution F () = F (0 ) exp ( ( 0 )). Hence, we have Fb = Fa exp (), and,
thus, = 1 ln (Fa /Fb ) for the extent of the slip arc. As the frictional traction q must
oppose the direction of the slip, the strip is moving faster than the pulley in the slip
arc. Therefore, the slip arc is located where the strip runs off the pulley, v a v > vb ,
whereas the sticking region occurring in the entry zone is characterized by v = vb = R,
see Figure 3.25.
The results obtained so far directly apply to the case of the entry bridle device. Concerning the exit bridle rolls, the subdivision of the pulley/strip contact arcs into sticking
and creep regions (see Figure 3.24) is obtained in a straight-forward manner. The extents
of the slip arcs are given as 1 = 1 ln (Fext,1 /F ) and 2 = 1 ln (F/Fext,2 ).
1 =
(3.92)
Remark 3.19 The steady-state creep ratio [Joh85] of the entry bridle device is =
12 /1 = (F Fext,2 ) / (1 + F ) > 0. Hence, in steady state, pulley 1 moves (slightly)
faster than pulley 2.
85
(3.93)
Again, as in the case of the entry bridle system, in steady-state the pulley 1 (tight side)
is moving faster than pulley 2 (slack side).
A Simplified Model
Motivated as a step towards the control design, a simplified model of the bridle roll
dynamics is obtained by taking the limit E , which amounts to neglecting the creep
effect, and, thus, the occurrence of slip arcs. Hence, v2 ' v2 for the entry bridle device,
and v1 ' v1 for the exit device. Then, we simply have
1 1 = M1 + (Fext,1 F ) R d1 1
2 2 = M2 + (F Fext,2 ) R d2 2
F = cR (1 2 ) ,
(3.94)
(3.95)
86
with containing the remaining parameters, see (3.82). For notational convenience we will
0 , 0 , 0 S, S Rdim( )
drop the dependence on whenever suitable. Given a point
0 , 0 , 0 = 0 and
0 , 0 , 0 nonsingular.
being C 1 on S,
Rdim() Rdim( ) ,
Then, by virtue of the implicit function theorem, there exist neighborhoods U Rdim()
0 and V Rdim() Rdim() of 0 , 0 such that for each , V the equation
of
, = 0 has a unique solution
U . This solution is denoted as
=
,
, with
the C 1 function
: V U.
The dynamics of the mill stand of Figure 3.2 (left)
the roll gap model (3.95) is to be
h0 . The motion of the main mill drive
combined with is given in (3.1), with Fr = Fr ,
(MMD) is simply modelled as
h 0 d R R ,
R R = MR Mr ,
(3.96)
with R as the moment of inertia of the rolls and the main mill drive transmission line,
MR as the torque supplied by the drive, and dR as a viscous friction coefficient.
The strip elements connecting the bridle devices to the roll gap, cf. Figure 3.1, are
modelled as mass-less linear-elastic springs with constant (nominal) cross sections Bh nom
en
and Bhnom
ex , respectively. Thus, by obeying the slip equations of the roll gap, the strip
tensions in the entry and exit section of the mill are given as
ES
R R ebr,1 Rebr ,
en ,
Len
ES
R R .
=
xbr,1 Rxbr ex ,
Lex
en =
ex
(3.97)
Thus, finally, the differential-algebraic dynamics of the mill are specified by (3.9) (or (3.13)
respectively, if the HGC is operating in force control mode), the bridle roll dynamics (3.92),
(3.93), and (3.95)(3.97). Clearly, the system has index 1, with the motion confined to
the manifold
roll gap model (3.95). The control inputs are given
rendered by the implicit
as uT = xdp , MR , Mebr,i , Mxbr,i , i = 1, 2. The coupling of the system to the periphery,
i.e., adjacent winders or other strip processing lines, is taken into account by means of
the external forces Feco = Febr,ext,2 and Fxco = Fxbr,ext,2 (not explicitely stated in the
equations) acting on the lower pulleys of the bridle roll devices, cf. Figure 3.1. To further
revealthe structure
of the mill dynamics, it is convenient to consider the decomposition
xT = wT , T , wT = [x1 , v1 , v0 , Fh , R , ebr,i , Febr , xbr,i , Fxbr ], i = 1, 2,
, w, u
w = f w ,
, w
(3.98)
= f ,
,
0=
87
roll gap model (3.95) at each time step of the numerical integration scheme. An alternative approach for the numerical integration of this system is given as follows. From
+
=
= 0 we have
=
1
f .
P > 0, to obtain
equation with the term
P ,
The simulation model reads
=
1
f P
.
, w, u
,
, w
,
(3.99)
1
=
f P
0 , 0 = 0. The attrac 0 , 0 chosen to meet the constraint
with the initial value
tiveness of the constraint manifold is easily seen by application of Lyapunovs theory. Let
T ,
then V =
T
=
TP
0.
V = 12
w = f w
= f
Example 3.8 (the parameters of the simulation model) The parameters used for the simulations of the rolling mill are given as follows. Mill stand model, cf. Figure 3.2 (left):
m1 = 105 kg, m0 = 23 103 kg, cg = 5.4 109 N/m, d1 = 107 N s/m, d0 = 1.14 107 N s/m;
hydraulic actuator, cf. Figure 3.3: A1 = 0.6567m2 , A2 = 0.2082m2 , Eoil = 1.6109 N/m2 ,
p2 = 80 105 N/m2 , V0 = 0.02m3 , pcm = 50. Main mill drive: R = 4723kgm2 ,
dR = 100N ms/rad; Bridle roll systems (identical configuration of the entry and the exit
bridle device): Rbr = 0.3m, br,1 = 235kgm2 , br,2 = 100kgm2 , dbr,1 = 10N ms/rad,
dbr,2 = 4.4N ms/rad; Distances between the entry and exit bridle roll centers and the
roll gap: Len = 2.2m, Lex = 2.4m; Strip parameters: ES = 2.1 1011 N/m2 , S =
7.7 103 kg/m3 ; strip width: BS = 0.5m. The roll gap parameters are due to Example 3.6.
Remark 3.21 (on the computational effort) Due to the stiffness of the mill dynamics (3.99) using the parameters as given in Example 3.8, which will be illustrated in the
subsequent Section 3.6.1, the numerical integration of (3.99) requires considerable computation time, typically about 250s per second real-time. This timing result is due to
using Matlab/Simulink [MS] with the ODE solver ode45 (Dormand-Prince; settings:
max. time step: 103 , tolerance bounds set to auto). See also Remark 3.18 addressing
the computational cost per evaluation of the roll gap model (including the calculation of
the Jacobian, and the solution of the linear system related to one iteration step of the
Newton method).
88
= [Fr , Mr , en , ex ].
+
=
= 0 we have
From d
=
1
with
en +1 en +3
h0
S
S
n
= 2 + nu + nv , and ek , eik = ki , as the elements of the canonical basis of Rn +4 . Then,
the sensitivities of y w.r.t. , are obtained as
1
+ y.
d y = ( y)
(3.100)
Example 3.9 (root loci, calculated at different mill speeds) Consider the parameters as
given in Example 3.8 and calculate the eigenvalues of the linearized mill dynamics, with the
mill speed vR taking values from 0.5m/s up to 20.5m/s. The Figures 3.26 and 3.27 show
the corresponding root loci, with the eigenvalues associated with v R = 0.5m/s indicated in
red color.
The eigenfrequencies related to the mill stand oszillations are clearly beyond the dynamic range of the hydraulic gap control under consideration, and, thus, they will be
discarded in view of the controller design. The motions of the bridle roll devices might
be decomposed into a high-frequency pulley/pulley counter-phase oszillation mode and a
lower-frequency oszillation of both pulleys (thought of as notionally glued together). As
the higher-frequency mode of oszillation is again beyond the dynamic range of the actuators, the pulley/pulley ozillations will also be discarded from the detailed bridle device
model. Figure 3.27 depicts a detailed view of Figure 3.26, containing the dynamics being
relevant for the control design.
89
2500
mill stand
2000
1500
1000
pulley/pulley osz.
Im
500
mill stand
bridle osz.
0
500
1000
1500
2000
2500
-300
-250
-200
-150
Re
-100
-50
Figure 3.26: (cf. Example 3.9) Root loci of the linearized mill dynamics for different mill
speeds (vR = 0.5m/s . . . 20.5m/s).
150
100
bridle osz.
Im
50
HGC
MMD
50
100
150
60
50
40
30
Re
20
10
Figure 3.27: Detail of Figure 3.26 (and comparison with the reduced-order model
(black/magenta) to be introduced in the subsequent chapter).
chapter
FOUR
Flatness-based Rolling Mill Control
rom the control point of view, particularly the handling of the acceleration and de-
F celeration procedures of the mill, which are intended to be kept as short as possible
for efficiency reasons by simultaneously meeting the tight tolerances for the elongation
coefficient, is known as a key objective in the industrial practice. Additionally, the control
system has to handle operator commands concerning the desired strip tensions en and
ex (see Figure 3.1), again by keeping the elongation coefficient within the given guarantee
values. Usually, control engineers face these problems by adjusting additional feed-forward
schemes during the implementation and startup phase of the mill which frequently points
out to be a very time-consuming task. Thus, besides the disturbance rejection performance of the control system, particular attention is to be paid to the trajectory planning
and the trajectory tracking problem.
91
see (3.1). From (4.1) and xp = x1 h0 l2, by arranging a change of coordinates from
h0 , z ,
Fh to z, see (3.5), we obtain x1 = ,
!
!
h0 + m0 g
z Fr ,
V0
x1 = h 0 + l 2 +
exp
1 ,
(4.2)
A1
Eoil A1
and
h0 cg ,
h0 , z l1 (m0 + m1 ) g = 0.
Fr ,
(4.3)
By assembling the dynamics (3.9) of the HGC-PCM, outlined in terms of z, and the
quasi-static mill stand model, we find
A1 d
z = pcm z + Eoil A1 ln 1 +
x + Fr , h0 m0 g
V0 p
"
#
(4.4)
h0 , en , ex ,
,
0=
.
h0 cg ,
h0 , z l1 (m0 + m1 ) g
Fr ,
This consideration entails the formulation of a third setting of the implicit algebraic roll
gap model besides (3.85) and (3.95), namely, the composition of the algebraic equations
of (4.4),
)
= 0,
,
(,
T = z en ex ,
(4.5)
T , h0 from (3.81).
with T =
Remark 4.1 (HGC-FCM) With the HGC operating in force control mode, cf. Section 3.3.2, we obtain
z = f cm Eoil Fhd Fr () + m0 g
(4.6)
)
,
0 = (,
for the assembly of the actuator dynamics and the quasi-static mill stand model.
of (4.5) meeting the conditions of the implicit function theorem, there
Again, with
),
defined on a neighborhood of (0 , 0 , 0 ), with
exists a C 1 function
, =
(,
0 , 0 , 0 ) = 0 and (0 , 0 , 0 ) regular. Thus, the DAE system of the reduced order
(
dynamics has again index 1.
From d
= ( )(
) +
= 0, we have =
Remark 4.2 (linearization of )
1 .
Due to the specific structure of (4.5), namely that
does not explicitly
( )
yields
depend on z and the roll force Fr is a function of only, the partial derivative
the particular simple representation
ex
0
en
=
.
zFr ()+m0 g
V0
exp
0
0
cg Eoil
Eoil A1
A2
1
92
(4.7)
nom
with the entry strip force Fen = en BS hnom
en and the spring coefficient cen = ES BS hen /Len .
Remark 4.3 Notice that the general case (with 2 /1 = , d2 /d1 = 6= ) can be traced
back to the dynamics of the symmetric configuration given above by setting 1 = ,
0
2 = , d1 = d, d2 = d, and introducing M1 = M1 + (/ 1) d1 with the new
0
input M1 .
Now, apply the input transformation
M1
0
R
u
=R
+
,
M2
Fext,2
R
w
(4.8)
see also [FSK03], [FGSK03]. The reason for this particular choice will become apparent
in a moment. For discarding the dynamics related to the pulley/pulley-oscillations, take
the limit c , i.e., lim0 F = R (1 2 ), = 1/c, rendered from the singular
perturbation point of view. Hence, we have 1 = 2 = , and, thus, the reduced-order
model representing the motion of (4.7) confined to the slow manifold reads
=
1
((1 + ) u + RFen (1 + ) d)
(1 + )
(4.9)
F en = cen (ven R)
and
1
Fen + w.
(4.10)
Especially, the choice = 1/ (1 + ) for the input transformation (4.8) yields a particular
appealing representation of the reduced order dynamics,
F =
(4.11)
93
nom
ebr = Bhnom
w, u), = f (, ,
w, u), 0 = (,
)
,
Let us consider the decomposition w = f w (, ,
T
of (4.12) with w = [R , ebr , xbr ], and as defined in (4.5). Then, the corresponding
explicit representation suitable for the simulation purpose, see also Remark 3.20, reads
w, u)
w = f w (, ,
w, u)
= f (, ,
1 ( )f
P
,
= ( )
with P > 0 introduced to asymptotically stabilize the systems motion with respect to
94
Proposition 4.1 The dynamics (4.12) are exact input/state linearizable via static state
feedback, and, thus, differentially flat. A (practically suitable) representative of the flat
output is given e.g. as
xbr Rxbr
T
(4.14)
y =
1 R en ex ,
ebr Rebr
entailing the relative degree r = (1, 1, 2, 2).
Proof. The static state feedback equivalence is easily seen e.g. by invoking the implicit
see Section 4.1.1. The
which (locally) gives =
function theorem on ,
(z, en , ex , ),
verification of the relative degree r = (1, 1, 2, 2) for (4.14) is straight-forward.
Remark 4.4 The function y 1 of (4.14) is the so-called elongation coefficient, which is
the crucial control variable in temper rolling, y 2 is the angular velocity of the main mill
drive, and y 3 , y 4 are given as the strip tensions. The representative (4.14) of the flat
output is particularly useful as it coincides with the set of control variables of this typcial
mill configuration.
Given a flat output, the theory of flatness based control, cf., e.g., [FLMR95], [Rot97],
provides a systematic way for the control system synthesis. The application of the nonlinear static state feedback entailing the exact linearization, y 1 = v 1 , y 2 = v 2 , y3 = v 3 ,
y4 = v 4 , with the new input v, yields a linear and time-invariant dynamics of the tracking
error e = y yd . Here, yd denotes the desired trajectory of the flat output. These tracking
error dynamics are easily shaped by means of linear techniques.
Example 4.2 Givem the parameters of Example 3.8, Figure 4.1 finally shows simulation
results of the flatness based tracking control. The quantization of the position signal of
xp is chosen as 2m, and the amplitude of the noise added to the strip tension signals is
0.4N/mm2 .
1.12
20
1.1
15
1.08
1.06
1.04
1.02
0
xp
xdp
-5
0.98
-10
75
140
120
100
80
60
40
20
0
-20
MR MRs (kNm)
65
60
55
50
45
85
30
en
75
70
65
60
ex
en , ex (N/mm2 )
80
95
10
70
R (rad/s)
xp xsp (m)
20
10
0
-10
-20
Mxbr
-30
Mebr
0 0.5 1 1.5 2 2.5 3 3.5 4
t (s)
Figure 4.1: Simulation results of the flatness-based rolling mill control, applied to the
detailed dynamics of Section 3.6. The figures on the left side represent the trajectories y,
yd of the flat output, and the figures on the right depict the associated control inputs. For
a better view, selected quantities are given as deviations from the selected steady state of
the mill (indicated by the superscript s).
chapter
FIVE
Non-linear Vehicle Dynamics Control A Flatness-based
Approach
his chapter proposes a novel approach for the non-linear vehicle dynamics control
which is essentially based on the observation that the dynamics of the planar holonomic
bicycle model depicted in Figure 5.1 are differentially flat. The contact between
PSfrag replacements
the tires and the road is modelled in terms of contact forces, which implies that the tires
are enabled to slip and slide on the road. The steering angle and the longitudinal tire
forces are regarded as control inputs.
v
Flv
v
y
C
Flh
Fsv (v, , r, )
x
r, Md
Y
Fsh (v, , r)
97
The bicycle model as introduced in [RS40] emerges from the four-wheel vehicle by
gluing together the front and the rear wheels to a single (mass-less) front and rear wheel,
respectively, located at the longitudinal axis of the car. This planar model, known as
a well-established basis for the design of vehicle dynamics control systems, see, e.g.,
[ABO99], [B
un98], [Rit04], is capable of rendering the longitudinal, lateral and yaw dynamics of the vehicle. The pitch and roll dynamics of a vehicle are clearly not involved
in the scope of this model.
This chapter is organized as follows: After a brief revisit of the bicycle dynamics in
Section 5.1, the discussion is concerned with the system analysis in paragraph 5.2 yielding
the flatness property, and, in particular, a physically relevant representative for the flat
output of the bicycle model as the main result. This representative of the flat output does
not depend on the particular choice of the vehicles actuation, i.e., it holds for the rear-,
front- and all-wheel driven car equivalently. Up to a certain family of functions which
have to be excluded, the system analysis does not refer to particular representatives for
the functions describing the lateral tire forces.
The task of (real-time) trajectory planning amounts to mapping the inputs supplied
by the driver, i.e., the current position of the throttle/brake pedal and the angle of the
steering wheel, to suitable trajectories for the flat output. The physical meaning of the
flat output is regarded to facilitate this objective significantly. Informations gathered by
scanning the environment, e.g., regarding the conditions of the road, or the position of a
detected obstacle, might also be incorporated for the real-time shaping of the trajectories.
The on-line trajectory shaping task (to also give a pleasant feeling for the driver) is an
open problem and not addressed in this thesis. Finally, to illustrate the proposed control
approach, simulation results given in Section 5.5 conclude this chapter.
98
Thus, we have
X = v cos ( + ) ,
Y = v sin ( + ) ,
= r,
(5.2)
with r denoting the yaw rate. Notice that in the scope of vehicle dynamics control (to
assist the driver in emergency situations), only the case v > 0 is considered.
In order to describe the tire/road contact via suitable mathematical models for the
lateral tire forces Fsv and Fsh , the literature offers a wide variety of modelling assumptions.
A very common and well-established assumption, cf. e.g. [BPL89], is that the lateral tire
forces, which allow for changes of the vehicles orientation, are regarded as functions of
the respective side-slip angles of the wheels. The side-slip angles h and v represent the
angles between the velocity vector at the rear/front wheel and the associated tire plane,
see Figure 5.1,
v sin lh r
h (v, , r) = arctan
,
v cos
(5.3)
v sin + lv r
.
v (v, , r, ) = arctan
v cos
However, in the scope of the system analysis to be given in Section 5.2, we do not
a-priori select a particular representative for the functions of the lateral tire forces, i.e.,
we will regard the functions Fsh (v, , r), Fsv (v, , r, ) as arbitrary smooth functions1 .
The dynamics of the bicycle model read
x = f (
x, u) ,
xT = v r
(5.4)
1
m (Fsv (v, , r, ) sin ( ) + Flv cos ( ) + Fsh (v, , r) sin + Flh cos )
F
(v,
,
r,
)
cos
(
F
sin
(
)
+
F
(v,
,
r)
cos
F
sin
sv
lv
sh
lh
.
f =
r +
mv
1
(lv (Fsv (v, , r, ) cos + Flv sin ) lh Fsh (v, , r) + Md )
J
(5.5)
The reason for selecting the representation (5.4), (5.5) is twofold: First, this particular
choice for the coordinates x implies that the functions f do not depend on the vehicles orientation (and the global position X, Y as well). Thus, these coordinates are
often referred to as vehicle coordinates. Secondly, as the vehicle dynamics control under
consideration is not concerned with position control of the car, but with control objectives depending on x only, the subsystem (5.2) has been dropped from the entire bicycle
dynamics.
1
This C assumption is arranged to avoid mathematical subtleties. Thus, it is a sufficient condition
in the scope of the following discussions.
99
Flv = (1 ) Fl ,
(5.6)
with [0, 1] referred to as the transmission ratio. This ratio is regarded either as
a function of the time, (t) : R 7 [0, 1], or as a function of the vehicles state x. The
first point of view amounts to regarding as an input given by the driver, the second
viewpoint, however, is understood as an action of a vehicle control system.
The central result for this type of vehicle is given in the following proposition.
Proposition 5.1 Let the lateral tire forces Fsh (v, , r), Fsv (v, , r, ) be arbitrary (smooth)
functions up to the requirements
J
(mlv )2 2
v sin cos + h v cos , v sin
r
(5.7)
Fsh (v, , r) 6=
J (lv + lh )
mlv
and
Fsv 6=
v (v, , r) (1 ) ( sin + (1 ) ) Fl
,
1 (1 cos )
(5.8)
J
r.
mlv
(5.9)
(5.10)
Proof. The exact linearizability is straight-forwardly verified with (5.9), (5.10). For
the (local) coordinates transformation z = (
x),
1 1
v cos
c (
x)
z
,
z 2 = c2 (
v sin Jr/ (mlv )
x) =
(5.11)
1
2
3
Lf c (
x)
z
(lv + lh ) (mlv ) Fsh (v, , r) vr cos
6= 0
m2 lv2
lv + l h
(5.12)
100
has to be fulfilled due to the implicit function theorem. The requirement (5.12) amounts
to excluding the family (5.7) of functions for the lateral rear tire force Fsh . Remark 5.5
provides a comment on the condition (5.7) to show that it does not imply a practically
relevant restriction on the choice of Fsh .
The static state feedback entailing the linear and time-invariant dynamics outlined in
the coordinates z, with the new input w T = [w1 , w2 ], is obtained (locally) as the solution
of
Lf c1 (
x, u) = w 1 ,
L2f c2 (
x, u) = w 2
(5.13)
(5.14)
as well as (5.12) is met. Thus, additionally to (5.7), we have to impose the requirement
(5.8). Again, Remark 5.5 provides a comment on this restriction.
101
Remark 5.6 Note that the transmission ratio , regarded as (t) or (v, , r), does not
explicitely appear in y 2 = z 3 = Lf c2 , see (5.11).
The point is interesting also from another point of view. To this end, let us calculate
the (x, y)-decomposition of the acceleration of a point Q located at the x-axis at the
distance x = from the center of mass C. Let R = SO (2), i.e., the group of rotations in
the plane,
cos sin
R () =
,
0 < 2
sin cos
denote the mapping from the inertial frame (X, Y ) to the vehicle coordinates (x, y). Variables augmented with the subscripts x, y are related to the vehicle coordinate system,
whereas the subscripts X, Y indicate the respective inertial frame representation. The
velocity of Q is given as
Q
Q
Q
vx
vx
v cos
vX
1
=
,
,
=R
Q
Q
vy
vyQ
v sin + r
vY
and the acceleration reads
Q Q
Q
d vX
vyQ
v x
ax
.
+
=
=R
v yQ
aQ
vxQ
dt vYQ
y
Thus, we have
1
(Flv cos + Flh Fsv (v, , r, ) sin ) r 2
(5.15)
m
for the longitudinal component, and
1
mlh
mlv
Q
ay =
1
Fsh (v, , r) + 1 +
(Flv sin + Fsv () cos ) + Md
m
J
J
J
(5.16)
for the lateral acceleration, with () = (v, , r, ).
aQ
x =
Fsh (v, , r) Md ,
1+
ay =
m
lv
lv
(5.17)
does not explicitly depend on the contact forces of the front wheel, i.e, F sv (v, , r, ) and
Flv .
Remark 5.8 For = J/ (mlh ), the lateral acceleration does not explicitly depend on the
lateral force Fsh (v, , r) of the rear wheel,
1
lv
1
P
ay =
(5.18)
1+
(Flv sin + Fsv (v, , r, ) cos ) + Md .
m
lh
lh
This point P, with coordinates (x, y) = (J/ (mlh ) , 0), occurs in the analysis of Ackermann
leading to the robustly decoupling control law of DLR [ABO99], [B
un98].
102
5.2.2 The Front- and the Rear-Wheel Driven Bicycle as Special Cases
The flatness property of the rear-wheel driven vehicle ( = 1) and the front-wheel driven
vehicle ( = 0) follow as special cases from Proposition 5.1. Again, let us emphasize
that the representative (5.9), (5.10) of the flat output does not depend on the vehicles
actuation. For the rear-wheel driven car, the restriction (5.14) reads Fsv tan Fsv 6= 0,
and, thus, the requirement
Fsv (v, , r, ) 6= v (v, , r) cos1 ,
(5.19)
is to be imposed, see (5.8). In the case of the front-wheel driven bicycle, the restriction
Fsv (v, , r, ) 6= v (v, , r) Fl
(5.20)
L[ ,f ] c2 = 0,
(5.21)
(5.22)
and
( Fsv (v, , r, ) sin ( ) + Fsv (v, , r, ) cos ( )) v c2 +
1
+ ( Fsv (v, , r, ) cos ( ) + Fsv (v, , r, ) sin ( )) c2 +
v
mlv
+
( Fsv (v, , r, ) cos + Fsv (v, , r, ) sin ) r c2 = 0. (5.23)
J
A solution to (5.22), (5.23) can be constructed by first observing that (5.22) admits the
solution
c2 (v, , r) = (v sin , r) = (, r)
103
with an arbitrary (smooth) function . This solution is substituted into the PDE (5.23)
yielding
A ( Fsv (v, , r, ) cos Fsv (v, , r, ) sin ) (v sin , r) +
+ ( Fsv (v, , r, ) cos Fsv (v, , r, ) sin ) r (v sin , r) = 0
and, thus,
A (, r) + r (, r) = 0,
(5.24)
(5.25)
104
(5.26)
with denoting the covariant derivative given by the Levi-Civit`a connection. The following result is formulated in intrinsic geometric terms (with T y : T Q R n1 denoting
the tangent map associated to y : U Q Rn1 ).
Proposition 5.2 [RM98] Let q be a point on Q, and U be an open neighborhood of q,
and suppose y : U Q Rn1 is a submersion. If (y 1 , . . . , y n1 ) is a configuration flat
output, then
g (ker T y, D) = 0.
(5.27)
Conversely, if g (ker T y, D) = 0 and if the regularity condition (stated below) holds at q,
then (y 1 , . . . , y n1 ) is a configuration flat output at q. The regularity condition is that the
ratios of functions in the following set should not all be the same at q,
(g (, Z)) : g (, Z) , (g (Z1 Z2 , )) : g (Z1 Z2 , ) , ( (V )) : (V ) ,
(5.28)
where Z, Z1 , Z2 are arbitrary vector fields around q that are y-related to some vector
field on Rn1 and , are non-vanishing vector fields such that P = span {} and
ker T y = span {}.
The proof can be found in [RM98]. The application of this proposition to check on
configuration flatness and to compute all possible configuration flat outputs might proceed
as follows. Given the control forces, and, hence, P, the distribution D is calculated via
105
1
(j glk + k glj l gjk ) g li ,
2
i, j, k = 1, . . . , n,
(5.29)
by linearity of .
If D = T Q, then the system is clearly not configuration flat: Given any y, one can
always find a vector field Z D = T Q such that g (ker T y, Z) 6= 0. If dim D < n,
then chose a one-dimensional distribution (spanned by a vector field ) that is orthogonal
to D, i.e., g (, Z) = 0, Z D. This one-dimensional distribution, with its annihilator
spanned by the 1-forms dy 1 , . . . , dy n1 , induces a foliation of the configuration manifold Q.
The functions (y 1 , . . . , y n1 ) qualify as a configuration flat output provided the regularity
conditions (5.28) hold.
In local coordinates, the regularity conditions take a particular appealing representation when carrying out the calculations in the coordinates q associated with the foliation,
i.e., q = (y 1 , . . . , y n1 , z). Let k = / qk . Then, (5.28) amounts to calculating the
following ratios of functions:
z (g (, )) : g (, ) ,
g k ,
: g k , ,
z ( (V )) : (V ) .
i = 1, . . . , n 1
i, k = 1, . . . , n 1
(5.30)
In case of identical equality of all these ratios (in a local neighborhood), the functions
y 1 , . . . , y n1 are differentially dependent, thus violating the requirements imposed on a
flat output. Below, we will follow this procedure for investigating configuration flatness
for three examples.
Example 5.1 (The planar rigid body actuated by two body-fixed forces). Consider the
rigid body (mass m, moment of inertia J) depicted in Figure 5.2 (left), which is moving
in the (vertical) plane under the action of (gravity and) two control forces {F 1 , F2 } being
body-fixed regarding the points of application {S1 , S2 } and the lines of action as well. The
lines of action intersect in point A, and a is the length of CA, with C as the center of
mass. The (clearly non-flat, as not controllable) case S1 = S2 = C is excluded. The
system is known to be configuration flat, see, e.g., [MRS95], [MMR03], with the PVTOL
being a well-known representative. The reason for discussing this example, in particular
via application of Proposition 5.2, is to reveal relations to the planar holonomic bicycle
1 A
2
F1
S1
y
F2
S2
F1
A
106
F2
C
(config. flat output)
F3
Fsh (v, , r)
Y
X
Figure 5.2: (cf. Example 5.1 and Remark 5.12) The planar rigid body actuated by two
body-fixed forces (left), and a link to the bicycle model (right).
dynamics, and the distinctions as well. Clearly, the system evolves on the configuration
manifold Q = SE (2) = R2 S 1 . In coordinates q = (X, Y, ), with (X, Y ) localizing C,
we have T = m2 (X 2 + Y 2 ) + 21 J 2 , and constant metric [gij ] = diag (m, m, J). The range
of control forces span the codistriubtion P,
P = span{ cos ( 1 ) dX + sin ( 1 ) dY a sin (1 ) d,
cos ( + 2 ) dX + sin ( + 2 ) dY + a sin (2 ) d },
whose annihilator P is spanned by
= a sin () X a cos () Y + .
Notice that does not depend on the angles 1 , 2 , hence, the same will hold for the flat
output. The distribution D of (5.26), (5.29),
D = span {, a cos () X + a sin () Y } ,
has rank n 1 = 2. The distribution orthogonal to D is spanned by the vector field
=
J
J
sin () X +
cos () Y + ,
ma
ma
J
cos () ,
ma
y2 = Y
J
sin () .
ma
(5.31)
Notice that up to now the potential V has not been involved. Indeed, V enters via the regularity conditions only, thus, interestingly, configuration flatness is primarily determined
by the metric g (and the way of which the control forces are applied, i.e., P, of course).
107
m
0
Ja sin (z)
h
i
k
l
J
0
m
cos (z) .
[
gij ] = 1 j 1 gkl =
a
J
J
J
a sin (z) a cos (z) J 1 + ma
2
The first set of regularity conditions of (5.30) reads
g , 1
: g , 1 = cos (z) : sin (z) ,
z
y
y
g , 2
: g , 2 = sin (z) : cos (z) ,
z
y
y
thus implying the qualification of (5.31), indicated as point in Figure 5.2 (left), as
configuration flat output by virtue of Proposition 5.2, regardless of the potential V .
Remark 5.9 (Huygens center of oscillation) The distinguished point , revealed as the
flat output of the planar rigid body, is historically known as center of oscillation from
planar pendulum dynamics (Christiaan Huygens, 16291695). Think of the planar rigid
body of Figure 5.2 (left) fixed at pivot A and left oscillating as a pendulum under the
action of gravity g, then the equation of motion reads
J + ma2 = mga cos () .
On the other hand, consider a (mathematical) pendulum, with pivot A, and the mass m
located at point , i.e.,
2
J
J
= mg a +
m a+
cos () ,
ma
ma
(5.32)
are not exact input/state-linearizable by static state feedback, but require for dynamic
(quasi-static) feedback.
108
Remark 5.11 (Differential parametrization of the rigid body dynamics (5.32) via the flat
output (5.31); Example 5.1 contd) From
1
J 2
1
y =
F1 cos (1 ) + F2 cos (2 ) + cos () ,
m
a
J 2
1
2
F1 cos (1 ) + F2 cos (2 ) + sin () ,
y =
m
a
we find
tan = y2 /
y1.
(5.33)
Hence, with the identity cos2 = (1 + tan2 ) and (5.31), we obtain the parametrization
of the center of mass C,
X = y1 +
J
y1
q
,
ma
2
2
1
2
(
y ) + (
y )
Y = y2 +
y2
J
q
.
ma
2
2
1
2
(
y ) + (
y )
(5.34)
J
m
y1
2
cos () a
J
F1 sin (1 ) F2 sin (2 ) =
a
F1 cos (1 ) + F2 cos (2 ) =
as
F1
F2
"
sin(2 )
sin(1 +2 )
sin(1 )
sin(1 +2 )
cos(2 )
sin(1 +2 )
cos(1 )
sin(
1 +2 )
#"
q
#
2
2
J 2
1
2
m (
y ) + (
y ) a
,
J
a
(5.35)
with and given in terms of y and its derivatives up to order 4, see (5.33).
Now, we will turn towards the bicycle dynamics by revealing certain links to the
planar rigid body example, and the distinctive features as well. In order to draw these
links even more illustrative, the nomenclature F1 , F2 , F3 for the tire forces (instead of Flv ,
Fsv and Flh ) has been introduced in Figure 5.2 (right). First, we will state the following
observation, though evolving from an unrealistic point of view (Fsh 0).
Remark 5.12 In a very first view, let Fsh 0 and regard the forces Fi , i = 1, 2, 3, as
control inputs. Obviously, F3 P, with P as the codistribution spanned by the front
tire forces F1 , F2 . (This is also reflected in (5.6) by using this dependency of the control
forces to distribute the aggregate engine/braking torque as longitudinal tire forces to the
front and the rear tire via the transmission ratio , i.e., F3 = Fl and F1 = (1 ) Fl ,
[0, 1]). Hence, with {F1 , F2 } as control inputs, this (simplified) bicycle model is
immediately identified as a special case of the rigid body example of Figure 5.2 (left).
109
Clearly, regarding the lateral tire forces Fsh (v, , r) and F2 = Fsv (v, , r, ) as functions of v, , r (and ) does not directly fit into the framework of configuration flatness of
Lagrangian systems as discussed above, as the force Fsh (v, , r) is not required to evolve
from a potential, and F2 does not depend on the configuration only.
The steering angle and the aggregate longitudinal tire force Fl (see Remark 5.12)
represent the control inputs, u = (, Fl ). Again, let Fsh 0. Then, (5.33) and (5.34) hold
for the differential parametrization of and X, Y . To solve for the contol inputs (, F l ),
consider the representation of the resulting tire force in the so-called vehicle coordinate
frame (x, y), i.e., F = Fx dx + Fy dy,
F = (Fl ( + (1 ) cos ) Fsv (v, , r, ) sin ) dx +
+ (Fl (1 ) sin + Fsv (v, , r, ) cos ) dy.
The independence of Fx and Fy requires for
F x F l F x
= Fsv sin (1 (1 cos )) ( Fsv + (1 ) Fl ) 6= 0.
det
F y F l F y
Notice that this condition appears as (5.14).
The check of qualification of (5.31) as the configuration flat output for the (exclusively
relevant) case of non-vanishing lateral rear tire force Fsh (v, , r) is somewhat more tedious
(as (5.33) and (5.34) do not hold in this case) and amounts to invoking the implicit function
theorem to validate the (local) existence of the differential parametrization of the bicycle
dynamics.
Remark 5.13 The vehicle dynamics control approach to be discussed in Section 5.5 rests
upon the qualification of the velocity components of as the flat output (of the dynamics
(5.4), (5.5) obtained by dropping the subsystem (5.2) related to the global position and
orientation, which is not involved in the control objective), cf. Proposition 5.1. However,
also the flat ouput (5.31) of (5.2), (5.4), (5.5), i.e., the global position of , might be
very valuable for applications: One might think, e.g., of the control objective of tracking a
vehicle (equipped with a global-positioning system) on a pre-specified path on a test track.
Fsv (v ) = cv v + Ov (2)
(5.36)
110
evolved at the points h = v = 0. Without loss of generality we will assume that the
tire stiffnesses ch and cv are related by cv = ch with 0 < < . With = 1/ch , the
substitution of (5.36) into the bicycle dynamics (5.2), (5.4)-(5.6) yields
v cos ( + )
X
Y =
(5.37)
v sin ( + )
and
(v (, ) sin ( ) + h () sin + b1 (, )) /m
= (v (, ) cos ( ) + h () cos + b2 (, )) / (mv)
(lv v (, ) cos lh h () + b3 (, )) /J
r
(5.38)
= 1, . . . , n
= 1, . . . , n
(5.39a)
(5.39b)
with the C 1 functions f , g and the so-called (small) perturbation parameter . For
= 0, the differential equations (5.39b) degenerate to the set of algebraic equations
0 = g (t, , , 0) .
(5.40)
Following [Kha96], the model (5.39) is referred to as the singular perturbation standard
form iff (5.40) exhibits k 1 isolated real roots
= hi (t, ) ,
i = 1, . . . , k
for each (t, ) [0, t1 ] D, with D Rn denoting some open connected set. Then, the
i-th root of (5.40) is associated with a welldefined reduced order model
= f (t, , hi (t, ) , 0)
evolving on a n dimensional manifold.
(5.41)
111
Clearly, the bicycle model (5.37), (5.38) exhibits the structure (5.39), or, more precisely,
= (X, Y, )
= f (, )
= g (, , Fl , )
= (v, , r) ,
0
v (v, , r, ) sin ( ) + h (v, , r) sin
0
lv v (v, , r, ) cos lh h (v, , r)
(5.42)
with the slip angles v , h due to (5.3), are functionally dependent, which is seen by
calculating the associated 1forms dg = g d yielding dg 1 dg 2 dg 3 = 0.
In order to reveal the relations between v, , r and imposed by (5.42), note that the
equations
g 1 (v, , r, ) cos g 2 (v, , r, ) sin
= v (v, , r, ) sin = 0
h (v, , r) sin
= 0
imply
v (v, , r, ) = arctan
and
h (v, , r) = arctan
v sin + lv r
v cos
v sin lh r
v cos
=0
= 0,
(5.43)
(5.44)
which are clearly attached with an obvious physical interpretation, namely the nonholonomic constraint of rolling contact without slipping.
With the velocity v at the center of mass serving as a free coordinate, the slip angle
and the yaw rate r are obtained from (5.43) and (5.44) as
lh
lh
v
= arctan
tan ,
r = sin arctan
tan
,
(5.45)
l
lh
l
arranging the abbreviation l = lh + lv . Thus, by substituting (5.45) into (5.37) the bicycle
model takes the form
lh
l
h
Y =
v
sin
arctan
tan
(5.46)
,
lh
v
sin arctan
tan
lh
l
PSfrag replacements
5. Non-linear Vehicle Dynamics Control
M
112
r =
v
y
vvlat = 0
x
C
r =
vh
(Xh , Yh )
Y
vhlat = 0
Figure 5.3: The nonholonomic bicycle, with M denoting the instantaneous center of
rotation.
representing the kinematic behavior of the system, with the variables v and regarded
as inputs. A more convenient representation of the system (5.46) is attained by applying
the diffeomorphic coordinates transformation
= () ,
namely,
Xh
X lh cos
Yh = Y lh sin ,
"
vh
uu = u (u) ,
v cos arctan
lh
tan
l
(5.47)
see also Figure 5.3. The application of the transformation (5.47) finally reveals the familiar
representation of the kinematic vehicle,
vh
X h = vh cos ,
Y h = vh sin ,
=
tan
(5.48)
l
with the inputs u = (vh , ). The following corollary summarizes some results on the
structure of the kinematic vehicle well-known in the literature.
Corollary 5.3 The kinematic vehicle (5.48) is differentially flat with the position y =
(Xh , Yh ) qualifying as a flat output. However, it is not static state feedback equivalent to
a linear timeinvariant system.
Proof. The property of differential flatness is easily shown by explicitly calculating
the system variables (Xh , Yh , , vh , ) in terms of the flat output y = (Xh , Yh ) and its time
derivatives, yielding
2
y
1
2
Xh = y ,
Yh = y ,
= arctan
y 1
and
vh =
(y 1 )2
(y 2 )2 ,
y 1 y2 y1 y 2
= arctan l
3/2
(y 1 )2 + (y 2 )2
113
The second statement of the corollary concerning the exact input/state linearizability via
static state feedback is obtained by the observation that the distribution
= span {[vh , f ] , [ , f ]} =
vh
tan
2
1 + tan
,
= span cos X + sin Y +
l
l
related to the existence of an output function c (X, Y, ) with relative degree 2 is not
involutive.
Another connection between the bicycle model of Figure 5.1 admitting side-slip of the
wheels and the non-holonomic vehicle is attained by invoking the Lagrange formalism with
the non-holonomic constraints of ideal rolling contact incorporated by means of Lagrange
multipliers.
To this end, let the systems motion be modelled on the bundle (E, pr1 , B) with local
coordinates (t) for the base manifold B R and (t, q ),
(q ) = (X, Y, , ) ,
(5.49)
S S 1 , and pr1
for the total configuration manifold E = B M, M = R2 S 1 S,
denoting the natural projection pr1 : E B. The velocities q are denoted as v = q ,
synonymously.
Due to the requirements imposed on the wheels not to slip in the lateral direction,
i.e., vhlat = vvlat = 0, see Figure 5.3, the systems motion is restricted by means of the
constraints
q1 sin q 3 + q2 cos q 3 lh q3
g (q, q)
=
= 0.
(5.50)
q1 sin (q 3 + q 4 ) + q2 cos (q 3 + q 4 ) + lv q3 cos q 4
However, more geometrical insight into the problem is achieved by considering the constraints (5.50) via the representation
1 = sin q 3 dq 1 + cos q 3 dq 2 lh dq 3
2 = sin q 3 + q 4 dq 1 + cos q 3 + q 4 dq 2 + lv cos q 4 dq 3
114
i.e, there does not exist a regular submanifold of the configuration manifold E the systems
motion is restricted to.
By introducing the first jet manifold J 1 (E) with local coordinates (t, q , q1 ), the kinetic
energy of the vehicle reads
2 1
2
1
1 1 2
2 2
T (q1 ) = m q1 + q1
(5.51)
+ J q13 + Js q14 ,
2
2
2
with Js denoting the moment of inertia of the front wheel, related to the Zaxis. The equations of motion of the constrained system are obtained by invoking the EulerLagrange
formalism
d1 1 L L = Q ,
= 1, . . . , 4
L (q, q,
) = L0 (q, q)
+ i gi (q, q)
= T (q)
+ i gi (q, q)
,
(5.52)
which have to be met additionally to the restrictions (5.50). The variables Q denote the
generalized forces, related to the coordinates q , i.e.,
Q1 = Flv cos (q 3 + q 4 ) + Flh cos q 3 ,
Q2 = Flv sin (q 3 + q 4 ) + Flh sin q 3 ,
Q3 = Flv lv sin q 4 + Md
Q 4 = Ms
(5.54)
with the steering torque Ms . With (5.50) and (5.51), the equations (5.53) finally give
mq21 11 sin q 3 21 sin (q 3 + q 4 ) 1 cos q 3 q13 2 cos (q 3 + q 4 ) (q13 + q14 )
= Q1
= Q2
= Q4 .
(5.55)
To further reveal the structure of the system (5.50), (5.55) it is illustrative to focus on
the state space representation
v
H11 H12 (q)
W (q, v, )
Q
q = v ,
+
=
(5.56)
0
0
g (q, v)
0
11 0
H
0 Js
11 = diag (m, m, J) ,
H
(5.57)
W
U
115
sin q 3 sin (q 3 + q 4 )
12 (q) = cos q 3
cos (q 3 + q 4 ) ,
H12 (q) =
,
H
(5.58)
lh
lv cos q 4
T
, U composed as
the function W T (q, v, ) = W
1 cos q 3 v 3 2 cos (q 3 + q 4 ) (v 3 + v 4 )
= 1 sin q 3 v 3 2 sin (q 3 + q 4 ) (v 3 + v 4 )
1 (v 1 cos q 3 + v 2 sin q 3 ) + 2 (v 1 cos (q 3 + q 4 ) + v 2 sin (q 3 + q 4 ) lv v 4 sin q 4 )
= 2 v 1 cos q 3 + q 4 + v 2 sin q 3 + q 4 + lv v 3 sin q 4
(5.59)
and
12 (q)
H
0
1 gi
C (q, v) = [ gi v ] =
21 0
H
21 (q) =
H
sin q 3
cos q 3
lh
3
4
3
4
sin (q + q ) cos (q + q ) lv cos q 4
v 1 v 3 cos q 3 v 2 v 3 sin q 3
(v 3 + v 4 ) (v 1 cos (q 3 + q 4 ) + v 2 sin (q 3 + q 4 )) lv v 3 v 4 sin q 4
and concluding that the DAE system exhibits index 1, as the matrix
H11
H12 (q)
H (q) =
H21 (q)
0
is regular for all q,
det H (q) = Js J + mlh2 (J mlv (2lh + lv )) cos2 q 4 > 0.
= H (q)
q = v ,
,
C (q, v)
(5.60)
(5.61)
where the initial conditions (q0 , v0 ) have to comply with the constraints (5.50) defining
the manifold C J 1 (E) the systems motion is restricted to. In order to render this
constraint manifold attractive for the purpose of numerical integration, the system (5.61)
is augmented as
v
W (q, v, ) + Q
1
q = v ,
= H (q)
(5.62)
C (q, v) P g (q, v)
116
with the positive definite matrix P . The asymptotic stability of (5.62) with respect to
the constraint manifold can be seen with the Lyapunov function
1
V (q, v) = g T g
2
yielding
11
H
0
q = v ,
= 1, 2, 3
4
q = u
(q, v, u, )
12 (q)
v
W
Q
H
+
=
g (q, v)
0
0
H (q) =
,
H21 (q)
0
W
1
(q)
=H
,
C (q, v, u) P g (q, v)
(5.63)
again augmented with P g (q, v), P > 0, in order to guarantee the asymptotic convergence
of the motion to the constraint manifold.
c2 (
x) = y 2 ,
Lf c1 (
x, u, ) = y 1 ,
Lf c2 (
x) = y 2 ,
L2f c2 (
x, u, ) = y2
}.
(5.64)
117
drivers demand
on the transmission ratio
driverss demand
on the long. and lat. dynamics
(throttle/brake ped. pos., steering angle )
real-time
trajectory
planning
yd , y d , .
z
nonlinear
tracking w
u
state feedback
controller
(5.66),(5.67)
(5.13)
x
bicycle
dynamics
(5.4)-(5.6)
x
z = (
x)
(5.11)
118
(5.65)
w2 = yd2 1 e2 2 e 2 2 = yd2 1 z 2 yd2 2 z 3 y d2 2
1 = e1 ,
2 = e2
(5.66)
(5.67)
and > 0,
> 0, and 1 , 2 , such that the characteristic polynomials are Hurwitz.
Thus, the error dynamics read e 1 = e1
1 , e2 = 1 e2 2 e 2 2 with (5.67).
In order to illustrate the proposed vehicle dynamics control approach, a close-to-reality
modelled sports car which is provided by the multi-body simulation program msc.adams
[MSC], is used as a testrig. This demo vehicle comprises fully-detailed models of the
suspension, the powertrain and the steering system, while the bodywork is considered
as a rigid body, spring-mounted on the chassis. The total number of degrees of freedom
amounts to 96.
The parameters of the corresponding bicycle model, which have been extracted from
the msc.adams vehicle, are given by m = 1529 kg, J = 1344 kgm2 , lh = 1.08 m and
lv = 1.481 m. The point , which is attached with particular meaning regarding the
flatness property, see Remark 5.2, is located at (x, y) = (0.594 m, 0).
The lateral forces Fsh , Fsv acting on the tires are modelled following the widely-used
model of Pacejka [BPL89],
Fsk (k ) = 2Ds sin (Cs arctan (Bs k Es (Bs k arctan (Bs k )))) ,
(5.68)
k {v, h}, often referred to as Pacejkas magic formula in the literature. Thus, the
lateral forces are given as functions of the respective side-slip angles, see also Remark 5.5.
For instance, the parameters of the rear tire model are chosen to be Bs = 13 rad1 ,
Cs = 1.65, Ds = 4789 N and Es = 0.68, see Figure 5.5.
The real-time trajectory planning as sketched in Remark 5.16 is an open problem.
For the current sake of illustration, the desired trajectory yd (t) : [0, T ] R2 of the flat
output is chosen as
3t2 T 2t3
(vT v0 )
(5.69)
yd1 (t) = v0 +
T3
for the longitudinal component of the velocity at , and
p (t t1 , a1 , t2 t1 ) , t [t1 , t2 )
2
p (t t2 , a2 , t3 t2 ) , t [t2 , t3 )
yd (t) =
(5.70)
0,
else
119
10
8
6
4
2
0
2
4
6
8
10
20 15 10 5 0
5
10
side-slip angle [ ]
5.6. Conclusions. . .
15
20
Figure 5.5: The lateral tire force as a function of the side-slip angle, see also Remark 5.5,
following the model (5.68) of Pacejka [BPL89].
for the lateral component, with
t3 ( t)3
p (t, a, ) = a
.
6
Here, v0 and vT denote the x-component of the velocity at the time t = 0 and t = T ,
respectively. Let t1 = 1.5 s, t2 = 2.5 s, t3 = 3.5 s, T = 5 s, v0 = 27.7 m/s, vT = 33.3 m/s,
a1 = 50 m/s, a2 = 57 m/s, then this trajectory yd (t) of (5.69), (5.70) implies a single
lane change maneuver, associated with an acceleration of the vehicle, see Figure 5.6.
Figure 5.7 finally depicts the simulation results of the proposed flatness-based control.
The coefficients of the tracking controller (5.66) are chosen as =
= 10, and 1 = 1200,
2 = 60, = 8000.
The following remarks provide comments on the simulation results given in Figure 5.7.
Remark 5.17 The difference between the actual rear tire longitudinal force Flh and the
predicted value (dashed line) is due to the fact that the msc.adams model involves aerodynamic forces and the road resistance, which have to be overcome to follow the desired
trajectory.
Remark 5.18 The deviation of the yaw rate and the steering input results from unequal
loading of the inner and the outer track during cornering. This unequal loading clearly
affects the lateral and longitudinal tire forces via the normal force.
5.6. Conclusions. . .
120
3.5
3
Y [m]
2.5
2
1.5
1
0.5
0
20
40
60
Figure 5.6: Single lane change maneuver due to (5.69), (5.70): Trajectory of the vehicles
center of gravity C.
available the steering system to the vehicle dynamics control offers additional potential to
support the driver in emergency situations. To this end, the differential flatness property
of the bicycle model may offer new perspectives to cope with the vehicle dynamics control
problem involving the longitudinal forces of the tires and the steering angle as control inputs. The flat output revealed in this contribution could be identified as the longitudinal
and the lateral component of the velocity of a certain point located on the vehicles longitudinal axis. The location of this distinguished point is determined in terms of the mass,
the moment of inertia and the distance between the front wheel and the center of mass,
which can be regarded as well-known parameters in practical applications. Additionally,
this flat output does not depend on the particular actuation of the vehicle, i.e., it holds
for the rear-, front- and all-wheel driven car equivalently.
Besides the objective of real-time trajectory shaping for the flat output, our future
research will be concerned with observer design and parameter identification of the tire
characteristics. While the first issue deals with drivers experience during demanding
maneuvers, the second point focuses on the fact that the automotive industry, in general,
is strongly reluctant to implement side-slip angle sensors. Finally, vehicle dynamics control
is required to work seamlessly under various road conditions.
5.6. Conclusions. . .
34.0
33.0
32.0
31.0
30.0
29.0
28.0
27.0
1.0
[ ]
0.5
0.0
-0.5
-1.0
-1.5
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
0.3
r [rad/s]
0.2
0.1
0.0
-0.1
-0.2
-0.3
2.0
1.5
1.0
0.5
0.0
-0.5
-1.0
-1.5
-2.0
8.0
6.0
4.0
2.0
0.0
-2.0
-4.0
-6.0
-8.0
121
1.5
1.5
1.0
[ ]
0.5
0.0
-0.5
-1.0
-1.5
Flh [kN]
aCy [m/s2 ]
h [ ]
y 2 [m/s]
y 1 [m/s]
3
t [s]
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
-0.5
t [s]
Figure 5.7: Simulation results of the proposed flatness based vehicle dynamics control
approach, using a close-to-reality modelled sports car provided by the multi-body simulation program msc.adams [MSC]. The desired trajectories are indicated with dashed
lines. See Remarks 5.17 and 5.18 for notes on these results.
BIBLIOGRAPHY
[ABO99]
J. Ackermann, T. B
unte, and D. Odenthal, Advantages of active steering
for vehicle dynamics control, 32nd International Symposium on Automotive
Technology and Automation (Vienna, Austria), 1999, pp. 263270. 3, 97,
101
[BF52]
D.R. Bland and H. Ford, Cold rolling with strip tension, part III: An approximate treatment of the elastic compression of the strip in rolling mills,
Journal of the Iron and Steel Institute (1952), 245249. 2, 39
[BPL89]
[B
un98]
T. B
unte, Beitrage zur robusten Lenkregelung von Personenkraftwagen,
Ph.D. thesis, Technische Hochschule Aachen, Aachen, 1998. 3, 97, 101
[CH93]
[DET94]
S.A. Domanti, W.J. Edwards, and P.J. Thomas, A model for foil and thin
strip rolling, AISE Annual Convention (Cleveland, USA), 1994. 2, 39, 44,
47, 56, 70, 78
[FGSK03]
S. Fuchshumer, G. Grabmair, K. Schlacher, and G. Keintzel, Automatisierungstechnik in der Mechatronik zwei Beispiele aus der Stahlindustrie,
e& Elektrotechnik und Informationstechnik 5 (2003), 164171. 92
[FJ87]
N.A. Fleck and K.L. Johnson, Towards a new theory of cold rolling thin foil,
Int. J. Mech. Sci. 29 (1987), no. 7, 507524. 2, 39, 70
[FJ03]
M. Fliess and C. Join, An algebraic approach to fault diagnosis for linear systems, Computational Engineering in Systems Applications (Lille, France),
2003. 4
122
BIBLIOGRAPHY
BIBLIOGRAPHY
123
[FJMZ92]
N.A. Fleck, K.L. Johnson, M.E. Mear, and L.C. Zhang, Cold rolling of foil,
Proc. Instn. Mech. Engrs, Part B: Journal of Engineering Manufacture 206
(1992), 119194. 2, 39, 44, 56, 70, 78
[FLMR92]
M. Fliess, J. Levine, Ph. Martin, and P. Rouchon, Sur les systemes non
lineaires differentiellement plats, C. R. Acad. Sci. Paris Ser. I Math. 315
(1992), 619624. 1
[FLMR95]
[FSR03]
M. Fliess and H. Sira-Ramrez, An algebraic framework for linear identification, ESAIM Contr. Opt. Calc. Variat. 9 (2003). 3, 4, 5, 7, 19, 23
[GvL96]
G.H. Golub and C.F. van Loan, Matrix computations, 3rd ed., The Johns
Hopkins University Press, Baltimore and London, 1996. 54
[Joh85]
[JOZ60]
D. Jortner, J.F. Osterle, and C.F. Zorowski, An analysis of cold strip rolling,
Int. J. Mech. Sci. 2 (1960), 179194. 2, 39, 43, 44, 47, 70, 78
[KB02]
P. Kohn and G. Baumgarten, Die Aktivlenkung das neue fahrdynamische Lenksystem von BMW, 11th Aachener Kolloquium Fahrzeug- und Motorentechnik, 2002. 3
[Kha96]
H.K. Khalil, Nonlinear systems, 2 ed., Prentice Hall, New Jersey, 1996. 76,
110
[KHS+ 00]
[KSK99]
A. Kugi, K. Schlacher, and G. Keintzel, Position control and active eccentricity compensation in rolling mills, atAutomatisierungstechnik (1999), no. 8,
342349. 35
BIBLIOGRAPHY
BIBLIOGRAPHY
124
[KSN01]
[Kug01]
A. Kugi, Non-linear control based on physical models, Lecture Notes in Control and Information Sciences, vol. 260, Springer, London, 2001. 34, 36, 37,
38
[LS01]
H.R. Le and M.P.F. Sutcliffe, A robust model for rolling of thin strip and
foil, Int. J. Mech. Sci. 43 (2001), 14051419. 2, 39, 44, 56, 70, 71, 72, 78
[Map]
Maple, http://www.maplesoft.com/. 78
[Mei01]
[MMR03]
[MRS95]
[MS]
[MSC]
[Mul]
[Pac]
[Par88]
H. Parkus, Mechanik der festen Korper, 2 ed., Springer, Wien, New York,
1988. 51
[PP00]
Ubungen,
Verlag Stahleisen GmbH, D
usseldorf, 2000. 41, 70, 72, 77
[Rit04]
T. Rittenschober, Fahrdynamikregelung mit differentialgeometrischen Methoden der Regelungstechnik, Diploma thesis, Johannes Kepler University,
Linz, Austria, 2004. 3, 97
[RM98]
M. Rathinam and R.M. Murray, Configuration flatness of Lagrangian systems underactuated by one control, SIAM J. Control Optim. 36 (1998), no. 1,
164179. 103, 104
BIBLIOGRAPHY
BIBLIOGRAPHY
125
[Rot97]
R. Rothfu, Anwendung der flachheitsbasierten Analyse und Regelung nichtlinearer Mehrgroensysteme, VDI Verlag, 1997. 94, 117
[RS40]
P. Riekert and T. Schunck, Zur Fahrmechanik des gummibereiften Kraftfahrzeugs, Ingenieur Archiv 11 (1940), 210224. 3, 97
[RSRF05]
[Rud03]
[SKN01]
K. Schlacher, A. Kugi, and R. Novak, Input to output linearization with constrained measurements, 5th IFAC Symposium on Nonlinear Control Systems
(NOLCOS 2001) (St.Petersburg, Russia), July 2001. 37
[TG70]
S.P. Timoshenko and J.N. Goodier, Theory of elasticity, 3rd ed., McGrawHill, Singapore, 1970. 44
[Zie92]
Curriculum Vitae
Personal data
name:
Austria
marital status:
address:
email:
stefan@fuchshumer.at
Education
09/197806/1982
09/198206/1990
10/199001/1997
diploma studies in Mechatronics at the Johannes Kepler University (JKU) of Linz, passed with distinction.
diploma thesis: Bestimmung der magnetischen Eigenschaften von
Elektroblech mit dem Tafeljoch; prepared at the Department of
Automatic Control and Control Systems Technology in cooperation with voestalpine Stahl Linz.
since 01/1999
Professional experience
02/199703/1997
04/199711/1997
military service
02/199810/1998
01/199912/2005
research assistant at the Christian Doppler Laboratory for Automatic Control of Mechatronic Systems in Steel Industries installed
at the Institute of Automatic Control and Control Systems Technology, JKU Linz, in cooperation with Voest-Alpine Industrieanlagenbau GmbH Linz (VAI)
since 1999
lecturer at the Institute of Automatic Control and Control Systems Technology, JKU Linz.
since 2002
Publications
Chapters of books:
S. Fuchshumer, K. Schlacher, A. Kugi: Mathematical Modelling and Nonlinear Control of a Temper Rolling Mill. Selected Topics in Structronics and Mechatronic Systems. Eds: A.K. Belyaev and A. Guran. Series on Stability, Vibration and Control
of Systems, vol. 3, World Scientific, 2003, pp. 175219.
K. Schlacher, S. Fuchshumer, J. Holl: Some Applications of Differential Geometry
in Control. Advanced Dynamics and Control of Structures and Machines. Eds: H.
Irschik and K. Schlacher. CISM Courses and Lectures No. 444, Springer, 2004,
pp. 249260.
Articles:
S. Fuchshumer, G. Grabmair, K. Schlacher, G. Keintzel: Automatisierungstechnik
in der Mechatronik: Zwei Beispiele aus der Stahlindustrie. e&i Elektrotechnik und
Informationstechnik, Heft 5, 2003, pp. 164171.
S. Fuchshumer, K. Schlacher, G. Grabmair, K. Straka: Flachheitsbasierte Folgeregelung des Labormodells Ball on the Wheel. e&i Elektrotechnik und Informationstechnik, Heft 9, 2004, pp. 301306.
S. Fuchshumer, K. Schlacher, T. Rittenschober: Ein Beitrag zur nichtlinearen Fahrdynamikregelung: Die differentielle Flachheit des Einspurmodells. e&i Elektrotechnik und Informationstechnik, Heft 9, 2005, pp. 319324.
Eidesstattliche Erklarung
Ich erklare an Eides statt, dass ich die vorliegende Dissertation selbstandig und ohne
fremde Hilfe verfasst, andere als die angegebenen Quellen und Hilfsmittel nicht benutzt
sowie die wortlich oder sinngema entnommenen Stellen als solche kenntlich gemacht
habe.
129