Substructuring in Engineering Dynamics: Emerging Numerical and Experimental Techniques Matthew S. Allen 2024 scribd download
Substructuring in Engineering Dynamics: Emerging Numerical and Experimental Techniques Matthew S. Allen 2024 scribd download
com
https://textbookfull.com/product/substructuring-in-
engineering-dynamics-emerging-numerical-and-experimental-
techniques-matthew-s-allen/
OR CLICK BUTTON
DOWNLOAD NOW
https://textbookfull.com/product/engineering-
dynamics-2-0-fundamentals-and-numerical-solutions-lester-w-schmerr/
textboxfull.com
https://textbookfull.com/product/emerging-security-algorithms-and-
techniques-ahmad/
textboxfull.com
https://textbookfull.com/product/building-dashboards-with-microsoft-
dynamics-gp-2016-allen/
textboxfull.com
https://textbookfull.com/product/advances-in-numerical-methods-in-
geotechnical-engineering-hany-shehata/
textboxfull.com
https://textbookfull.com/product/numerical-analysis-for-applied-
science-2nd-edition-myron-b-allen-iii/
textboxfull.com
CISM International Centre for Mechanical Sciences 594
Courses and Lectures
Matthew S. Allen
Daniel Rixen
Maarten van der Seijs
Paolo Tiso
Thomas Abrahamsson
Randall L. Mayes
Substructuring
in Engineering
Dynamics
Emerging Numerical and Experimental
Techniques
International Centre
for Mechanical Sciences
CISM International Centre for Mechanical
Sciences
Volume 594
Managing Editor
Paolo Serafini, CISM—International Centre for Mechanical Sciences, Udine, Italy
Series Editors
Elisabeth Guazzelli, IUSTI UMR 7343, Aix-Marseille Université, Marseille, France
Franz G. Rammerstorfer, Institut für Leichtbau und Struktur-Biomechanik, TU
Wien, Vienna, Wien, Austria
Wolfgang A. Wall, Institute for Computational Mechanics, Technical University
Munich, Munich, Bayern, Germany
Bernhard Schrefler, CISM—International Centre for Mechanical Sciences, Udine,
Italy
For more than 40 years the book series edited by CISM, “International Centre for
Mechanical Sciences: Courses and Lectures”, has presented groundbreaking
developments in mechanics and computational engineering methods. It covers
such fields as solid and fluid mechanics, mechanics of materials, micro- and
nanomechanics, biomechanics, and mechatronics. The papers are written by
international authorities in the field. The books are at graduate level but may include
some introductory material.
Substructuring
in Engineering Dynamics
Emerging Numerical and Experimental
Techniques
123
Matthew S. Allen Daniel Rixen
Department of Engineering Physics Lehrstuhl für Angewandte Mechanik
University of Wisconsin–Madison Technische Universität München
Madison, WI, USA Munich, Bayern, Germany
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
v
vi Preface
vii
viii Contents
Abstract “Divide and Conquer” is a paradigm that helped Julius Cesar to dominate
on the wide Roman Empire. The power of dividing systems, then analyze them as
parts before combining them in an assembly, is also an approach often followed in
science and engineering. In this introductory chapter, we shortly discuss the main idea
behind domain decomposition and substructuring applied to mechanical systems. —
Chapter Author: Daniel Rixen
pioneered by mathematicians like Courant (1943), Courant and Hilbert (1953), can in
fact be seen as a clever application of the Rayleigh–Ritz technique on small region of
the computational domain, resulting therefore in not only a function decomposition
of the problem within the element, but also a spatial decomposition of the problem
as the physical domain is broken into elements.
Today when we consider substructuring, most often we envision dividing a finite
element model into subcomponents that each contain many elements. This substruc-
turing approach was developed in the 60s, motivated by the need to solve structural
problems in aerospace and aeronautics of several tens of nodes (considered to be very
large problems at that time, given the fact that computers were just arising). Hence,
the idea was to add another level of decomposition and reduction by subdividing the
mesh of a finite element model into substructures and to represent the dynamics per
substructure in a reduced and approximate way. Some of the substructuring tech-
niques still commonly used today originate from that time where memory and CPU
power were very limited.
Later, in the 80s and especially in the 90s, solving very large problems using
domain decomposition became a very active research topic. In order to efficiently
use the computational power of new multiprocessor machines, it was essential to min-
imize the communications between processing units and to ensure that each CPU
was fed with enough work that could be done independently. Domain Decomposi-
tion is the paradigm to achieve this, by letting each CPU solve the local problems
of each domain and searching for the global solution by iterating on the interface
solutions (Toselli and Widlund 2006). Nowadays, high-performance computing tech-
niques solve engineering problems of close to a billion degrees of freedom on several
hundreds of thousands of processors thanks to that paradigm (Toivanen et al. 2018).
The last wave of intensive research in substructuring came at the beginning of
the twenty-first century, this time not directly powered by exponential grows of
computing power, but by significant progress in techniques to accurately measure
the dynamics of components. Experimental substructuring aims to build models of
assemblies from the data measured on components, possibly in combination with
numerical models. Although the theory of experimental dynamic substructuring is
essentially the same as its analytical counterpart, many novel techniques needed to
be developed in order to alleviate the adverse effects of measurement errors (even if
strongly reduced thanks to novel sensing and acquisition techniques) when building
a model based on experimentally characterized components.
In mechanical engineering practice, substructuring techniques are useful for the fol-
lowing cases:
• When working on a large project (for example, the design of an aircraft), different
groups and departments work on different components. In that case, it is essential
1.2 Advantages of Substructuring in Mechanical Engineering 3
that each team can concentrate on the modeling of their part, and then have an
efficient means for assembling to other components in order to predict the overall
dynamics and optimize their design accordingly.
• In industry, it is common that models of subsystems must be shared. A classical
example is the coupled load analysis of the assembly of a satellite and a launcher:
in that case, the satellite company needs to share its model with the launcher
operator in order to perform a global dynamic analysis and guarantee that vibration
limits will not be crossed. In that case, exchanging information can be efficient if
the different components have been modeled separately in such a way that their
complexity (number of degrees of freedom) is low and such that they can be
assembled on their interface.
• In many engineering problems, parts of the system to be analyzed are significantly
more complex than the rest. In structural dynamics, for instance, parts of the model
could be highly nonlinear whereas the remainder can be considered as linear. In that
case, it is very advantageous to substructure the problem in such a way that only
the parts that are nonlinear are treated with the appropriate techniques, whereas
the linear parts are, for instance, condensed once and for all. Such a subdivision of
the problem into substructures of different complexities can greatly improve the
analysis speed and give better insight into the dynamic behavior.
• When optimizing the design during the development of a new product, or during an
improvement of an old design, one often wants to modify only a small number of
parts (for instance, change the bushings connecting the power train of a car to the
chassis in order to modify the noise and vibration harshness). When the parts that
will not be modified are available as substructures and already precomputed (e.g.,
their dynamics on the interface is known), then recomputing the global behavior
when only the components under optimization are changed requires a significantly
reduced cost, enabling much faster design cycles.
• In experimental testing, it is not always possible to test the full system; often
system-level analysis is needed before all of the components have been manufac-
tured. In other cases, it is impossible to test the full assembly because appropriately
exciting a large structure would require forces that are beyond the capabilities of
existing shakers. It is then very advantageous to experimentally characterize the
structure part by part. This allows troubleshooting problems arising from the local
dynamics in those parts, but also to build a full model by assembling the measured
parts using experimental substructuring techniques.
• When parts of the system have not yet been built, it is possible to combine the
measured dynamics of hardware components with other parts that are modeled
only numerically (hybrid substructuring). In some cases, the combination in real
time of the transient response of a hardware component (the physical substructure)
with the dynamics of a numerical substructure is essential to predict the dynamics
of the hardware part in realistic conditions. Such hardware-in-the-loop tests (also
called Real-Time Substructuring, Hybrid Testing or Cyber-Physics) allow testing
very complex systems by having only a part in the lab, the rest of the system being
co-simulated in real time.
4 1 Introduction and Motivation
• Substructuring methods open new opportunities for efficient analyses and design,
including the development of digital twins, or a model that can be used to monitor
the health of a system.
These are only a few examples of cases in which substructuring can be advan-
tageous. For these and other reasons, substructuring has attracted a lot of attention
over the years and is yet today a very active research field.
References
Courant, R. (1943). Variational methods for the solution of problems of equilibrium and vibrations.
Bulletin of the American Mathematical Society, 49(1), 23.
Courant, R., & Hilbert, D. (1953). Methods of Mathematical Physics (Vol. 1). New York: Interscience
Publishers.
Géradin, M., & Rixen, D. (2015). Mechanical vibrations. Theory and Application to Structural
Dynamics (3rd ed.). Chichester: Wiley.
Rayleigh, J. (1896). The Theory of Sound (2nd ed.). Mineola: Dover.
Ritz, W. (1909). Uber eine neue methode zur lösung gewisser variations probleme der mathema-
tishen physik. Journal für die Reine und Angewandte Mathematik, 135, 1–61.
Schwarz, H. A. (1890). Gesammelte Mathematische Abhandlungen (Vol. 2, pp. 133–143). Berlin:
Springer; First published in Vierteljahrsschrift der Naturforschenden Gesellschaft in Zürich
(1870) (Vol. 15, pp. 272–286).
Toivanen, J., Avery, P., & Farhat, C. (2018). A multilevel feti-dp method and its performance for
problems with billions of degrees of freedom. International Journal for Numerical Methods in
Engineering, 116(10–11), 661–682.
Toselli, A., & Widlund, O. (2006). Domain decomposition methods-algorithms and theory (Vol.
34). Berlin: Springer Science & Business Media.
Chapter 2
Preliminaries: Primal and Dual
Assembly of Dynamic Models
In these lecture notes, we assume that the problem has already1 been discretized
(using, for instance, an appropriate Finite Element or Boundary Element formula-
tion). The discretized problem describing the dynamics in a substructure Ω (s) can
be written in the form
(s) (s)
M(s) ü(s) + f (s)
nl u̇ , u = f (s) + g(s) . (2.1)
The superscript (s) indicates that the equation is written for a given substructure s.
M denotes the discretized mass matrix,2 f nl the discretized nonlinear internal force
Fig. 2.1 Schematic overview of substructuring domains (van der Seijs 2016)
2.1 The Dynamics of a Substructure: Domains of Representation 7
The description outlined in (2.2) expresses the dynamics of the substructure using
the displacement at a specific nodal location and is referred to as a physical repre-
sentation. Note that representation in the physical domain can be either discretized
(as assumed in this text) or continuous. In the later case, the unknown is the physical
response field and the associated equations are partial differential equations describ-
ing the (nonlinear) continuous dynamics in a body. This will not be considered here.
where ωi(s) are the eigenfrequencies of the free–free substructure and φ i(s) the asso-
ciated eigenmodes that have the fundamental property of being mass and stiffness
orthogonal, namely,
T 2
φ i(s) K(s) φ (s) (s)
j = ωi δi j , (2.4)
T
φ i(s) M(s) φ (s)
j = δi j , (2.5)
where we have assumed that the modes amplitudes have been chosen to be mass-
normalized and where δi j is the Kronecker symbol such that
δi j = 1 if i = j.
δi j = 0 if i = j.
where n (s) is the number of degrees of freedom in substructure (s). Here, ηi(s) are
the amplitudes of the modal component of the response and are often called modal
coordinate of substructure (s). Often, we will use a matrix notation as in the second
equality of (2.6), where Φ (s) is a matrix containing in its columns the vibration
8 2 Preliminaries: Primal and Dual Assembly of Dynamic Models
modes and η (s) is a uni-column matrix containing all modal coordinates. Usually,
only a subset of modes is considered in order to have an approximated but reduced
representation of the substructure. This will be discussed in later chapters.
In general, the response u(s) can be represented as a combination of n (s) indepen-
dent vectors and we write
(s)
n
u (s)
= vi(s) qi(s) = V(s) q(s) , (2.7)
i=1
where V(s) is a square matrix containing the basis vectors for the desired change of
T
variables. Substituting in the dynamic equation (2.2) and premultiplying by V(s) to
project the equations onto the same space, we obtain
T T T T T
V(s) M(s) V(s) q̈(s) + V(s) C(s) V(s) q̇(s) + V(s) K(s) V(s) q(s) = V(s) f (s) + V(s) g(s) ,
(2.8)
which is usually written as
(s) (s) (s) (s)
M̃ q̈(s) + C̃ q̇(s) + K̃ q(s) = f̃ + g̃(s) , (2.9)
where the tilde superscript indicates that the matrices and vectors pertain now to a
representation in a transformed space. The representation vectors stored in V(s) can
be any set of independent vectors, in particular, they can be chosen as the vibration
(s) (s)
modes Φ (s) , in which case the transformed mass and stiffness matrices M̃ and K̃
will be diagonal.
The representation (2.9) will often be referred to as the modal representation, even
when the base vectors are not vibration modes Φ (s) but general representation modes
V(s) . The associated degrees of freedom q̇(s) are then called generalized degrees
of freedom or modal coordinates and do in general not represent the solution at a
particular physical location.
In case an incomplete basis is used for the representation, namely, when fewer
modes than the number n (s) of degrees of freedom in the substructure are used, the
modal representation represents the dynamics in a reduced subspace and in general
only in an approximate way (this will be discussed in detail in Chap. 3). We will then
call the representation a reduced modal representation.
In the form of Eq. (2.2), the unknown dynamic response is considered a function of
time and we say that, from a spectral point of view, the equations are expressed in
the time domain.
2.1 The Dynamics of a Substructure: Domains of Representation 9
Similar to the decomposition of the spatial response in component modes, the time
dependency of the response can also be decomposed into a combination of time
contributions. The most classical one is the Fourier decomposition4 that writes the
time function of the response in terms of harmonic functions. Using complex number
notations, the Fourier decomposition can be written as
∞
u(s) (t) = ū(s) (ω) e−iωt dω, (2.10)
−∞
(s)
where f̄ and ḡ(s) are the Fourier components of the forces, for instance,
∞
(s) 1
f̄ = f (s) (t) eiωt dt. (2.12)
2π −∞
The dynamic equation in the frequency domain (2.14) is also often written as
(s)
Z(s) ū(s) = f̄ + ḡ(s) (2.13)
3 For an interesting matrix description of time discretization see Rixen and van der Valk (2013),
van der Valk and Rixen (2014).
4 Note that other base functions in time can be used (such as wavelets), but this will not be discussed
here.
10 2 Preliminaries: Primal and Dual Assembly of Dynamic Models
where
Z(s) (ω) = −ω 2 M(s) + iωC(s) + K(s) (2.14)
Z(s) is a dynamic stiffness matrix and is a function of the frequency ω. In this form, ū(s)
is the complex amplitude of the harmonic displacement response (or equivalently the
Fourier component of the transient response). A similar equation can be written for the
amplitude of the harmonic velocity or accelerations in which case the operator Z(ω)
is commonly called the mechanical impedance or the apparent mass, respectively.
The dynamic relation can also be inverted and written
(s)
ū(s) = Y(s) f̄ + ḡ(s) , (2.15)
where −1
Y(s) (ω) = Z(s) (ω)−1 = −ω 2 M(s) + iωC(s) + K(s) (2.16)
Y(s) is a Frequency Response Function (FRF) matrix and is often called the admit-
tance or dynamic flexibility, or more specifically receptance, mobility, or acceler-
ance/inertance if ū(s) are displacements, velocities or accelerations respectively.
Obviously, in practice, the harmonic components are calculated only for a finite
discrete number of frequencies ω, and (2.10) is approximated by the Discrete Fourier
Decomposition
Nω
u(s) (t) = ū(s)
k eiωk (2.17)
k=−Nω
choosing a frequency range covering the spectral range of the excitation. It is note-
worthy that the decomposition in (2.17) is comparable to the decomposition of the
space function of the response in (2.6) and can also be seen as a reduction of the
transient response in the time domain. It is an approximation unless the excitation
can be exactly represented by a finite combination of harmonics.
Laplace Domain
Another often used representation of the time evolution of the dynamic response is
in terms of the Laplace components. The idea is to look for the dynamic response
when modulated with a decreasing exponential function, namely,
∞
ū(s) (s) = L u(s) (t) = e−st u(s) (t) dt (2.18)
0
Clearly, there is a similarity between Laplace and Fourier transforms since (2.18)
becomes a Fourier transform if s is taken as imaginary. The main difference is that the
inverse transform is trivial for the Fourier domain (leading to the frequency domain
decomposition (2.10) or (2.17)), whereas finding the inverse Laplace transform is far
more difficult and not general. In structural dynamics, Laplace transforms are used
for highly transient problems that cannot efficiently be represented by harmonic
superposition, such as impact responses and shock propagations (see for instance
Sect. 4.3.2 in Géradin and Rixen (2015)).
Displacement Space
In addition to representing the space and spectral behavior of the system in dif-
ferent domains as explained above, the very state of the system can be described in
mainly two different manners: either one sees the displacements as the only indepen-
dent unknowns (velocities and accelerations being dependent on the displacement
through derivatives) or the velocities are seen as additional independent variables
for which the derivative relation to the displacement is explicitly expressed in the
formulation. For the first approach, often used in structural dynamics, the dynamics
of the system are described by a single set of second-order equations as in (2.1) (for
nonlinear structures) or (2.2) (for linear ones).
State-Space Representation
In the second case, namely, the velocities are seen as additional independent variables,
the state of the system is described both by the displacements and the velocities:
u(s) (t)
x(s) (t) = (2.19)
u̇(s) (t)
and the associated linear dynamic equation can for instance be written as
(s)
I 0 (s) 0 I(s) (s) 0
ẋ = x (t) + (2.20)
0 M(s) −K(s) −C(s) f (s) + g(s) .
In this state-space representation the number of equations has doubled, but the order
of the differential is now reduced to one. This representation is often used especially
in control. This form is also commonly used in structural dynamics when strong
damping is present since the concept of modes of vibration properly generalizes only
when writing the system in the State Space (see, for instance, Sect. 3.3 in Géradin
and Rixen (2015)).
12 2 Preliminaries: Primal and Dual Assembly of Dynamic Models
From the short summary of the formulation of the structural dynamics problem, it
is clear that many variants to describe the problem, combining a spatial, spectral,
and state representation, can be constructed. In Fig. 2.1, the different aspects are
shown graphically and relations between them are summarized. Depending on the
representation chosen, numerical and experimental techniques in substructuring can
significantly differ as will be seen in these lecture notes.
Let us consider again the linearized dynamic equilibrium equation (2.11) of a sub-
structure in the physical space and in the frequency domain5
(s)
Z(s) ū(s) = f̄ + ḡ(s) s = 1 . . . N sub , (2.21)
5 Expressing the coupling of substructures in other domains (modal, time, state space …) will be
discussed in later chapters and use exactly the same approach.
6 These lecture notes deal with structural problems. Nevertheless, the general theory is also applicable
The interface equilibrium requires that the interface forces, ḡ(s) , which are internal
forces between the substructures, sum to zero when assembled. This is merely a
manifestation of Newton’s “action–reaction" principle. Considering, for instance, an
interface Γ (sr ) between two substructures s and r , one could express this condition
as7
g(s) (r )
b + gb = 0 on Γ (sr ) , (2.24)
gi(s) =0 gi(r ) = 0, (2.25)
where the subscript b indicates a restriction of the DOF to the boundary and where
we assumed that the DOF are numbered in the same manner on both sides of the
interface. The subscript i denotes DOF that are not on a boundary and are thus internal
DOF. On internal DOF, no connecting forces should exist.
In practice, the numbering of the DOF on the interface will not match across the
interfaces and in addition more than two substructures can intersect on an interface
(so-called cross-points in 2D and 3D, and edges in 3D). Hence in general, the interface
equilibrium condition needs to be expressed using Boolean localization matrices
T
L(s) of dimension n × n s that combine the forces on either side of the interface to
satisfy force equilibrium. Interestingly, these localization matrices also map the DOF
of substructure s to a global and unique set of n global DOF, as will be elaborated
later. In general, the interface equilibrium thus is written as
sub
N
T
L(s) ḡ(s) = 0 . (2.26)
s=1
This equilibrium
s condition can also be written using the block matrix LT of dimension
n b × s n acting on the set of all substructure interface forces
sub T
LT ḡ = 0 where LT = L(1) · · · L(N )
T
(2.27)
7 Ingeneral, we will assume that the decomposition in substructures generates an interface dis-
cretization that is conforming and matching, namely, that the shape functions used on either side of
the interface are identical and that the nodes coincide. In case of nonconforming or nonmatching
interfaces, the theory used in these lecture notes are still generally applicable, but the assembly
operators are then no longer Boolean. Indeed, going back to the variational principle underlying
the discretized problem, the part related to the compatibility condition over an interface Γ can be
written as
μT vdΓ
Γ
where μ and v are the field. More details about nonmatching interfaces can be found, for instance,
in Rixen (1997).
14 2 Preliminaries: Primal and Dual Assembly of Dynamic Models
To illustrate the notation, consider the examples in Fig. 2.2. The green num-
bers indicate the global nodes for which the interface force conditions are written
(i.e., the row number in the localization matrices).
For the beam example in Fig. 2.2a, the localization matrices are
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
1 100 0
L(1) = ⎣0⎦ L(2) = ⎣0 1 0⎦ L(2) = ⎣0 ⎦
T T T
0 001 1
For the second example in Fig. 2.2b, we have two degrees of freedom per
node and the localization matrix can be written as
LT = L(1) L(2) L(3)
2.2 Interface Conditions for Coupled Substructures 15
⎡⎡ ⎤ ⎡ ⎤⎡ ⎤⎤
1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
⎢⎢0
⎢⎢ 1 0 0 0 0 0 0⎥⎥
⎢0
⎢ 0 0 0 0 0 0 0⎥⎥⎢
⎢0 0 0 0 0 0 0 0⎥⎥⎥
⎥
⎢⎢0
⎢⎢ 0 1 0 0 0 0 0⎥⎥
⎢1
⎢ 0 0 0 0 0 0 0⎥ ⎢
⎥ ⎢0 0 0 0 0 0 0 0⎥⎥
⎥
⎥
⎢⎢0
⎢⎢ 0 0 1 0 0 0 0⎥⎥
⎢0
⎢ 1 0 0 0 0 0 0⎥ ⎢
⎥
⎢0 0 0 0 0 0 0 0⎥⎥⎥
⎥
⎢⎢0
⎢⎢ 0 0 0 1 0 0 0⎥⎥
⎢0
⎢ 0 0 0 0 0 0 0⎥ ⎢
⎥ ⎢0 0 0 0 0 0 0 0⎥⎥
⎥
⎥
⎢⎢0
⎢⎢ 0 0 0 0 1 0 0⎥⎥
⎢0
⎢ 0 0 0 0 0 0 0⎥ ⎢
⎥
⎢0 0 0 0 0 0 0 0⎥⎥⎥
⎥
⎢⎢0
⎢⎢ 0 0 0 0 0 1 0⎥⎥
⎢0
⎢ 0 0 0 1 0 0 0⎥ ⎢
⎥ ⎢1 0 0 0 0 0 0 0⎥⎥
⎥
⎥
⎢⎢0 0 0 0 0 0 0 1⎥⎥
⎢0
⎢ 0 0 0 0 1 0 0⎥ ⎢
⎥
⎢0 1 0 0 0 0 0 0⎥ ⎥
LT = ⎢ ⎢ ⎥⎥
⎢⎢0
⎢⎢ 0 0 0 0 0 0 0⎥⎥
⎢0
⎢ 0 1 0 0 0 0 0⎥ ⎢
⎥ ⎢0 0 0 0 0 0 0 0⎥⎥
⎥
⎥
⎢⎢0
⎢⎢ 0 0 0 0 0 0 0⎥⎥
⎢0
⎢ 0 0 1 0 0 0 0⎥ ⎢
⎥
⎢0 0 0 0 0 0 0 0⎥⎥⎥
⎥
⎢⎢0
⎢⎢ 0 0 0 0 0 0 0⎥⎥
⎢0
⎢ 0 0 0 0 0 1 0⎥ ⎢
⎥ ⎢0 0 1 0 0 0 0 0⎥⎥
⎥
⎥
⎢⎢0
⎢⎢ 0 0 0 0 0 0 0⎥⎥
⎢0
⎢ 0 0 0 0 0 0 1⎥ ⎢
⎥
⎢0 0 0 1 0 0 0 0⎥⎥⎥
⎥
⎢⎢0
⎢⎢ 0 0 0 0 0 0 0⎥⎥
⎢0
⎢ 0 0 0 0 0 0 0⎥ ⎢
⎥ ⎢0 0 0 0 1 0 0 0⎥⎥
⎥
⎥
⎢⎢0
⎢⎢ 0 0 0 0 0 0 0⎥⎥
⎢0
⎢ 0 0 0 0 0 0 0⎥ ⎢
⎥
⎢0 0 0 0 0 1 0 0⎥⎥⎥
⎥
⎣⎣0 0 0 0 0 0 0 0⎦ ⎣0 0 0 0 0 0 0 0 ⎦ ⎣0 0 0 0 0 0 1 0 ⎦
⎦
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1
These matrices are in fact identical to the localization matrices used in Finite
Element codes to assemble elementary matrices in the global system, however here
the localization is not written for one element but for one substructure (that one
could consider as a super-element or macro-element). Obviously, it is not efficient
to store these Boolean matrices as written above, but rather one should store them
as sparse matrices, or even better one should construct the mapping tables based on
the connectivity of the substructures over the interfaces.
The second condition that needs to be satisfied on the interface is that DOF pertaining
to some structural node have the same response on both sides of the interface, or in
other words that the DOF are compatible on the interface. Considering the DOF of
two substructures s and r coupled on the interface Γ (sr ) , the compatibility condition
becomes
ū(s) (r )
b − ūb = 0 on Γ (sr )
where, as before, the subscript b indicates that the compatibility is written for the
boundary DOF and where we assumed that the DOF are numbered identically on
both sides of the interface.
In general, the numbering on the interface does not coincide and therefore the
compatibility conditions are expressed using signed Boolean matrices B(s) . When
operating on ū(s) , these operators extract the interfaces DOF and give them an oppo-
site sign on each side of the interface. The interface compatibility can then be written
16 2 Preliminaries: Primal and Dual Assembly of Dynamic Models
sub
N
B(s) ū(s) = 0 (2.29)
s=1
One can use a block matrix notation to write this condition also in the form
sub
Bū = 0 where B = B(1) · · · B(N ) . (2.30)
The side on which the entry in B is positive and negative can be chosen freely.
The interpretation of B and its associated Lagrange multipliers is depicted in
Fig. 2.3a.
For the second example in Fig. 2.2, the constraint matrix can be written as
B = B(1) B(2) B(3) .
⎡⎡ ⎤⎡ ⎤⎡ ⎤⎤
0 0 1 0 0 0 0 0 −1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
⎢⎢0
⎢⎢ 0 0 1 0 0 0 0⎥ ⎢
⎥⎢0 −1 0 0 0 0 0 0⎥ ⎢
⎥⎢0 0 0 0 0 0 0 0⎥⎥⎥
⎥
⎢⎢0
⎢⎢ 0 0 0 0 0 1 0⎥ ⎢
⎥⎢0 0 0 0 −1 0 0 0⎥ ⎢
⎥⎢0 0 0 0 0 0 0 ⎥
0 ⎥⎥⎥
⎢⎢0
⎢⎢ 0 0 0 0 0 0 1⎥ ⎢
⎥
⎢0 0 0 0 0 −1 0 0⎥ ⎢
⎥
⎢0 0 0 0 0 0 0 0⎥⎥⎥
⎥
⎢⎢0 0 0 0 0 0 0 0⎥ ⎢
⎥⎢0 0 0 0 1 0 0 0⎥ ⎢
⎥ ⎢−1 0 0 0 0 0 0 ⎥
0 ⎥⎥
=⎢ ⎢ ⎥
⎢⎢0
⎢⎢ 0 0 0 0 0 0 0⎥ ⎢
⎥
⎢0 0 0 0 0 1 0 0⎥ ⎢
⎥
⎢0 −1 0 0 0 0 0 0⎥⎥⎥
⎥
⎢⎢0
⎢⎢ 0 0 0 0 0 0 0⎥ ⎢
⎥⎢0 0 0 0 0 0 1 0⎥ ⎢
⎥⎢0 0 −1 0 0 0 0 ⎥
0 ⎥⎥⎥
⎢⎢0
⎢⎢ 0 0 0 0 0 0 0⎥ ⎢
⎥
⎢0 0 0 0 0 0 0 1⎥ ⎢
⎥
⎢0 0 0 −1 0 0 0 0⎥⎥⎥
⎥
⎣⎣0 0 0 0 0 0 1 0⎦ ⎣ 0 0 0 0 0 0 0 0⎦ ⎣−1 0 0 0 0 0 0 0⎦⎦
0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 −1 0 0 0 0 0 0
As for the localization matrix L(s) , the constraint matrices B(s) are, in practice, not
stored as full but as sparse, or only connectivity information is used to apply them
to a vector.
To summarize this section, an assembly of substructures in the physical frequency
domain (and similarly in other domains) is obtained by imposing compatibility and
interface equilibrium conditions, leading to the set of equations
⎧ (s)
⎪
⎪ Z(s) ū(s) = f̄ + ḡ(s) s = 1 . . . N sub
⎪
⎪
⎪
⎪ (s) (s)
sub
⎪ N
⎨ B ū = 0
(2.32)
⎪ s=1
⎪ N sub
⎪
⎪
⎪
⎪ T
⎪
⎩ L(s) ḡ(s) = 0
s=1
⎧
⎨ Zū = f̄ + ḡ
Bū = 0 (2.33)
⎩ T
L ḡ = 0
The form (2.32) (or equivalently (2.33)) of the coupled problem uses two interface
fields, namely, the primal unknowns u per substructure (i.e., on each side of the
interfaces) and the substructure interface forces g called dual unknowns.8 Solving
the dynamic problem of the assembly in the form (2.32) can be expensive since many
interface unknowns need to be resolved. Hence, these equations can be rearranged in
order to eliminate the interface forces and write the problem in terms of unique inter-
face displacements (primal assembly) or by introducing interface forces satisfying
the interface equilibrium (dual assembly).
We can define as primal unknowns for the interface a set of DOF ug that are global
and uniquely defined for the entire structure. The DOF of each substructure are then
obtained by mapping the global set ug to the local DOF of each substructure u(s) .
Such a mapping was already introduced in the previous section, (2.26), to map the
local interface forces to a global set. The same mapping, but now from the global
DOF to the local ones can be used to write
If the substructure DOF are obtained from a unique set as described above, they
automatically satisfy the compatibility conditions that matching interface DOF must
be equal. Hence
BLug = 0 ∀ ug . (2.35)
This relation mathematically means that L represents the nullspace of the constraint
matrix B:
L = null(B) (2.36)
8 Inthese lecture notes, we will introduce the coupling conditions using basically a two field
approach. A more general three-field approach can also be considered but will not be discussed
here. See for instance Voormeeren (2012).
Discovering Diverse Content Through
Random Scribd Documents
The Project Gutenberg eBook of Remusen
kotiripitykset
This ebook is for the use of anyone anywhere in the United States
and most other parts of the world at no cost and with almost no
restrictions whatsoever. You may copy it, give it away or re-use it
under the terms of the Project Gutenberg License included with this
ebook or online at www.gutenberg.org. If you are not located in the
United States, you will have to check the laws of the country where
you are located before using this eBook.
Language: Finnish
Kirj.
G. von Moser
Amalia. Kiitos kaunis, hyvä Evert! — Sinä olet aina niin kiltti.
(Ojentaa hänelle kätensä.)
Evert. Amerikanko?
Amalia. No vaikkapa Amerikankin, sama se minusta — mutta
ummelleen vuosi sitten — kenties muistatkin?
(Palvelija tulee)
Amalia. Niin juuri, sinä itse. Siinä ei ollut tarpeeksi, että sain niin
paljon vaivata itseäni, vieläpä ihan turhaan — muistuttaakseni sinua
niistä hetkistä, joita itse sanoit mitä ihanimmiksi — minun täytyy
vielä kärsiä sitäkin, että paljon paremmin muistat jonkun pihvin kuin
noita aikoja —
Evert. Minä pyydän sinua, Ami kulta, olehan nyt toki järkevä.
(Asettuu niin että hän estää tulijoita näkemästä Amaliaa.)
Nauravathan he meille. (Vaunun kolinaa, kellonhelinää, vihellyksiä.)
TOINEN KOHTAUS.
KOLMAS KOHTAUS.
Hilarius. Serafina.
Serafina. Niin, niin — sinä olet uhri! Sinä kärsit kuin lammas!
Sinulla on paljon vaivaa saada kaikki taasen oikeaan järjestykseen,
mitä minä pelmutan. Ja sinä olet aina oikeassa, ja jos minä heikko
vaimo kärsin, niin se ei liikuta sinua; jos murhe murtaa sydämmeni
ja sieluni, niin sinulle on se samantekevä.
NELJÄS KOHTAUS.
Evert. Niin!
VIIDES KOHTAUS.
Mimmi. Voin nyt vallan hyvin ja kiitän sinua vieläkin siitä, että otit
minut kerran mukaasi. Tämä matka on antanut minulle paljon huvia.
KUUDES KOHTAUS.
Remunen. Ja kun minä nain toisen kerran, niin ajattelin minä: nyt
ei sellaista enää tapahdu! Ensimmäiseen vaimoon nähden täytyi
minun aina olla alakynnessä ja sain täyttää kaikki hänen oikkunsa. —
Kun nyt tiesin keinon, niin ajattelin itsekseni: nyt rupean minä
puolestani sitä käyttämään. Siinä koko temppu. — Nyt pidän minä
kotiripityksiä vaimolleni.
SEITSEMÄS KOHTAUS.
KAHDEKSAS KOHTAUS.
YHDEKSÄS KOHTAUS.
Evert. Minulle?
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade
Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.
textbookfull.com