Você está na página 1de 151

Dynamic Analysis of Bridge Girders Subjected to Moving Loads

Numerical and Analytical Beam Models Considering Warping Effects

João Manuel Alves Serra

Thesis to obtain the Master of Science Degree in

Civil Engineering

Supervisors: Ricardo José de Figueiredo Mendes Vieira, PhD


Francisco Baptista Esteves Virtuoso, PhD

Examination Committee
Chairperson: Fernando Manuel Fernandes Simões, PhD
Supervisor: Ricardo José de Figueiredo Mendes Vieira, PhD
Members of the Committee: Luís Manuel Coelho Guerreiro, PhD

October 2014
ii
À minha família.

iii
iv
Resumo

Os tabuleiros de pontes ferroviárias de via dupla são solicitados em torção por cargas significativas. No
caso de se tratarem de vigas de parede fina a resposta estrutural deve ser alvo de especial cuidado nas várias
fases de projecto, nomeadamente no que respeita ao efeito do empenamento das secções transversais.

Na presente dissertação, são desenvolvidos um modelo numérico e um modelo analítico com vista à
realização de análises dinâmicas de pontes de comboios de alta velocidade com tabuleiro contínuo e tendo
em conta o efeito do empenamento das secções.

As equações de equilíbrio acopladas de vigas de parede fina são deduzidas com base na teoria de Vlasov,
aplicando-se para tal o princípio de Hamilton. Estas equações são resolvidas com recurso a uma análise
modal, sendo as frequências próprias e os respectivos modos de vibração determinados através dos
modelos desenvolvidos. Para efeitos de comparação desenvolve-se também um modelo em elementos
finitos de casca com recurso a um software comercial.

No modelo numérico o comportamento estrutural é descrito através de um elemento finito linear, ao qual
se adiciona um grau de liberdade suplementar representativo da deformabilidade por empenamento.
Apresentam-se as matrizes de massa e de rigidez elementares para secções transversais arbitrárias.

O modelo analítico proposto permite resolver a forma homogénea das equações de equilíbrio de forma
exacta. Apresentam-se expressões explícitas para a determinação dos modos de vibração.

Os comboios são modelados como conjuntos de massas móveis excêntricas ao eixo do tabuleiro, sendo
contabilizado o efeito inercial respectivo.

PALAVRAS CHAVE: vigas de parede fina; análise dinâmica; alta velocidade; empenamento; massas
móveis; pontes.

v
vi
Abstract

In the case of railway bridges traversed by eccentric moving loads, important torsion moments are
introduced. Therefore, special consideration must be given to the torsional response of thin walled girders
and warping of the cross sections cannot be neglected in structural analysis and design.

In the present work, a numerical and an analytical beam models to perform efficient dynamic analysis of
continuous high-speed railway bridges are proposed.

By applying the Hamilton’s principle along with the Vlasov’s beam theory the coupled governing
equations of thin walled beams are derived. These equations are solved by considering a modal analysis
along with a direct integration scheme. The natural frequencies and vibration mode shapes of each
problem are determined through the developed models and compared with the results given by a
commercial shell finite element model.

In the developed numerical model, the behaviour of the bridge girder is described by a three-dimensional
Euler-Bernoulli beam finite element, where an additional degree of freedom is considered to describe the
warping displacements. Explicit expressions for the stiffness and the mass element matrices are presented
for arbitrary cross sections.

The proposed analytical model allows to solve the governing equations in an exact sense. The mode
shapes of vibration are calculated by analytical explicit expressions, yielding the exact solution for the
problem of coupled bending-torsional vibrations.

The trains are modelled as series of moving masses acting eccentrically at constant velocity. Thus, the
corresponding inertial effects of the moving loads are considered besides the gravitational force of mass.

KEYWORDS: thin walled beams; dynamic analysis; high-speed; warping; moving masses; bridges.

vii
viii
Acknowledgments
Agradecimentos

I would like to express my gratitude for my supervisors Professor Ricardo Vieira and Professor Francisco
Virtuoso. Their help, confidence and knowledge were a valuable contribution to this dissertation.

To my friends and colleagues, I would thank all the support and patience transmitted throughout the last
years. Thank you all for your motivation and encouragement given.

My deepest thankfulness goes to my family. Obrigado aos meus pais e avós.

ix
x
Contents

Resumo v
Abstract vii
Acknowledgments ix
Contents xi
List of Figures xiii
List of Tables xv
List of Symbols xvii

1 Introduction 1
1.1 The aim of the present work 1
1.2 Layout of the present work 3
1.3 Literature review 3
2 On bending and torsion of thin walled beams 9
2.1 Warping of thin walled beams 9
2.2 Beam kinematics 11
2.3 Variational formulation of the governing equations 16
2.3.1 The governing equations 17
2.4 Change of reference axes 22
3 Development of a finite element solution 27
3.1 The finite element equations 28
3.1.1 Assembly of the element matrices 36
3.1.2 The complete system of equations 39
3.2 The dynamic response 40
3.2.1 Vibration frequencies and mode shapes 40
3.2.2 Viscous damping matrix 41
3.2.3 Direct integration of the governing equations 42
4 Development of an analytical solution 45
4.1 The dynamic response 45
4.1.1 Vibration frequencies and mode shapes 47
4.1.2 Viscous damping 53
4.1.3 Direct integration of the governing equations 56

xi
5 Dynamic response of continuous beams under moving loads 57
5.1 Simply supported beam 59
5.1.1 Undamped free vibration analysis 59
5.1.2 Undamped forced vibration analysis 70
5.2 Continuous beam 72
5.2.1 Undamped free vibration analysis 73
5.2.2 Damped forced vibration analysis 79
6 Conclusions and future developments 85
6.1 Conclusions 85
6.2 Future developments 86
References 89
Appendix I 93
Appendix II 111

xii
List of Figures

Figure 2.1 – vector decomposition of an eccentric load acting in the longitudinal direction of a 10
thin walled I beam.
Figure 2.2 – vector representation of the pair of moments 𝑀0 and rotation 𝜙 of the cross section. 11
Figure 2.3 – general geometry of a beam with a thin walled cross section. 12
Figure 2.4 – relative orientation between the global and the local coordinate systems. 13
Figure 2.5 – geometrical relations between the quantities 𝑟, 𝑠 and 𝜃. 13
Figure 2.6 – cross-sectional displacements and torsion rotation. 14
Figure 2.7 – geometrical meaning of the sectorial coordinate 𝜔 as an area. 15
Figure 2.8 – element loading and internal forces. 19
Figure 2.9 – changes of the position functions 𝑞 and 𝑟 due to a pole 𝑃̅ moving. 22
Figure 2.10 – global coordinate system transformation. 24
Figure 3.1 – thin walled beam element and the corresponding seven nodal generalized 28
displacements referred to the centre of gravity.
Figure 3.2 – (a) linear approximation functions 𝒩𝜉𝑒𝑥 and (b) finite element approximation of the 30
axial displacements 𝜉𝑥𝑒 .
Figure 3.3 – (a) Hermite cubic approximation functions 𝒩𝜂𝑒 and (b) finite element approximation 31
of the generalized displacements 𝜂𝑒 .
Figure 3.4 – the constant average acceleration method. 44
Figure 4.1 – model of the 𝑁-span bridge considered in the analytical formulation 46
Figure 4.2 – roots of the fourth-order polynomial (4.15). 49
Figure 4.3 – vibration mode shapes and inertial forces of the 𝑚- and the 𝑛-th modes. 54
Figure 5.1 – general geometry of the cross sections considered in the practical examples (m). 58
Figure 5.2 – longitudinal model of the single-span bridge considered in the practical examples. 59
Figure 5.3 – vertical bending vibration mode shapes of the double-T cross section girder 61
(including warping).
Figure 5.4 – vertical bending vibration mode shapes of the double-T cross section girder 61
(neglecting warping).
Figure 5.5 – lateral bending-torsional vibration mode shapes of the double-T cross section girder 62
(including warping) (Part 1/2).
Figure 5.6 – lateral bending-torsional vibration mode shapes of the double-T cross section girder 63
(including warping) (Part 2/2).
Figure 5.7 – lateral bending-torsional vibration mode shapes of the double-T cross section girder 63
(neglecting warping) (Part 1/2).
Figure 5.8 – lateral bending-torsional vibration mode shapes of the double-T cross section girder 64
(neglecting warping) (Part 2/2).
Figure 5.9 – vertical bending vibration mode shapes of the box cross section girder (including 65
warping) (Part 1/2).

xiii
Figure 5.10 – vertical bending vibration mode shapes of the box cross section girder (including 66
warping) (Part 2/2).
Figure 5.11 – vertical bending vibration mode shapes of the box cross section girder (neglecting 66
warping).
Figure 5.12 – lateral bending-torsional vibration mode shapes of the box cross section girder 67
(including warping) (Part 1/2).
Figure 5.13 – lateral bending-torsional vibration mode shapes of the box cross section girder 68
(including warping) (Part 2/2).
Figure 5.14 – lateral bending-torsional vibration mode shapes of the box cross section girder 68
(neglecting warping) (Part 1/2).
Figure 5.15 – lateral bending-torsional vibration mode shapes of the box cross section girder 69
(neglecting warping) (Part 2/2).
Figure 5.16 – dynamic influence lines of the 𝑧-acceleration at the track’s midpoint. 71
Figure 5.17 – longitudinal model of the three-span bridge considered in the practical examples 72
(m).
Figure 5.18 – general dimensions and load magnitudes of the high-speed universal trains proposed 72
by [EN 1991-2].
Figure 5.19 – vertical bending vibration mode shapes of the double-T cross section girder. 74
Figure 5.20 – lateral bending-torsional vibration mode shapes of the double-T cross section girder 74
(Part 1/2).
Figure 5.21 – lateral bending-torsional vibration mode shapes of the double-T cross section girder 75
(Part 2/2).
Figure 5.22 – vertical bending vibration mode shapes of the box cross section girder (Part 1/2). 76
Figure 5.23 – vertical bending vibration mode shapes of the box cross section girder (Part 2/2). 77
Figure 5.24 – lateral bending-torsional vibration mode shapes of the box cross section girder (Part 77
1/2).
Figure 5.25 – lateral bending-torsional vibration mode shapes of the box cross section girder (Part 78
2/2).
Figure 5.26 – dynamic influence lines of the 𝑧-displacement at the midpoint of the central span. 79
Figure 5.27 – dynamic influence lines of the 𝜙-rotation at the midpoint of the central span. 80
Figure 5.28 – dynamic influence lines of the 𝑧-acceleration at the midpoint of the central span. 80
Figure 5.29 – dynamic influence lines of the 𝜙-acceleration at the midpoint of the central span. 81
Figure 5.30 – dynamic influence lines of the 𝑧-displacement at the track’s midpoint of the central 81
span.
Figure 5.31 – dynamic influence lines of the 𝑧-acceleration at the track’s midpoint of the central 82
span.
Figure 5.32 – influence of the mass effect on the dynamic influence lines of the 𝑧-displacement at 83
the track’s midpoint of the central span.
Figure 5.33 – influence of the mass effect on the dynamic influence lines of the 𝑧-acceleration at 83
the track’s midpoint of the central span.
Figure 5.34 – period elongation and amplitude decay errors. 84

xiv
List of Tables

Table 5.1 – material and geometrical properties of the cross sections considered in the practical 58
examples.
Table 5.2 – prescribed boundary conditions at the supports of the single-span bridge. 59
Table 5.3 – natural frequencies of vibration of the double-T cross section girder. 60
Table 5.4 – natural frequencies of vibration of the box cross section girder. 65
Table 5.5 – prescribed boundary conditions at the supports. 72
Table 5.6 – natural frequencies of vibration of the double-T cross section girder. 73
Table 5.7 – natural frequencies of vibration of the box cross section girder. 76

xv
List of Symbols

COORDINATE SYSTEMS AND SPECIAL POINTS

(𝑥, 𝑦, 𝑧) global Cartesian coordinate system


(𝑥, 𝑛, 𝑠) local curvilinear coordinate system
(𝑥̂, 𝑛̂, 𝑠̂ ) auxiliary Cartesian coordinate system which origin coincides with the sectorial pole
𝑥̂ sectorial pole axis
𝐶𝐺 , 𝐶𝐺(𝑥𝐶𝐺 , 𝑦𝐶𝐺 , 𝑧𝐶𝐺 ) centre of gravity of the cross section
𝑆𝐶 , 𝑆𝐶(𝑥𝑆𝐶 , 𝑦𝑆𝐶 , 𝑧𝑆𝐶 ) shear centre of the cross section

LATIN LETTERS

𝑎 , 𝑏, 𝑐 , 𝑑positive non-dimenional parameters


𝑐 velocity of the moving loads
𝑒 number of the finite element
𝑒𝑐𝑐 eccentricity of a moving load
𝑓𝑖 linearly distributed applied force in the direction of the 𝑖 axis (𝑖 = 𝑥, 𝑦, 𝑧)
𝑓𝑆 shear flow due to Saint-Venant torsion
𝒇𝑒 element external force vector
𝒇, 𝒇(𝑡) global external force vector
𝕗, 𝕗(𝑡) global external force vector written for the normalized modal coordinates
𝒇𝑒𝛺 element external body force vector
𝒇𝑒𝛤 element external boundary force vector
𝒇𝑒𝒟 contribution to 𝒇𝑒 associated with the 𝒟 motion (𝒟 = 𝜉𝑥 , 𝜉𝑧 , 𝜉𝑦 , 𝜙)
ℎ𝑤 height of an I beam’s web
𝑖 imaginary unit
𝑖, 𝑗, 𝑘, 𝑙 , 𝑚, 𝑛, 𝑟, 𝑠 integer; constant
𝑙𝑠 length of the 𝑠-th span of a continuous beam
𝑚𝑖 linearly distributed applied bending moment about the 𝑖 axis (𝑖 = 𝑦, 𝑧)
𝑚𝑥 linearly distributed applied torsion moment about the 𝑥 axis
𝑚𝜔 linearly distributed applied warping moment

xvii
𝑛 normal coordinate at a generic point
𝑛𝑥 cosine of the angle between the 𝑥 axis and the normal to the boundary
𝑛𝑑𝑜𝑓 number of degrees of freedom
𝑛𝑒𝑙 number of elements of a finite element mesh
𝑛𝑙𝑜𝑎𝑑 number of loads
𝑝 circular frequency
𝑝𝑛 𝑛-th natural frequency of vibration
𝑞𝑖 𝑖-th generalized coordinate of a discrete system
𝑞 , 𝑞(𝑠) 𝑠̂ coordinate at a generic point placed in the centreline of a cross section
𝑟, 𝑟(𝑠) 𝑛̂ coordinate at a generic point placed in the centreline of a cross section
𝑠 tangential coordinate at a generic point
𝑡, 𝑡(𝑠) wall thickness
𝑡, 𝑡𝑖 time; instant of time
𝑢𝑛 (𝑥, 𝑠) normal in-plane displacement of a point at the centreline of a cross section
𝑢𝑛 (𝑥, 𝑛, 𝑠) normal in-plane displacement of a generic point
𝑢𝑠 (𝑥, 𝑠) tangential in-plane displacement of a point at the centreline of a cross section
𝑢𝑠 (𝑥, 𝑛, 𝑠) tangential in-plane displacement of a generic point
𝑢𝑥 (𝑥, 𝑠) out-of-plane displacement of a point at the centreline of a cross section
𝑢𝑥 (𝑥, 𝑛, 𝑠), 𝑢𝑥 (𝑥, 𝑦, 𝑧) out-of-plane displacement of a generic point
𝑢(𝑥, 𝑡) displacement field variable (𝑢 = 𝜉𝑥 , 𝜉𝑧 , 𝜉𝑦 , 𝜙)
𝑢𝒩 (𝑥, 𝑡) polynomial approximation of the solution 𝑢(𝑥, 𝑡) of each displacement field variable
𝑢𝑗 (𝑡) 𝑗-th nodal generalized displacement
𝒖, 𝒖(𝑥, 𝑛, 𝑠) displacement vector of a generic point
𝒖, 𝒖(𝑡) nodal displacements vector
𝒖𝑡 nodal displacements vector at time 𝑡
𝒖𝑒 , 𝒖𝑒 (𝑡) element nodal displacements vector
𝒗𝑛 , 𝒗𝑛 (𝑥) 𝑛-th mode shape of vibration
𝑥 longitudinal space coordinate of a generic point
𝑥𝑒 longitudinal space coordinate of the 𝑒-th global node
𝑥𝑗𝑒 longitudinal space coordinate of the 𝑗-th node of the 𝑒-th finite element (𝑗 = 1,2)
𝑥𝑘 (𝑡) position of the 𝑘-th load at time 𝑡
𝑦, 𝑦(𝑛, 𝑠) space coordinate of a generic point
𝑦(𝑠) 𝑦 coordinate of a generic point placed in the centreline of a cross section
𝑧, 𝑧(𝑛, 𝑠) space coordinate of a generic point
𝑧(𝑠) 𝑧 coordinate of a generic point placed in the centreline of a cross section

𝐴 cross-sectional area
𝐴𝑚 enclosed area
𝒜𝑗 , ℬ𝑗 , 𝒞𝑗 , 𝒰𝑗 𝑗-th constant
𝐵(𝑥, 𝑦, 𝑧) generic point
𝐵(𝑥̂, 𝑟, 𝑞), 𝐵 (𝑥, 𝑛, 𝑠 ) generic point at the centreline of a cross section

xviii
origin of the global Cartesian coordinate system
𝐶(𝑥𝐶 , 𝑦𝐶 , 𝑧𝐶 )
𝑪 element damping matrix
𝑒

𝑪 global damping matrix


ℂ global damping matrix written for the modal coordinates
𝐶𝑧𝑛 , 𝐶𝑦𝜙𝑛 generalized damping coefficients associated with the 𝑛-th mode of vibration
𝐷 differential operator
𝐸 Young’s modulus
𝓕𝑡 effective loading vector at time 𝑡
ℱ𝑡 effective modal loading at time 𝑡
𝐺 shear modulus
Η Heaviside step function
𝐽 torsion constant
𝐼𝑖𝑗 second moment of inertia (𝑖, 𝑗 = 𝑦, 𝑧, 𝜔)
𝐼𝑖𝑖𝑃 second moment of inertia referred to the sectorial pole 𝑃 (𝑖 = 𝑦, 𝑧)
𝐼𝜔𝜔 warping constant
𝑲𝑒 element stiffness matrix
𝑲 global stiffness matrix
𝑲𝐺 global stiffness matrix written for the modal coordinates
𝕂 global stiffness matrix written for the normalized modal coordinates
𝓚 effective stiffness matrix
𝒦 effective modal stiffness
𝑲𝑒𝒟 contribution to 𝑲𝑒 associated with the 𝒟 motion (𝒟 = 𝜉𝑥 , 𝜉𝑧 , 𝜉𝑦 , 𝜙)
𝐿 length
𝑀𝑖 internal bending moment about the 𝑖 axis (𝑖 = 𝑦, 𝑧)
𝑀𝑥 internal torsion moment
𝑀𝑥𝑆 Saint-Venant torsion moment
𝑀𝑥𝜔 warping torsion moment contribution
𝑀𝜔 internal warping moment; bimoment
𝑀0 pair of moments that constitute the bimoment
𝑴𝑒 element mass matrix
𝑴 global mass matrix
𝑴𝐺 global mass matrix written for the modal coordinates
𝕄 global mass matrix written for the normalized modal coordinates
𝑴𝑒𝒟 contribution to 𝑴𝑒 associated with the 𝒟 motion (𝒟 = 𝜉𝑥 , 𝜉𝑧 , 𝜉𝑦 , 𝜙)
𝑀𝑧𝑛 , 𝑀𝑦𝜙𝑛 generalized mass associated with the 𝑛-th mode of vibration
𝑁 internal axial force
𝓝𝑒𝒟 element approximation functions matrix (𝒟 = 𝜉𝑥 , 𝜉𝑧 , 𝜉𝑦 , 𝜙)
𝒩𝑗 (𝑥) 𝑗-th
linear independent polynomial approximation function
𝑂(𝑥𝑂 , 𝑦𝑂 , 𝑧𝑂 ) sectorial origin of the curvilinear coordinate 𝑠
𝑃(𝑥𝑃 , 𝑦𝑃 , 𝑧𝑃 ) sectorial pole
𝑃𝑧𝑛 (𝑡), 𝑃𝑦𝜙𝑛 (𝑡) generalized loading associated with the 𝑛-th mode of vibration
𝑄 eccentric concentrated load acting in the beam’s axial direction

xix
𝑅(𝑠) radius of curvature of the centreline
𝑆𝑖 first moment of inertia (𝑖 = 𝑦, 𝑧, 𝜔)
𝑆𝑖𝑃 first moment of inertia referred to the sectorial pole 𝑃 (𝑖 = 𝑦, 𝑧)
𝑆𝜔 first sectorial moment
𝑉 volume
𝑉𝑖 internal shear force in the direction of the 𝑖 axis (𝑖 = 𝑦, 𝑧)
𝑽 mode shape matrix

GREEK LETTERS

𝛼 angle between the 𝑦 and 𝑧 global axes and the 𝑦̅ and 𝑧̅ global axes, respectively
𝛼, 𝛽 , 𝛾, 𝛿 real values
𝛾, 𝛾𝑥𝑠 (𝑥, 𝑛, 𝑠) shear strain over the wall thickness
𝛾𝑥𝑠 (𝑥, 𝑠), 𝛾𝑠𝑛 (𝑥, 𝑠) shear strain evaluated at a centreline’s point
𝑆
𝛾𝑥𝑠 Saint-Venant shear strain
𝜔
𝛾𝑥𝑠 shear strain due to warping
𝛿 variation
δ(𝑡) Dirac delta function
𝛿𝑠𝑟 Kronecker delta
𝜀 , 𝜀𝑥𝑥 (𝑥, 𝑛, 𝑠) normal strain over the wall thickness
𝜁 element natural coordinate
𝜂(𝑥, 𝑡) displacement field variable (𝜂 = 𝜉𝑦 , 𝜉𝑧 , 𝜙)
𝜂𝑗 𝑗-th root of a polynomial
𝜃, 𝜃(𝑠) angle between the 𝑛 and 𝑠 local axes and the 𝑦 and 𝑧 global axes, respectively
𝜗 weight function
𝜅𝑗 constant of proportionality between ℬ𝑖 and 𝒞𝑖 (𝑗 = 𝛼, 𝛽, 𝛾, 𝛿)
𝜆 frequency parameter
𝜉𝑖 , 𝜉𝑖 (𝑥), 𝜉𝑖 (𝑥, 𝑡) 𝑖-th translation component (𝑖 = 𝑥, 𝑦, 𝑧)
𝜎𝑥𝑥 longitudinal normal stresses
𝜔
𝜎𝑥𝑥 longitudinal normal stresses due to cross-sectional warping
𝜍𝑛 viscous damping ratio associated with the 𝑛-th mode of vibration
𝜌 mass per unit volume
𝜏𝑥𝑠 in-plane shear stresses
𝑆
𝜏𝑥𝑠 shear stresses due to Saint-Venant torsion
𝜔
𝜏𝑥𝑠 shear stresses due to cross-sectional warping
𝜙, 𝜙(𝑥), 𝜙(𝑥, 𝑡) rotation about the 𝑥̂ axis; twist angle
𝝍𝑛 , 𝝍𝑛 (𝑥) 𝑛-th normalized mode shape of vibration
𝜔(𝑠) sectorial coordinate of a point at the centreline of a cross section
𝜔 , 𝜔(𝑛, 𝑠) sectorial coordinate of a generic point

xx
𝛤 centerline; boundary
𝛤ℒ static boundary (ℒ = 𝑁, 𝑉𝑧 , 𝑉𝑦 , 𝑀𝑦 , 𝑀𝑧 , 𝑀𝑥 , 𝑀𝜔 )
𝛤𝒟 kinematic boundary (𝒟 = 𝜉𝑥 , 𝜉𝑧 , 𝜉𝑦 , 𝜉𝑧′ , 𝜉𝑦′ , 𝜙, 𝜙 ′ )
𝛬 kinetic energy of a body
𝛯𝑖 , 𝛯𝑖 (𝑥) shape function of the 𝑖-th translation component (𝑖 = 𝑦, 𝑧)
𝛱 total potential energy of a body
𝛱𝑖 internal strain energy
𝛱𝑒 potential energy of the external applied loading
𝛶𝑛 , 𝛶𝑛 (𝑡), 𝛹𝑛 , 𝛹𝑛 (𝑡) modal amplitude of the 𝑛-th mode shape
𝛷, 𝛷(𝑥) shape function of the rotation about the 𝑥̂ axis
𝛹, 𝛹(𝑥), 𝛹𝑖 (𝑥) torsion variable for closed profiles
𝜳 normalized mode shape matrix
𝛺 one-dimensional domain

xxi
xxii
1
Introduction

T he expansion of the european railway network in the last few years was funded by many countries,
being the high-speed trains seen nowadays as a competitive mean of transport. This gives rise to an
increasingly need to provide efficient design methods that take into account the dynamic effect of high-
speed loading, with particular interest to railway bridges. The models traditionally considered in design
codes use dynamic impact factors to simulate the expected dynamic amplification of the response.
However, these models do not take into account resonance effects, which can be caused by arrays of
moving loads.

Thin walled beam-like structural elements have been efficiently adopted in civil engineering given that
weight economy and cost are design prime variables. Continuing developments and improvements in
design and structural materials, such as in steel and high-strength concrete, enable the construction of
bridges with more slender girders, which increases their vulnerability to moving loads.

In the case of railway bridges traversed by eccentric moving loads, important torsion moments are
introduced. Therefore, special consideration must be given to the torsional response of thin walled girders
and warping of the cross sections cannot be neglected in structural analysis and design.

1.1 THE AIM OF THE PRESENT WORK

Performing dynamic analyses of bridges with shell finite element software allows to obtain accurate
results, as long as a detailed description of the geometrical and material properties of the structure is
known. In general, using such models at an early design stage is time consuming, which renders the
evaluation of several structural solutions cumbersome. The use of beam models can minimize the time

1
needed to perform a dynamic analysis and provide good agreement with the results of a shell finite
element commercial software.

The present work aims to propose a numerical and an analytical beam models to perform the dynamic
analysis of multi-span high-speed railway bridges. The behaviour of the bridge girder is described by a
three-dimensional Euler-Bernoulli beam element, where an additional degree of freedom is considered at
each beam end to describe the warping of the cross sections. The warping displacements are determined
according to the Vlasov’s beam theory for thin walled open sections and calculated to closed cross
sections according to the von Kármán theory. The trains are modelled as series of moving point loads
acting eccentrically at constant velocity. The inertial effects of the moving loads are investigated by
considering the corresponding masses.

The dynamic equilibrium equations are solved by considering a modal analysis. The natural frequencies
and vibration mode shapes of each problem are determined through the two different developed beam
models. On the one hand, it is applied the finite element method, so widely used in the dynamic analysis
of structures. The Hermite’s polynomial interpolation is used and explicit expressions for the stiffness, the
damping and the mass matrices are presented for arbitrary cross sections. On the other hand, the same
goal is achieved in an exact sense by deriving an analytical solution for the governing differential equations
of the beam element. The mode shapes of vibration are calculated by analytical explicit expressions,
yielding the exact solution for the free vibration problem. The corresponding dynamic response due to the
travelling loads is obtained by direct integration of the governing equations, being used the Newmark’s β
method.

The dynamic analyses performed by the developed thin walled beam models are compared with a shell
model implemented in the finite element commercial software SAP2000. Dynamic influence lines for the
torsion rotation, the vertical and the lateral displacements of the midspan point for different train
velocities are presented.

The time needed to define the bridge and the loading properties and run the beam models does not
exceed a couple of minutes in a common portable computer. This is substantially less in comparison with
the time needed to model the same structure in a commercial software and get ready to perform a similar
dynamic analysis, i.e., using a shell finite element model.

2
1.2 LAYOUT OF THE PRESENT WORK

This dissertation is organized in 6 chapters.

In the current chapter, a general introduction to the present work is made. A literature review is presented
with the focus on previous works done in dynamic analysis of beams under free and forced vibrations and
its application for engineering purposes.

In Chapter 2, the importance of including the cross-sectional warping in the kinematic description of thin
walled structural elements is stated. The theories developed by Vlasov and von Kármán for thin walled
beams, respectively for open and closed cross sections, are briefly reviewed. A complete kinematical
description of thin walled beams is provided and the coupled governing differential equations are derived
through an energetic approach.

A numerical solution for the governing equations is deduced in Chapter 3. A beam finite element model is
developed in order to consider the cross-sectional warping. The element matrices are derived and the
global finite element equations obtained through a direct assembly procedure. In Chapter 4, an analytical
solution of the governing equations is derived, being presented expressions to determine the natural
frequencies and mode shapes of vibration of continuous beams. The dynamic response due to eccentric
moving loads is obtained considering a modal analysis along with the direct integration method, being
used the Newmark’s β method.

In Chapter 5, the developed beam models are firstly used to perform the dynamic analysis of a single-span
bridge subjected to a single eccentric load. The results obtained are compared with the Fryba’s analytical
solution. Afterwards, the dynamic response of a multi-span bridge due to the passage of high-speed trains
acting eccentrically is investigated. The dynamic influence lines of the accelerations and displacements of
the midspan point are obtained, and the mass effect of the train is discussed.

In Chapter 6, general remarks on the presented work are made and conclusions are drawn.

1.3 LITERATURE REVIEW

In this section, the accredited knowledge related to the main concepts of the present work is summarized.

Thin Walled Beams

The structural behaviour of thin walled beams has been a topic of investigation for many years. The
general theory for thin walled beams of open cross section was derived by [Vlasov, 1961], including
already the dynamic analysis of such structures. [Gjelsvik, 1981] used the theory of cylindrical shells, see

3
[Flügge, 1966], to formulate the same static equilibrium equations as Vlasov, and [Ahmad, 1984] used the
same theory to include the dynamic effect in the equilibrium equations. [Von Kármán and Christensen,
1994] developed an approximate theory for thin walled beams of closed sections. The Vlasov’s governing
equations were derived along with a modified warping parameter.

Although firmly established and extensively used, the Vlasov’s theory considers that shear strains do not
contribute to the beam flexibility. This can lead to important errors in dynamic analysis associated with
higher modes of vibration, even in the case of slender beams, [Timoshenko et al., 1974]. [Tso, 1965]
extended Vlasov’s differential free vibration equations to include shear deformations. In what closed cross
sections is concerned, shear deformations were considered by [Benscoter, 1954] in his thin walled beam
theory. The author characterizes the warping degree of freedom by an independent function different
from the gradient of the twist angle.

Free and Forced Vibration Problems

The vibration analysis of beam-like structural elements is a fundamental subject in structural analysis,
embracing a wide class of engineering problems. Depending on the assumptions, the kind of analysis and
the loading properties, numerous publications on free and forced vibrations have been reported in the
literature.

A uniform beam is characterized by three parallel axes through the centre of mass, the shear centre and
the geometric centre of its cross section. When these axes do not coincide coupling of bending, torsional
and tensional vibrations occurs. In the present work, only homogeneous cross-sectional beams, i.e., with
coincident mass and geometric axes, are considered, being the tensional vibrations decoupled.

Previous work related to free vibrations of homogeneous beams can be classified into three groups
depending on the cross-sectional symmetry, which influences the locations of the centre of mass and the
shear centre.

For uniform beams possessing double symmetric cross sections, the centre of mass and the shear centre
are coincident, being the bending and the torsional displacements decoupled. Several theoretical
approaches have been proposed to determine exact natural frequencies and the associated mode shapes of
vibration of Euler-Bernoulli beams, e.g. [Bishop and Johnson, 1979]. The effect of warping in free
decoupled torsional vibrations of thin walled beams of double symmetric cross sections is considered by
[Gere, 1954] by solving exactly the differential governing equations.

In the case of beams having single cross-sectional symmetry, torsional vibrations are coupled with
bending vibrations in the perpendicular direction of the axis of symmetry. In this field, [Mei, 1970] used
the finite element method with beam elements to determine the coupled natural frequencies of thin walled
beams of open cross section. A consistent mass matrix based on cubic polynomials as approximation
functions was evaluated and the frequency results were compared with experiments. The dynamic
stiffness method was used by [Hallauer and Liu, 1982] to obtain exact natural frequencies of a cantilever

4
beam. Unlike the finite element method, where mass and stiffness matrices are separately derived to
enable vibration analysis of coupled systems to be made, the dynamic stiffness matrix method relies on
one single frequency-dependent matrix, which combines both mass and stiffness properties. The accuracy
of this exact beam theory can be combined with the finite element technique to solve single or continuous
beams. Using this theory, long physical beams do not need to be modelled as an assemblage of short
mathematical beams to obtain sufficient accuracy in the performed analysis as it often does when the
standard finite element method is used. The Euler-Bernoulli beam theory was used in this work and thus
no allowance is made for the cross-sectional warping. Large errors in the calculation of natural frequencies
may be incurred, particularly when thin walled beams of open cross section are considered. [Bishop et al.,
1989] demonstrated this fact by extending the exact study of [Dokumaci, 1987] to include the effect of
warping in the vibration analysis of cantilever beams having single cross-sectional symmetry. To this end,
Bishop et al. solved exactly the governing differential equations to determine the free coupled bending-
torsional vibration frequencies.

For arbitrary cross sections coupling between torsional vibrations and bending vibrations in two
perpendicular directions occurs. This situation was termed triple coupling by [Timoshenko et al., 1974].
[Gere and Lin, 1958] considered the triple coupled free vibrations of thin walled single-span beams with
various boundary conditions by employing the Rayleigh-Ritz technique with decoupled modes as
coordinate functions. Considering the Euler-Bernoulli beam theory, [Friberg, 1983] and [Banerjee, 1989]
proposed numerical procedures that lead to the so-called dynamic stiffness matrix. To achieve exact
results on coupled vibrations successive matrix operations are performed. [Friberg, 1985] presented a
numerical procedure to derive exact dynamic stiffness matrices of beams possessing no cross-sectional
symmetry, based on the Vlasov’s beam theory. However, no explicit expressions for the dynamic matrix
elements are available in this work. Later, [Banerjee et al., 1996] extended the previous approach of
[Banerjee, 1989] to make allowance for warping stiffness. By making use of symbolic algebra software,
Banerjee investigates the effect of warping on natural frequencies of thin walled open cross section beams
for several boundary conditions. The author yielded algebraic expressions for the dynamic matrix
elements of coupled beams, which fills the gap left by Friberg. [Tanaka and Bercin, 1997] presented a
beam finite element formulation capable of determining free vibrational frequencies of asymmetric cross
section beams including warping stiffness and rotary inertia of the cross sections. Explicit expressions for
the stiffness and mass matrices are presented. Later, the same authors, [Tanaka and Bercin, 1999], studied
triple coupling of open cross section uniform beams using the symbolic algebra software Mathematica. As
an extension of the approach of [Bishop et al., 1989], the authors presented a closed form solution
procedure for the boundary value problem involving Vlasov’s coupled differential equations for beams of
asymmetric cross section.

The study of the dynamic behaviour of structures that can be simulated as beams under the action of
moving forces or masses, such as bridges on railways, has been greatly enhanced since the very first
construction of railway bridges. The first dynamic analyses of structures acted upon by moving loads

5
involved two extreme approximations in modelling the physical problem: the inertial effects of either the
load or the structure were neglected. It were [Stokes, 1849] and [Zimmermann, 1896] who first studied
and approximately solved the problem of forced vibrations of structures with negligible mass, compared
to the mass of a single moving load of constant magnitude. [Krylov, 1905], [Timoshenko, 1908] and
[Lowan, 1935] considered the case of a moving load with negligible mass, compared to the mass of the
girder, commonly called a moving force approximation problem. A closed form solution can be derived
when the force travels with constant velocity. Analytical solutions of the complete problem, i.e., when
both inertial effects are taken into account, were derived by [Inglis, 1934] and [Biggs et al., 1959].
[Hillerborg, 1951] used the Fourier’s method to evaluate the dynamic response of simply supported
beams.

More recently, modern computational techniques have overcome some limitations and simplifications of
the previous studies, since only simple structures were amenable to analytical treatment. The principle of
virtual displacements and the variational principles enable the development of methods that are suitable
for automatic computation. As a consequence, more precise and complex models are possible to be
developed. The finite element method enables the discretization of complicated structures into small
mathematical elements, and combined with certain numerical integration schemes in the time domain
repeated finite element analyses at each time step can be performed in order to obtain dynamic responses.

The problem of the dynamic response of bridges subjected to moving loads was reviewed in detail by
several authors, namely [Timoshenko, 1953], [Kolousek, 1956] and [Fryba, 1972]. Analytical solutions for
simple problems of simply supported and continuous beams of uniform cross section are presented in
Fryba’s book. Extensive references to the literature on the subject can also be found. Based on his text,
[Fryba, 1976] and [Fryba, 1980] studied the effect of speed and damping on the beam’s dynamic response.
Later, the author published another monograph with numerous reports on bridge dynamics therein,
[Fryba, 1996], most of them treating uniform simply supported beams.

Vehicle-bridge interaction problems were recently reviewed by [Yang et al., 2004]. [Olsson, 1991]
presented analytical and finite element solutions of a simply supported beam under a constant force
travelling at constant velocity. His analytical solution consists on a series solution, similar to the one given
by [Fryba, 1972]. [Cai et al., 1994] investigated the dynamic interaction between train and guideway of a
Maglev system by modelling the vehicle as a two-degree of freedom model. Numerous papers discuss
beam vibration excitation by a moving mass, e.g. [Michaltsos et al., 1996] who obtained a closed form
solution for a single-span bridge under the action of a moving mass of constant magnitude and velocity. A
series solution for beam deflection is derived by using the modal superposition method and by
considering as a first approximation the solution of the corresponding vibration problem without the
effect of the mass. The influence of mass effect on the dynamic response of single-span beams under
moving loads was also studied by [Foda and Abduljabbar, 1998] by considering the method of dynamic
Green function. Their study extended the one of [Ting et al., 1974] that considered a static Green
function. The possibility of separation of the moving mass from the beam during the motion was

6
investigated by [Lee, H.P., 1996(1)] and [Lee, H.P., 1996(2)] for Euler-Bernoulli and Timoshenko single-
span beams, respectively. [Zhu and Law, 2001] have used the Hamilton’s principle to determine the
dynamic response of a continuous beam under moving loads. To this end, the Newmark time integration
method was used and high precision results achieved.

Reports on dynamic problems of multi-span bridges are limited. [Yang et al., 1995] proposed impact
formulas for vehicles traversing simple and continuous bridges. [Lee, H.P., 1996(3)] investigated the
dynamic response of multi-span Euler-Bernoulli beams subjected to a moving mass. [Ichikawa et al., 1999]
performed a similar study but disregarding the mass effect. Nevertheless, using the modal analysis, the
effects of acceleration or deceleration of the moving load were estimated on the impact factor for a
symmetric three-span beam. Later, [Ichikawa et al., 2000] used the modal analysis to study continuous
beams traversed by a moving mass at constant velocity. The solution is found through the direct
integration method applied to the coupled ordinary differential equation of second-order for the
generalized coordinate in which the Euler-Bernoulli equation of transverse motion is transformed.
Amplification factors were evaluated for different magnitudes and velocities of the travelling mass. Non-
uniform Euler-Bernoulli continuous beams under moving loads were studied by [Dugush and
Eisenberger, 2002]. The modal analysis and the direct integration method were used to obtain exact results
on natural frequencies and mode shapes of vibration by deriving the exact dynamic stiffness matrices
along the beams.

Various dynamic problems caused by moving loads are briefly overviewed by [Ouyang, 2011], such as
vehicle-bridge interaction, train-track interaction, separation and reattachment of the moving vehicles
onto the supporting structure, and structural damage identification and health monitoring by vehicle-
induced vibrations.

The moving load problem on beams resting on elastic foundation is investigated by [Fryba et al., 1993]
and [Thambiratnam and Zhuge, 1996], and the identification problem associated with the moving force
may be found in [Law et al., 1997] and [Chan and Ashebo, 2006]. Closed form solutions of the response
of Euler-Bernoulli beams with general boundary conditions traversed by a single deterministic moving
force with different types of motion are derived by [Hilal and Zibdeh, 2000]. Several damping cases are
analysed and comparisons with known solutions are made.

The dynamic behaviour of long-span box girder bridges was evaluated by [Lee, S.Y., and Yhim, 2005],
having been carried out numerical and experimental studies. A multi-span simply supported prestressed
concrete bridge located on the high-speed railway line between Brussels and Paris was experimentally
tested by [Xia et al., 2003]. The moving loads were from high-speed Thalys trains.

Coupled Beams under Moving Loads

The number of works on coupled lateral bending-torsional vibrations of beams under moving loads is
rather limited. Relevant studies are those of [Sophianopoulos and Michaltsos, 1999], [Stayridis and

7
Michaltsos, 1999] and [Michaltsos et al., 2005]. In the last work, a numerical study on the subject is
presented, being considered simply supported thin walled open cross section steel beams traversed by
single eccentric moving loads. The analysed structures simulate highway bridges with eccentric lanes.
Simplified single vehicles are taken into account and the structural damping is neglected. Through a modal
analysis, the free and forced vibrations are investigated by utilizing symbolic algebra software.

Recently, [Lisi, 2011] developed a beam finite element model for the static and the dynamic analysis of
thin walled beams under the action of single eccentric moving forces. His work constitutes a powerful
contribution to the present work, which aims to extend Lisi’s study in order to include sets of moving
masses.

8
2
On Bending and
Torsion of
Thin Walled Beams

I n this chapter, the behaviour of straight, prismatic, thin walled beams is introduced. The cross
section’s geometry is arbitrary, being the wall thickness 𝑡(𝑠) considered to be constant along the
longitudinal axis of the structural element.

The kinematical description of thin walled beams is presented according to the theories developed by
Vlasov and von Kármán, respectively for open and closed cross sections. Afterwards, the corresponding
governing differential equations are derived through an energetic approach. Some simplification of these
coupled equations is finally made by changing the position and orientation of the reference axes.

2.1 WARPING OF THIN WALLED BEAMS

The cross-sectional warping is a distinctive behaviour of thin walled structures, which are extensively used
in civil engineering. The practical importance of this phenomenon has led to a need of provide an
understanding of the static and the dynamic behaviour of such structures.

Since the cross sections of a thin walled beam subjected to non-uniform torsion experience different
warping, the torsional stiffness of the element has two contributions: one from the classical Saint-Venant
torsional stiffness, proportional to the shear modulus 𝐺 , and another contribution due to constraining the

9
warping of the cross sections. Therefore, when a thin walled beam is loaded in torsion with a variable rate
of twist not only Saint-Venant shear stresses 𝜏𝑥𝑠
𝑆
, but also normal stresses 𝜎𝑥𝑥
𝜔
in the longitudinal direction
and a second set of shear stresses 𝜏𝑥𝑠
𝜔
arise in the cross sections. These two warping stresses, say 𝜎𝑥𝑥
𝜔
and
𝜔,
𝜏𝑥𝑠 can play an important role since they can be significant in relation to the bending stresses, and so they
cannot be neglected.

The non-uniform warping of the cross sections in the longitudinal direction results in a bimoment 𝑀𝜔 .
The practical significance of a bimoment can be understood by considering an I-shaped beam carrying an
eccentric load 4𝑄 acting parallel to the longitudinal axis, as shown in Figure 2.1. This load can be
decomposed into four sets of loads, which represent axial loading, bending moments, and the bimoment.

The bimoment 𝑀𝜔 can be represented by a pair of moments 𝑀0 of equal magnitude, acting about the same
axis with opposite orientation and separated by a distance equal to the height of the beam ℎ𝑤 :

𝑀𝜔 = ℎ𝑤 𝑀0 . (2.1)

(a) – longitudinal (b) – bending about (c) – bending about (d) – warping
force 𝑦 axis 𝑧 axis

Figure 2.1 – vector decomposition of an eccentric load acting in the longitudinal direction of a thin walled I beam.

By separating the flanges and the web from one another, and loading the free end of the beam with the
two moments 𝑀0 each flange will bend separately, as illustrated in Figure 2.2. To restore the integrity of
the beam, both flanges and web have to be twisted. This can be obtained by setting a Saint-Venant torsion
moment 𝑀𝑥𝑆 or a rotation 𝜙. Thus, as the web is flexible in torsion, the cross section is permitted to exhibit
out-of-plane displacements, and the Euler-Bernoulli hypothesis of plane cross sections after deformation
is no longer valid. This implies that the moment vectors 𝑀0 cannot be added vectorially and their effects
must be considered in the kinematical description of a thin walled beam.

The general theory of thin walled beams of open cross section was developed by [Vlasov, 1961], being
considered an extension of the Euler-Bernoulli beam theory to include non-uniform torsion and the
corresponding warping displacements. The cross sections are therefore permitted to warp, being however
assumed to be in-plane rigid and with null distortion 𝛾𝑥𝑠 in the middle surface.

10
Figure 2.2 – vector representation of the pair of moments 𝑀0 and rotation 𝜙 of the cross section.

An approximate theory for closed profiles was developed by [von Kármán and Christensen, 1944], having
the same resulting governing equations as the Vlasov’s theory along with different section properties. The
crucial hypothesis for closed profiles is that, having a variable torsion moment, the distortions 𝛾𝑥𝑠
𝜔 due to

warping are small and can be neglected. This implies that the associated warping shear stresses 𝜏𝑥𝑠
𝜔 cannot

be determined from the corresponding constitutive relation


𝜔 = 𝐺𝛾 𝜔 ,
𝜏𝑥𝑠 𝑥𝑠 (2.2)
and an equilibrium condition in the longitudinal direction must be considered, [Murray, 1984].

Some hollow sections experience large warping distortions 𝛾𝑥𝑠


𝜔 , which means that the shear stresses 𝜏 𝜔 also
𝑥𝑠

have to be considered. A more accurate theory is required, such as the one introduced by [Benscoter,
1954].

2.2 BEAM KINEMATICS

The beam governing differential equations are derived in the next sections considering a variational
approach. To this end, the potential and the kinetic energies associated with the displacement field of a
three-dimensional beam element are defined.

In this regard, a complete description of the beam kinematics is made in the present section. The middle
surface of the thin walled beam is cylindrical, as represented in Figure 2.3, being considered two different
types of right-handed coordinate systems: one global Cartesian coordinate system (𝑥, 𝑦, 𝑧), positioned with
the 𝑥 axis parallel to the middle surface’s axis, and one local curvilinear coordinate system (𝑥, 𝑛, 𝑠) placed in
the middle surface. A plane normal to the 𝑥 axis cuts the middle surface in a line called the centreline of
the cross section.

At first, both coordinate systems are placed arbitrarily, and then some changes in the reference axes will
be made in order to simplify the governing equations.

11
Figure 2.3 – general geometry of a beam with a thin walled cross section.

The present formulation for bending and torsion of thin walled beams considers the theory of linear
elasticity, being considered the following assumptions:

(i) the cross sections are in-plane rigid, i.e., the distortion 𝛾𝑠𝑛 is small enough to be disregarded. In
general, this assumption is made in non-uniform torsion theories of thin walled bars and its
accuracy depends on the wall thickness, the shape of the centreline, the existence of rigid
diaphragms and on the applied loading;
(ii) the only significant stresses are the longitudinal normal stresses 𝜎𝑥𝑥 and the in-plane shear stresses
𝜏𝑥𝑠 . As far as open profiles is concerned, it is admitted the assumption introduced by [Vlasov,
1961], which states that the distortion 𝛾𝑥𝑠 is null at the middle surface. In a more general way, it is
assumed that the shear strain 𝛾𝑥𝑠 has the same distribution in the centreline as in the Saint-Venant
torsion, that holds for both open and closed profiles;
(iii) the walls behave as thin shells, being considered the Kirchhoff assumption, which states that
straight lines remain normal to the middle surface during a deformation;
(iv) as a first-order theory the equilibrium is based on the undeformed geometry.

Let (𝑥̂, 𝑛̂, 𝑠̂ ) be an auxiliary Cartesian coordinate system with its origin in the so-called sectorial pole
𝑃(𝑥𝑃 , 𝑦𝑃 , 𝑧𝑃 ). The 𝑥̂ axis is referred to as the sectorial pole axis, being 𝑛̂ and 𝑠̂ parallel to the 𝑛 and 𝑠 local
axes, respectively, at a generic point 𝐵(𝑥̂, 𝑟, 𝑞) placed in the centreline of a cross section, as shown in Figure
2.4. The 𝑛 and 𝑠 local axes make an angle 𝜃(𝑠) − 𝜋/2 with the 𝑦 and 𝑧 global axes, where 𝜃(𝑠) is measured
with respect to 𝑦 and 𝑠, and denotes the relative orientation between both coordinate systems. Notice that
the local curvilinear coordinate 𝑠 is measured around the profile following the centreline from a sectorial
origin 𝑂(𝑥𝑂 , 𝑦𝑂 , 𝑧𝑂 ).

In general, the kinematic quantities are associated with the middle surface. When any ambiguity is
possible, the independent variables of a function are given explicitly.

As the middle surface is cylindrical the quantities 𝑟, 𝑞 and 𝜃 can be defined as functions of 𝑠:

𝑟 = 𝑟(𝑠) , 𝑞 = 𝑞(𝑠) and 𝜃 = 𝜃(𝑠) . (2.3)

12
Figure 2.4 – relative orientation between the global and the local coordinate systems.

Considering two neighboring points 𝐵 and 𝐵′ placed a distance 𝑑𝑠 apart, see Figure 2.5, the following
relations are established:

𝑑𝑟 = 𝑞𝑑𝜃 , (2.4a)
𝑑𝑠 = 𝑑𝑞 + 𝑟𝑑𝜃 . (2.4b)

Figure 2.5 – geometrical relations between the quantities 𝑟, 𝑠 and 𝜃.

By comparison with Figure 2.4, it can be verified that

𝑑𝑧 = sin 𝜃 𝑑𝑠 , 𝑑𝑦 = cos 𝜃 𝑑𝑠 , (2.5)

and some other geometrical relations are obtained:

𝑧(𝑠) − 𝑧𝑃 = −𝑟 cos 𝜃 + 𝑞 sin 𝜃 , (2.6a)

𝑦(𝑠) − 𝑦𝑃 = 𝑟 sin 𝜃 + 𝑞 cos 𝜃 . (2.6b)

Recall that according to assumption (i) the cross section does not deform in its own plane. Thus, the
centreline and the sectorial pole 𝑃 will describe the same rigid body motion as the (𝑦, 𝑧) plane. This
enables the definition of the plane displacements as functions of the beam’s translation and rotation about
the pole axis 𝑥̂.

13
The displacements of a point at the centreline are described in the local coordinate system (𝑥, 𝑛, 𝑠) by
𝑢𝑛 (𝑥, 𝑠), 𝑢𝑠 (𝑥, 𝑠) and 𝑢𝑥 (𝑥, 𝑠). The in-plane displacements 𝑢𝑛 and 𝑢𝑠 are therefore defined in terms of the
cross section’s generalized displacements as follows:

𝑢𝑛 (𝑥, 𝑠) = −𝜉𝑧 (𝑥) cos 𝜃(𝑠) + 𝜉𝑦 (𝑥) sin 𝜃(𝑠) − 𝜙(𝑥)𝑞(𝑠) , (2.7a)
𝑢𝑠 (𝑥, 𝑠) = 𝜉𝑧 (𝑥) sin 𝜃(𝑠) + 𝜉𝑦 (𝑥) cos 𝜃(𝑠) + 𝜙(𝑥)𝑟(𝑠) , (2.7b)

where the variables 𝜉𝑖 represent the 𝑖-th cross-sectional translation component, being 𝑖 either 𝑥, 𝑦 or 𝑧, and
𝜙 represents a rotation about the 𝑥̂ axis, as illustrated in Figure 2.6.

Figure 2.6 – cross-sectional displacements and torsion rotation.

In the case of open profiles, the out-of-plane displacements 𝑢𝑥 (𝑥, 𝑠) are obtained from the Vlasov
assumption (ii), where the distortion 𝛾𝑥𝑠 (𝑥, 𝑠) is set equal to zero, as follows:

𝜕𝑢𝑠 (𝑥, 𝑠) 𝜕𝑢𝑥 (𝑥, 𝑠)


𝛾𝑠𝑥 (𝑥, 𝑠) = + =0. (2.8)
𝜕𝑥 𝜕𝑠

Hence, substituting for 𝑢𝑠 from eqn (2.7b), integrating the resulting equation between the origin 𝑂 and a
generic point 𝐵(𝑥, 𝑛, 𝑠), and taking into account the relations (2.5), one may write

𝑢𝑥 (𝑥, 𝑠) = 𝜉𝑥 (𝑥) − 𝜉𝑧′ (𝑥)𝑧(𝑠) − 𝜉𝑦′ (𝑥)𝑦(𝑠) − 𝜙 ′ (𝑥)𝜔(𝑠) , (2.9)

where the primes indicate differentiation with respect to 𝑥 and 𝜔(𝑠) represents the sectorial coordinate,
being defined for an open cross section as follows:
𝑠
𝜔(𝑠) = ∫ 𝑟(𝑠) 𝑑𝑠 . (2.10)
0

The term 𝜉𝑥 in eqn (2.9) is an integration constant that represents the average longitudinal displacement of
the cross section, being given by

𝜉𝑥 (𝑥) = 𝑢𝑥 (𝑥, 0) + 𝜉𝑧′ (𝑥)𝑧𝑂 + 𝜉𝑦′ (𝑥)𝑦𝑂 . (2.11)

14
The sectorial coordinate 𝜔(𝑠) can be interpreted geometrically as twice the area 𝐴𝑚 (2.12) swept out by the
⃗⃗⃗⃗ when the generic point 𝐵 moves along the centreline, see Figure 2.7,
position vector 𝑃𝐵

𝜔(𝑠) = 2𝐴𝑚 . (2.12)

Figure 2.7 – geometrical meaning of the sectorial coordinate 𝜔 as an area.

For closed cross sections, the component 𝑢𝑥 (𝑥, 𝑠) of deflection along the 𝑥 direction can be obtained by
considering the theory of von Kármán through the integration of the expression

𝜕𝑢𝑠 (𝑥, 𝑠) 𝜕𝑢𝑥 (𝑥, 𝑠) 𝜏𝑆


𝑆 (𝑥, 𝑠) = 𝑥𝑠
𝛾𝑥𝑠 (𝑥, 𝑠) = + ≅ 𝛾𝑥𝑠 , (2.13)
𝜕𝑥 𝜕𝑠 𝐺

being the distortion 𝛾𝑥𝑠𝑆 defined according to the Saint-Venant torsion.

Considering an infinitesimal torsion moment 𝑑𝑀𝑥𝑆 about the pole axis 𝑥̂ acting on an element of width 𝑑𝑠,
defined by

𝑑𝑀𝑥𝑆 = 𝑓𝑆 𝑟𝑑𝑠 , (2.14)

where 𝑓𝑆 is the shear flow obtained from the Saint-Venant torsion, and performing its integration around
the whole closed profile the resultant value of the Saint-Venant torsion moment 𝑀𝑥𝑆 is given by

𝑀𝑥𝑆 = ∮ 𝑑𝑀𝑥𝑆 = 2𝑓𝑆 𝐴𝑚 , (2.15)

being 𝐴𝑚 the enclosed area by the cross section’s centreline.

The variation of the twist angle 𝜙 along the beam axis, of length 𝐿, is given by equating the strain energy
𝛱𝑖 and the symmetric of the external potential energy 𝛱𝑒 , respectively

𝑆 2
𝐿 𝑡𝜏𝑥𝑠 1
𝛱𝑖 = ∮ 𝑑𝑠 and 𝛱𝑒 = − 𝑀𝑥𝑆 𝜙 , (2.16)
2 𝐺 2

as follows:

𝜙 𝑓𝑆 𝑑𝑠
𝜙′ = = ∮ . (2.17)
𝐿 2𝐺𝐴𝑚 𝑡

For closed cross sections with more than one cell the total torsion moment 𝑀𝑥𝑆 is given by

𝑀𝑥𝑆 = ∑ [2𝑓𝑆 𝐴𝑚 ]𝑖 , (2.18)


𝑖

being the term in square brackets the contribution of the 𝑖-th cell.

15
For convenience of exposure a new variable 𝛹 is introduced, such that the resulting Saint-Venant torsion
moment 𝑀𝑥𝑆 is defined as a function of the variation of the twist angle 𝜙, [Murray, 1984]. Therefore,

[𝑓𝑆 ]𝑖
𝛹𝑖 (𝑥) = . (2.19)
𝐺𝜙 ′

The variation of 𝜙 is given by

1 𝑑𝑠 𝑑𝑠
𝜙′ = ([𝑓𝑆 ∮ ] − [𝑓𝑆 ∫ ] ) , (2.20)
2𝐺[𝐴𝑚 ]𝑖 𝑡 𝑖 𝑡 𝑗

where the subscript 𝑗 represents the adjacent cells of 𝑖. Finally, the torsion moment 𝑀𝑥𝑆 (2.18) is written as

𝑀𝑥𝑆 = 𝐺𝐽𝜙 ′ , (2.21)


being the torsion constant 𝐽 defined as follows:

𝐽 = ∑ [2𝑓𝑆 𝐴𝑚 ]𝑖 . (2.22)
𝑖

Similarly, for a profile with a single cell the function 𝛹 is given by

𝑓𝑆 2𝐴𝑚
𝛹(𝑥) = ′ =
𝑑𝑠
. (2.23)
𝐺𝜙
∮ 𝑡

The out-of-plane displacements 𝑢𝑥 (𝑥, 𝑠) of the points at the centreline of closed profiles are obtained
through the integration of the eqn (2.13) and by considering the result (2.23), being given by the equation

𝑢𝑥 (𝑥, 𝑠) = 𝜉𝑥 (𝑥) − 𝜉𝑧′ (𝑥)𝑧(𝑠) − 𝜉𝑦′ (𝑥)𝑦(𝑠) − 𝜙 ′ (𝑥)𝜔(𝑠) , (2.9)

where 𝜉𝑥 was defined in (2.11) and the sectorial coordinate 𝜔(𝑠) for a closed profile is now defined as
𝑠
𝛹𝑖 (𝑥)
𝜔(𝑠) = ∫ [𝑟(𝑠) − ] 𝑑𝑠 . (2.24)
0 𝑡(𝑠)

Notice that the sectorial coordinate 𝜔(𝑠) is defined differently wherever it applies to an open or a closed
profile, see eqns (2.10) and (2.24). Despite the same symbol is used for both cases no ambiguity is possible
since there is no shear flow 𝑓𝑆 in an open profile and the term in 𝛹 vanishes.

2.3 VARIATIONAL FORMULATION OF THE


GOVERNING EQUATIONS

In the present section, the thin walled beam’s governing differential equations are developed by applying
the Hamilton’s principle, which can be considered as a generalization of the principle of virtual works to
time-dependent responses, i.e., to dynamic problems.

16
For an elastic body, the Hamilton’s principle is represented by the equation
𝑡2
∫ 𝛿(𝛬 − 𝛱) 𝑑𝑡 = 0 , (2.25)
𝑡1

where 𝛬 and 𝛱 are the kinetic energy and total potential energy of the body, respectively, and 𝛿 represents
a variation. Once the energies are formulated in terms of the displacement and strain fields, only the
kinematic description of the element is required.

For the sake of finding the equations of motion of a discrete system with 𝑛𝑑𝑜𝑓 degrees of freedom,
consider both the potential energy 𝛱 and the kinetic energy 𝛬 described in terms of a discrete number of
generalized coordinates 𝑞1 , 𝑞2 , ..., 𝑞𝑛𝑑𝑜𝑓 and their derivatives, such that

𝛬 = 𝛬(𝑞𝑖 , 𝑞𝑖̇ , 𝑞̇ 𝑖′ ) and 𝛱 = 𝛱(𝑞𝑖 , 𝑞𝑖′ , 𝑞𝑖′′ ) , (2.26)


where 𝑖 is an integer between 1 and 𝑛𝑑𝑜𝑓 .

Hence, the eqn (2.25) becomes


𝑡2
𝜕𝛬 𝜕𝛬 𝜕𝛬 𝜕𝛱 𝜕𝛱 𝜕𝛱
∫ ( 𝛿𝑞𝑖 + 𝛿𝑞𝑖̇ + ′ 𝛿𝑞̇ 𝑖′ − 𝛿𝑞𝑖 − ′ 𝛿𝑞𝑖′ − ′′ 𝛿𝑞𝑖′′ ) 𝑑𝑡 = 0 . (2.27)
𝑡1 𝜕𝑞𝑖 𝜕𝑞𝑖̇ 𝜕𝑞𝑖
̇ 𝜕𝑞𝑖 𝜕𝑞𝑖 𝜕𝑞𝑖

Integrating by parts the first time derivative terms in the above, gives
𝑡2 𝑡2 𝑡2
𝜕𝛬 𝜕𝛬 𝜕 𝜕𝛬
∫ ( 𝛿𝑞𝑖̇ ) 𝑑𝑡 = [ 𝛿𝑞𝑖 ] − ∫ ( ) 𝛿𝑞𝑖 𝑑𝑡 , (2.28a)
𝑡1 𝜕𝑞𝑖̇ 𝜕𝑞𝑖̇ 𝑡 𝑡1 𝜕𝑡 𝜕𝑞𝑖̇
1

𝑡2 𝑡2 𝑡
𝜕𝛬 𝜕𝛬 2
𝜕 𝜕𝛬
∫ ( ′ 𝛿𝑞̇ 𝑖′ ) 𝑑𝑡 = [ ′ 𝛿𝑞𝑖′ ] − ∫ ( ′ ) 𝛿𝑞𝑖′ 𝑑𝑡 . (2.28b)
𝑡1 𝜕𝑞𝑖
̇ 𝜕𝑞𝑖
̇ 𝑡1 𝜕𝑡 𝜕𝑞̇ 𝑖
𝑡 1

An admissible variation of the integrand implies zero variation for each generalized coordinate 𝑞𝑖 and their
first derivatives in 𝑥 at the initial and final times, [Reddy, 2002], so that

𝛿𝑞𝑖 (𝑡1 ) = 𝛿𝑞𝑖 (𝑡2 ) = 𝛿𝑞𝑖′ (𝑡1 ) = 𝛿𝑞𝑖′ (𝑡2 ) = 0 . (2.29)


Thus, after substituting eqns (2.28) and (2.29) into (2.27) and rearranging terms, the Hamilton’s principle
reduces to

𝜕 𝜕𝛬 𝜕𝛬 𝜕𝛱 𝜕 𝜕𝛬 𝜕𝛱 𝜕𝛱
[ ( )− + ] 𝛿𝑞𝑖 + [ ( ′ ) + ′ ] 𝛿𝑞𝑖′ + [ ′′ ] 𝛿𝑞𝑖′′ = 0 . (2.30)
𝜕𝑡 𝜕𝑞𝑖̇ 𝜕𝑞𝑖 𝜕𝑞𝑖 𝜕𝑡 𝜕𝑞̇ 𝑖 𝜕𝑞𝑖 𝜕𝑞𝑖

2.3.1 THE GOVERNING EQUATIONS

The total potential energy 𝛱 is defined as the sum of the potential energy of the external loads 𝛱𝑒 and the
internal strain energy 𝛱𝑖 , eqn (2.31). The former represents the capacity of the applied loading to do work
through a deformation of the structure, whereas the latter is the capacity of the internal stresses to do
work through the structure’s strains.

17
Thus,

𝛱 = 𝛱𝑖 + 𝛱𝑒 . (2.31)

The internal strain energy 𝛱𝑖 is given by the integration of the strain energy of a unit volume 𝑑𝛱𝑖 over the
entire volume domain 𝑉. Considering the Hooke’s law, the strain energy 𝛱𝑖 becomes

𝐸 𝐺
𝛱𝑖 = ∫ 𝜀 2 𝑑𝑉 + ∫ 𝛾 2 𝑑𝑉 , (2.32)
2 𝑉 2 𝑉

where 𝐸 is the Young’s modulus, 𝜀 is constituted only from 𝜀𝑥𝑥 (𝑥, 𝑛, 𝑠), and 𝛾 corresponds to the shear
strain component 𝛾𝑥𝑠 (𝑥, 𝑛, 𝑠) over the wall thickness:

𝜕𝑢𝑥 (𝑥, 𝑛, 𝑠)
𝜀𝑥𝑥 (𝑥, 𝑛, 𝑠) = , (2.33a)
𝜕𝑥

𝜕𝑢𝑠 (𝑥, 𝑛, 𝑠) 𝜕𝑢𝑥 (𝑥, 𝑛, 𝑠)


𝛾𝑥𝑠 (𝑥, 𝑛, 𝑠) = + . (2.33b)
𝜕𝑥 𝜕𝑠

The displacements 𝑢𝑠 (𝑥, 𝑛, 𝑠) and 𝑢𝑥 (𝑥, 𝑛, 𝑠) of a generic point in the cross section are given by

𝑢𝑠 (𝑥, 𝑛, 𝑠) = 𝑢𝑠 (𝑥, 𝑠) + 𝜙(𝑥)𝑛 , (2.34a)


𝜕𝑢𝑛 (𝑥, 𝑠)
𝑢𝑥 (𝑥, 𝑛, 𝑠) = 𝑢𝑥 (𝑥, 𝑠) − 𝑛 , (2.34b)
𝜕𝑥

where the Kirchhoff assumption (iii) was retained.

Substituting 𝑢𝑛 (𝑥, 𝑠), 𝑢𝑠 (𝑥, 𝑠) and 𝑢𝑥 (𝑥, 𝑠) given by eqns (2.7) and (2.9) into eqns (2.34) gives

𝑢𝑠 (𝑥, 𝑛, 𝑠) = 𝜉𝑧 (𝑥) sin 𝜃(𝑠) + 𝜉𝑦 (𝑥) cos 𝜃(𝑠) + 𝜙(𝑥)[𝑟(𝑠) + 𝑛] , (2.35a)


𝑢𝑥 (𝑥, 𝑛, 𝑠) = 𝜉𝑥 (𝑥) − 𝜉𝑧′ (𝑥)[𝑧(𝑠) − 𝑛 cos 𝜃(𝑠)] − 𝜉𝑦′ (𝑥)[𝑦(𝑠) + 𝑛 sin 𝜃(𝑠)] − 𝜙 ′ (𝑥)[𝜔(𝑠) − 𝑛𝑞(𝑠)] . (2.35b)

Considering the position of a cross section’s generic point, the global coordinates 𝑦 , 𝑧 and 𝜔 can be
written as functions of 𝑠 and 𝑛, yielding

𝑧 ≡ 𝑧(𝑛, 𝑠) = 𝑧(𝑠) − 𝑛 cos 𝜃(𝑠) , (2.36a)


𝑦 ≡ 𝑦(𝑛, 𝑠) = 𝑦(𝑠) + 𝑛 sin 𝜃(𝑠) , (2.36b)
𝜔 ≡ 𝜔(𝑛, 𝑠) = 𝜔(𝑠) − 𝑛𝑞(𝑠) , (2.36c)
and eqn (2.35b) can be rewritten as follows:

𝑢𝑥 (𝑥, 𝑦, 𝑧) = 𝜉𝑥 (𝑥) − 𝜉𝑧′ (𝑥)𝑧 − 𝜉𝑦′ (𝑥)𝑦 − 𝜙 ′ (𝑥)𝜔 . (2.37)

From eqn (2.36c) it becomes clear that the sectorial coordinate 𝜔(𝑛, 𝑠) of a generic point in the cross
section is constituted of two contributions: one due to the warping of the centreline 𝜔(𝑠) and another one
due to secondary warping over the wall thickness, defined by the second term of the righ-hand side of eqn
(2.36c).

Considering that the radius of curvature of the centreline 𝑅(𝑠) is defined by

𝑑𝜃(𝑠) (2.38)
𝑅(𝑠) = ,
𝑑𝑠

18
and taking into account the relations (2.35), (2.36c), (2.4b) and (2.5), the eqns (2.33) reduce to

𝜀𝑥𝑥 (𝑥, 𝑛, 𝑠) = 𝜉𝑥′ (𝑥) − 𝜉𝑧′′ (𝑥)𝑧 − 𝜉𝑦′′ (𝑥)𝑦 − 𝜙 ′′ (𝑥)𝜔 , (2.39a)
𝑛 𝑟(𝑠) 𝛹𝑖 (𝑥)
𝛾𝑥𝑠 (𝑥, 𝑛, 𝑠) = −[𝜉𝑧′ (𝑥) sin 𝜃(𝑠) + 𝜉𝑦′ (𝑥) cos 𝜃(𝑠)] + 𝜙 ′ (𝑥) {𝑛 [2 − ]− } , (2.39b)
𝑅(𝑠) 𝑅(𝑠) 𝑡(𝑠)

where the sectorial coordinate 𝜔(𝑠) is defined in (2.10) or (2.24) for an open or a closed cross section,
respectively.

Considering the loading of the beam constituted by loads, bending, torsion and warping moments linearly
distributed along the longitudinal axis 𝑥, the potential energy of the external loads 𝛱𝑒 is expressed by

𝛱𝑒 = − ∫ (𝜉𝑥 𝑓𝑥 + 𝜉𝑧 𝑓𝑧 + 𝜉𝑦 𝑓𝑦 + 𝜙𝑚𝑥 − 𝜉𝑧′ 𝑚𝑦 + 𝜉𝑦′ 𝑚𝑧 + 𝜙 ′ 𝑚𝜔 ) 𝑑𝑥 . (2.40)


𝐿

As shown in Figure 2.8, the applied torsion moment 𝑚𝑥 acts about the pole axis 𝑥̂ and the transverse loads
𝑓𝑧 and 𝑓𝑦 act through it. On the other hand, the bending moments 𝑚𝑦 and 𝑚𝑧 , and the axial force 𝑓𝑥 are
referred to in the (𝑥, 𝑦, 𝑧) coordinate system. The warping moment 𝑚𝜔 has no point of application, but it is
shown lying along the pole axis 𝑥̂ for consistency with the bimoment 𝑀𝜔 representation.

Substituting eqns (2.32) and (2.40) into (2.31) the total potential energy 𝛱 becomes

1
𝛱 = ∫ ∫ (𝐸𝜀 2 + 𝐺𝛾 2 ) 𝑑𝐴𝑑𝑥 − ∫ (𝜉𝑥 𝑓𝑥 + 𝜉𝑧 𝑓𝑧 + 𝜉𝑦 𝑓𝑦 + 𝜙𝑚𝑥 − 𝜉𝑧′ 𝑚𝑦 + 𝜉𝑦′ 𝑚𝑧 + 𝜙 ′ 𝑚𝜔 ) 𝑑𝑥 , (2.41)
2 𝐿 𝐴 𝐿

being the beam element considered to have constant cross-sectional area 𝐴, and the differential volume 𝑑𝑉
has been replaced by the product of 𝑑𝐴 and 𝑑𝑥 .

Figure 2.8 – element loading and internal forces.

Substituting eqns (2.39) into eqn (2.41), the total potential energy is obtained as a function of the
element’s generalized displacements, so that

1 2 2 2 2
𝛱 = ∫ [ 𝐸(𝐴𝜉𝑥′ + 𝐼𝑦𝑦 𝜉𝑧′′ + 𝐼𝑧𝑧 𝜉𝑦′′ + 𝐼𝜔𝜔 𝜙 ′′ − 2𝑆𝑦 𝜉𝑥′ 𝜉𝑧′′ − 2𝑆𝑧 𝜉𝑥′ 𝜉𝑦′′ − 2𝑆𝜔 𝜉𝑥′ 𝜙 ′′ + 2𝐼𝑦𝑧 𝜉𝑧′′ 𝜉𝑦′′ + 2𝐼𝑦𝜔 𝜉𝑧′′ 𝜙 ′′
𝐿 2
1 2
+ 2𝐼𝑧𝜔 𝜉𝑦′′ 𝜙 ′′ ) + 𝐺𝐽𝜙 ′ − (𝜉𝑥 𝑓𝑥 + 𝜉𝑦 𝑓𝑦 + 𝜉𝑧 𝑓𝑧 + 𝜙𝑚𝑥 + 𝜉𝑦′ 𝑚𝑧 − 𝜉𝑧′ 𝑚𝑦 + 𝜙 ′ 𝑚𝜔 )] 𝑑𝑥 , (2.42)
2

19
Adopting the most common cross-sectional layouts for civil engineering purposes, the wall’s curvature
𝑅(𝑠) is infinite and the corresponding terms in eqn (2.39) vanish. Therefore, the torsion constant 𝐽 is
expressed for both open and closed cross sections by

𝑡(𝑠)3
𝐽=∫ 𝑑𝑠 + ∑ [2𝐴𝑚 𝛹(𝑥)]𝑖 , (2.43)
𝛤 3 𝑖

where 𝑖 is the 𝑖-th cell of the cross section.

The remaining geometrical properties of the cross section are defined as follows:

𝐴 = ∫ 𝑑𝐴 = ∫ 𝑡(𝑠) 𝑑𝑠 , (2.44a)
𝐴 𝛤

𝑆𝑦 = ∫ 𝑧(𝑛, 𝑠) 𝑑𝐴 = ∫ 𝑡(𝑠)𝑧(𝑠) 𝑑𝑠 , (2.44b)


𝐴 𝛤

𝑆𝑧 = ∫ 𝑦(𝑛, 𝑠) 𝑑𝐴 = ∫ 𝑡(𝑠)𝑦(𝑠) 𝑑𝑠 , (2.44c)


𝐴 𝛤

𝑆𝜔 = ∫ 𝜔(𝑛, 𝑠) 𝑑𝐴 = ∫ 𝑡(𝑠)𝜔(𝑠) 𝑑𝑠 , (2.44d)


𝐴 𝛤

𝑡 2 (𝑠)
𝐼𝑦𝑦 = ∫ 𝑧 2 (𝑛, 𝑠) 𝑑𝐴 = ∫ 𝑡(𝑠) [𝑧 2 (𝑠) + cos2 𝜃(𝑠)] 𝑑𝑠 , (2.44e)
𝐴 𝛤 12

𝑡 2 (𝑠) 2
𝐼𝑧𝑧 = ∫ 𝑦 2 (𝑛, 𝑠) 𝑑𝐴 = ∫ 𝑡(𝑠) [𝑦 2 (𝑠) − sin 𝜃(𝑠)] 𝑑𝑠 , (2.44f)
𝐴 𝛤 12

𝑡 2 (𝑠) 2
𝐼𝜔𝜔 = ∫ 𝜔2 (𝑛, 𝑠) 𝑑𝐴 = ∫ 𝑡(𝑠) [𝜔2 (𝑠) + 𝑞 (𝑠)] 𝑑𝑠 , (2.44g)
𝐴 𝛤 12

𝑡 2 (𝑠)
𝐼𝑦𝑧 = 𝐼𝑧𝑦 = ∫ 𝑦(𝑛, 𝑠)𝑧(𝑛, 𝑠) 𝑑𝐴 = ∫ 𝑡(𝑠) [𝑦(𝑠)𝑧(𝑠) − sin 𝜃(𝑠) cos 𝜃(𝑠)] 𝑑𝑠 , (2.44h)
𝐴 𝛤 12

𝑡 2 (𝑠)
𝐼𝑦𝜔 = 𝐼𝜔𝑦 = ∫ 𝑧(𝑛, 𝑠)𝜔(𝑛, 𝑠) 𝑑𝐴 = ∫ 𝑡(𝑠) [𝑧(𝑠)𝜔(𝑠) + 𝑞(𝑠) cos 𝜃(𝑠)] 𝑑𝑠 , (2.44i)
𝐴 𝛤 12

𝑡 2 (𝑠)
𝐼𝑧𝜔 = 𝐼𝜔𝑧 = ∫ 𝑦(𝑛, 𝑠)𝜔(𝑛, 𝑠) 𝑑𝐴 = ∫ 𝑡(𝑠) [𝑦(𝑠)𝜔(𝑠) − 𝑞(𝑠) sin 𝜃(𝑠)] 𝑑𝑠 . (2.44j)
𝐴 𝛤 12

where 𝛤 represents the centreline domain.

The kinetic energy 𝛬 of a body is given by

1 𝑇 1
𝛬=∫ 𝒖̇ 𝜌𝒖̇ 𝑑𝑉 = ∫ 𝜌[𝑢̇ 𝑥 2 (𝑥, 𝑛, 𝑠) + 𝑢̇ 𝑛 2 (𝑥, 𝑛, 𝑠) + 𝑢̇ 𝑠 2 (𝑥, 𝑛, 𝑠)] 𝑑𝑉 , (2.45)
𝑉 2 𝑉 2

where 𝒖(𝑥, 𝑛, 𝑠) is the displacement vector of a generic point, 𝜌 is the mass per unit volume and dots
represent derivatives with respect to time 𝑡.

20
Substituting for 𝑢𝑥 (𝑥, 𝑛, 𝑠), 𝑢𝑛 (𝑥, 𝑛, 𝑠) and 𝑢𝑠 (𝑥, 𝑛, 𝑠) from eqns (2.7a) and (2.34) into eqn (2.45) gives

1 2 2 2 2
𝛬=∫ 𝜌 (𝐴𝜉𝑥̇ + 𝐼𝑦𝑦 𝜉𝑧̇ ′ + 𝐼𝑧𝑧 𝜉𝑦̇ ′ + 𝐼𝜔𝜔 𝜙̇ ′ − 2𝑆𝑦 𝜉𝑥̇ 𝜉𝑧̇ ′ − 2𝑆𝑧 𝜉𝑥̇ 𝜉𝑦̇ ′ − 2𝑆𝜔 𝜉𝑥̇ 𝜙̇ ′ + 2𝐼𝑦𝑧 𝜉𝑧̇ ′ 𝜉𝑦̇ ′ + 2𝐼𝑦𝜔 𝜉𝑧̇ ′ 𝜙̇ ′ + 2𝐼𝑧𝜔 𝜉𝑦̇ ′ 𝜙̇ ′ ) 𝑑𝑥
𝐿 2
1 2 2
+∫ 𝜌 [𝐴𝜉𝑧̇ + 𝐴𝜉𝑦̇ + 2𝑆𝑧𝑃 𝜉𝑧̇ 𝜙̇ − 2𝑆𝑦𝑃 𝜉𝑦̇ 𝜙̇ + (𝐼𝑧𝑧
𝑃 + 𝐼𝑃 )𝜙̇ 2 ] 𝑑𝑥 .
𝑦𝑦 (2.46)
𝐿 2

where

𝑆𝑦𝑃 = ∫ [𝑧(𝑛, 𝑠) − 𝑧𝑃 ] 𝑑𝐴 , 𝑆𝑧𝑃 = ∫ [𝑦(𝑛, 𝑠) − 𝑦𝑃 ] 𝑑𝐴 , (2.47a)


𝐴 𝐴

𝑃 = ∫ [𝑧(𝑛, 𝑠) − 𝑧 ]2 𝑑𝐴 , 𝐼𝑃 = ∫ [𝑦(𝑛, 𝑠) − 𝑦 ]2 𝑑𝐴 .
𝐼𝑦𝑦 𝑃 𝑧𝑧 𝑃 (2.47b)
𝐴 𝐴

The governing differential equations of a thin walled beam dynamic problem are obtained by applying the
Hamilton’s principle, which results in the application of eqn (2.30) for each generalized coordinate, say 𝜉𝑥 ,
𝜉𝑧 , 𝜉𝑦 and 𝜙.

The axial extension 𝜉𝑥 is firstly considered, and the corresponding form of eqn (2.30) is written as

𝛿𝜉𝑥 ∫ [𝜌(𝐴𝜉𝑥̈ − 𝑆𝑦 𝜉𝑧̈ ′ − 𝑆𝑧 𝜉𝑦̈ ′ − 𝑆𝜔 𝜙̈ ′ ) − 𝑓𝑥 ] 𝑑𝑥 + 𝛿𝜉𝑥′ ∫ 𝐸(𝐴𝜉𝑥′ − 𝑆𝑦 𝜉𝑧′′ − 𝑆𝑧 𝜉𝑦′′ − 𝑆𝜔 𝜙 ′′ ) 𝑑𝑥 = 0 . (2.48)


𝐿 𝐿

Integrating by parts the second term of the left-hand side yields

𝛿𝜉𝑥′ ∫ 𝐸(𝐴𝜉𝑥′ − 𝑆𝑦 𝜉𝑧′′ − 𝑆𝑧 𝜉𝑦′′ − 𝑆𝜔 𝜙 ′′ ) 𝑑𝑥 = −𝛿𝜉𝑥 ∫ 𝐸(𝐴𝜉𝑥′′ − 𝑆𝑦 𝜉𝑧′′′ − 𝑆𝑧 𝜉𝑦′′′ − 𝑆𝜔 𝜙 ′′′ ) 𝑑𝑥 . (2.49)
𝐿 𝐿

Substituting eqn (2.49) into eqn (2.48), rearranging terms and noticing that it has to hold whatever 𝛿𝜉𝑥 is,
the governing equation for the axial extension problem follows as

𝜌(𝐴𝜉𝑥̈ − 𝑆𝑦 𝜉𝑧̈ ′ − 𝑆𝑧 𝜉𝑦̈ ′ − 𝑆𝜔 𝜙̈ ′ ) − 𝐸(𝐴𝜉𝑥′′ − 𝑆𝑦 𝜉𝑧′′′ − 𝑆𝑧 𝜉𝑦′′′ − 𝑆𝜔 𝜙 ′′′ ) = 𝑓𝑥 . (2.50a)

Following a similar procedure as it was done to obtain eqn (2.50a) the remaining governing equations are

𝜌(𝐴𝜉𝑧̈ + 𝑆𝑧𝑃 𝜙̈) − 𝜌(−𝑆𝑦 𝜉𝑥̈ ′ + 𝐼𝑦𝑦 𝜉𝑧̈ ′′ + 𝐼𝑦𝑧 𝜉𝑦̈ ′′ + 𝐼𝑦𝜔 𝜙̈ ′′ ) + 𝐸(−𝑆𝑦 𝜉𝑥′′′ + 𝐼𝑦𝑦 𝜉𝑧′′′′ + 𝐼𝑦𝑧 𝜉𝑦′′′′ + 𝐼𝑦𝜔 𝜙 ′′′′ ) = 𝑓𝑧 + 𝑚𝑦′ , (2.50b)

𝜌(𝐴𝜉𝑦̈ − 𝑆𝑦𝑃 𝜙̈) − 𝜌(−𝑆𝑧 𝜉𝑥̈ ′ + 𝐼𝑧𝑦 𝜉𝑧̈ ′′ + 𝐼𝑧𝑧 𝜉𝑦̈ ′′ + 𝐼𝑧𝜔 𝜙̈ ′′ ) + 𝐸(−𝑆𝑧 𝜉𝑥′′′ + 𝐼𝑧𝑦 𝜉𝑧′′′′ + 𝐼𝑧𝑧 𝜉𝑦′′′′ + 𝐼𝑧𝜔 𝜙 ′′′′ ) = 𝑓𝑦 − 𝑚𝑧′ , (2.50c)

𝜌[𝑆𝑧𝑃 𝜉𝑧̈ − 𝑆𝑦𝑃 𝜉𝑦̈ + (𝐼𝑧𝑧


𝑃 + 𝐼𝑃 )𝜙̈ ] − 𝜌(−𝑆 𝜉̈ ′ + 𝐼 𝜉 ′′
𝑦𝑦 𝜔 𝑥
̈ ̈ ′′ ̈ ′′
𝜔𝑦 𝑧 + 𝐼𝜔𝑧 𝜉𝑦 + 𝐼𝜔𝜔 𝜙 )
′ ,
+ 𝐸 (−𝑆𝜔 𝜉𝑥′′′ + 𝐼𝜔𝑦 𝜉𝑧′′′′ + 𝐼𝜔𝑧 𝜉𝑦′′′′ + 𝐼𝜔𝜔 𝜙 ′′′′ ) − 𝐺𝐽𝜙 ′′ = 𝑚𝑥 − 𝑚𝜔 (2.50d)

or in matrix form

0 0 0 0 𝜉𝑥̈ 𝐴 −𝑆𝑦 −𝑆𝑧 −𝑆𝜔 𝜉𝑥̈ ′ 𝐴 −𝑆𝑦 −𝑆𝑧 −𝑆𝜔 𝜉𝑥′′′


0 𝑆𝑧𝑃
0 𝐴 𝜉𝑧̈ −𝑆𝑦 𝐼𝑦𝑦 𝐼𝑦𝑧 𝐼𝑦𝜔 𝜉𝑧̈ ′′ −𝑆𝑦 𝐼𝑦𝑦 𝐼𝑦𝑧 𝐼𝑦𝜔 𝜉𝑧′′′′
𝜌 𝐴 −𝑆𝑦𝑃 −𝜌 +𝐸 −
0 0 𝜉𝑦̈ −𝑆𝑧 𝐼𝑧𝑦 𝐼𝑧𝑧 𝐼𝑧𝜔 𝜉𝑦̈ ′′ −𝑆𝑧 𝐼𝑧𝑦 𝐼𝑧𝑧 𝐼𝑧𝜔 𝜉𝑦′′′′
𝑃
[0 𝑆𝑧 −𝑆𝑦𝑃 𝑃 𝑃
𝐼𝑧𝑧 + 𝐼𝑦𝑦 ] { 𝜙̈ } [ −𝑆𝜔 𝐼𝜔𝑦 𝐼𝜔𝑧 𝐼𝜔𝜔 ] {𝜙̈ ′′ } [ −𝑆𝜔 𝐼𝜔𝑦 𝐼𝜔𝑧 𝐼𝜔𝜔 ′′′′
] {𝜙 }

0 0 0 0 𝜉𝑥
′ −𝑓𝑥′
0 0 0 0 𝜉𝑧
′′ 𝑓𝑧 + 𝑚𝑦′
−𝐺 [ ]
0 0 𝜉𝑦′′ = , (2.51)
0 0 𝑓𝑦 − 𝑚𝑧′
0 0 0 𝐽 {𝜙 ′′ } ′
{ 𝑚𝑥 − 𝑚𝜔 }

21
2.4 CHANGE OF REFERENCE AXES

Some simplifications of the governing differential equations (2.49) will be made in the following by
changing the position and the orientation of the global coordinate axes (𝑥, 𝑦, 𝑧) , the position of the
sectorial pole 𝑃 and the position of the sectorial origin 𝑂. By referring all the geometrical quantities to the
principal coordinate systems, the eqns (2.49) can be decoupled in their static terms.

To begin with, consider that the pole 𝑃(𝑥𝑃 , 𝑦𝑃 , 𝑧𝑃 ) and the origin 𝑂(𝑥𝑂 , 𝑦𝑂 , 𝑧𝑂 ), both relative to the arbitrary
references, are shifted to new positions 𝑃̅(𝑥𝑃̅ , 𝑦𝑃̅ , 𝑧𝑃̅ ) and 𝑂̅(𝑥𝑂̅ , 𝑦𝑂̅ , 𝑧𝑂̅ ), respectively. The position functions 𝑟̅
and 𝑞̅, as illustrated in Figure 2.9, are now written as follows:

𝑟̅ (𝑠̅ ) = 𝑟(𝑠) − (𝑦𝑃̅ − 𝑦𝑃 ) sin 𝜃(𝑠) + (𝑧𝑃̅ − 𝑧𝑃 ) cos 𝜃(𝑠) , (2.52a)


𝑞̅(𝑠̅ ) = 𝑞(𝑠) − (𝑦𝑃̅ − 𝑦𝑃 ) cos 𝜃(𝑠) − (𝑧𝑃̅ − 𝑧𝑃 ) sin 𝜃(𝑠) . (2.52b)
To calculate the change in the value of the sectorial coordinate, the Leibnitz sectorial formula, [Murray,
1984], is used and the following definitions can be made:

𝑑𝜔(𝑠) = [𝑦(𝑠) − 𝑦𝑃 ]𝑑𝑧 − [𝑧(𝑠) − 𝑧𝑃 ]𝑑𝑦 , (2.53a)


̅(𝑠̅ ) = [𝑦(𝑠̅ ) − 𝑦𝑃̅ ]𝑑𝑧 − [𝑧(𝑠̅ ) − 𝑧𝑃̅ ]𝑑𝑦 ,
𝑑𝜔 (2.53b)
and thus,

̅(𝑠̅ ) − 𝜔(𝑠)] = (𝑦𝑃 − 𝑦𝑃̅ )𝑑𝑧 − (𝑧𝑃 − 𝑧𝑃̅ )𝑑𝑦 .


𝑑[𝜔 (2.54)

Figure 2.9 – changes of the position functions 𝑞 and 𝑟 due to a pole 𝑃̅ moving.

Integrating the eqn (2.54) over 𝑠̅, the sectorial coordinate 𝜔̅(𝑠̅ ) relative to the new positions of the pole 𝑃̅
and the origin 𝑂̅ becomes

̅(𝑠̅ ) = 𝜔(𝑠) − 𝜔(𝑠𝑂̅ ) + (𝑦𝑃 − 𝑦𝑃̅ )[𝑧(𝑠̅ ) − 𝑧𝑂̅ ] − (𝑧𝑃 − 𝑧𝑃̅ )[𝑦(𝑠̅ ) − 𝑦𝑂̅ ] ,
𝜔 (2.55)
where, by definition, the sectorial coordinates 𝜔̅(𝑠̅ ) is zero at the origin 𝑂̅, and
𝑠=𝑠𝑂
̅
𝜔(𝑠𝑂̅ ) = ∫ 𝑑𝜔 . (2.56)
𝑠=0

22
Due to this transformation of the sectorial coordinates, the cross-sectional sectorial properties given in
(2.44) are rewritten as follows:

𝑆𝜔̅ = 𝑆𝜔 − 𝜔(𝑠𝑂̅ )𝐴 + (𝑦𝑃 − 𝑦𝑃̅ )(𝑆𝑦 − 𝑧𝑂̅ 𝐴) − (𝑧𝑃 − 𝑧𝑃̅ )(𝑆𝑧 − 𝑦𝑂̅ 𝐴) , (2.57a)

𝐼𝑧𝜔̅ = 𝐼𝜔̅𝑧 = 𝐼𝑧𝜔 − 𝜔(𝑠𝑂̅ )𝑆𝑧 + (𝑦𝑃 − 𝑦𝑃̅ )(𝐼𝑧𝑦 − 𝑧𝑂̅ 𝑆𝑧 ) − (𝑧𝑃 − 𝑧𝑃̅ )(𝐼𝑧𝑧 − 𝑦𝑂̅ 𝑆𝑧 ) , (2.57b)
𝐼𝑦𝜔̅ = 𝐼𝜔̅𝑦 = 𝐼𝑦𝜔 − 𝜔(𝑠𝑂̅ )𝑆𝑦 + (𝑦𝑃 − 𝑦𝑃̅ )(𝐼𝑦𝑦 − 𝑧𝑂̅ 𝑆𝑦 ) − (𝑧𝑃 − 𝑧𝑃̅ )(𝐼𝑦𝑧 − 𝑦𝑂̅ 𝑆𝑦 ) , (2.57c)

𝐼𝜔̅𝜔̅ = 𝐼𝜔𝜔 + 𝜔2 (𝑠𝑂̅ )𝐴 − 2𝜔(𝑠𝑂̅ )𝑆𝜔 + (𝑦𝑃 − 𝑦𝑃̅ )2 (𝐼𝑦𝑦 − 2𝑧𝑂̅ 𝑆𝑦 + 𝑧𝑂̅ 2 𝐴) + (𝑧𝑃 − 𝑧𝑃̅ )2 (𝐼𝑧𝑧 − 2𝑦𝑂̅ 𝑆𝑧 + 𝑦𝑂̅ 2 𝐴)
+ 2(𝑦𝑃 − 𝑦𝑃̅ )[𝐼𝜔𝑦 − 𝑧𝑂̅ 𝑆𝜔 − 𝜔(𝑠𝑂̅ )𝑆𝑦 + 𝜔(𝑠𝑂̅ )𝑧𝑂̅ 𝐴]
− 2(𝑧𝑃 − 𝑧𝑃̅ )[𝐼𝜔𝑧 − 𝑦𝑂̅ 𝑆𝜔 − 𝜔(𝑠𝑂̅ )𝑆𝑧 + 𝜔(𝑠𝑂̅ )𝑦𝑂̅ 𝐴]
− 2(𝑧𝑃 − 𝑧𝑃̅ )(𝑦𝑃 − 𝑦𝑃̅ )(𝐼𝑦𝑧 − 𝑦𝑂̅ 𝑆𝑦 − 𝑧𝑂̅ 𝑆𝑧 + 𝑦𝑂̅ 𝑧𝑂̅ 𝐴) . (2.57d)

The set of equations (2.57) can be simplified by changing the global coordinate axes (𝑥, 𝑦, 𝑧) from the
arbitrary position to one that decouples the static terms of the governing equations. The set of axes
(𝑥̅ , 𝑦̅, 𝑧̅) is thus chosen in a way that makes

𝑆𝑦̅ = 𝑆𝑧̅ = 𝐼𝑦̅𝑧̅ = 𝐼𝑦̅𝜔̅ = 𝐼𝑧̅𝜔̅ = 0 . (2.58)

Therefore the eqns (2.57) can be written as

𝑆𝜔̅ = 𝑆𝜔 − 𝐴[𝜔(𝑠𝑂̅ ) + (𝑦̅𝑃 − 𝑦̅𝑃̅ )𝑧̅𝑂̅ − (𝑧̅𝑃 − 𝑧̅𝑃̅ )𝑦̅𝑂̅ ] , (2.59a)


𝐼𝑧̅𝜔̅ = 𝐼𝜔̅𝑧̅ = 𝐼𝑧̅𝜔 − (𝑧̅𝑃 − 𝑧̅𝑃̅ )𝐼𝑧̅𝑧̅ , (2.59b)
𝐼𝑦̅𝜔̅ = 𝐼𝜔̅𝑦̅ = 𝐼𝑦̅𝜔 + (𝑦̅𝑃 − 𝑦̅𝑃̅ )𝐼𝑦̅𝑦̅ , (2.59c)
𝐼𝜔̅𝜔̅ = 𝐼𝜔𝜔 + 𝜔2 (𝑠𝑂̅ )𝐴 − 2𝜔(𝑠𝑂̅ )𝑆𝜔 + (𝑦̅𝑃 − 𝑦̅𝑃̅ )2 (𝐼𝑦̅𝑦̅ + 𝑧̅𝑂̅ 2 𝐴) + (𝑧̅𝑃 − 𝑧̅𝑃̅ )2 (𝐼𝑧̅𝑧̅ + 𝑦̅𝑂̅ 2 𝐴)
+ 2(𝑦̅𝑃 − 𝑦̅𝑃̅ )[𝐼𝜔𝑦̅ − 𝑧̅𝑂̅ 𝑆𝜔 + 𝜔(𝑠𝑂̅ )𝑧̅𝑂̅ 𝐴] − 2(𝑧̅𝑃 − 𝑧̅𝑃̅ )[𝐼𝜔𝑧̅ − 𝑦̅𝑂̅ 𝑆𝜔 + 𝜔(𝑠𝑂̅ )𝑦̅𝑂̅ 𝐴]
− 2(𝑧̅𝑃 − 𝑧̅𝑃̅ )(𝑦̅𝑃 − 𝑦̅𝑃̅ )(𝑦̅𝑂̅ 𝑧̅𝑂̅ 𝐴) . (2.59d)
Notice that the properties 𝐼𝑧̅𝜔̅ and 𝐼𝑦̅𝜔̅ have the same value regardless of the position of 𝑂̅ and the
orientation of the global axes (𝑥̅ , 𝑦̅, 𝑧̅), being only dependent upon the position of the pole 𝑃̅. Whether the
pole 𝑃̅ is located at its principal position, i.e., coincident with the shear centre 𝑆𝐶 of the cross section, it
allows to consider

𝐼𝑦̅𝜔̅ = 𝐼𝑧̅𝜔̅ = 0 . (2.60)

This leads to the 𝑦̅𝑃̅ and 𝑧̅𝑃̅ global coordinates of the principal sectorial pole 𝑃̅ in terms of the arbitrary
pole 𝑃 and origin 𝑂. Therefore, from eqns (2.59b) and (2.59c), it follows that

𝐼𝑧̅𝜔 𝐼𝑦̅𝜔
𝑧̅𝑃̅ = 𝑧̅𝑃 − and 𝑦̅𝑃̅ = 𝑦̅𝑃 + . (2.61)
𝐼𝑧̅𝑧̅ 𝐼𝑦̅𝑦̅

The principal position of 𝑂̅ is such that the magnitude of 𝑆𝜔̅ is set to zero, see eqn (2.59a). Thus,
considering that the pole 𝑃 is placed coincident with the principal pole 𝑃̅, the principal position of 𝑂̅ is
such that the following equation is satisfied:

𝑆𝜔
𝜔(𝑠𝑂̅ ) = . (2.62)
𝐴

23
It is important to emphasize that both the principal sectorial pole 𝑃̅ and the principal sectorial origin 𝑂̅ are
section properties, as they do not depend upon the arbitrary position of 𝑃 and 𝑂.

The transformed warping constant 𝐼𝜔̅𝜔̅ from eqn (2.59d), along with eqn (2.62), reduces to
2
𝑆𝜔
𝐼𝜔̅𝜔̅ = 𝐼𝜔𝜔 − + (𝑦̅𝑃 − 𝑦̅𝑃̅ )2 (𝐼𝑦̅𝑦̅ + 𝑧̅𝑂̅ 2 𝐴) + (𝑧̅𝑃 − 𝑧̅𝑃̅ )2 (𝐼𝑧̅𝑧̅ + 𝑦̅𝑂̅ 2 𝐴) + 2(𝑦̅𝑃 − 𝑦̅𝑃̅ )𝐼𝜔𝑦̅
𝐴
− 2(𝑧̅𝑃 − 𝑧̅𝑃̅ )𝐼𝜔𝑧̅ − 2(𝑧̅𝑃 − 𝑧̅𝑃̅ )(𝑦̅𝑃 − 𝑦̅𝑃̅ )(𝑦̅𝑂̅ 𝑧̅𝑂̅ 𝐴) . (2.63)

Further discussion involving the transformation of the global coordinate system is given in the following.
Consider the transformation equations for the global coordinates 𝑦̅(𝑠) and 𝑧̅(𝑠):

𝑦̅(𝑠) = [𝑦(𝑠) − 𝑦𝐶 ] cos 𝛼 + [𝑧(𝑠) − 𝑧𝐶 ] sin 𝛼 , (2.64a)


𝑧̅(𝑠) = −[𝑦(𝑠) − 𝑦𝐶 ] sin 𝛼 + [𝑧(𝑠) − 𝑧𝐶 ] cos 𝛼 , (2.64b)
where 𝐶(𝑥𝐶 , 𝑦𝐶 , 𝑧𝐶 ) is the origin of the transformed global axes (𝑥̅ , 𝑦̅, 𝑧̅), which coordinates are relative to the
arbitrary system (𝑥, 𝑦, 𝑧) , as shown in Figure 2.10. The relative orientation between both coordinate
systems is given by

𝛼 = 𝜃(𝑠) − 𝜃̅ (𝑠) (2.65)

Figure 2.10 – global coordinate system transformation.

The transformation of the first and the second moments of inertia caused by the global system
transformation are equated by invoking the definitions (2.44), (2.64) and (2.65), as follows:

𝑆𝑦̅ = (𝑆𝑦 − 𝑧𝐶 𝐴) cos 𝛼 − (𝑆𝑧 − 𝑦𝐶 𝐴) sin 𝛼 , (2.66a)

𝑆𝑧̅ = (𝑆𝑧 − 𝑦𝐶 𝐴) cos 𝛼 + (𝑆𝑦 − 𝑧𝐶 𝐴) sin 𝛼 , (2.66b)

𝐼𝑦̅𝜔 = 𝐼𝜔𝑦̅ = (𝐼𝑦𝜔 − 𝑧𝐶 𝑆𝜔 ) cos 𝛼 − (𝐼𝑧𝜔 − 𝑦𝐶 𝑆𝜔 ) sin 𝛼 , (2.66c)

𝐼𝑧̅𝜔 = 𝐼𝜔𝑧̅ = (𝐼𝑧𝜔 − 𝑦𝐶 𝑆𝜔 ) cos 𝛼 + (𝐼𝑦𝜔 − 𝑧𝐶 𝑆𝜔 ) sin 𝛼 , (2.66d)

𝐼𝑦̅𝑦̅ = (𝐼𝑧𝑧 − 2𝑦𝐶 𝑆𝑧 + 𝑦𝐶 2 𝐴) sin2 𝛼 + (𝐼𝑦𝑦 − 2𝑧𝐶 𝑆𝑦 + 𝑧𝐶 2 𝐴) cos2 𝛼 − 2(𝐼𝑦𝑧 − 𝑦𝐶 𝑆𝑦 − 𝑧𝐶 𝑆𝑧 + 𝑦𝐶 𝑧𝐶 𝐴) cos 𝛼 sin 𝛼 , (2.66e)

𝐼𝑧̅𝑧̅ = (𝐼𝑧𝑧 − 2𝑦𝐶 𝑆𝑧 + 𝑦𝐶 2 𝐴) cos2 𝛼 + (𝐼𝑦𝑦 − 2𝑧𝐶 𝑆𝑦 + 𝑧𝐶 2 𝐴) sin2 𝛼 − 2(𝐼𝑧𝑦 − 𝑦𝐶 𝑆𝑦 − 𝑧𝐶 𝑆𝑧 + 𝑦𝐶 𝑧𝐶 𝐴) cos 𝛼 sin 𝛼 , (2.66f)

𝐼𝑧̅𝑦̅ = 𝐼𝑦̅𝑧̅ = (𝐼𝑦𝑦 − 𝐼𝑧𝑧 − 2𝑧𝐶 𝑆𝑦 + 2𝑦𝐶 𝑆𝑧 + 𝑧𝐶 2 𝐴 − 𝑦𝐶 2 𝐴) cos 𝛼 sin 𝛼


+ (𝐼𝑦𝑧 − 𝑧𝐶 𝑆𝑧 − 𝑦𝐶 𝑆𝑦 + 𝑦𝐶 𝑧𝐶 𝐴)(cos2 𝛼 − sin2 𝛼) . (2.66g)

24
The principal position of the global origin 𝐶 coincides with centre of gravity 𝐶𝐺(𝑥𝐶𝐺 , 𝑦𝐶𝐺 , 𝑧𝐶𝐺 ) of the cross
section, regardless of the angle 𝛼, being given by imposing

𝑆𝑦̅ = 𝑆𝑧̅ = 0 . (2.67)

Thus, by taking into account the eqns (2.66a) and (2.66b), the 𝐶𝐺 coordinates become

𝑆𝑧̅ 𝑆𝑦̅
𝑦𝐶𝐺 = and 𝑧𝐶𝐺 = . (2.68)
𝐴 𝐴

Therefore, by placing the global coordinate system’s origin 𝐶 coincident with 𝐶𝐺, eqns (2.66e) to (2.66g)
are rewritten as follows:

𝐼𝑦̅𝑦̅ = (𝐼𝑧𝑧 + 𝑦𝐶 2 𝐴) sin2 𝛼 + (𝐼𝑦𝑦 + 𝑧𝐶 2 𝐴) cos2 𝛼 − 2(𝐼𝑦𝑧 + 𝑦𝐶 𝑧𝐶 𝐴) cos 𝛼 sin 𝛼 , (2.69a)

𝐼𝑧̅𝑧̅ = (𝐼𝑧𝑧 + 𝑦𝐶 2 𝐴) cos2 𝛼 + (𝐼𝑦𝑦 + 𝑧𝐶 2 𝐴) sin2 𝛼 − 2(𝐼𝑧𝑦 + 𝑦𝐶 𝑧𝐶 𝐴) cos 𝛼 sin 𝛼 , (2.69b)

𝐼𝑧̅𝑦̅ = 𝐼𝑦̅𝑧̅ = (𝐼𝑦𝑦 − 𝐼𝑧𝑧 + 𝑧𝐶 2 𝐴 − 𝑦𝐶 2 𝐴) cos 𝛼 sin 𝛼 + (𝐼𝑦𝑧 + 𝑦𝐶 𝑧𝐶 𝐴)(cos2 𝛼 − sin2 𝛼) . (2.69c)

In order to complete the definition of the principal global reference (𝑥̅ , 𝑦̅, 𝑧̅), the orientation of the axes is
now evaluated. The (𝑦̅, 𝑧̅) plane must be rotated through an angle 𝛼, so that

𝐼𝑧̅𝑦̅ = 𝐼𝑦̅𝑧̅ = 0 , (2.70)

and performing a trigonometric treatment, the eqn (2.69c) reduces to

2(𝐼𝑦𝑧 − 𝑦𝐶𝐺 𝑧𝐶𝐺 𝐴)


tan 2𝛼 = , (2.71)
𝐼𝑧𝑧 − 𝐼𝑦𝑦 + 𝑦𝐶𝐺 2 𝐴 − 𝑧𝐶𝐺 2 𝐴

which expression would also arise from Mohr’s circle.

Finally, after placing the global coordinate axes (𝑥̅ , 𝑦̅, 𝑧̅), the sectorial pole 𝑃̅ and the sectorial origin 𝑂̅ at
their principal positions, the coupled governing equations (2.51) can now be stated in a more simplistic
manner as

0 0 0 0 𝜉𝑥̈ 𝐴 0 0 0 𝜉𝑥̈ ′ 𝐴 0 0 0 𝜉𝑥′′′


0 𝐴 0 𝑆𝑧̅𝑃 𝜉𝑧̈ 0 𝐼𝑦̅𝑦̅ 0 0 𝜉𝑧̈ ′′ 0 𝐼𝑦̅𝑦̅ 0 0 𝜉𝑧
′′′′
𝜌 𝐴 −𝑆𝑦𝑃̅ −𝜌[ ] ̈ ′′ + 𝐸 [ ] ′′′′
0 0 𝜉𝑦̈ 0 0 𝐼𝑧̅𝑧̅ 0 𝜉𝑦 0 0 𝐼𝑧̅𝑧̅ 0 𝜉𝑦
𝑃
[0 𝑆𝑧̅ −𝑆𝑦𝑃̅ 𝑃 𝑃 0 𝐼𝜔̅𝜔̅ 𝜙̈ ′′ 0 𝐼𝜔̅𝜔̅ {𝜙 ′′′′ }
𝐼𝑧̅𝑧̅ + 𝐼𝑦𝑦 ] { 𝜙̈ } 0 0 { } 0 0

0 0 0 0 𝜉𝑥
′ −𝑓𝑥′
0 0 0 0 𝜉𝑧
′′ 𝑓𝑧 + 𝑚𝑦′
−𝐺[ ]
0 0 𝜉𝑦′′ = . (2.72)
0 0 𝑓𝑦 − 𝑚𝑧′
0 0 0 𝐽 {𝜙 ′′ } ′
{𝑚𝑥 − 𝑚𝜔 }

25
3
Development of a
Finite Element
Solution

A numerical solution for the coupled governing differential equations of thin walled beams of
arbitrary cross section is presented in this chapter. A beam finite element is derived for the static
and the dynamic analysis of a thin walled girder considering the cross-sectional warping. The proposed
formulation deals with the dynamic response of multi-span bridges due to the passage of eccentric loads,
considering both the mass associated with the bridge’s girder and with the moving-vehicles.

The solution of the problem consists in determining the beam’s generalized displacements, which are used
to evaluate the stresses and the strains of the structural element. The finite element method yields a set of
algebraic equations that describes the unknown quantities at a discrete number of points, instead of
solving the differential equations for the entire domain. This is performed by a two-step procedure.
Firstly, the bridge’s girder is divided into finite elements, which are connected to each other by nodes. The
Galerkin’s method is applied carrying all time-dependent terms in the formulation, which results in a
semidiscrete system of equations, i.e., a set of ordinary differential equations in time. Secondly, the
semidiscrete equations are integrated in time using Newmark’s β method, which allows their conversion
into a set of algebraic equations written for the nodal values at different time steps, [Reddy, 2006].

The beam finite element that will be developed considers the six nodal degrees of freedom of a three-
dimensional Euler-Bernoulli beam and an additional warping degree of freedom, which represents the
out-of-plane displacements of the cross sections when a beam is loaded in torsion, as depicted in Figure
3.1. The assembly of the seven degrees of freedom will be referred to the centre of gravity, which
motivates the presence of coupling terms in the element matrices.

27
Figure 3.1 – thin walled beam element and the corresponding seven nodal generalized
displacements referred to the centre of gravity.

3.1 THE FINITE ELEMENT EQUATIONS

The developed finite element approximation takes into account a space-time decoupled formulation
where spatial and time variations are treated separately. Therefore, the polynomial approximation 𝑢𝒩 of
the solution 𝑢 of each displacement field variable is considered to be of the form
𝑛𝑑𝑜𝑓
𝑢(𝑥, 𝑡) ≈ 𝑢𝒩 (𝑥, 𝑡) = ∑ 𝒩𝑗 (𝑥)𝑢𝑗 (𝑡) , (3.1)
𝑗=1

where 𝒩𝑗 are linear independent polynomial approximation functions in the spatial coordinate 𝑥 , 𝑢𝑗
represents the nodal generalized displacements which depend upon time 𝑡 , and 𝑛𝑑𝑜𝑓 is the number of
degrees of freedom of the physical problem. Notice that, for linear problems, the convergence to the
exact results depends upon the refinement of the finite element mesh, provided that the degree of the
polynomial approximation functions is sufficient to comply with the compatibility conditions. Some
fundamentals on the mathematical basis for convergence can be found in [Reddy, 2002].

In order to obtain the discrete finite element equations written for the displacement and force nodal
values, the partial differential equations (2.50) are restated in an integral form called the weak form, which
is equivalent to the governing equations and boundary conditions. The polynomial approximation
functions 𝒩𝑗 of the displacement field must satisfy the displacement boundary conditions and have a
degree such that continuity requirements over the domain are verified. This implies that the derivatives of
the approximation functions have to be at least 𝐶 0 functions and therefore 𝒩𝑗 must be differentiable with
respect to 𝑥 as many times as called for in the weak form.

After performing some mathematical treatments, described in more detail in the following, the weak form
leads to the element equations, written in the matrix form as

𝑴𝑒 𝒖̈ 𝑒 + 𝑪𝑒 𝒖̇ 𝑒 + 𝑲𝑒 𝒖𝑒 = 𝒇𝑒 , (3.2)

28
being 𝑴𝑒 the element mass matrix, 𝑪𝑒 the element damping matrix, 𝑲𝑒 the element stiffness matrix, and 𝒇𝑒
the element external force vector. These element vectorial quantities are firstly derived for a finite element
by taking into account each nodal degree of freedom at a time. Afterwards, the resulting matrices and
vectors are assembled.

Prior to further considerations, some remarks concerning the boundary conditions and the approximation
functions are provided. Let the one-dimensional domain 𝛺 be subjected to a set of specified boundary
conditions at the two end nodes. These conditions can consist of specified applied loads ℒ̅ and prescribed
̅ . The portion of the boundary 𝛤 where the displacements are specified is referred to as
displacements 𝒟
the kinematic boundary, and it is denoted by 𝛤𝒟 . Its complement is called the static boundary, being
denoted by 𝛤ℒ . Considering that the developed beam finite element has seven degrees of freedom per
node, the following boundary conditions are defined:

𝑛𝑥 ℒ|𝛤ℒ = ℒ̅ with ℒ = 𝑁, 𝑉𝑧 , 𝑉𝑦 , 𝑀𝑦 , 𝑀𝑧 , 𝑀𝑥 , 𝑀𝜔 , (3.3a)


̅
𝒟|𝛤𝒟 = 𝒟 with 𝒟 = 𝜉𝑥 , 𝜉𝑧 , 𝜉𝑦 , 𝜉𝑧′ , 𝜉𝑦′ , 𝜙, 𝜙 ′ , (3.3b)

where 𝑛𝑥 is the cosine of the angle between the 𝑥 axis and the normal to the boundary. In this case, 𝑛𝑥
equals either -1, at the initial node, or 1, at the final node of the spatial subdomain. Notice that prescribed
quantities are identified with overbars.

At each node seven boundary conditions are required, but only some combinations are possible. Consider
the following complementary relations:

𝛤𝑁 ∩ 𝛤𝜉𝑥 = 𝛤𝑉𝑧 ∩ 𝛤𝜉𝑧 = 𝛤𝑉𝑦 ∩ 𝛤𝜉𝑦 = 𝛤𝑀𝑦 ∩ 𝛤𝜉𝑧′ = 𝛤𝑀𝑧 ∩ 𝛤𝜉𝑦′ = 𝛤𝑀𝑥 ∩ 𝛤𝜙 = 𝛤𝑀𝜔 ∩ 𝛤𝜙′ = ∅ , (3.4a)

𝛤𝑁 ∪ 𝛤𝜉𝑥 ∪ 𝛤𝑉𝑧 ∪ 𝛤𝜉𝑧 ∪ 𝛤𝑉𝑦 ∪ 𝛤𝜉𝑦 ∪ 𝛤𝑀𝑦 ∪ 𝛤𝜉𝑧′ ∪ 𝛤𝑀𝑧 ∪ 𝛤𝜉𝑦′ ∪ 𝛤𝑀𝑥 ∪ 𝛤𝜙 ∪ 𝛤𝑀𝜔 ∪ 𝛤𝜙′ = 𝛤 . (3.4b)
𝑒
The approximate solution 𝑢𝒩 for the element nodal generalized displacements, say 𝜉𝑥𝑒 , 𝜉𝑧𝑒 , 𝜉𝑦𝑒 and 𝜙 𝑒 , and
the corresponding derivatives when required, are written in the form
𝑒 (𝑥,
𝑢𝒩 𝑡) = 𝓝𝑒𝑢 (𝑥)𝒖𝑒𝑢 (𝑡) , (3.5)

being the element nodal displacements vectors 𝒖𝑒𝑢 defined by

𝒖𝑒𝜉𝑥 (𝑡) = [𝜉𝑥1


𝑒 (𝑡) 𝑒 (𝑡)]𝑇
𝜉𝑥2 or (3.6a)
𝑇
𝑑𝜂1𝑒 𝑑𝜂2𝑒
𝒖𝑒𝜂 (𝑡) = [𝜂1𝑒 (𝑡) (𝑡) 𝜂2𝑒 (𝑡) (𝑡)] with 𝜂𝑒 (𝑡) = 𝜉𝑧𝑒 (𝑡), 𝜉𝑦𝑒 (𝑡), 𝜙 𝑒 (𝑡) , (3.6b)
𝑑𝑥 𝑑𝑥

where the subscripts 1 and 2 represent the initial and the final element node, respectively.

The element approximation functions matrices 𝓝𝑒𝑢 differ from each other in dimension and order of their
functions. Considering the axial extension of a one-dimensional linear element, the derivatives in the weak
form are of order one. For this reason, the approximation of the axial displacement field is only possible
whether the approximate functions are 𝐶1 continuously differentiable, as they are required to be
differentiable at least once in the element’s domain. The corresponding element approximation functions
𝒩𝜉𝑒𝑥 are then composed by Lagrange interpolation functions, since they only have to interpolate function

29
values and not their derivatives, see Figure 3.2. A linear polynomial approximation is suitable, and
therefore,
𝑒
𝓝𝑒𝜉𝑥 (𝑥) = [𝒩𝜉𝑥 1 (𝑥) 𝒩𝜉𝑒𝑥 2 (𝑥)] = 1 [1 − 𝜁 1 + 𝜁] , (3.7a)
2

𝑑𝓝𝑒𝜉𝑥 𝑑𝒩 𝑒 𝑑𝒩𝜉𝑒𝑥 2 2 𝑑𝒩𝜉𝑒𝑥 1 𝑑𝒩𝜉𝑒𝑥 2 1


(𝑥) = [ 𝜉𝑥 1 (𝑥) (𝑥)] = 𝑒 [ ]= [−1 1] , (3.7b)
𝑑𝑥 𝑑𝑥 𝑑𝑥 𝐿 𝑑𝜁 𝑑𝜁 𝐿𝑒

where an element natural coordinate 𝜁 was introduced. The transformation mapping between 𝜁 and the
local axial coordinate system 𝑥 𝑒 for each element is

2𝑥 𝑒
𝜁= −1 with 𝑥 𝑒 ∈ [0, 𝐿𝑒 ] and thus, 𝜁 ∈ [−1,1] . (3.8)
𝐿𝑒

𝜉𝑥 (𝑥, 𝑡)
𝜉𝑥 (𝑥, 𝑡) 𝜉𝑥𝑒 (𝑥, 𝑡) = 𝒩𝜉𝑒𝑥1 (𝑥)𝜉𝑥1
𝑒 (𝑡)
+ 𝒩𝜉𝑒𝑥2 (𝑥)𝜉𝑥2
𝑒 (𝑡)
𝒩𝜉𝑒𝑥1 𝒩𝜉𝑒𝑥2

𝒩𝜉𝑒𝑥1 (𝑥)𝜉𝑥1
𝑒 (𝑡)
𝒩𝜉𝑒𝑥2 (𝑥)𝜉𝑥2
𝑒 (𝑡)

𝑒 (𝑡)
𝜉𝑥2
𝑒 (𝑡)
𝜉𝑥1

𝑥
(a) (b)
Figure 3.2 – (a) linear approximation functions 𝒩𝜉𝑒𝑥 and
(b) finite element approximation of the axial displacements 𝜉𝑥𝑒 .

Considering the bending and torsion of a thin walled beam, the approximation functions are required to
be twice differentiable in the element’s domain, since both the displacements and the derivatives have to
be continuous at the nodes. As shown in Figure 3.3, the Hermite cubic interpolation functions are suitable
𝐶2 functions and can be expressed in terms of the 𝜁 coordinate as
𝑇
2(2 + 𝜁)(1 − 𝜁)2
𝑒 𝑒 1 𝐿𝑒 (1 + 𝜁)(1 − 𝜁)2
𝓝𝑒𝜉𝑧 (𝑥) = 𝓝𝑒𝜉𝑦 (𝑥) = 𝓝𝑒𝜙 (𝑥) = [𝒩𝜉𝑧 1 (𝑥) 𝒩𝜉𝑧′ 1 (𝑥) 𝒩𝜉𝑒𝑧 2 (𝑥) 𝒩𝜉𝑒𝑧′ 2 (𝑥)] = , (3.9a)
8 2(2 − 𝜁)(1 + 𝜁)2
𝑒 2
{𝐿 (−1 + 𝜁)(1 + 𝜁) }

being the corresponding first and second derivatives with respect to 𝑥 given by
𝑇
2(−3 + 3𝜁 2 )
𝑑𝓝𝑒𝜉𝑧 𝑑𝓝𝑒𝜉𝑦 𝑑𝓝𝑒𝜙 2 𝑑𝒩𝜉𝑒𝑧 1 𝑑𝒩𝜉𝑒𝑧′ 1 𝑑𝒩𝜉𝑒𝑧 2 𝑑𝒩𝜉𝑒𝑧′ 21 𝐿𝑒 (−1 − 2𝜁 + 3𝜁 2 )
(𝑥) = (𝑥) = (𝑥) = [ ]= 𝑒 , (3.9b)
𝑑𝑥 𝑑𝑥 𝑑𝑥 𝐿𝑒 𝑑𝜁 𝑑𝜁 𝑑𝜁 𝑑𝜁 4𝐿 2(3 − 3𝜁 2 )
𝑒 (−1 + 2𝜁 + 3𝜁 2 )
{𝐿 }
𝑇
6𝜁
𝑑2 𝓝𝑒𝜉𝑧 𝑑2 𝓝𝑒𝜉𝑦 𝑑2 𝓝𝑒𝜙 4 𝑑2 𝒩𝜉𝑒𝑧 1 𝑑2 𝒩𝜉𝑒𝑧′ 1 𝑑2 𝒩𝜉𝑒𝑧 2 𝑑2 𝒩𝜉𝑒𝑧′ 2 1 𝐿𝑒 (3𝜁
− 1)
= = = [ ] = 𝑒2 { } . (3.9b)
𝑑𝑥 2 𝑑𝑥 2 𝑑𝑥 2 𝐿𝑒 2 𝑑𝜁 2 𝑑𝜁 2 𝑑𝜁 2 𝑑𝜁 2 𝐿 −6𝜁
𝑒
𝐿 (3𝜁 + 1)

30
𝜂(𝑥, 𝑡) 𝜂𝑒 (𝑥, 𝑡) = 𝒩𝜂1
𝑒 (𝑥)𝜂 𝑒 (𝑡) 𝑒 (𝑥)𝜂′𝑒 (𝑡)
+ 𝒩𝜂′1 𝑒 (𝑥)𝜂 𝑒 (𝑡)
+ 𝒩𝜂2 𝑒 (𝑥)𝜂′𝑒 (𝑡)
+ 𝒩𝜂′2
𝑒 1 1 2 2
𝒩𝜂′1

𝜂(𝑥, 𝑡)
𝑒 (𝑥)𝜂 𝑒 (𝑡) 𝑒 (𝑥)𝜂 𝑒 (𝑡)
𝒩𝜂1 1 𝒩𝜂2 2
𝑒
𝒩𝜂1 𝑒 (𝑥)𝜂′𝑒 (𝑡)
𝑒
𝒩𝜂′2 𝒩𝜂′1 1
𝑒
𝒩𝜂2 𝜂1𝑒 (𝑡)

𝜂2𝑒 (𝑡)
𝜂′1𝑒 (𝑡)

𝜂′𝑒2 (𝑡) 𝑥
𝑒 (𝑥)𝜂′𝑒 (𝑡)
𝒩𝜂′2 2

(a) (b)
Figure 3.3 – (a) Hermite cubic approximation functions 𝒩𝜂𝑒 and
(b) finite element approximation of the generalized displacements 𝜂𝑒 .

The weak formulation is presented in the sequel considering separately each governing equation. The axial
extension problem is firstly developed. Thus, considering the eqn (2.50a), multiplying it with a weight
function 𝜗 and integrating over the domain 𝛺, the weighted-integral statement of the governing equation
is given by

𝜕𝜉𝑧̈ 𝜕𝜉𝑦̈ 𝜕𝜙̈ 𝜕𝑁


∫ 𝜗 (𝜌𝐴𝜉𝑥̈ − 𝜌𝑆𝑦 − 𝜌𝑆𝑧 − 𝜌𝑆𝜔 − − 𝑓𝑥 ) 𝑑𝑥 = 0 , ∀𝜗 , (3.10)
𝛺 𝜕𝑥 𝜕𝑥 𝜕𝑥 𝜕𝑥

where

𝜕𝑁 𝜕 2 𝜉𝑥 𝜕 3 𝜉𝑧 𝜕 3 𝜉𝑦 𝜕3𝜙
= 𝐸𝐴 2
− 𝐸𝑆𝑦 3
− 𝐸𝑆𝑧 3
− 𝐸𝑆𝜔 3 . (3.11)
𝜕𝑥 𝜕𝑥 𝜕𝑥 𝜕𝑥 𝜕𝑥

Integrating by parts the term in 𝑁 in eqn (3.10), in order to trade differentiation from the displacement
field variables 𝑢 to the weight function 𝜗, it follows that

𝜕𝜉𝑧̈ 𝜕𝜉𝑦̈ 𝜕𝜙̈ 𝜕𝜗


∫ 𝜗 (𝜌𝐴𝜉𝑥̈ − 𝜌𝑆𝑦 − 𝜌𝑆𝑧 − 𝜌𝑆𝜔 ) 𝑑𝑥 − [𝜗𝑛𝑥 𝑁]𝛤𝑁 + ∫ 𝑁 𝑑𝑥 − ∫ 𝜗𝑓𝑥 𝑑𝑥 = 0 , ∀𝜗 , (3.12)
𝛺 𝜕𝑥 𝜕𝑥 𝜕𝑥 𝛺 𝜕𝑥 𝛺

where the term in square brackets represents the boundary conditions at the two end nodes of the beam.
This procedure not only weakens the continuity requirements on the approximation functions 𝒩𝑗 that will
be considered, but also leads to symmetric element matrices, [Fish and Belytschko, 2007].

Consider that an axial force 𝑁̅ is applied at the static boundary 𝛤𝑁 . At the kinematic boundary 𝛤𝜉𝑥 , null axial
displacements are prescribed, being the weight function 𝜗 required to be null as well. The corresponding
equilibrium and compatibility conditions are given by

̅,
𝑛𝑥 𝑁|𝛤𝑁 = 𝑁 (3.13a)
𝜉𝑥 |𝛤𝜉𝑥 = 𝜗|𝛤𝜉𝑥 = 0 . (3.13b)

Applying the same procedure as described for the governing equation (3.10), the static boundary
condition (3.13a) is weighted as follows:

̅ )]𝛤 = 0 , ∀𝜗 .
[𝜗(𝑛𝑥 𝑁 − 𝑁 (3.14)
𝑁

31
Using the eqns (3.12) and (3.14), the weak form of the differential equation (2.50a) can be rewritten as

𝜕𝜉𝑧̈ 𝜕𝜉𝑦̈ 𝜕𝜙̈ 𝜕𝜗


∫ 𝜗 (𝜌𝐴𝜉𝑥̈ − 𝜌𝑆𝑦 − 𝜌𝑆𝑧 − 𝜌𝑆𝜔 ) 𝑑𝑥 + ∫ ̅ ]𝛤 = 0 , ∀𝜗: 𝜗|𝛤 = 0 .
𝑁 𝑑𝑥 − ∫ 𝜗𝑓𝑥 𝑑𝑥 − [𝜗𝑁 (3.15)
𝛺 𝜕𝑥 𝜕𝑥 𝜕𝑥 𝛺 𝜕𝑥 𝛺
𝑁 𝜉𝑥

Consider a mesh of linear elements and a generic element’s domain 𝛺𝑒 located between the global nodes
𝑥𝑒 and 𝑥𝑒+1. The integrations in eqn (3.15) for the whole domain 𝛺 must be carried out over the element
domains ]𝑥1𝑒 , 𝑥2𝑒 [ as a sum of 𝑛𝑒𝑙 integrals, being 𝑛𝑒𝑙 the number of finite elements in which the beam is
discretized. Thus, one obtains

𝑛𝑒𝑙 𝜕𝜉𝑧̈ 𝑒 𝜕𝜉𝑦̈ 𝑒 𝜕𝜙̈ 𝑒 𝜕𝜗 𝑒 𝑒 𝑒 𝜕𝜉𝑥𝑒


∑ {∫ 𝜗 𝑒 𝜌𝑒 𝐴𝑒 𝜉𝑥̈ 𝑒 𝑑𝑥 − ∫ 𝜗 𝑒 𝜌𝑒 𝑆𝑦𝑒 𝑑𝑥 − ∫ 𝜗 𝑒 𝜌𝑒 𝑆𝑧𝑒 𝑑𝑥 − ∫ 𝜗 𝑒 𝜌𝑒 𝑆𝜔
𝑒
𝑑𝑥 + ∫ 𝐸 𝐴 𝑑𝑥
𝑒=1 𝛺𝑒 𝛺𝑒 𝜕𝑥 𝛺𝑒 𝜕𝑥 𝛺𝑒 𝜕𝑥 𝛺𝑒 𝜕𝑥 𝜕𝑥

𝜕𝜗 𝑒 𝑒 𝜕 2 𝜉𝑧𝑒
𝑒
𝜕𝜗 𝑒 𝑒 𝑒 𝜕 2 𝜉𝑦𝑒 𝜕𝜗 𝑒 𝑒 𝑒 𝜕 2 𝜙 𝑒
−∫ 𝐸 𝑆𝑦 𝑑𝑥 − ∫ 𝐸 𝑆𝑧 𝑑𝑥 − ∫ 𝐸 𝑆𝜔 𝑑𝑥 − ∫ 𝜗 𝑒 𝑓𝑥𝑒 𝑑𝑥
𝛺𝑒 𝜕𝑥 𝜕𝑥 2 𝛺𝑒 𝜕𝑥 𝜕𝑥 2 𝛺𝑒 𝜕𝑥 𝜕𝑥 2 𝛺𝑒

̅ 𝑒 ]𝛤 𝑒 } = 0 , ∀𝜗𝑒 : 𝜗 𝑒 |𝛤 𝑒 = 0 .
− [𝜗 𝑒 𝑁 (3.16)
𝑁 𝜉 𝑥

Notice that hereafter in this chapter, element numbers are denoted by numerical superscripts, whereas a
subscript represents the element-related node number.

By making use of the Galerkin’s method, the weight functions 𝜗 are selected to be the same functions
used for the approximated solutions 𝑢𝒩 , i.e.,
𝑇
𝜗 𝑒 (𝑥, 𝑡) = 𝓝𝑒𝜉𝑥 (𝑥)𝝑𝑒𝜉𝑥 (𝑡) with 𝝑𝑒𝜉𝑥 = [𝜗𝜉𝑒𝑥1(𝑡) 𝜗𝜉𝑒𝑥2(𝑡)] , (3.17)

and therefore the eqn (3.16) reduces to

𝑛𝑒𝑙
𝑇 𝑇 𝑑2 𝒖𝑒𝜉𝑥 𝑇 𝑑𝓝𝑒𝜉𝑧 𝑑2 𝒖𝑒𝜉𝑧 𝑇
𝑑𝓝𝑒𝜉𝑦 𝑑2 𝒖𝑒𝜉𝑦
∑ 𝝑𝑒𝜉𝑥 {∫ 𝓝𝑒𝜉𝑥 𝜌𝑒 𝐴𝑒 𝓝𝑒𝜉𝑥 𝑑𝑥 − ∫ 𝓝𝑒𝜉𝑥 𝜌𝑒 𝑆𝑦𝑒 𝑑𝑥 − ∫ 𝓝𝑒𝜉𝑥 𝜌𝑒 𝑆𝑧𝑒 𝑑𝑥
𝑒=1 𝛺𝑒 𝑑𝑡 2 𝛺𝑒 𝑑𝑥 𝑑𝑡 2 𝛺𝑒 𝑑𝑥 𝑑𝑡 2
𝑇
𝑇 𝑑𝓝𝑒𝜙 𝑑2 𝒖𝑒𝜙 𝑑𝓝𝑒𝜉𝑥 𝑑𝓝𝑒𝜉𝑥
− ∫ 𝓝𝑒𝜉𝑥 𝜌𝑒 𝑆𝜔
𝑒 𝑑𝑥 +∫ ( ) 𝐸 𝑒 𝐴𝑒 𝑑𝑥𝒖𝑒𝜉𝑥
𝛺𝑒 𝑑𝑥 𝑑𝑡 2 𝛺𝑒 𝑑𝑥 𝑑𝑥
𝑇 𝑇
𝑑𝓝𝑒𝜉𝑥 𝑑2 𝓝𝑒𝜉𝑧 𝑑𝓝𝑒𝜉𝑥 𝑑2 𝓝𝑒𝜉𝑦
𝑒
−∫ ( ) 𝐸 𝑒 𝑆𝑦𝑒 𝑑𝑥𝒖𝜉𝑧 − ∫ ( ) 𝐸 𝑒 𝑆𝑒
𝑧 𝑑𝑥𝒖𝑒𝜉𝑦
𝛺𝑒 𝑑𝑥 𝑑𝑥 2 𝛺𝑒 𝑑𝑥 𝑑𝑥 2
𝑇
𝑑𝓝𝑒𝜉𝑥 𝑑2 𝓝𝑒𝜙 𝑒 𝑒 𝑇 𝑒 𝑒 ̅𝑒
2 𝑑𝑥𝒖𝜙 − ∫ 𝑒 𝓝𝜉𝑥 𝑓𝑥 𝑑𝑥 − [𝓝𝜉𝑥 𝑁 ]𝛤𝑒 } = 0 .
−∫ ( ) 𝐸 𝑒 𝑆𝜔
𝑒
(3.18)
𝛺 𝑒 𝑑𝑥 𝑑𝑥 𝛺 𝑁

For consistency, the weight function terms were substituted in (3.18) for its transpose, as 𝜗 is now a
vectorial quantity.

The contributions 𝑴𝑒𝜉𝑥 , 𝑲𝑒𝜉𝑥 and 𝒇𝑒𝜉𝑥 to the element mass matrix 𝑴𝑒 , to the element stiffness matrix 𝑲𝑒 , and
to the element external force vector 𝒇𝑒 , respectively, can be identified from eqn (3.18), being written as
follows:

𝑑𝓝𝑒𝜉𝑧
𝑴𝑒𝜉𝑥 = [∫ 𝓝𝑒𝜉𝑥 𝑇 𝜌𝑒 𝐴𝑒 𝓝𝑒𝜉𝑥 𝑑𝑥 − ∫ 𝓝𝑒𝜉𝑥 𝜌𝑒 𝑆𝑦𝑒
𝑇
𝑑𝑥 ⋯
𝛺𝑒 𝛺𝑒 𝑑𝑥
𝑑𝓝𝑒𝜉𝑦 𝑑𝓝𝑒𝜙
≪ ⋯ − ∫ 𝓝𝑒𝜉 𝑇 𝜌𝑒 𝑆𝑧𝑒 𝑑𝑥 − ∫ 𝓝𝑒𝜉𝑥 𝜌𝑒 𝑆𝜔
𝑒 𝑇
𝑑𝑥 ] , (3.19a)
𝑥
𝛺𝑒 𝑑𝑥 𝛺𝑒 𝑑𝑥

32
and

𝑇 𝑇
𝑑𝓝𝑒𝜉𝑥 𝑑𝓝𝑒𝜉𝑥 𝑑𝓝𝑒𝜉𝑥 𝑑2 𝓝𝑒𝜉𝑧
𝑲𝑒𝜉𝑥 = [∫ ( ) 𝐸 𝑒 𝐴𝑒 𝑑𝑥 −∫ ( ) 𝐸 𝑒 𝑆𝑦𝑒 𝑑𝑥 ⋯
𝛺𝑒 𝑑𝑥 𝑑𝑥 𝛺𝑒 𝑑𝑥 𝑑𝑥 2
𝑇 𝑇
𝑑𝓝𝑒𝜉𝑥 𝑑2 𝓝𝑒𝜉𝑦 𝑑𝓝𝑒𝜉𝑥 𝑑2 𝓝𝑒𝜙
<⋯ −∫ ( ) 𝐸 𝑒 𝑆𝑧𝑒 𝑑𝑥 −∫ ( ) 𝐸 𝑒 𝑆𝜔
𝑒
𝑑𝑥 ] , (3.19b)
𝛺𝑒 𝑑𝑥 𝑑𝑥 2 𝛺𝑒 𝑑𝑥 𝑑𝑥 2

𝑇 ̅𝑒]
𝒇𝑒𝜉𝑥 = 𝒇𝑒𝜉𝑥 𝛺 + 𝒇𝑒𝜉𝑥 𝛤 = ∫ 𝓝𝑒𝜉𝑥 𝑓𝑥𝑒 𝑑𝑥 + [𝓝𝑒𝜉𝑥 𝑁 . (3.19c)
𝛤𝑁𝑒
𝛺𝑒

Considering the approximation functions defined in (3.7) and (3.9), the contribution to the element mass
matrix 𝑴𝑒 (3.19a) becomes
𝑇
2(−3 + 3𝜁 2 )
𝜁=+1 𝜁=+1
1 1 − 𝜁 𝑒 𝑒 1 1 − 𝜁 𝑇 𝐿𝑒 1 1 − 𝜁 𝑒 𝑒 1 𝐿 (−1 − 2𝜁 + 3𝜁 2 ) 𝐿𝑒
𝑒
𝑴𝑒𝜉𝑥 = ∫ { }𝜌 𝐴 { } 𝑑𝜁 −∫ { } 𝜌 𝑆𝑦 𝑒 𝑑𝜁 ⋯
𝜁=−1 2 1+𝜁 2 1+𝜁 2 𝜁=−1 2
1+𝜁 4𝐿 2(3 − 3𝜁 2 ) 2
𝑒 2
[ {𝐿 (−1 + 2𝜁 + 3𝜁 )}
𝑇
2(−3 + 3𝜁 2 )
𝜁=+1
1 1 − 𝜁 𝑒 𝑒 1 𝐿 (−1 − 2𝜁 + 3𝜁 2 ) 𝐿𝑒
𝑒
< ⋯ −∫ { } 𝜌 𝑆𝑧 𝑒 𝑑𝜁 ⋯<
𝜁=−1 2
1+𝜁 4𝐿 2(3 − 3𝜁 2 ) 2
𝑒 2
{𝐿 (−1 + 2𝜁 + 3𝜁 )}
𝑇
2(−3 + 3𝜁 2 )
𝜁=+1
1 1 − 𝜁 𝑒 𝑒 1 𝐿 (−1 − 2𝜁 + 3𝜁 2 ) 𝐿𝑒
𝑒
< ⋯ −∫ { } 𝜌 𝑆𝜔 𝑒 𝑑𝜁 , (3.20)
𝜁=−1 2
1+𝜁 4𝐿 2(3 − 3𝜁 2 ) 2
𝑒 2
{𝐿 (−1 + 2𝜁 + 3𝜁 )} ]

and performing the integrations with respect to the element natural coordinate 𝜁 one obtains

𝐿𝑒 𝜌𝑒 𝐴𝑒 𝜌𝑒 𝑆𝑦𝑒 𝜌𝑒 𝑆𝑧𝑒 𝜌𝑒 𝑆𝜔
𝑒
𝑴𝑒𝜉𝑥 = [ 𝒎1 − 𝒎2 − 𝒎2 − 𝒎2 ] , (3.21)
6 12 12 12

where 𝒎1 and 𝒎2 are given by

2 1 𝐿𝑒 −𝐿𝑒
𝒎1 = [ ] and 𝒎2 = [−6 6
] . (3.22)
1 2 −6 −𝐿𝑒 6 𝐿𝑒

Regarding the contribution to the element stiffness matrix 𝑲𝑒 and to the element external body and
boundary force vectors, respectively 𝒇𝑒𝛺 and 𝒇𝑒𝛤 , the following expressions are obtained:

𝐸 𝑒 𝐴𝑒 𝐸 𝑒 𝑆𝑦𝑒 𝐸 𝑒 𝑆𝑧𝑒 𝐸 𝑒 𝑆𝜔
𝑒
𝑲𝑒𝜉𝑥 = [ 𝑒 𝒌1 − 𝒌 − 𝒌 − 𝒌2 ] , (3.23)
𝐿 𝐿𝑒 2 𝐿𝑒 2 𝐿𝑒
𝑇
𝑓𝑥𝑒 𝐿𝑒 𝑓𝑥𝑒 𝐿𝑒
𝒇𝑒𝜉𝑥 = 𝒇𝑒𝜉𝑥 𝛺 + 𝒇𝑒𝜉𝑥 𝛤 = [ − 𝑁1𝑒 + 𝑁2𝑒 ] , (3.24)
2 2

being 𝒌1 and 𝒌2 defined by

1 −1
𝒌1 = [ ] and 𝒌2 = [0 1 0 −1
]. (3.25)
−1 1 0 −1 0 1

33
The procedure to develop the weak forms of the remaining governing equations (2.50) is similar to the
one that has been used so far in the axial extension problem. Considering the torsion partial differential
equation (2.50d) multiplied by an arbitrary weight function 𝜗 and integrating the product over the domain
gives

𝜕 𝜕𝜉𝑧̈ 𝜕𝜉𝑦̈ 𝜕𝜙̈


∫ 𝜗[𝜌𝑆𝑧𝑃 𝜉𝑧̈ − 𝜌𝑆𝑦𝑃 𝜉𝑦̈ + 𝜌(𝐼𝑦𝑦
𝑃 + 𝐼𝑃 )𝜙̈ ] 𝑑𝑥 + ∫
𝑧𝑧 (𝜗𝜌𝑆𝜔 𝜉𝑥̈ − 𝜗𝜌𝐼𝜔𝑦 − 𝜗𝜌𝐼𝜔𝑧 − 𝜗𝜌𝐼𝜔𝜔 ) 𝑑𝑥
𝛺 𝛺 𝜕𝑥 𝜕𝑥 𝜕𝑥 𝜕𝑥
𝜕𝜗 𝜕𝜉𝑧̈ 𝜕𝜉𝑦̈ 𝜕𝜙̈ 𝜕 𝜕𝑀𝜔
−∫ (𝜌𝑆𝜔 𝜉𝑥̈ − 𝜌𝐼𝜔𝑦 − 𝜌𝐼𝜔𝑧 − 𝜌𝐼𝜔𝜔 ) 𝑑𝑥 + ∫ (𝜗 ) 𝑑𝑥
𝛺 𝜕𝑥 𝜕𝑥 𝜕𝑥 𝜕𝑥 𝛺 𝜕𝑥 𝜕𝑥

𝜕𝜗 𝜕𝑀𝜔 𝜕 𝜕𝜗 𝑆
−∫ 𝑑𝑥 − ∫ (𝜗𝑀𝑥𝑆 ) 𝑑𝑥 + ∫ 𝑀𝑥 𝑑𝑥 − ∫ 𝜗𝑚𝑥 𝑑𝑥
𝛺 𝜕𝑥 𝜕𝑥 𝛺 𝜕𝑥 𝛺 𝜕𝑥 𝛺

𝜕 𝜕𝜗
+∫ (𝜗𝑚𝜔 ) 𝑑𝑥 − ∫ 𝑚𝜔 𝑑𝑥 = 0 , ∀𝜗 , (3.26)
𝛺 𝜕𝑥 𝛺 𝜕𝑥

where 𝑀𝑥𝑆 was defined in eqn (2.21), and

𝜕 2 𝑀𝜔 𝜕 3 𝜉𝑥 𝜕 4 𝜉𝑧 𝜕 4 𝜉𝑦 𝜕4𝜙
= −𝐸𝑆𝜔 + 𝐸𝐼𝜔𝑦 + 𝐸𝐼𝜔𝑧 + 𝐸𝐼𝜔𝜔 . (3.27)
𝜕𝑥 2 𝜕𝑥 3 𝜕𝑥 4 𝜕𝑥 4 𝜕𝑥 4

The static boundary conditions 𝛤𝑀𝑥 and 𝛤𝑀𝜔 are weighted as follows:

̅𝑥 )]𝛤 = 0 , ∀𝜗 ,
[𝜗(𝑛𝑥 𝑀𝑥 − 𝑀 (3.28a)
𝑀𝑥

𝜕𝜗
̅𝜔 )]
[ (𝑛𝑥 𝑀𝜔 − 𝑀 = 0 , ∀𝜗 . (3.28b)
𝜕𝑥 𝛤𝑀𝜔

Considering the conditions (3.28) and that

𝑀𝑥 = 𝑀𝑥𝑆 + 𝑀𝑥𝜔 , (3.29a)


𝜕𝑀𝜔
= −𝑀𝑥𝜔 − 𝑚𝜔 , (3.29b)
𝜕𝑥

the weak form for the torsion problem is stated as follows:

𝜕𝜗 𝜕𝜉𝑥̈ 𝜕 2 𝜉𝑧̈ 𝜕 2 𝜉𝑦̈ 𝜕 2 𝜙̈


∫ 𝜗[𝜌𝑆𝑧𝑃 𝜉𝑧̈ − 𝜌𝑆𝑦𝑃 𝜉𝑦̈ + 𝜌(𝐼𝑧𝑧
𝑃 + 𝐼 𝑃 )𝜙̈ ] 𝑑𝑥 − ∫
𝑦𝑦 (𝜌𝑆𝜔 − 𝜌𝐼𝜔𝑦 2
− 𝜌𝐼𝜔𝑧 2
− 𝜌𝐼𝜔𝜔 2 ) 𝑑𝑥 (3.30)
𝛺 𝛺 𝜕𝑥 𝜕𝑥 𝜕𝑥 𝜕𝑥 𝜕𝑥

𝜕2𝜗 𝜕𝜗 𝑆 𝜕𝜗 𝜕𝜗
+∫ 𝑀 𝑑𝑥 + ∫ 𝑀𝑥 𝑑𝑥 − ∫ 𝜗𝑚𝑥 𝑑𝑥 − ∫ ̅𝑥 ]𝛤 − [ 𝑀
𝑚𝜔 𝑑𝑥 − [𝜗𝑀 ̅ ]
𝛺 𝜕𝑥 2 𝜔 𝛺 𝜕𝑥 𝛺 𝛺 𝜕𝑥 𝑀𝑥
𝜕𝑥 𝜔 𝛤𝑀𝜔

𝜕𝜉𝑧̈̅ 𝜕𝜉𝑦̈̅ 𝜕𝜙̅̈


+ [𝜗𝜌𝑆𝜔 𝜉𝑥̈̅ ] − [𝜗𝜌𝐼𝜔𝑦 ] − [𝜗𝜌𝐼𝜔𝑧 ] − [𝜗𝜌𝐼𝜔𝜔 ] = 0 , ∀𝜗: 𝜗|𝛤𝜙 = 𝜗|𝛤𝜙′ = 0 .
𝛤𝜉̈
𝑥
𝜕𝑥 𝜕𝑥 𝜕𝑥
𝛤𝜉̈′ 𝛤𝜉̈′ 𝛤𝜙̈′
𝑧 𝑦

Since there are no boundary conditions on applied nodal accelerations, the last four terms of eqn (3.30) in
square brackets vanish.

34
Integrating the eqn (3.30) over the element domains ]𝑥1𝑒 , 𝑥2𝑒 [ as a sum of 𝑛𝑒𝑙 integrals gives

𝑛𝑒𝑙 𝜕𝜗 𝑒 𝑒 𝑒 𝑒
∑ {∫ 𝜗 𝑒 𝜌𝑒 𝑆𝑧𝑃𝑒 𝜉𝑧̈ 𝑒 𝑑𝑥 − ∫ 𝜗 𝑒 𝜌𝑒 𝑆𝑦𝑃𝑒 𝜉𝑦̈ 𝑒 𝑑𝑥 + ∫ 𝜗 𝑒 𝜌𝑒 (𝐼𝑧𝑧
𝑃𝑒 𝑃𝑒
+ 𝐼𝑦𝑦 )𝜙̈ 𝑒 𝑑𝑥 − ∫ 𝜌 𝑆𝜔 𝜉𝑥̈ 𝑑𝑥
𝑒=1 𝛺𝑒 𝛺𝑒 𝛺𝑒 𝛺𝑒 𝜕𝑥
𝜕𝜗 𝑒 𝑒 𝑒 𝜕𝜉𝑧̈ 𝑒 𝜕𝜗 𝑒 𝑒 𝑒 𝜕𝜉𝑦̈ 𝑒 𝜕𝜗 𝑒 𝑒 𝑒 𝜕𝜙̈ 𝑒
+∫ 𝜌 𝐼𝜔𝑦 𝑑𝑥 + ∫ 𝜌 𝐼𝜔𝑧 𝑑𝑥 + ∫ 𝜌 𝐼𝜔𝜔 𝑑𝑥
𝛺𝑒 𝜕𝑥 𝜕𝑥 𝛺𝑒 𝜕𝑥 𝜕𝑥 𝛺𝑒 𝜕𝑥 𝜕𝑥
𝜕𝜗 𝑒 𝑒 𝑒 𝜕𝜉𝑥𝑒 𝜕 2 𝜗 𝑒 𝑒 𝑒 𝜕 2 𝜉𝑧𝑒 𝜕 2 𝜗 𝑒 𝑒 𝑒 𝜕 2 𝜉𝑦𝑒
−∫ 𝐸 𝑆𝜔 𝑑𝑥 + ∫ 2
𝐸 𝐼𝜔𝑦 𝑑𝑥 + ∫ 𝐸 𝐼𝜔𝑧 𝑑𝑥
𝛺𝑒 𝜕𝑥 𝜕𝑥 𝛺𝑒 𝜕𝑥 𝜕𝑥 2 𝛺𝑒 𝜕𝑥
2 𝜕𝑥 2
𝜕2 𝜗𝑒 𝑒 𝑒 𝜕2 𝜙𝑒 𝜕𝜗 𝑒 𝑒 𝑒 𝜕𝜙 𝑒 𝜕𝜗 𝑒 𝑒
+∫ 2 𝐸 𝐼𝜔𝜔 2 𝑑𝑥 + ∫ 𝐺 𝐽 𝑑𝑥 − ∫ 𝜗 𝑒 𝑚𝑥𝑒 𝑑𝑥 − ∫ 𝑚𝜔 𝑑𝑥
𝛺𝑒 𝜕𝑥 𝜕𝑥 𝛺𝑒 𝜕𝑥 𝜕𝑥 𝛺𝑒 𝛺𝑒 𝜕𝑥

𝜕𝜗 𝑒 𝑒
̅𝑥𝑒 ]𝛤 𝑒 − [
− [𝜗 𝑒 𝑀 ̅ ] } = 0 , ∀𝜗 𝑒 : 𝜗 𝑒 |𝛤𝑒 = 𝜗 𝑒 |𝛤 𝑒 = 0 .
𝑀 (3.31)
𝑀 𝑥 𝜕𝑥 𝜔 𝛤𝑀𝑒 𝜙 𝜙′
𝜔

Recalling that the Galerkin’s method is being used, the weight functions 𝜗 are such that
𝑒 𝑒 𝑇
𝑑𝜗𝜙1 𝑑𝜗𝜙2 (3.32)
𝜗 𝑒 (𝑥, 𝑡) = 𝓝𝑒𝜙 (𝑥)𝝑𝑒𝜙 (𝑡) with 𝝑𝑒𝜙 (𝑡) = [𝜗𝜙1
𝑒 (𝑡) (𝑡) 𝜗𝜙2
𝑒 (𝑡) (𝑡)] ,
𝑑𝑥 𝑑𝑥

and the contributions to the element matrices result


𝑇 𝑇
𝑑𝓝𝑒𝜙 𝑑𝓝𝑒𝜙 𝑑𝓝𝑒𝜉𝑧
𝑴𝑒𝜙 = [− ∫ ( 𝑒 𝓝𝑒 𝑑𝑥
) 𝜌𝑒 𝑆𝜔 𝜉𝑥
𝑇
∫ 𝓝𝑒𝜙 𝜌𝑒 𝑆𝑧𝑃𝑒 𝓝𝑒𝜉𝑧 𝑑𝑥 + ∫ ( ) 𝜌𝑒 𝐼𝜔𝑦
𝑒 𝑑𝑥 ⋯
𝛺𝑒 𝑑𝑥 𝛺𝑒 𝛺𝑒 𝑑𝑥 𝑑𝑥
𝑇 𝑒
𝑑𝓝𝑒𝜙 𝑑𝓝𝜉𝑦
< ⋯ − ∫ 𝓝𝑒𝜙 𝑇 𝜌𝑒 𝑆𝑦𝑃𝑒 𝓝𝑒𝜉 𝑑𝑥 + ∫ ( ) 𝜌𝑒 𝐼𝜔𝑧
𝑒
𝑑𝑥 ⋯
𝛺𝑒
𝑦
𝛺𝑒 𝑑𝑥 𝑑𝑥
𝑇
𝑑𝓝𝑒𝜙 𝑑𝓝𝑒𝜙
< ⋯ ∫ 𝓝𝑒𝜙 𝑇 𝜌𝑒 (𝐼𝑧𝑧
𝑃𝑒 𝑃𝑒
+ 𝐼𝑦𝑦 )𝓝𝑒𝜙 𝑑𝑥 + ∫ ( ) 𝜌𝑒 𝐼𝜔𝜔
𝑒
𝑑𝑥 ] , (3.33a)
𝛺𝑒 𝛺𝑒 𝑑𝑥 𝑑𝑥

𝑇 𝑇 𝑇 𝑒
𝑑2 𝓝𝑒𝜙 𝑑𝓝𝑒𝜉𝑥 𝑑2 𝓝𝑒𝜙 𝑑2 𝓝𝑒𝜉𝑧 𝑑2 𝓝𝑒𝜙 𝑑2 𝓝𝜉𝑦
𝑲𝑒𝜙 = [− ∫ ( ) 𝐸 𝑒 𝑆𝜔
𝑒
𝑑𝑥 ∫ ( ) 𝐸 𝑒 𝐼𝜔𝑦
𝑒
𝑑𝑥 ∫ ( ) 𝐸 𝑒 𝐼𝜔𝑧
𝑒
𝑑𝑥 ⋯
𝑑𝑥 2 𝑑𝑥 𝑑𝑥 2 𝑑𝑥 2 𝑑𝑥 2 𝑑𝑥 2
𝛺𝑒 𝛺𝑒 𝛺𝑒
𝑇 𝑇
𝑑2 𝓝𝑒𝜙 𝑑2 𝓝𝑒𝜙 𝑑𝓝𝑒𝜙 𝑑𝓝𝑒𝜙
+⋯ ∫ ( ) 𝐸 𝑒 𝐼𝑒
𝜔𝜔 𝑑𝑥 + ∫ ( ) 𝐺 𝑒 𝐽𝑒 𝑑𝑥 ] , (3.33b)
𝛺𝑒 𝑑𝑥 2 𝑑𝑥 2 𝛺𝑒 𝑑𝑥 𝑑𝑥

𝑇 𝑇
𝑇 𝑑𝓝𝑒𝜙 𝑒 𝑑𝑥} − {[𝓝𝑒 𝑇 𝑀
𝑑𝓝𝑒𝜙
𝒇𝑒𝜙 = 𝒇𝑒𝜙𝛺 + 𝒇𝑒𝜙𝛤 = − {∫ 𝓝𝑒𝜙 𝑚𝑥𝑒 𝑑𝑥 + ∫ ( ) 𝑚𝜔 𝜙
̅ 𝑒]
𝑥 + [( ̅𝜔
) 𝑀 𝑒]
} . (3.33c)
𝛺𝑒 𝛺𝑒 𝑑𝑥 𝑒
𝛤𝑀 𝑥 𝑑𝑥 𝑒
𝛤𝑀 𝜔

Proceeding through the integrations as it was performed in the axial motion case, the following matrices
are obtained:

𝜌𝑒 𝑆𝜔
𝑒
𝜌𝑒 𝑆𝑧𝑃𝑒 𝐿𝑒 𝜌𝑒 𝐼𝜔𝑦
𝑒
𝜌𝑒 𝑆𝑦𝑃𝑒 𝐿𝑒 𝜌𝑒 𝐼𝜔𝑧
𝑒 𝜌𝑒 (𝐼𝑧𝑧
𝑃𝑒 + 𝐼𝑃𝑒 )𝐿𝑒
𝜌𝑒 𝐼𝜔𝜔
𝑒
𝑴𝑒𝜙 = [ 𝒎3 𝒎4 + 𝒎 − 𝒎4 + 𝒎
𝑦𝑦
𝒎4 + 𝒎 ] , (3.34a)
12 420 30𝐿𝑒 5 420 30𝐿𝑒 5 420 30𝐿𝑒 5

𝐸𝑒 𝑆𝑒 𝐸 𝑒 𝐼𝜔𝑦
𝑒
𝐸 𝑒 𝐼𝜔𝑧
𝑒
𝐸 𝑒 𝐼𝜔𝜔
𝑒
𝐺 𝑒 𝐽𝑒
𝑲𝑒𝜙 = [ 𝑒 𝜔 𝒌3 𝒌4 𝒌4 𝒌4 + 𝒌 ] , (3.34b)
𝐿 𝐿 𝑒3 𝐿 𝑒3 𝐿𝑒3 30𝐿𝑒 5
𝑇
𝑚𝑒 𝐿𝑒 𝑚𝑥𝑒 𝐿𝑒 𝑚𝑥𝑒 𝐿𝑒 𝑚𝑥𝑒 𝐿𝑒
𝒇𝑒𝜙 = 𝒇𝑒𝜙𝛺 + 𝒇𝑒𝜙𝛤 = [ 𝑥 − 𝑚𝜔
𝑒 − 𝑀𝑒
𝑥1
𝑒
− 𝑀𝜔1 𝑒 + 𝑀𝑒
+ 𝑚𝜔 𝑥2 − 𝑒 ]
+ 𝑀𝜔2 , (3.34c)
2 12 2 12

35
where

6 6 156 22𝐿𝑒 54 −13𝐿𝑒 36 3𝐿𝑒 −36 3𝐿𝑒


𝒎3 = [ −𝐿 𝑒
𝐿𝑒
] , 𝒎4 = [ 22𝐿𝑒 4𝐿𝑒 2 13𝐿𝑒 −3𝐿𝑒 2 ] , 𝒎5 = [ 3𝐿 𝑒 4𝐿𝑒 2 −3𝐿𝑒 −𝐿𝑒 2 ] ,
−6 −6 54 13𝐿𝑒 156 −22𝐿𝑒 −36 −3𝐿𝑒 36 −3𝐿𝑒 (3.35a)
𝐿𝑒 −𝐿𝑒 −13𝐿𝑒 −3𝐿𝑒 2 −22𝐿 𝑒
4𝐿𝑒 2 3𝐿𝑒 −𝐿𝑒 2 −3𝐿𝑒 4𝐿𝑒 2

0 0 12 6𝐿𝑒 −12 6𝐿𝑒 36 3𝐿𝑒 −36 3𝐿𝑒


𝑒 4𝐿𝑒 2
𝒌3 = [−1 1 ] , 𝒌4 = [ 6𝐿 −6𝐿𝑒 2𝐿𝑒 2 ] , 𝒌5 = [ 3𝐿 𝑒 4𝐿𝑒 2
−3𝐿𝑒 −𝐿𝑒 2 ] .
0 0 −12 −6𝐿𝑒 12 −6𝐿𝑒 −36 −3𝐿𝑒 36 −3𝐿𝑒 (3.35b)
1 −1 6𝐿𝑒 2𝐿𝑒 2 −6𝐿 𝑒
4𝐿𝑒 2 3𝐿𝑒 −𝐿𝑒 2 −3𝐿𝑒 4𝐿𝑒 2

The contributions to the element matrices regarding the bending about the 𝑦 axis problem are written as

𝜌𝑒 𝑆𝑦𝑒 𝜌𝑒 𝐴𝑒 𝐿𝑒 𝜌𝑒 𝐼𝑦𝑦
𝑒
𝜌𝑒 𝐼𝑦𝑧
𝑒
𝜌𝑒 𝑆𝑧𝑃𝑒 𝐿𝑒 𝜌𝑒 𝐼𝑦𝜔
𝑒
𝑴𝑒𝜉𝑧 = [ 𝒎3 𝒎4 + 𝒎 𝒎 𝒎4 + 𝒎 ] , (3.36a)
12 420 30𝐿𝑒 5 30𝐿𝑒 5 420 30𝐿𝑒 5

𝐸 𝑒 𝑆𝑦𝑒 𝐸 𝑒 𝐼𝑦𝑦
𝑒
𝐸 𝑒 𝐼𝑦𝑧
𝑒
𝐸 𝑒 𝐼𝑦𝜔
𝑒
𝑲𝑒𝜉𝑧 = [ 𝑒 𝒌3 𝒌4 𝒌4 𝒌4 ] , (3.36b)
𝐿 𝐿𝑒 3 𝐿𝑒 3 𝐿𝑒 3
𝑇
𝑓 𝑒 𝐿𝑒 𝑓𝑧𝑒 𝐿𝑒 𝑓𝑧𝑒 𝐿𝑒 𝑓𝑧𝑒 𝐿𝑒
𝒇𝑒𝜉𝑧 = 𝒇𝑒𝜉𝑧 𝛺 + 𝒇𝑒𝜉𝑧 𝛤 = [ 𝑧 + 𝑚𝑦𝑒 − 𝑉𝑧1
𝑒 𝑒
+ 𝑀𝑦1 𝑒
− 𝑚𝑦𝑒 + 𝑉𝑧2 − 𝑒 ]
− 𝑀𝑦2 , (3.36c)
2 12 2 12

and for bending about the 𝑧 axis one could write

𝜌𝑒 𝑆𝑧𝑒 𝜌𝑒 𝐼𝑧𝑦
𝑒
𝜌𝑒 𝐴𝑒 𝐿𝑒 𝜌𝑒 𝐼𝑧𝑧
𝑒 𝜌𝑒 𝑆𝑦𝑃𝑒 𝐿𝑒 𝜌𝑒 𝐼𝑧𝜔
𝑒
𝑴𝑒𝜉𝑦 = [ 𝒎3 𝒎 𝒎4 + 𝒎 − 𝒎4 + 𝒎 ] , (3.37a)
12 30𝐿𝑒 5 420 30𝐿𝑒 5 420 30𝐿𝑒 5

𝐸 𝑒 𝑆𝑧𝑒 𝐸 𝑒 𝐼𝑧𝑦
𝑒
𝐸 𝑒 𝐼𝑧𝑧
𝑒
𝐸 𝑒 𝐼𝑧𝜔
𝑒
𝑲𝑒𝜉𝑦 = [ 𝒌 𝒌4 𝒌4 𝒌4 ] , (3.37b)
𝐿𝑒 3 𝐿𝑒 3 𝐿𝑒 3 𝐿𝑒 3
𝑇
𝑓𝑦𝑒 𝐿𝑒 𝑓𝑦𝑒 𝐿𝑒 2 𝑓𝑦𝑒 𝐿𝑒 𝑓𝑦𝑒 𝐿𝑒 2 (3.37c)
𝒇𝑒𝜉𝑦 = 𝒇𝑒𝜉𝑦𝛺 + 𝒇𝑒𝜉𝑦𝛤 = [ 𝑒
− 𝑚𝑧𝑒 − 𝑉𝑦1 𝑒
− 𝑀𝑧1 𝑒
+ 𝑚𝑧𝑒 + 𝑉𝑦2 − 𝑒 ]
+ 𝑀𝑧2 .
2 12 2 12

3.1.1 ASSEMBLY OF THE ELEMENT MATRICES

The results obtained in the preceding section have to be assembled to develop the element mass and the
element stiffness matrices, 𝑴𝑒 and 𝑲𝑒 , including all deformation effects. For this purpose, the axial, the
bending, and the torsional motions are included in the same finite element, as shown in Figure 3.1, which
has seven degrees of freedom per node.

The element nodal displacements vector 𝒖𝑒 is conveniently defined as


𝑒 𝑒 𝑇
𝑒
𝑑𝜉𝑧1 𝑑𝜉𝑦1 𝑑𝜙1𝑒 𝑒
𝑑𝜉𝑧2 𝑒 𝑑𝜉𝑦2 𝑑𝜙2𝑒
𝒖𝑒 (𝑡) = [𝜉𝑥1
𝑒 𝑒
𝜉𝑧1 𝑒
𝜉𝑦1 𝜙1𝑒 𝑒
𝜉𝑥2 𝑒
𝜉𝑧2 𝜉𝑦2 𝜙2𝑒 ] . (3.38)
𝑑𝑥 𝑑𝑥 𝑑𝑥 𝑑𝑥 𝑑𝑥 𝑑𝑥

Performing a direct superposition of the mass matrices obtained from the axial extension, the bending and
the torsional motions, the element mass matrix 𝑴𝑒 for a thin walled beam with arbitrary cross section
yields

36
2𝑚11 𝑚12 −𝑚13 𝑚14 −𝑚15 𝑚16 −𝑚17 𝑚11 −𝑚12 𝑚13 −𝑚14 𝑚15 −𝑚16 𝑚17
𝑚31 𝑚32 𝑚21 𝑚22 𝑚41 𝑚42 𝑚12 𝑚51 𝑚52 −𝑚21 𝑚22 𝑚61 𝑚62
𝑚33 𝑚22 𝑚23 𝑚42 𝑚43 𝑚13 −𝑚52 𝑚53 −𝑚22 −𝑚24 −𝑚62 𝑚63
𝑚34 𝑚35 𝑚44 𝑚45 𝑚14 −𝑚21 𝑚22 𝑚54 𝑚55 𝑚64 𝑚65
𝑚36 𝑚45 𝑚46 𝑚15 −𝑚22 −𝑚24 −𝑚55 𝑚56 −𝑚65 𝑚66
𝑚37 𝑚38 𝑚16 𝑚61 𝑚62 𝑚64 𝑚65 𝑚57 𝑚58
𝑚39 𝑚17 −𝑚62 𝑚63 −𝑚65 𝑚66 −𝑚58 𝑚59
𝑴𝑒 = 2𝑚11 −𝑚12 −𝑚13 −𝑚14 −𝑚15 −𝑚16 −𝑚17 ,
𝑚31 −𝑚32 𝑚21 −𝑚22 𝑚41 −𝑚42
𝑚33 −𝑚22 𝑚23 −𝑚42 𝑚43
𝑚34 −𝑚35 𝑚44 −𝑚45
symmetric
𝑚36 −𝑚45 𝑚46
𝑚37 −𝑚38
[ 𝑚39 ]
(3.39)
where

1 1 1 𝑒 𝑒 𝑒 1
𝑚11 = 𝜌𝑒 𝐴𝑒 𝐿𝑒 , 𝑚12 = 𝜌𝑒 𝑆𝑦𝑒 , 𝑚13 = 𝜌 𝑆𝑦 𝐿 , 𝑚14 = 𝜌𝑒 𝑆𝑧𝑒 ,
6 2 12 2

1 𝑒 𝑒 𝑒 1 1 𝑒 𝑒 𝑒
𝑚15 = 𝜌 𝑆𝑧 𝐿 , 𝑚16 = 𝜌𝑒 𝑆𝜔
𝑒 , 𝑚
17 = 𝜌 𝑆𝜔 𝐿 ,
12 2 12

6 𝜌𝑒 𝐼𝑦𝑧
𝑒
1 𝑒 𝑒 2 𝑒 𝑒 𝑒 1 𝑒 𝑒 𝑒
𝑚21 = 𝑒
, 𝑚22 = 𝜌 𝐼𝑦𝑧 , 𝑚23 = 𝜌 𝐼𝑦𝑧 𝐿 , 𝑚24 = 𝜌 𝐼𝑦𝑧 𝐿 ,
5 𝐿 10 15 30

13 𝑒 𝑒 𝑒 6 𝜌𝑒 𝐼𝑦𝑦
𝑒
11 𝑒 𝑒 𝑒 2 1 𝑒 𝑒 1 𝑒 𝑒 𝑒3 2 𝑒 𝑒 𝑒
𝑚31 = 𝜌 𝐴 𝐿 + , 𝑚32 = 𝜌 𝐴 𝐿 + 𝜌 𝐼𝑦𝑦 , 𝑚33 = 𝜌 𝐴 𝐿 + 𝜌 𝐼𝑦𝑦 𝐿 ,
35 5 𝐿𝑒 210 10 105 15

13 𝑒 𝑒 𝑒 6 𝜌𝑒 𝐼𝑧𝑧
𝑒
11 𝑒 𝑒 𝑒 2 1 𝑒 𝑒 1 𝑒 𝑒 𝑒3 2 𝑒 𝑒 𝑒
𝑚34 = 𝜌 𝐴 𝐿 + , 𝑚35 = 𝜌 𝐴 𝐿 + 𝜌 𝐼𝑧𝑧 , 𝑚36 = 𝜌 𝐴 𝐿 + 𝜌 𝐼𝑧𝑧 𝐿 ,
35 5 𝐿𝑒 210 10 105 15

13 𝑒 𝑃𝑒 6 𝜌𝑒 𝐼𝜔𝜔
𝑒
11 𝑒 𝑃𝑒 𝑃𝑒 )𝐿𝑒 2 +
1 𝑒 𝑒
𝑚37 = 𝑃𝑒 )𝐿𝑒 +
𝜌 (𝐼𝑧𝑧 + 𝐼𝑦𝑦 , 𝑚38 = 𝜌 (𝐼𝑧𝑧 + 𝐼𝑦𝑦 𝜌 𝐼𝜔𝜔 ,
35 5 𝐿𝑒 210 10

1 𝑒 𝑃𝑒 𝑃𝑒 )𝐿𝑒 3 +
2 𝑒 𝑒 𝑒
𝑚39 = 𝜌 (𝐼𝑧𝑧 + 𝐼𝑦𝑦 𝜌 𝐼𝜔𝜔 𝐿 ,
105 15

13 𝑒 𝑃𝑒 𝑒 6 𝜌𝑒 𝐼𝑦𝜔
𝑒
11 𝑒 𝑃𝑒 𝑒 2 1 𝑒 𝑒 1 𝑒 𝑃𝑒 𝑒 3 2 𝑒 𝑒 𝑒
𝑚41 = 𝜌 𝑆𝑧 𝐿 + , 𝑚42 = 𝜌 𝑆𝑧 𝐿 + 𝜌 𝐼𝑦𝜔 , 𝑚43 = 𝜌 𝑆𝑧 𝐿 + 𝜌 𝐼𝑦𝜔 𝐿 ,
35 5 𝐿𝑒 210 10 105 15

13 𝑒 𝑃𝑒 𝑒 6 𝜌𝑒 𝐼𝑧𝜔
𝑒
11 𝑒 𝑃𝑒 𝑒 2 1 1 𝑒 𝑃𝑒 𝑒 3 2 𝑒 𝑒 𝑒
𝑚44 = − 𝜌 𝑆𝑦 𝐿 + , 𝑚45 = − 𝑒 , 𝑚
𝜌 𝑆𝑦 𝐿 + 𝜌𝑒 𝐼𝑧𝜔 46 = − 𝜌 𝑆𝑦 𝐿 + 𝜌 𝐼𝑧𝜔 𝐿 ,
35 5 𝐿𝑒 210 10 105 15

9 𝑒 𝑒 𝑒 6 𝜌𝑒 𝐼𝑦𝑦
𝑒
13 𝑒 𝑒 𝑒 2 1 𝑒 𝑒 1 𝑒 𝑒 𝑒3 1 𝑒 𝑒 𝑒
𝑚51 = 𝜌 𝐴 𝐿 − , 𝑚52 = − 𝜌 𝐴 𝐿 + 𝜌 𝐼𝑦𝑦 , 𝑚53 = − 𝜌 𝐴 𝐿 − 𝜌 𝐼𝑦𝑦 𝐿 ,
70 5 𝐿𝑒 420 10 140 30

9 𝑒 𝑒 𝑒 6 𝜌𝑒 𝐼𝑧𝑧
𝑒
13 𝑒 𝑒 𝑒 2 1 𝑒 𝑒 1 𝑒 𝑒 𝑒3 1 𝑒 𝑒 𝑒
𝑚54 = 𝜌 𝐴 𝐿 − , 𝑚55 = − 𝜌 𝐴 𝐿 + 𝜌 𝐼𝑧𝑧 , 𝑚56 = − 𝜌 𝐴 𝐿 − 𝜌 𝐼𝑧𝑧 𝐿 ,
70 5 𝐿𝑒 420 10 140 30

9 𝑒 𝑃𝑒 6 𝜌𝑒 𝐼𝜔𝜔
𝑒
13 𝑒 𝑃𝑒 𝑃𝑒 )𝐿𝑒 2 +
1 𝑒 𝑒
𝑚57 = 𝑃𝑒 )𝐿𝑒 −
𝜌 (𝐼𝑧𝑧 + 𝐼𝑦𝑦 , 𝑚58 = − 𝜌 (𝐼𝑧𝑧 + 𝐼𝑦𝑦 𝜌 𝐼𝜔𝜔 ,
70 5 𝐿𝑒 420 10

1 𝑒 𝑃𝑒 𝑃𝑒 )𝐿𝑒 3 −
1 𝑒 𝑒 𝑒
𝑚59 = − 𝜌 (𝐼𝑧𝑧 + 𝐼𝑦𝑦 𝜌 𝐼𝜔𝜔 𝐿 ,
140 30

9 𝑒 𝑃𝑒 𝑒 6 𝜌𝑒 𝐼𝑦𝜔
𝑒
13 𝑒 𝑃𝑒 𝑒 2 1 𝑒 𝑒 1 𝑒 𝑃𝑒 𝑒 3 1
𝑚61 = 𝜌 𝑆𝑧 𝐿 − , 𝑚62 = − 𝜌 𝑆𝑧 𝐿 + 𝜌 𝐼𝑦𝜔 , 𝑚63 = − 𝑒 𝐿𝑒 ,
𝜌 𝑆𝑧 𝐿 − 𝜌𝑒 𝐼𝑦𝜔
70 5 𝐿𝑒 420 10 140 30

9 𝑒 𝑃𝑒 𝑒 6 𝜌𝑒 𝐼𝑧𝜔
𝑒
13 𝑒 𝑃𝑒 𝑒 2 1 1 𝑒 𝑃𝑒 𝑒 3 1
𝑚64 = − 𝜌 𝑆𝑦 𝐿 − , 𝑚65 = 𝑒 , 𝑚
𝜌 𝑆𝑦 𝐿 + 𝜌𝑒 𝐼𝑧𝜔 66 =
𝑒 𝐿𝑒 .
𝜌 𝑆𝑦 𝐿 − 𝜌𝑒 𝐼𝑧𝜔
70 5 𝐿𝑒 420 10 140 30

37
Following an analogous procedure, the element stiffness matrix 𝑲𝑒 is given by

−𝑘12 −𝑘13 −𝑘14 −𝑘11 𝑘12 𝑘13 𝑘14


𝑘11
𝑘22 𝑘32 𝑘41 𝑘42 −𝑘21 𝑘22 −𝑘31 𝑘32 −𝑘41 𝑘42
𝑘21 𝑘31
2𝑘23 2𝑘33 𝑘42 2𝑘43 𝑘12 −𝑘22 𝑘23 −𝑘32 𝑘33 −𝑘42 𝑘43
𝑘32
𝑘52 𝑘61 𝑘62 −𝑘31 𝑘32 −𝑘51 𝑘52 −𝑘61 𝑘62
𝑘51
2𝑘53 𝑘62 2𝑘63 𝑘13 −𝑘32 𝑘33 −𝑘52 𝑘53 −𝑘62 𝑘63
𝑘71 + 𝑘81 𝑘72 + 𝑘82 −𝑘41 𝑘42 −𝑘61 𝑘62 −𝑘71 − 𝑘81 𝑘72 + 𝑘82
2𝑘73 + 4𝑘83 𝑘14 −𝑘42 𝑘43 −𝑘62 𝑘63 −𝑘72 − 𝑘82 𝑘73 − 𝑘83
𝑲𝑒 =
−𝑘13 −𝑘14
,
𝑘11 −𝑘12
𝑘21 −𝑘22 𝑘31 −𝑘32 𝑘41 −𝑘42
2𝑘23 −𝑘32 2𝑘33 −𝑘42 2𝑘43
symmetric 𝑘51 −𝑘52 𝑘61 −𝑘62
2𝑘53 −𝑘62 2𝑘63
𝑘71 + 𝑘81 −𝑘72 − 𝑘82
[ 2𝑘73 + 4𝑘83 ]

where (3.40)

𝐸 𝑒 𝐴𝑒 𝐸 𝑒 𝑆𝑦𝑒 𝐸 𝑒 𝑆𝑧𝑒 𝐸 𝑒 𝑆𝜔
𝑒
𝑘11 = , 𝑘12 = , 𝑘13 = , 𝑘14 = ,
𝐿𝑒 𝐿𝑒 𝐿𝑒 𝐿𝑒

𝐸 𝑒 𝐼𝑦𝑦
𝑒 𝐸 𝑒 𝐼𝑦𝑦
𝑒 𝐸 𝑒 𝐼𝑦𝑦
𝑒
𝑘21 = 12 , 𝑘22 = 6 , 𝑘23 = 2 ,
𝐿𝑒 3 𝐿𝑒 2 𝐿𝑒

𝐸 𝑒 𝐼𝑦𝑧
𝑒
𝐸 𝑒 𝐼𝑦𝑧
𝑒
𝐸 𝑒 𝐼𝑦𝑧
𝑒
𝑘31 = 12 , 𝑘32 = 6 , 𝑘33 = 2 ,
𝐿𝑒 3 𝐿𝑒 2 𝐿𝑒

𝐸 𝑒 𝐼𝑦𝜔
𝑒 𝐸 𝑒 𝐼𝑦𝜔
𝑒 𝐸 𝑒 𝐼𝑦𝜔
𝑒
𝑘41 = 12 , 𝑘42 = 6 , 𝑘43 = 2 ,
𝐿𝑒 3 𝐿𝑒 2 𝐿 𝑒

𝐸 𝑒 𝐼𝑧𝑧
𝑒
𝐸 𝑒 𝐼𝑧𝑧
𝑒
𝐸 𝑒 𝐼𝑧𝑧
𝑒
𝑘51 = 12 𝑒3
, 𝑘52 = 6 𝑒2
, 𝑘53 = 2 𝑒
,
𝐿 𝐿 𝐿

𝐸 𝑒 𝐼𝑧𝜔
𝑒
𝐸 𝑒 𝐼𝑧𝜔
𝑒
𝐸 𝑒 𝐼𝑧𝜔
𝑒
𝑘61 = 12 , 𝑘62 = 6 , 𝑘63 = 2 ,
𝐿𝑒 3 𝐿𝑒 2 𝐿𝑒

𝐸 𝑒 𝐼𝜔𝜔
𝑒
𝐸 𝑒 𝐼𝜔𝜔
𝑒
𝐸 𝑒 𝐼𝜔𝜔
𝑒
𝑘71 = 12 , 𝑘72 = 6 , 𝑘73 = 2 ,
𝐿𝑒 3 𝐿𝑒 2 𝐿 𝑒

36 𝐺 𝑒 𝐽𝑒 3 𝑒 𝑒 1 𝑒 𝑒 𝑒
𝑘81 = , 𝑘82 = 𝐺 𝐽 , 𝑘83 = 𝐺 𝐽 𝐿 .
30 𝐿𝑒 30 30

Finally, the element external force vector 𝒇𝑒 for a thin walled beam is defined as follows:

𝑓𝑥𝑒 𝐿𝑒 𝑓𝑧𝑒 𝐿𝑒 𝑓𝑧𝑒 𝐿𝑒 𝑓𝑦𝑒 𝐿𝑒 𝑓𝑦𝑒 𝐿𝑒 2 𝑚𝑥𝑒 𝐿𝑒


𝒇𝑒 = [ − 𝑁1𝑒 𝑒
+ 𝑚𝑦𝑒 − 𝑉𝑧1 𝑒
+ 𝑀𝑦1 𝑒
− 𝑚𝑧𝑒 − 𝑉𝑦1 𝑒
− 𝑀𝑧1 − 𝑚𝜔𝑒 − 𝑀𝑒
2 2 12 2 12 𝑥1 ⋯
2
𝑚𝑥𝑒 𝐿𝑒 𝑓𝑥𝑒 𝐿𝑒 𝑓𝑧𝑒 𝐿𝑒 𝑓𝑧𝑒 𝐿𝑒 𝑓𝑦𝑒 𝐿𝑒
<⋯ 𝑒
− 𝑀𝜔1 + 𝑁2𝑒 𝑒
− 𝑚𝑦𝑒 + 𝑉𝑧2 − 𝑒
− 𝑀𝑦2 𝑒
+ 𝑚𝑧𝑒 + 𝑉𝑦2 ⋯
2 2 2 12 2
𝑇
𝑓𝑦𝑒 𝐿𝑒 2 𝑚𝑥𝑒 𝐿𝑒 𝑚𝑥𝑒 𝐿𝑒
<⋯ − 𝑒
+ 𝑀𝑧2
𝑒 + 𝑀𝑒
+ 𝑚𝜔 𝑥2 − 𝑒
+ 𝑀𝜔2 ] . (3.41)
12 2 2

As shown in Figure 3.1, the developed finite element is described in terms of a single axis. In this case, the
displacement fields are referred to the 𝑥 axis by considering the principal global coordinate system (𝑥, 𝑦, 𝑧)
and by placing the sectorial pole 𝑃 coincident with the cross-sectional centre of gravity 𝐶𝐺. For the case of
cross sections without double symmetry, it is not possible to decouple the bending and the torsional
dynamic behaviours due to the inertial forces. However, this procedure allows that some matrix terms
vanish from the element matrices (3.39) and (3.40).

38
The resulting element mass matrix 𝑴𝑒𝐶𝐺 is written as follows:

2𝑚11 𝑚16 −𝑚17 𝑚11 −𝑚16 𝑚17


𝑚31 𝑚32 𝑚71 𝑚72 𝑚51 𝑚52 −𝑚71 𝑚72
𝑚33 𝑚72 𝑚73 −𝑚52 𝑚53 −𝑚72 −𝑚73 /4
𝑚34 𝑚35 𝑚74 𝑚75 𝑚54 𝑚55 −𝑚74 𝑚75
𝑚36 𝑚75 𝑚76 −𝑚55 𝑚56 −𝑚75 −𝑚76 /4
𝑚37 𝑚38 𝑚16 −𝑚71 𝑚72 −𝑚74 𝑚75 𝑚57 𝑚58
𝑚39 𝑚17 −𝑚72 −𝑚73 /4 −𝑚75 −𝑚76 /4 −𝑚58 𝑚59
𝑴𝑒𝐶𝐺 = 2𝑚11 −𝑚16 −𝑚17
𝑚31 −𝑚32 𝑚71 −𝑚72
𝑚33 −𝑚72 𝑚73
𝑚34 −𝑚35 𝑚74 −𝑚75
symmetric −𝑚75 𝑚76
𝑚36
𝑚37 −𝑚38
[ 𝑚39 ]

(3.42)
where some of the matrix elements were defined in (3.39) and the remaining ones are given by

6 𝜌𝑒 𝐼𝑦𝜔
𝑒
1 𝑒 𝑒 2 𝑒 𝑒 𝑒
𝑚71 = 𝑒
, 𝑚72 = 𝜌 𝐼𝑦𝜔 , 𝑚73 = 𝜌 𝐼𝑦𝜔 𝐿 ,
5 𝐿 10 15

6 𝜌𝑒 𝐼𝑧𝜔
𝑒 1 𝑒 𝑒 2 𝑒 𝑒 𝑒
𝑚74 = 𝑒 , 𝑚75 = 𝜌 𝐼𝑧𝜔 , 𝑚76 = 𝜌 𝐼𝑧𝜔 𝐿 .
5 𝐿 10 15

The element stiffness matrix 𝑲𝑒𝐶𝐺 for a pole 𝑃 coincident with the centre of gravity 𝐶𝐺 yields

−𝑘14 −𝑘11 𝑘14


𝑘11
𝑘41 𝑘42 −𝑘21 𝑘22 −𝑘41 𝑘42
𝑘21 𝑘22
2𝑘23 𝑘42 2𝑘43 −𝑘22 𝑘23 −𝑘42 𝑘43
𝑘51 𝑘52 𝑘61 𝑘62 −𝑘51 𝑘52 −𝑘61 𝑘62
2𝑘53 𝑘62 2𝑘63 −𝑘52 𝑘53 −𝑘62 𝑘63
𝑘71 + 𝑘81 𝑘72 + 𝑘82 −𝑘41 𝑘42 −𝑘61 𝑘62 −𝑘71 − 𝑘81 𝑘72 + 𝑘82
2𝑘73 + 4𝑘83 𝑘14 −𝑘42 𝑘43 −𝑘62 𝑘63 −𝑘72 − 𝑘82 𝑘73 − 𝑘83
𝑲𝑒𝐶𝐺 =
𝑘11 −𝑘14
𝑘21 −𝑘22 𝑘41 −𝑘42
2𝑘23 −𝑘42 2𝑘43
symmetric 𝑘51 −𝑘52 −𝑘62
𝑘61
2𝑘53 −𝑘62 2𝑘63
𝑘71 + 𝑘81 −𝑘72 − 𝑘82
[ 2𝑘73 + 4𝑘83 ]

being the matrix entries defined in (3.40). (3.43)

3.1.2 THE COMPLETE SYSTEM OF EQUATIONS

The system of linear algebraic equations obtained for each element has twice as much unknowns than the
number of equations available. This fact means that the definition of the field variables is only possible
with additional conditions. Some of these conditions are provided by assembling the element equations
and imposing the corresponding boundary conditions, i.e., by ensuring the interelement static and
kinematic compatibility at common nodes.

The global finite element equations are written in the matrix form as follows:

𝑴𝒖̈ + 𝑪𝒖̇ + 𝑲𝒖 = 𝒇 , (3.44)

39
where the global mass matrix 𝑴, the global damping matrix 𝑪, the global stiffness matrix 𝑲, and the global
external force vector 𝒇 are derived through a direct assembly procedure, [Fish and Belytschko, 2007].

The eqn (3.44) represents a set of equations discretized not only in space, but also in time, since for a
given dynamic system the displacement field must be evaluated at different time increments, and thus a
discretization with respect to time is also needed. The numerical integration in time will be made using the
Newmark’s β method.

3.2 THE DYNAMIC RESPONSE

The response of a structural element with a dynamic behaviour can be adequately described by the
governing equations (3.44). This is a two-fold problem. On the one hand, free vibrations are analysed,
being sought the natural frequencies of vibration and the corresponding mode shapes. On the other hand,
forced vibrations are evaluated, and the structure’s response to a prescribed loading is determined.

3.2.1 VIBRATION FREQUENCIES AND MODE SHAPES

The natural frequencies and the corresponding vibration mode shapes are obtained by neglecting the
damping effect and considering a null applied loading. Therefore, the governing equations for a freely
vibrating undamped system with several degrees of freedom are obtained by omitting the damping matrix
𝑪 and the externally applied force vector 𝒇 from eqn (3.44), yielding

𝑴𝒖̈ + 𝑲𝒖 = 𝟎 . (3.45)

Considering a harmonic solution of eqn (3.45), the dynamic response can be expressed as follows:

𝒖(𝑥, 𝑡) = 𝒗(𝑥) sin(𝑝𝑡) . (3.46)

The vibration mode shapes 𝒗 of the system, which are constant in time, and the corresponding natural
frequencies 𝑝 can be computed from the associated generalized eigenvalue problem

[𝑲 − 𝑝2 𝑴]𝒗 = 𝟎 , (3.47)

where 𝑝2 and 𝒗 represent the eigenvalues and the eigenvectors of the problem, respectively.

Since 𝒗 is non-trivial, i.e., the displacement vectors 𝒗 are non-null, the following relation must be verified:

det (𝑲 − 𝑝2 𝑴) = 0 . (3.48)

40
There is a different solution of 𝒗 for each of the 𝑛𝑑𝑜𝑓 roots of eqn (3.48), being 𝑛𝑑𝑜𝑓 the number of
degrees of freedom of the physical system. Once the natural frequencies have been calculated, the 𝑛-th
mode shape 𝒗𝑛 , in which the structure is considered to respond when vibrates with a frequency 𝑝𝑛 , is
obtained by

[𝑲 − 𝑝𝑛 2 𝑴]𝒗𝑛 = 𝟎 . (3.49)

3.2.2 VISCOUS DAMPING MATRIX

In the vibration of a structure there are energy losses due to friction and viscous deformations. This may
have significant influence on the response of a dynamically loaded structure. However, the damping
properties of a structure are difficult to quantify, and data from dynamic experiments on similar structures
are generally unavailable.

There are several ways to formulate the damping matrix 𝑪, since the damping values may not be constant
for all the vibration modes, [Clough and Penzien, 1995]. One way to quantify the damping properties,
when damping mechanisms are considered distributed throughout the structure, is by establishing a
proportional relation regarding the modal mass properties.

The modal mass properties are determined by considering the orthogonality of the vibration modes with
respect to mass, which allows the decoupling of the governing equations (3.44). To derive the
orthogonality conditions with respect to the mass matrix 𝑴, consider the following two solutions, 𝑟 and 𝑠,
of the eigenvalue problem (3.47), such that one can write:

𝑲𝒗𝑟 = 𝑝𝑟 2 𝑴𝒗𝑟 , (3.50a)


𝑲𝒗𝑠 = 𝑝𝑠 2 𝑴𝒗𝑠 . (3.50b)
By premultiplying both sides of eqn (3.50a) by 𝒗𝑇𝑠 , postmultiplying the transpose of (3.50b) by 𝒗𝑟 , and
subtracting the resulting equations from one another, the following relation is obtained:

(𝑝𝑟 2 − 𝑝𝑠 2 )𝒗𝑠 𝑇 𝑴𝒗𝑟 = 0 . (3.51)

Since the natural frequencies are, in general, distinct, the eqn (3.51) implies that

𝒗𝑠 𝑇 𝑴𝒗𝑟 = 0 , (3.52)

which means that the vibration mode shapes are orthogonal with respect to the mass matrix.

By assembling the free-vibrational mode shapes as

𝑽 = [𝒗1 𝒗2 ⋯ 𝒗𝑛 ⋯ 𝒗𝑛𝑑𝑜𝑓 ] , (3.53)

the diagonal modal matrices 𝑴𝐺 and 𝑲𝐺 , written for the modal coordinates, are given by

𝑴𝐺 = 𝑽𝑇 𝑴𝑽 and 𝑲𝐺 = 𝑽𝑇 𝑲𝑽 . (3.54)

41
Considering the 𝑛-th normalized mode shape 𝝍𝑛 given by
𝒗𝑛
𝝍𝑛 = , (3.55)
√𝒗𝑛 𝑇 𝑴𝒗𝑛

the mass and stiffness matrices, 𝑴 and 𝑲, may be rewritten for a new set of coordinates, corresponding to
the normalized modes, as follows:

𝕄 = 𝜳𝑇 𝑴𝜳 and 𝕂 = 𝜳𝑇 𝑲𝜳 , (3.56)

where 𝜳 is defined as

𝜳 = [𝝍1 𝝍2 ⋯ 𝝍𝑛 ⋯ 𝝍𝑛𝑑𝑜𝑓 ] , (3.57)

The elements of the diagonal matrices 𝕄 and 𝕂 are respectively expressed by

𝝍𝑠 𝑇 𝑴𝝍𝑟 = 𝛿𝑠𝑟 and 𝝍𝑠 𝑇 𝑲𝝍𝑟 = 𝛿𝑠𝑟 𝑝𝑠2 , (3.58)

being 𝛿𝑠𝑟 the Kronecker delta,

1 ,𝑠 = 𝑟
𝛿𝑠𝑟 = {
0 ,𝑠 ≠ 𝑟
. (3.59)

It is thus concluded that, by choosing the modal coordinates, the governing equations for a linear
structural system can be rendered decoupled. In other words, the governing equations can be expressed in
terms of different coordinates, such that neither stiffness coupling nor dynamic coupling are exhibited.
The response calculations are simplified, being each of the eqns (3.44) transformed into a single degree of
freedom governing equation.

The decoupled differential equations (3.44) are rewritten as

𝕄𝒖̈ + ℂ𝒖̇ + 𝕂𝒖 = 𝕗 , (3.60)

in which 𝕄 and 𝕂 are the diagonal matrices of eqns (3.56). The modal loading is given by

𝕗 = 𝜳𝑇 𝒇 , (3.61)
and ℂ is a diagonal matrix which elements are expressed by

2𝜍𝑠 𝑝𝑟 𝛿𝑠𝑟 , (3.62)

where 𝜍𝑠 represents the viscous damping ratio corresponding to the 𝑠-th vibration mode.

3.2.3 DIRECT INTEGRATION OF THE GOVERNING EQUATIONS

The response of a linear structural system to a prescribed time-dependent loading 𝒇(𝑡) involves the
solution of the ordinary differential equation (3.60) in time. In the following discussion, the modal
superposition method and the direct integration of the governing equations are described.

42
After solving the linear system (3.60), the modal participation factor of each vibration mode should be
checked in order to reduce computation cost by neglecting vibration modes of minor relevance. Let the
nodal displacements vector 𝒖 be obtained by summing the 𝑛𝑑𝑜𝑓 products of the mode shape vectors 𝝍𝑛
and the modal amplitudes 𝛶𝑛 as follows:
𝑛𝑑𝑜𝑓
𝒖=∑ 𝝍𝑛 𝛶𝑛 . (3.63)
𝑛=1

Towards the definition of the modal amplitudes 𝛶𝑛 both sides of eqn (3.63) are premultiplied by 𝝍𝑛 𝑇 𝑴.
Due to the orthogonality properties of the mode shapes with respect to mass, this operation results in

𝝍𝑛 𝑇 𝑴𝒖 = 𝝍𝑛 𝑇 𝑴𝝍𝑛 𝛶𝑛 , (3.64)
and therefore, the 𝑛-th modal amplitude is defined by

𝝍𝑛 𝑇 𝑴𝒖
𝛶𝑛 = . (3.65)
𝝍𝑛 𝑇 𝑴𝝍𝑛

The equation (3.63) is solved by adopting the Newmark’s β method, which consists in a direct integration
method that uses step-by-step numerical integration. Its versality is evidenced by its adoption in numerous
commercially available computer programs for purposes of structural dynamic analysis, [Logan, 2007].

In the Newmark’s formulation, two parameters 𝛾 and 𝛽 are introduced to express the contribution of the
acceleration for the velocity and the displacement at the end of the time interval. The displacement and
velocity at time 𝑡 + Δ𝑡 can then be obtained, [Wunderlich and Pilkey, 2003], by considering the equations

𝒖𝑡+Δ𝑡 = 𝒖𝑡 + Δ𝑡𝒖̇ 𝑡 + (1⁄2 − 𝛽)(Δ𝑡)2 𝒖̈ 𝑡 + 𝛽(Δ𝑡)2 𝒖̈ 𝑡+Δ𝑡 , (3.66a)

𝒖̇ 𝑡+Δ𝑡 = 𝒖̇ 𝑡 + Δ𝑡𝒖̇ 𝑡 + (1 − 𝛾)Δ𝑡𝒖̈ 𝑡 + 𝛽Δ𝑡𝒖̈ 𝑡+Δ𝑡 . (3.66b)


It is interesting to notice that the parameter 𝛾 introduces a linear varying weighting between the influence
of the initial and the final accelerations on the change of velocity. On the other hand, 𝛽 provides for
weighting the contributions of the initial and final accelerations to the change of displacement, [Clough
and Penzien, 1995]. By choosing 𝛽 = 1/4 the method can be termed as the constant average acceleration
method, see Figure 3.4.

By adopting 𝛾 = 1/2 and 𝛽 = 1/4, the method is unconditionally stable, which enables the convergence to
the solution regardless the initial conditions and the time step Δ𝑡, [Newmark, 1959].

The implicit recursive relation for 𝒖𝑡+Δ𝑡 can be converted to an explicit form by substituting eqns (3.66)
into the governing equations (3.60) at time 𝑡 + Δ𝑡, which result is

𝓚𝒖𝑡+Δ𝑡 = 𝓕𝑡+Δ𝑡 , (3.67)

where the effective stiffness matrix 𝓚 and the effective loading vector 𝓕𝑡+Δ𝑡 are respectively defined as

1 𝛾
𝓚= 𝕄+ ℂ+𝕂, (3.68a)
𝛽(Δ𝑡)2 𝛽Δ𝑡

1 1 1 𝛾 𝛾 𝛾 Δ𝑡
𝓕𝑡+Δ𝑡 = [ 𝒖𝑡 + 𝒖̇ 𝑡 + ( − 1) 𝒖̈ 𝑡 ] 𝕄 + [ 𝒖𝑡 + ( − 1) 𝒖̇ 𝑡 + ( − 2) 𝒖̈ ] ℂ + 𝕗𝑡+Δ𝑡 . (3.68b)
𝛽(Δ𝑡)2 𝛽Δ𝑡 2𝛽 𝛽Δ𝑡 𝛽 𝛽 2 𝑡

43
The vector 𝕗𝑡+Δ𝑡 contains the magnitude of the nodal loads 𝒇 at time 𝑡 + Δ𝑡 transformed into modal
coordinates. Whenever a point load is placed in the domain of an element 𝑒, the values of the considered
loading at the nodes 𝑥1𝑒 and 𝑥2𝑒 are approximated by a linear interpolation.

Since 𝒖𝑡+Δ𝑡 is calculated using eqn (3.67), the velocity and the acceleration at the corresponding time step
become

𝛾 1
𝒖̇ 𝑡+Δ𝑡 = 𝒖̇ 𝑡 + (1 − 𝛾)Δ𝑡𝒖̈ 𝑡 + [𝒖 − 𝒖𝑡 − Δ𝑡𝒖̇ 𝑡 − ( − 𝛽) (Δ𝑡)2 𝒖̈ 𝑡 ] , (3.69a)
𝛽Δ𝑡 𝑡+Δ𝑡 2

1 1
𝒖̈ 𝑡+Δ𝑡 = [𝒖 − 𝒖𝑡 + Δ𝑡𝒖̇ 𝑡 + ( − 𝛽) (Δ𝑡)2 𝒖̈ 𝑡 ] . (3.69b)
𝛽(Δ𝑡)2 𝑡+Δ𝑡 2

1
𝒖̈ (𝜏) = (𝒖̈ 𝑡 + 𝒖̈ 𝑡+Δ𝑡 )
2
Acceleration (constant)
𝒖̈ 𝑡+Δ𝑡
𝒖̈ 𝑡

𝜏
𝒖̇ (𝜏) = 𝒖̇ 𝑡 + (𝒖̈ 𝑡 + 𝒖̈ 𝑡+Δ𝑡 )
Velocity (linear) 2

𝒖̇ 𝑡+Δ𝑡

𝒖̇ 𝑡

𝜏2
𝒖(𝜏) = 𝒖𝑡 + 𝒖̇ 𝑡 𝜏 + (𝒖̈ + 𝒖̈ 𝑡+Δ𝑡 )
4 𝑡
Displacement (quadratic)
𝒖𝑡+Δ𝑡

𝒖𝑡

𝑡 Δ𝑡 𝑡 + Δt
𝜏

Figure 3.4 – the constant average acceleration method.

In [ERRI D214, 1999] it is recommended that the modelling of the vibration mode with the highest
frequency 𝑓𝑚𝑎𝑥 (Hz) should consider at least ten time steps in the corresponding wavelength. Therefore,
the integration scheme is accurate enough if

1
Δ𝑡 = . (3.70)
10𝑓𝑚𝑎𝑥

44
4
Development of an
Analytical Solution

I n the present chapter, the dynamic response of a multi-span bridge traversed by moving vehicles is
studied through an analytical model. The bridge’s girder is modelled as a continuous thin walled beam
with mono-symmetric cross section. The exact natural frequencies and the corresponding mode shapes of
vibration are obtained for the cases of decoupled vertical bending motion and coupled lateral bending-
torsional motion. To this end, explicit expressions are derived.

The dynamic response due to eccentric moving loads with constant velocity is obtained by considering a
modal analysis along with the direct integration method. In Chapter 5, the mass effect of the moving
vehicles is also introduced in the theoretical formulation.

4.1 THE DYNAMIC RESPONSE

An analytical solution for the problem of coupled bending-torsional vibrations is developed in the sequel.
The formulation is based on the work done by [Chan and Ashebo, 2006]. In that work, the bending
natural frequencies and the corresponding modes shapes of vibration of an Euler-Bernoulli continuous
beam are met exactly, being the shear deformation and the cross-sectional rotary inertia neglected. This
formulation was also used by [Graça, 2011] to model the dynamic behaviour of continuous bridges due to
the passage of high-speed trains, despite coupled bending-torsional motion was not considered.

45
The proposed formulation considers an 𝑁-span bridge of total length 𝐿𝑡𝑜𝑡 , as illustrated in Figure 4.1. Each
span is considered to have uniform geometrical and material properties along its length. For clarity of
notation, the properties of the generic span 𝑠 are denoted by the subscript 𝑠, being the local coordinate 𝑥𝑠
related to the global coordinate 𝑥 by
𝑟
𝑥𝑠 = 𝑥 − ∑ 𝑙𝑠 . (4.1)
𝑠=1

𝑉𝑧𝑘 𝑉𝑧2 𝑉𝑧1

𝑥1 𝑥2 𝑥𝑟 𝑥𝑠 𝑥𝑠+1 𝑥𝑁

𝑙1 𝑙2 𝑙𝑟 𝑙𝑠 𝑙𝑠+1 𝑙𝑁
𝐿𝑡𝑜𝑡
Figure 4.1 – model of the 𝑁-span bridge considered in the analytical formulation

The dynamic equilibrium differential equations of thin walled beams were obtained for a generic cross
section in Chapter 2, eqn (2.50). For mono-symmetric cross sections, being 𝑧 the axis of symmetry, and
considering the principal global coordinate system (𝑥, 𝑦, 𝑧), the principal sectorial pole 𝑃 and the principal
sectorial origin 𝑂, the dynamic equilibrium of the bridge’s girder is given by

𝜌𝐴𝜉𝑧̈ + 𝐸𝐼𝑦𝑦 𝜉𝑧′′′′ = 𝑓𝑧 , (4.2a)

𝜌𝐴𝜉𝑦̈ − 𝜌𝑆𝑦𝑃 𝜙̈ + 𝐸𝐼𝑧𝑧 𝜉𝑦′′′′ = 𝑓𝑦 , (4.2b)

−𝜌𝑆𝑦𝑃 𝜉𝑦̈ + 𝜌(𝐼𝑧𝑧


𝑃 + 𝐼𝑃 )𝜙̈ + 𝐸𝐼
𝑦𝑦 𝜔𝜔 𝜙
′′′′ − 𝐺𝐽𝜙 ′′ = 𝑚 ,
𝑥 (4.2c)

where the warping effect is included and the rotary inertia is neglected. Considering the sectorial pole 𝑃
coincident with the centre of gravity 𝐶𝐺 of the cross section, the eqns (4.2) are rewritten as follows:

𝜌𝐴𝜉𝑧̈ + 𝐸𝐼𝑦𝑦 𝜉𝑧′′′′ = 𝑓𝑧 , (4.3a)

𝜌𝐴𝜉𝑦̈ + 𝐸𝐼𝑧𝑧 𝜉𝑦′′′′ + 𝐸𝐼𝑧𝜔 𝜙 ′′′′ = 𝑓𝑦 , (4.3b)

𝜌(𝐼𝑧𝑧 + 𝐼𝑦𝑦 )𝜙̈ + 𝐸𝐼𝜔𝑧 𝜉𝑦′′′′ + 𝐸𝐼𝜔𝜔 𝜙 ′′′′ − 𝐺𝐽𝜙 ′′ = 𝑚𝑥 . (4.3c)

Both sets of equations, (4.2) and (4.3), will be required in the next discussion.

Since the present formulation deals with arrays of moving point loads, 𝑓𝑧 , 𝑓𝑦 , and 𝑚𝑥 are of the form
𝑛𝑙𝑜𝑎𝑑
𝑓𝑧 (𝑥, 𝑡) = ∑ δ(𝑥 − 𝑥𝑘 (𝑡)) 𝑉𝑧𝑘 (𝑡) , (4.4a)
𝑘

𝑛𝑙𝑜𝑎𝑑
𝑓𝑦 (𝑥, 𝑡) = ∑ δ(𝑥 − 𝑥𝑘 (𝑡)) 𝑉𝑦𝑘 (𝑡) , (4.4b)
𝑘

𝑛𝑙𝑜𝑎𝑑
𝑚𝑥 (𝑥, 𝑡) = ∑ δ(𝑥 − 𝑥𝑘 (𝑡)) 𝑀𝑥𝑘 (𝑡) , (4.4c)
𝑘

being 𝑥𝑘 (𝑡) the position of the 𝑘-th load, δ(𝑡) the Dirac delta function, and 𝑛𝑙𝑜𝑎𝑑 the number of loads.

46
4.1.1 VIBRATION FREQUENCIES AND MODE SHAPES

The homogeneous solution of the governing equations (4.2) is obtained by considering harmonic
functions for the displacement components, say 𝜉𝑧 , 𝜉𝑦 and 𝜙, which can be expressed as follows:

𝜉𝑧 (𝑥, 𝑡) = 𝛯𝑧 (𝑥) sin(𝑝𝑡) , (4.5a)


𝜉𝑦 (𝑥, 𝑡) = 𝛯𝑦 (𝑥) sin(𝑝𝑡) , (4.5b)
𝜙(𝑥, 𝑡) = 𝛷(𝑥) sin(𝑝𝑡) , (4.5c)
being 𝛯𝑧 , 𝛯𝑦 and 𝛷 the respective shape functions of the vertical displacement 𝜉𝑧 , lateral displacement 𝜉𝑦
and torsional rotation 𝜙, and 𝑝 is the corresponding circular frequency.

Since the modal superposition method is used, the mode shapes of vibration must be previously
determined. Thus, by substituting eqns (4.5) into the homogeneous form of eqns (4.2) and noticing that
sin(𝑝𝑡) can be non-null leads to

−𝜌𝐴𝑝2 𝛯𝑧 (𝑥) + 𝐸𝐼𝑦𝑦 𝛯𝑧′′′′ (𝑥) = 0 , (4.6a)

−𝜌𝐴𝑝2 𝛯𝑦 (𝑥) + 𝜌𝑆𝑦𝑃 𝑝2 𝛷(𝑥) + 𝐸𝐼𝑧𝑧 𝛯𝑦′′′′ (𝑥) = 0 , (4.6b)

𝜌𝑆𝑦𝑃 𝑝2 𝛯𝑦 (𝑥) − 𝜌(𝐼𝑧𝑧


𝑃 + 𝐼𝑃 )𝑝 2 𝛷(𝑥) + 𝐸𝐼
𝑦𝑦 𝜔𝜔 𝛷
′′′′ (𝑥) − 𝐺𝐽𝛷 ′′ (𝑥) = 0 . (4.6c)

For the case of decoupled vertical bending motion, the equation (4.6a) is simplified to the fourth-order
ordinary differential equation

−𝜆4 𝛯𝑧 (𝑥) + 𝛯𝑧′′′′ (𝑥) = 0 , (4.7)

by introducing a frequency parameter 𝜆 given by

𝜌𝐴𝑝2
𝜆4 = . (4.8)
𝐸𝐼𝑦𝑦

In a free vibration analysis the natural frequencies are obtained by calculating the frequency parameter 𝜆.

The solution 𝛯𝑧 of the eqn (4.7) may be expressed in trigonometric and hyperbolic terms as

𝛯𝑧 (𝑥) = 𝒜1 cosh(𝜆𝑥) + 𝒜2 sinh(𝜆𝑥) + 𝒜3 cos(𝜆𝑥) + 𝒜4 sin(𝜆𝑥) , (4.9)

where the 𝒜𝑗 constants must be such that the boundary conditions at the ends of the beam are satisfied,
[Clough and Penzien, 1995]. This allows expressing three of the constants 𝒜𝑗 as functions of the fourth,
which represents an arbitrary amplitude of the vibration mode shape.

For the case of coupled lateral bending-torsional motion, the solutions 𝛯𝑦 and 𝛷 of the eqns (4.6b) and
(4.6c) involve a more cumbersome mathematical treatment. To start with, consider that these two
equations are combined into one equation by eliminating either 𝛯𝑦 or 𝛷, [Banerjee et al, 1996].

47
This operation leads to the eighth-order ordinary differential equation
𝑃 𝑃
𝐸𝐼𝜔𝜔 𝑑8 𝜒 𝑑6 𝜒 𝜌𝑝2 𝐴𝐸𝐼𝜔𝜔 𝐼𝑧𝑧 + 𝐼𝑦𝑦 𝑑4 𝜒 𝜌𝐴𝑝 2 𝑑2 𝜒 (𝜌𝑝2 )2 𝑃 2 𝑃 + 𝐼𝑃 )] 𝜒(𝑥) = 0 ,
(𝑥) − 6 (𝑥) − ( + ) 4 (𝑥) + (𝑥) − [𝑆 − 𝐴(𝐼𝑧𝑧
𝐺𝐽 𝑑𝑥 8 𝑑𝑥 𝐺𝐽 𝐺𝐽𝐸𝐼𝑧𝑧 𝐺𝐽 𝑑𝑥 𝐸𝐼𝑧𝑧 𝑑𝑥 2 𝐺𝐽𝐸𝐼𝑧𝑧 𝑦 𝑦𝑦

(4.10)
where 𝜒 is either 𝛯𝑦 or 𝛷, depending on the variable that was eliminated.

The eqn (4.10) can be written in a more compact form as

[𝑑𝐷8 − 𝐷6 − (𝑎 + 𝑏𝑑)𝐷4 + 𝑏𝐷2 + 𝑎𝑏𝑐]𝜒(𝑥) = 0 , (4.11)


being 𝐷 a differential operator and 𝑎, 𝑏, 𝑐 and 𝑑 non-dimensional parameters, defined as follows:

𝑑
𝐷= , (4.12a)
𝑑𝑥
2
𝜌𝑝2 𝜌𝐴𝑝2 𝑆𝑦𝑃 𝐸𝐼𝜔𝜔
𝑃 + 𝐼𝑃 )
𝑎 = (𝐼𝑧𝑧 𝑦𝑦 , 𝑏= , 𝑐 =1− and 𝑑 = . (4.12b)
𝐺𝐽 𝐸𝐼𝑧𝑧 𝑃
𝐴(𝐼𝑧𝑧 𝑃
+ 𝐼𝑦𝑦 ) 𝐺𝐽

Considering an exponential solution for the differential equation (4.11) of the form

𝜒(𝑥) = 𝑒 𝓈𝑥 (4.13)

leads to the characteristic equation

𝑑𝓈 8 − 𝓈 6 − (𝑎 + 𝑏𝑑)𝓈 4 + 𝑏𝓈 2 + 𝑎𝑏𝑐 = 0 . (4.14)

The eqn (4.14) reduces to a quartic polynomial with real coefficients by substituting 𝓈2 by 𝜂 , as follows:

𝑑𝜂4 − 𝜂3 − (𝑎 + 𝑏𝑑)𝜂2 + 𝑏𝜂 + 𝑎𝑏𝑐 = 0 . (4.15)

Since, for a physical problem, 𝑎, 𝑏, 𝑐 and 𝑑 are always positive reals, being 𝑐 less than one, it is possible to
show, [Bishop et al., 1989], that all four roots of the equation are reals and they are within four distinct
intervals, namely

]−∞, −√𝑏[ , ]−√𝑏, 0[ , ]0, √𝑏[ and ]√𝑏, +∞[ . (4.16)

Considering 𝜂1 , 𝜂2 , -𝜂3 and -𝜂4 to be the four real roots, as represented in Figure 4.2, the eight roots of the
characteristic equation (4.14) are

𝛼 , −𝛼 , 𝛽 , −𝛽 , 𝑖𝛾 , −𝑖𝛾 , 𝑖𝛿 , and −𝑖𝛿 , (4.17)

with

𝛼 = √𝜂1 , 𝛽 = √𝜂2 , 𝛾 = √𝜂3 , 𝛿 = √𝜂4 , and 𝑖 = √−1 . (4.18)


Substituting each of these roots into eqn (4.14) separately and adding the resulting eight terms, one
obtains the complete solution of the eqn (4.11) as

𝜒(𝑥) = 𝒰1 𝑒 𝛼𝑥 + 𝒰2 𝑒 −𝛼𝑥 + 𝒰3 𝑒 𝛽𝑥 + 𝒰4 𝑒 −𝛽𝑥 + 𝒰5 𝑒 𝑖𝛾𝑥 + 𝒰6 𝑒 −𝑖𝛾𝑥 + 𝒰7 𝑒 𝑖𝛿𝑥 + 𝒰8 𝑒 −𝑖𝛿𝑥 , (4.19)

in which the constants 𝒰𝑗 must satisfy the boundary conditions at the ends of the beam.

48
f(𝜂)

−√𝑏 √𝑏

Figure 4.2 – roots of the fourth-order polynomial (4.15).

By expressing the exponential functions of eqn (4.19) in the respective trigonometric and hyperbolic
forms, the solutions for the lateral bending displacement 𝛯𝑦 and for the torsional rotation 𝛷 are given by

𝛯𝑦 (𝑥) = ℬ1 cosh(𝛼𝑥) + ℬ2 sinh(𝛼𝑥) + ℬ3 cosh(𝛽𝑥) + ℬ4 sinh(𝛽𝑥) + ℬ5 cos(𝛾𝑥)


+ ℬ6 sin(𝛾𝑥) + ℬ7 cos(𝛿𝑥) + ℬ8 sin(𝛿𝑥) , (4.20a)
𝛷(𝑥) = 𝒞1 cosh(𝛼𝑥) + 𝒞2 sinh(𝛼𝑥) + 𝒞3 cosh(𝛽𝑥) + 𝒞4 sinh(𝛽𝑥) + 𝒞5 cos(𝛾𝑥)
+ 𝒞6 sin(𝛾𝑥) + 𝒞7 cos(𝛿𝑥) + 𝒞8 sin(𝛿𝑥) . (4.20b)

Notice that in eqn (4.19) 𝜒 represents either 𝛯𝑦 or 𝛷, depending on the variable eliminated when the eqns
(4.6b) and (4.6c) were combined into one equation, (4.10).

The relations between the two sets of constants ℬ𝑗 and 𝒞𝑗 are established by substituting eqns (4.20) into
eqn (4.6b) or (4.6c), being given by

𝒞1 = 𝜅𝛼 ℬ1 , 𝒞2 = 𝜅𝛼 ℬ2 , (4.21a)
𝒞3 = 𝜅𝛽 ℬ3 , 𝒞4 = 𝜅𝛽 ℬ4 , (4.21b)
𝒞5 = 𝜅𝛾 ℬ5 , 𝒞6 = 𝜅𝛾 ℬ6 , (4.21c)
𝒞7 = 𝜅𝛿 ℬ7 , 𝒞8 = 𝜅𝛿 ℬ8 , (4.21d)

where the coefficients 𝜅𝛼 , 𝜅𝛽 , 𝜅𝛾 and 𝜅𝛿 are defined as follows:

𝜌𝐴𝑝2 − 𝐸𝐼𝑧𝑧 𝛼 4 𝜌𝐴𝑝2 − 𝐸𝐼𝑧𝑧 𝛽4


𝜅𝛼 = , 𝜅𝛽 = , (4.22a)
𝜌𝑆𝑦𝑃 𝑝2 𝜌𝑆𝑦𝑃 𝑝2

𝜌𝐴𝑝2 − 𝐸𝐼𝑧𝑧 𝛾 4 𝜌𝐴𝑝2 − 𝐸𝐼𝑧𝑧 𝛿 4


𝜅𝛾 = , 𝜅𝛿 = . (4.22b)
𝜌𝑆𝑦𝑃 𝑝2 𝜌𝑆𝑦𝑃 𝑝2

At this point, the analytical solutions of the governing differential equations (4.6) have been derived for a
thin walled beam with single cross-sectional symmetry, being given by eqns (4.9), (4.20a) and (4.20b).

For a general case of continuous beam, as represented in Figure 4.1, an analytical solution is now
proposed. The developed formulation is based on the work of [Chan and Ashebo, 2006], where the free
bending vibration of an Euler-Bernoulli continuous beam is considered, being neglected the shear
deformation and the rotary inertia of cross sections. In order to calculate the vertical bending natural

49
frequencies and the respective mode shapes of vibration, the solution (4.9) is applied separately to each
span 𝑠 and the continuity requirements at the inner supports are imposed afterwards. More recently, this
procedure was also used by [Graça, 2011] to perform dynamic analyses of continuous bridges traversed by
high-speed trains.

In the present work, the Chan’s assembly procedure is extended to include coupled lateral bending-
torsional vibrations of thin walled beams, and thus the cross-sectional warping is considered in the
dynamic analysis.

Consider that each span 𝑠 has uniform geometrical and material properties along its length 𝑙𝑠 . Therefore,
the 𝑛-th vibration mode shape of the 𝑠-th span is given by

𝛯𝑧𝑛𝑠 (𝑥𝑠 ) = 𝒜1𝑛𝑠 cosh(𝜆𝑛𝑠 𝑥𝑠 ) + 𝒜2𝑛𝑠 sinh(𝜆𝑛𝑠 𝑥𝑠 ) + 𝒜3𝑛𝑠 cos(𝜆𝑛𝑠 𝑥𝑠 ) + 𝒜4𝑛𝑠 sin(𝜆𝑛𝑠 𝑥𝑠 ) , (4.23a)
𝛯𝑦𝑛𝑠 (𝑥𝑠 ) = ℬ1𝑛𝑠 cosh(𝛼𝑛𝑠 𝑥𝑠 ) + ℬ2𝑛𝑠 sinh(𝛼𝑛𝑠 𝑥𝑠 ) + ℬ3𝑛𝑠 cosh(𝛽𝑛𝑠 𝑥𝑠 ) + ℬ4𝑛𝑠 sinh(𝛽𝑛𝑠 𝑥𝑠 )
+ ℬ5𝑛𝑠 cos(𝛾𝑛𝑠 𝑥𝑠 ) + ℬ6𝑛𝑠 sin(𝛾𝑛𝑠 𝑥𝑠 ) + ℬ7𝑛𝑠 cos(𝛿𝑛𝑠 𝑥𝑠 ) + ℬ8𝑛𝑠 sin(𝛿𝑛𝑠 𝑥𝑠 ) , (4.23b)

𝛷𝑛𝑠 (𝑥𝑠 ) = 𝒞1𝑛𝑠 cosh(𝛼𝑛𝑠 𝑥𝑠 ) + 𝒞2𝑛𝑠 sinh(𝛼𝑛𝑠 𝑥𝑠 ) + 𝒞3𝑛𝑠 cosh(𝛽𝑛𝑠 𝑥𝑠 ) + 𝒞4𝑛𝑠 sinh(𝛽𝑛𝑠 𝑥𝑠 )
+ 𝒞5𝑛𝑠 cos(𝛾𝑛𝑠 𝑥𝑠 ) + 𝒞6𝑛𝑠 sin(𝛾𝑛𝑠 𝑥𝑠 ) + 𝒞7𝑛𝑠 cos(𝛿𝑛𝑠 𝑥𝑠 ) + 𝒞8𝑛𝑠 sin(𝛿𝑛𝑠 𝑥𝑠 ) , (4.23c)

where the frequency parameter 𝜆𝑛𝑠 is obtained from

4 𝜌𝑠 𝐴𝑠 𝑝𝑛 2
𝜆𝑛𝑠 = √ . (4.24)
𝐸𝑠 𝐼𝑦𝑦𝑠

Similarly to eqns (4.21), the constants 𝒞𝑗𝑛𝑠 are related with ℬ𝑗𝑛𝑠 by

𝜌𝑠 𝐴𝑠 𝑝𝑛 2 − 𝐸𝑠 𝐼𝑧𝑧𝑠 𝛼𝑛𝑠 4 𝜌𝑠 𝐴𝑠 𝑝𝑛 2 − 𝐸𝑠 𝐼𝑧𝑧𝑠 𝛽𝑛𝑠 4


𝜅𝛼𝑛𝑠 = 𝑃 𝑝 2 , 𝜅𝛽𝑛𝑠 = 𝑃 𝑝 2 , (4.25a)
𝜌𝑠 𝑆𝑦𝑠 𝑛 𝜌𝑠 𝑆𝑦𝑠 𝑛

𝜌𝑠 𝐴𝑠 𝑝𝑛 2 − 𝐸𝑠 𝐼𝑧𝑧𝑠 𝛾𝑛𝑠 4 𝜌𝑠 𝐴𝑠 𝑝𝑛 2 − 𝐸𝑠 𝐼𝑧𝑧𝑠 𝛿𝑛𝑠 4


𝜅𝛾𝑛𝑠 = 𝑃 𝑝 2 , 𝜅𝛿𝑛𝑠 = 𝑃 𝑝 2 , (4.25b)
𝜌𝑠 𝑆𝑦𝑠 𝑛 𝜌𝑠 𝑆𝑦𝑠 𝑛

while 𝛼𝑛𝑠 , 𝛽𝑛𝑠 , 𝛾𝑛𝑠 and 𝛿𝑛𝑠 are defined according to (4.18) considering the non-dimensional parameters
2
𝜌𝑠 𝑝𝑛 2 𝜌𝑠 𝐴𝑠 𝑝𝑛 2 𝑃
𝑆𝑦𝑠 𝐸𝑠 𝐼𝜔𝜔𝑠
𝑃 + 𝐼𝑃 )
𝑎𝑠 = (𝐼𝑧𝑧𝑠 𝑦𝑦𝑠 , 𝑏𝑠 = , 𝑐𝑠 = 1 − , and 𝑑𝑠 = . (4.26)
𝐺𝑠 𝐽𝑠 𝐸𝑠 𝐼𝑧𝑧𝑠 𝑃
𝐴𝑠 (𝐼𝑧𝑧𝑠 𝑃
+ 𝐼𝑦𝑦𝑠 ) 𝐺𝑠 𝐽𝑠

The constants 𝒜𝑗𝑛𝑠 and ℬ𝑗𝑛𝑠 , since 𝒞𝑗𝑛𝑠 depends upon ℬ𝑗𝑛𝑠 , are calculated by imposing the boundary
conditions at the ends of the 𝑠-th span. For a continuous beam, the kinematic boundary conditions of the
𝑠-th span require that

𝑑𝛯𝑧𝑛𝑠 (𝑥𝑠 ) 𝑑𝛯𝑧𝑛𝑟 (𝑥𝑟 )


𝛯𝑧𝑛𝑠 (0) = 0 , 𝛯𝑧𝑛𝑠 (𝑙𝑠 ) = 0 , | = | , (4.27a)
𝑑𝑥𝑠 𝑥
𝑑𝑥𝑟
𝑠 =0 𝑥 𝑟 =𝑙𝑟

𝑑𝛯𝑦𝑛𝑠 (𝑥𝑠 ) 𝑑𝛯𝑦𝑛𝑟 (𝑥𝑟 )


𝛯𝑦𝑛𝑠 (0) = 0 , 𝛯𝑦𝑛𝑠 (𝑙𝑠 ) = 0 , | = | , (4.27b)
𝑑𝑥𝑠 𝑥
𝑑𝑥𝑟
𝑠 =0 𝑥 𝑟 =𝑙𝑟

𝑑𝛷𝑛𝑠 (𝑥𝑠 ) 𝑑𝛷𝑛𝑟 (𝑥𝑟 )


𝛷𝑛𝑠 (0) = 0 , 𝛷𝑛𝑠 (𝑙𝑠 ) = 0 , | = | . (4.27c)
𝑑𝑥𝑠 𝑥
𝑑𝑥𝑟
𝑠 =0 𝑥 𝑟 =𝑙𝑟

50
From continuity over the inner supports, the static boundary conditions imply the following relations:

𝑑2 𝛯𝑧𝑛𝑠 (𝑥𝑠 ) 𝑑2 𝛯𝑧𝑛𝑟 (𝑥𝑟 )


𝐸𝑠 𝐼𝑦𝑦𝑠 | = 𝐸𝑟 𝐼𝑦𝑦𝑟 | , (4.28a)
𝑑𝑥𝑠 2 𝑥
𝑑𝑥𝑟 2
𝑠 =0 𝑥 𝑟 =𝑙𝑟

𝑑2 𝛯𝑦𝑛𝑠 (𝑥𝑠 ) 𝑑2 𝛯𝑦𝑛𝑟 (𝑥𝑟 )


𝐸𝑠 𝐼𝑧𝑧𝑠 | = 𝐸𝑟 𝐼𝑧𝑧𝑟 | , (4.28b)
𝑑𝑥𝑠 2 𝑥
𝑑𝑥𝑟 2
𝑠 =0 𝑥 𝑟 =𝑙𝑟

𝑑2 𝛷𝑛𝑠 (𝑥𝑠 ) 𝑑2 𝛷𝑛𝑟 (𝑥𝑟 )


𝐸𝑠 𝐼𝜔𝜔𝑠 | = 𝐸𝑟 𝐼𝜔𝜔𝑟 | . (4.28c)
𝑑𝑥𝑠 2 𝑥
𝑑𝑥𝑟 2
𝑠 =0 𝑥 𝑟 =𝑙𝑟

The eqns (4.27) imply that the vertical and the lateral displacements, 𝛯𝑧𝑛𝑠 and 𝛯𝑦𝑛𝑠 respectively, as well as
the torsional rotation 𝛷𝑛𝑠 vanish at the two supports of the 𝑠-th span. The third terms of these equations
state that the slopes of two adjacent spans must be equal at the common support. The eqns (4.28) ensure
that the bending moments and the bimoments are continuous over the inner supports.

For an easier understanding of the assembly procedure presented by Chan, which is being extended to
coupled vibrations, the free bending decoupled motion about the 𝑦 axis is firstly considered.

Substituting the boundary conditions (4.27a) and (4.28a) in the equation of motion (4.23a), the constants
𝒜𝑗𝑛𝑠 can be obtained from explicit expressions. Considering that one of the four 𝒜𝑗𝑛1 constants, relative
to the first span, represents an arbitrary amplitude of the shape function 𝛯𝑧𝑛1 , one can obtain for the first
span

sinh(𝜆𝑛1 𝑙1 )
𝒜1𝑛1 = 0 , 𝒜2𝑛1 = 1 , 𝒜3𝑛1 = 0 , 𝒜4𝑛1 = − , (4.29)
sin(𝜆𝑛1 𝑙1 )

and for the 𝑠-th inner span

1 𝐸𝑟 𝐼𝑦𝑦𝑟 𝜆𝑛𝑟 2
𝒜1𝑛𝑠 = [𝒜 cosh(𝜆𝑛𝑟 𝑙𝑟 ) + 𝒜2𝑛𝑟 sinh(𝜆𝑛𝑟 𝑙𝑟 ) − 𝒜3𝑛𝑟 cos(𝜆𝑛𝑟 𝑙𝑟 ) − 𝒜4𝑛𝑟 sin(𝜆𝑛𝑟 𝑙𝑟 )] , (4.30a)
2 𝐸𝑠 𝐼𝑦𝑦𝑠 𝜆𝑛𝑠 2 1𝑛𝑟

𝜆𝑛𝑟
𝒜2𝑛𝑠 = [𝒜 sinh(𝜆𝑛𝑟 𝑙𝑟 ) + 𝒜2𝑛𝑟 cosh(𝜆𝑛𝑟 𝑙𝑟 ) − 𝒜3𝑛𝑟 sin(𝜆𝑛𝑟 𝑙𝑟 )
𝜆𝑛𝑠 1𝑛𝑟
sin(𝜆𝑛𝑠 𝑙𝑠 ) cos(𝜆𝑛𝑠 𝑙𝑠 ) − cosh(𝜆𝑛𝑠 𝑙𝑠 )
+ 𝒜4𝑛𝑟 cos(𝜆𝑛𝑟 𝑙𝑟 )] − 𝒜1𝑛𝑠 , (4.30b)
sin(𝜆𝑛𝑠 𝑙𝑠 ) − sinh(𝜆𝑛𝑠 𝑙𝑠 ) sin(𝜆𝑛𝑠 𝑙𝑠 ) − sinh(𝜆𝑛𝑠 𝑙𝑠 )

𝒜3𝑛𝑠 = −𝒜1𝑛𝑠 , (4.30c)


𝜆𝑛𝑟
𝒜4𝑛𝑠 = [𝒜 sinh(𝜆𝑛𝑟 𝑙𝑟 ) + 𝒜2𝑛𝑟 cosh(𝜆𝑛𝑟 𝑙𝑟 ) − 𝒜3𝑛𝑟 sin(𝜆𝑛𝑟 𝑙𝑟 ) + 𝒜4𝑛𝑟 cos(𝜆𝑛𝑟 𝑙𝑟 )] − 𝒜2𝑛𝑠 . (4.30d)
𝜆𝑛𝑠 1𝑛𝑟

Then, by imposing the boundary condition on the bending moment at the end of the last span 𝑁,

𝑑2 𝛯𝑧𝑛𝑁 (𝑥𝑁 )
| =0, (4.31)
𝑑𝑥𝑁 2 𝑥 𝑁 =𝑙𝑁

the natural frequencies of vibration for the decoupled bending motion about the 𝑦 axis are given by the
roots of the following equation:

𝒜1𝑛𝑁 cosh(𝜆𝑛𝑁 𝑙𝑁 ) + 𝒜2𝑛𝑁 sinh(𝜆𝑛𝑁 𝑙𝑁 ) − 𝒜3𝑛𝑁 cos(𝜆𝑛𝑁 𝑙𝑁 ) − 𝒜4𝑛𝑁 sin(𝜆𝑛𝑁 𝑙𝑁 ) = 0 . (4.32)

51
Notice that the equation (4.32) has not been presented in [Chan and Ashebo, 2006]. In fact, similar
defining expressions for the constants 𝒜𝑗𝑛𝑠 have been derived, but the procedure of finding the natural
frequencies is unclear.

The formulation that has been presented is now extended to coupled lateral bending-torsional vibrations
of the continuous beam illustrated in Figure 4.1.

The eight constants ℬ𝑗𝑛1 of the first span are obtained by substituting the remaining boundary conditions
(4.27) and (4.28) in the equations (4.23b) and (4.23c), yielding

ℬ1𝑛1 = 0 , ℬ3𝑛1 = 0 , ℬ5𝑛1 = 0 , ℬ7𝑛1 = 0 , (4.33a)

ℬ2𝑛1 sinh(𝛼𝑛1 𝑙1 ) (𝜅𝛿1 − 𝜅𝛼1 ) + ℬ4𝑛1 sinh(𝛽𝑛1 𝑙1 ) (𝜅𝛿1 − 𝜅𝛽1 )


ℬ6𝑛1 = , (4.33b)
sin(𝛾𝑛1 𝑙1 ) (𝜅𝛾1 − 𝜅𝛿1 )

ℬ2𝑛1 sinh(𝛼𝑛1 𝑙1 ) (𝜅𝛼1 − 𝜅𝛾1 ) + ℬ4𝑛1 sinh(𝛽𝑛1 𝑙1 ) (𝜅𝛽1 − 𝜅𝛾1 )


ℬ8𝑛1 = . (4.33c)
sin(𝛿𝑛1 ) (𝜅𝛾1 − 𝜅𝛿1 )

For the generic inner span 𝑠 the expressions of the eight constants ℬ𝑗𝑛𝑠 can be easily obtained using
symbolic algebra software, such as Mathematica. Although these expressions are quite large, see Appendix
I, they hold for any inner span. Therefore, by computing the explicit expressions for the constants ℬ𝑗𝑛𝑠 , it
is straightforward to apply the presented analytical procedure to girders with any number of spans.

The natural frequencies 𝑝𝑛 for the coupled lateral bending-torsional vibrations are obtained by imposing
the static boundary conditions at the end of the 𝑁-th span,

𝑑2 𝛯𝑦𝑛𝑁 (𝑥𝑁 ) 𝑑2 𝛷𝑛𝑁 (𝑥𝑁 )


| =0 and | =0, (4.34)
𝑑𝑥𝑁 2 𝑥
𝑑𝑥𝑁 2
𝑁 =𝑙𝑁 𝑥 𝑁 =𝑙𝑁

and finding the roots that satisfy both equations.

This procedure would not be feasible to compute by symbolic algebra software, since it is not a priori
possible to express six of the eight constants ℬ𝑗𝑛1 in terms of the seventh and the eighth constants and
give them a numerical value. This means that the 𝑠 -th span constants ℬ𝑗𝑛𝑠 must be evaluated as
functionals, i.e., functions of ℬ𝑗𝑛𝑟 , which can be considered functions of ℬ𝑗𝑛1 . Thus, very large expressions
are obtained, leading to long computation time.

To overcome this problem, a very simple technique may be employed. Consider that the boundary
conditions (4.34) are written, in the matrix form, only as functions of ℬ2𝑛1 and ℬ4𝑛1 as follows:

𝑆22 𝑆24 ℬ2𝑛1


[ ]{ }=𝟎, (4.35)
𝑆42 𝑆44 ℬ4𝑛1

where the numerical value of each component 𝑆𝑘𝑙 can be readily obtained by making ℬ𝑘𝑛1 and ℬ𝑙𝑛1 equal
to zero and one, respectively, eqns (4.33).

The crux of this development has now been reached: by solving the matrix equation (4.35) it is possible to
express seven of the eight constants ℬ𝑗𝑛1 in terms of the eighth. The eighth constant cannot be evaluated

52
directly in a free vibration analysis because it represents an arbitrary amplitude of the mode shape
components 𝛯𝑦𝑛𝑠 and 𝛷𝑛𝑠 .

An implicit expression from which the natural frequencies are determined can be derived by giving ℬ2𝑛1 a
numerical value, say unity, and therefore the constants ℬ𝑗𝑛1 (4.33) become

ℬ1𝑛1 = 0 , ℬ3𝑛1 = 0 , ℬ5𝑛1 = 0 , ℬ7𝑛1 = 0 , (4.36a)


𝑆22
ℬ2𝑛1 = 1 , ℬ4𝑛1 = − , (4.36b)
𝑆24

𝑆24 sinh(𝛼𝑛1 𝑙1 ) (𝜅𝛿1 − 𝜅𝛼1 ) + 𝑆22 sinh(𝛽𝑛1 𝑙1 ) (𝜅𝛽1 − 𝜅𝛿1 )


ℬ6𝑛1 = , (4.36c)
𝑆24 sin(𝛾𝑛1 𝑙1 ) (𝜅𝛾1 − 𝜅𝛿1 )

𝑆24 sinh(𝛼𝑛1 𝑙1 ) (𝜅𝛼1 − 𝜅𝛾1 ) + 𝑆22 sinh(𝛽𝑛1 𝑙1 ) (𝜅𝛾1 − 𝜅𝛽1 )


ℬ8𝑛1 = . (4.36d)
𝑆24 sin(𝛿𝑛1 𝑙1 ) (𝜅𝛾1 − 𝜅𝛿1 )

Finally, the natural frequencies 𝑝𝑛 for the coupled bending-torsional motion are calculated by seeking for
the roots of the following equation:

𝑆22 𝑆24 (4.37)


det ([ ]) = 0 .
𝑆42 𝑆44

Once the frequencies of vibration have been calculated, the 𝑚- or the 𝑛 -th mode shape, in which the
entire continuous beam is said to respond when vibrates with a frequency 𝑝𝑚 or 𝑝𝑛 , is established through
the assembly of the 𝑚- or the 𝑛-th mode shape functions calculated to each span, as follows:
𝑁 𝑠−1 𝑠
𝛯𝑧𝑚 (𝑥) = ∑ [Η (𝑥 − ∑ 𝑙𝑟 ) − Η (𝑥 − ∑ 𝑙𝑟 )] 𝛯𝑧𝑚𝑠 (𝑥𝑠 ) , (4.38a)
𝑠=1 𝑟=1 𝑟=1

𝑁 𝑠−1 𝑠
𝛯𝑦𝑛 (𝑥) = ∑ [Η (𝑥 − ∑ 𝑙𝑟 ) − Η (𝑥 − ∑ 𝑙𝑟 )] 𝛯𝑦𝑛𝑠 (𝑥𝑠 ) , (4.38b)
𝑠=1 𝑟=1 𝑟=1

𝑁 𝑠−1 𝑠
𝛷𝑛 (𝑥) = ∑ [Η (𝑥 − ∑ 𝑙𝑟 ) − Η (𝑥 − ∑ 𝑙𝑟 )] 𝛷𝑛𝑠 (𝑥𝑠 ) , (4.38c)
𝑠=1 𝑟=1 𝑟=1

being Η the Heaviside step function.

4.1.2 VISCOUS DAMPING

As it was mentioned in Section 3.2.2, damping values may not be constant for all vibration modes,
[Clough and Penzien, 1995]. One way to quantify the damping properties, when they are considered
distributed throughout the structural member, is by establishing a proportional relation to the modal mass
properties.

A procedure for determining the modal mass properties is by considering the orthogonality of the
vibration mode shapes with respect to mass, which allows the decoupling of the governing equations (4.3).

53
The orthogonality conditions, equivalent to those defined in Section 3.2.2 for a discrete system, can be
demonstrated for a continuum system through the application of the Bettis’s law. In fact, the work of the
inertial forces of mode 𝑛 on the displacements of mode 𝑚 is equal to the work of the 𝑚-th mode forces
on the 𝑛-th mode deflection.

Considering that the beam’s generalized displacements 𝜉𝑧 , 𝜉𝑦 and 𝜙 are defined using the modal
superposition method by

𝜉𝑧 (𝑥, 𝑡) = ∑ 𝛯𝑧𝑘 (𝑥)𝛹𝑘 (𝑡) , (4.39a)
𝑘


𝜉𝑦 (𝑥, 𝑡) = ∑ 𝛯𝑦𝑙 (𝑥)𝛶𝑙 (𝑡) , (4.39b)
𝑙


𝜙 (𝑥, 𝑡) = ∑ 𝛷𝑙 (𝑥)𝛶𝑙 (𝑡) , (4.39c)
𝑙

where 𝛹𝑘 and 𝛶𝑙 are the time-varying amplitudes of the 𝑘-th and the 𝑙-th vibration modes, see Figure 4.3.

Substituting the eqns (4.39) in the governing equations (4.3), the Betti’s law stands for
𝐿 𝐿
∫ 𝛯𝑧𝑚 (𝑥)𝛹𝑚 (𝑡)𝜌𝐴𝛯𝑧𝑛 (𝑥)𝑝𝑛 2 𝛹𝑛 (𝑡) 𝑑𝑥 = ∫ 𝛯𝑧𝑛 (𝑥)𝛹𝑛 (𝑡)𝜌𝐴𝛯𝑧𝑚 (𝑥)𝑝𝑚 2 𝛹𝑚 (𝑡) 𝑑𝑥 , (4.40a)
0 0

𝐿
∫ [𝛯𝑦𝑚 (𝑥)𝛶𝑚 (𝑡)𝜌𝐴𝛯𝑦𝑛 (𝑥)𝑝𝑛 2 𝛶𝑛 (𝑡) + 𝛷𝑚 (𝑥)𝛶𝑚 (𝑡)𝜌(𝐼𝑧𝑧 + 𝐼𝑦𝑦 )𝛷𝑛 (𝑥)𝑝𝑛 2 𝛶𝑛 (𝑡)] 𝑑𝑥
0
𝐿
= ∫ [𝛯𝑦𝑛 (𝑥)𝛶𝑛 (𝑡)𝜌𝐴𝛯𝑦𝑚 (𝑥)𝑝𝑚 2 𝛶𝑚 (𝑡) + 𝛷𝑛 (𝑥)𝛶𝑛 (𝑡)𝜌(𝐼𝑧𝑧 + 𝐼𝑦𝑦 )𝛷𝑚 (𝑥)𝑝𝑚 2 𝛶𝑚 (𝑡)] 𝑑𝑥 . (4.40b)
0

Mode 𝑚 Mode 𝑛
𝜉𝑛 𝜉𝑧𝑚 (𝑥, 𝑡) = 𝛯𝑧𝑚 (𝑥)𝛹𝑚 (𝑡) 𝜉𝑛 𝜉𝑧𝑛 (𝑥, 𝑡) = 𝛯𝑧𝑛 (𝑥)𝛹𝑛 (𝑡)

𝑥 𝑥

𝐿 𝐿

inertial forces relative to the 𝑚-th mode


inertial forces relative to the 𝑛-th mode

Figure 4.3 – vibration mode shapes and inertial forces of the 𝑚- and the 𝑛-th modes.

Subtracting both sides of eqns (4.40) from one another the following orthogonality conditions are
obtained:
𝐿
∫ 𝛯𝑧𝑚 (𝑥)𝛯𝑧𝑛 (𝑥) 𝑑𝑥 = 0 , 𝑝𝑛 ≠ 𝑝𝑚 (4.41a)
0

𝐿
∫ [𝜌𝐴𝛯𝑦𝑚 (𝑥)𝛯𝑦𝑛 (𝑥) + 𝜌(𝐼𝑧𝑧 + 𝐼𝑦𝑦 )𝛷𝑚 (𝑥)𝛷𝑛 (𝑥)] 𝑑𝑥 = 0 , 𝑝𝑛 ≠ 𝑝𝑚 (4.41b)
0

54
The decoupling of the governing equations (4.3) is accomplished by considering the modal coordinate
expressions (4.39), which leads to

∑ [𝜌𝐴𝛯𝑧𝑘 (𝑥)𝛹̈𝑘 (𝑡) + 𝐸𝐼𝑦𝑦 𝛯𝑧𝑘
′′′′ (𝑥)𝛹 (𝑡)]
𝑘 = 𝑓𝑧 (𝑥, 𝑡) , (4.42a)
𝑘


∑ [𝜌𝐴𝛯𝑦𝑙 (𝑥)𝛶𝑙̈ (𝑡) + 𝐸𝐼𝑧𝑧 𝛯𝑦𝑙
′′′′ (𝑥)𝛶 (𝑡)
𝑙 + 𝐸𝐼𝑧𝜔 𝛷𝑙′′′′ (𝑥)𝛶𝑙 (𝑡)] = 𝑓𝑦 (𝑥, 𝑡) , (4.42b)
𝑙


∑ [𝜌(𝐼𝑧𝑧 + 𝐼𝑦𝑦 )𝛷𝑙 (𝑥)𝛶𝑙̈ (𝑡) + 𝐸𝐼𝜔𝑧 𝛯𝑦𝑙
′′′′ (𝑥)𝛶 (𝑡)
𝑙 + 𝐸𝐼𝜔𝜔 𝛷𝑙′′′′ (𝑥)𝛶𝑙 (𝑡) − 𝐺𝐽𝛷𝑙′′ (𝑥)𝛶𝑙 (𝑡)] = 𝑚𝑥 (𝑥, 𝑡) . (4.42c)
𝑙

Multiplying the eqn (4.42a) by 𝛯𝑧𝑚 (𝑥) , integrating with respect to 𝑥 , and applying the orthogonality
condition (4.41a) yields

𝑀𝑧𝑚 𝛹̈𝑚 (𝑡) + 𝑝𝑚 2 𝑀𝑧𝑚 𝛹𝑚 (𝑡) = 𝑃𝑧𝑚 (𝑡) , (4.43)


where the generalized mass 𝑀𝑧𝑚 and the generalized loading 𝑃𝑧𝑚 associated with the 𝑚 -th mode of
vibration are given by
𝐿
𝑀𝑧𝑚 = 𝜌𝐴 ∫ 𝛯𝑧𝑚 2 (𝑥) 𝑑𝑥 , (4.44a)
0

𝐿
𝑃𝑧𝑚 (𝑡) = ∫ 𝛯𝑧𝑚 (𝑥)𝑓𝑧 (𝑥, 𝑡) 𝑑𝑥 . (4.44b)
0

A similar procedure is now applied to the eqns (4.42b) and (4.42c). Multiplying these equations by 𝛯𝑦𝑛 and
𝛷𝑛 , respectively, integrating with respect to 𝑥, adding them to one another and applying the orthogonality
condition (4.41b) gives

𝑀𝑦𝜙𝑛 𝛶𝑛̈ (𝑡) + 𝑝𝑛 2 𝑀𝑦𝜙𝑛 𝛶𝑛 (𝑡) = 𝑃𝑦𝜙𝑛 (𝑡) , (4.45)

being the generalized mass 𝑀𝑦𝜙𝑛 and the genralized loading 𝑃𝑦𝜙𝑛 associated with the 𝑛-th mode:
𝐿
𝑀𝑦𝜙𝑛 = ∫ [𝜌𝐴𝛯𝑦𝑛 2 (𝑥) + 𝜌(𝐼𝑧𝑧 + 𝐼𝑦𝑦 )𝛷𝑛 2 (𝑥)] 𝑑𝑥 , (4.46a)
0

𝐿
𝑃𝑦𝜙𝑛 (𝑡) = ∫ [𝛯𝑦𝑛 (𝑥)𝑓𝑦 (𝑥, 𝑡) + 𝛷𝑛 (𝑥)𝑚𝑥 (𝑥, 𝑡)] 𝑑𝑥 . (4.46b)
0

The viscous damping is now obtained in an equivalent way to that used in eqn (3.60) and (3.62). Thus, the
decoupled modal equations are written as follows:

𝑀𝑧𝑚 𝛹̈𝑚 (𝑡) + 𝐶𝑧𝑚 𝛹̇𝑚 (𝑡) + 𝑝𝑚 2 𝑀𝑧𝑚 𝛹𝑚 (𝑡) = 𝑃𝑧𝑚 (𝑡) , (4.47a)
𝑀𝑦𝜙𝑛 𝛶𝑛̈ (𝑡) + 𝐶𝑦𝜙𝑛 𝛶𝑛̇ (𝑡) + 𝑝𝑛 2 𝑀𝑦𝜙𝑛 𝛶𝑛 (𝑡) = 𝑃𝑦𝜙𝑛 (𝑡) , (4.47b)

where the generalized damping coefficients 𝐶𝑧𝑚 and 𝐶𝑦𝜙𝑛 are given by

𝐶𝑧𝑚 = 2𝜍𝑧𝑚 𝑝𝑚 𝑀𝑧𝑚 , (4.48a)


𝐶𝑦𝜙𝑛 = 2𝜍𝑦𝜙𝑛 𝑝𝑛 𝑀𝑦𝜙𝑛 , (4.48b)
being 𝜍𝑧𝑚 and 𝜍𝑦𝜙𝑛 the modal viscous damping ratios of each vibration mode.

55
4.1.3 DIRECT INTEGRATION OF THE GOVERNING EQUATIONS

As in the discrete case, the Newmark’s 𝛽 method is used for solving the governing equations (4.47). The
step-by-step integration equations for the generalized displacements and velocities at time 𝑡 + Δ𝑡 are given
by

𝑢𝑡+Δ𝑡 = 𝑢𝑡 + Δ𝑡𝑢̇ 𝑡 + (1⁄2 − 𝛽)(Δ𝑡)2 𝑢̈ 𝑡 + 𝛽(Δ𝑡)2 𝑢̈ 𝑡+Δ𝑡 , (4.49a)

𝑢̇ 𝑡+Δ𝑡 = 𝑢̇ 𝑡 + Δ𝑡𝑢̇ 𝑡 + (1 − 𝛾)Δ𝑡𝑢̈ 𝑡 + 𝛽Δ𝑡𝑢̈ 𝑡+Δ𝑡 , (4.49b)


where 𝑢 represents the modal amplitudes 𝛹𝑚 or 𝛶𝑛 .

The implicit recurrence relation for 𝑢𝑡+Δ𝑡 can be converted into an explicit form by substituting the eqns
(4.49) into (4.47) at time 𝑡 + Δ𝑡, which result is written as follows:

𝒦𝑢𝑡+Δ𝑡 = ℱ𝑡+Δ𝑡 , (4.50)

being the effective modal stiffness 𝒦 and the effective modal loading ℱ𝑡+Δ𝑡 respectively defined as

1 𝛾
𝒦= 𝑀+ 𝐶+𝐾, (4.51a)
𝛽(Δ𝑡)2 𝛽Δ𝑡

1 1 1 𝛾 𝛾 𝛾 Δ𝑡
ℱ𝑡+Δ𝑡 = [ 𝑢𝑡 + 𝑢̇ 𝑡 + ( − 1) 𝑢̈ 𝑡 ] 𝑀 + [ 𝑢𝑡 + ( − 1) 𝑢̇ 𝑡 + ( − 2) 𝑢̈ ] 𝐶 + 𝑃𝑡+Δ𝑡 , (4.51b)
𝛽(Δ𝑡)2 𝛽Δ𝑡 2𝛽 𝛽Δ𝑡 𝛽 𝛽 2 𝑡

in which 𝑀 , 𝐶 and 𝐾 represent the modal mass 𝑀𝑧𝑚 or 𝑀𝑦𝜙𝑛 , the modal damping 𝐶𝑧𝑚 or 𝐶𝑦𝜙𝑛 , and the
modal stiffness 𝑝𝑚 2 𝑀𝑧𝑚 or 𝑝𝑛 2𝑀𝑦𝜙𝑛 , respectively. The generalized loading 𝑃 represents either 𝑃𝑧𝑚 or 𝑃𝑦𝜙𝑛 ,
where the magnitude of the applied loads at time 𝑡 + Δ𝑡 are transformed into modal coordinates.

Since 𝑢𝑡+Δ𝑡 is calculated using eqn (4.50), the velocity and the acceleration at the same time step become

𝛾 1
𝑢̇ 𝑡+Δ𝑡 = 𝑢̇ 𝑡 + (1 − 𝛾)Δ𝑡𝑢̈ 𝑡 + [𝑢𝑡+Δ𝑡 − 𝑢𝑡 − Δ𝑡𝑢̇ 𝑡 − ( − 𝛽) (Δ𝑡)2 𝑢̈ 𝑡 ] , (4.52a)
𝛽Δ𝑡 2

1 1
𝑢̈ 𝑡+Δ𝑡 = [𝑢 − 𝑢𝑡 + Δ𝑡𝑢̇ 𝑡 + ( − 𝛽) (Δ𝑡)2 𝑢̈ 𝑡 ] . (4.52b)
𝛽(Δ𝑡)2 𝑡+Δ𝑡 2

By adopting 𝛾 = 1/2 and 𝛽 = 1/4, the method is unconditionally stable, which enables the convergence to
the solution regardless the initial conditions and the time step Δ𝑡.

56
5
Dynamic Response of
Continuous Beams
under Moving Loads

I n this chapter, the numerical and the analytical beam models, developed in Chapter 3 and Chapter 4, are
applied in order to perform the dynamic analysis of multi-span bridges eccentrically traversed by
moving vehicles.

Two different mono-symmetric cross sections are considered: one open double-T cross section, and one
closed box cross section. The corresponding geometries are sketched in Figure 5.1, being the material and
the geometrical properties presented in Table 5.1. In both cases, the bridge’s girder is modelled as a thin
walled beam, and the influence of warping is discussed in the sequel.

A simply supported beam subjected to a single moving load is firstly analysed by the developed beam
models. The results obtained are compared with the exact results calculated by the analytical formulation
proposed by [Fryba, 1999]. Afterwards, the study is extended to a three-span bridge traversed by a set of
high-speed moving loads.

The effectiveness of the developed thin walled beam models is verified by comparison with a shell finite
element model implemented in the commercial software SAP2000. The selected thin-shell elements use a
four-node formulation that combines separately the membrane and the bending behaviour. To simulate
the cross-sectional in-plane indeformability of the bridge’s girder, body constraints are assigned to the
element nodes.

The natural frequencies and mode shapes of vibration are obtained for the cases of decoupled vertical
bending motion and coupled lateral bending-torsional motion. Forced vibrations due to eccentric moving

57
loads with constant velocity are evaluated during the period of the load moving along the structure, being
the mass effect of the moving vehicles also taken into account.

A MATLAB code is used for the numerical treatment of both the finite element and the analytical solutions
derived in the previous chapters. The time needed to define the bridge and the loading features and run
the developed beam models does not exceed a couple of minutes in a common portable computer. This is
substantially less in comparison with the time needed to model the same structure using shell finite
elements in commercial software and get ready to perform a similar dynamic analysis.

(a) – Double-T Cross Section


𝐶𝐶
3.33 6.65 3.33 𝑧̃𝐶𝐶

𝑧̃𝐶𝐺
𝐶𝐺
𝑦 0.35
3.03
0.80

𝑧
(b) – Box Cross Section
3.00 7.30 3.00

𝑧̃𝐶𝐺 𝐶𝐺 𝑧̃𝐶𝐶 0.35


𝑦 𝐶𝐶 2.18
0.50

0.20
𝑧

Figure 5.1 – general geometry of the cross sections considered in the practical examples (m).

Table 5.1 – material and geometrical properties of the cross sections considered in the practical examples.
Property Double-T Cross Section Box Cross Section
𝐸 (kN/m2) 33 000 33 000
𝐺 (kN/m2) 13 750 13 750
𝜌 (ton/m3) 2.5484 2.5484
𝐴 (m2) 9.5065 8.2950
𝐴𝑚 (m2) – 15.9140
𝑧̃𝐶𝐺 (m) 0.7726 0.6702
𝑧̃𝐶𝐶 (m) 0.6636 0.8174
𝐽 (m4) 1.2245 15.4166
𝛹 (m2) – 0.4817
𝐼𝑦𝑦 (m4) 9.1618 6.6665
𝐼𝑧𝑧 (m4) 122.3711 104.1452
𝐼𝑧𝜔 (m5) 175.7444 -15.3345
𝐼𝜔𝜔 (m6) 362.5405 23.2883

58
5.1 SIMPLY SUPPORTED BEAM

This section deals with the undamped dynamic response of a simply supported beam, see Figure 5.2. The
structure is subjected to a single moving load 𝑃𝑧 with magnitude of 1000 kN travelling eccentrically at a
constant velocity of 100 ms-1. The prescribed boundary conditions at the supports are listed in Table 5.2.

𝑒𝑐𝑐 =2.5 m 𝑒𝑐𝑐 =2.5 m


𝑃𝑧 𝑃𝑧 𝑃𝑧
𝑆1 𝑆2

𝐶𝐺 𝐶𝐺
𝑦 𝑦

𝐿=40 m
𝑧 𝑧
𝑥
Figure 5.2 – longitudinal model of the single-span bridge considered in the practical examples.

Table 5.2 – prescribed boundary conditions at the supports of the single-span bridge.
Support Nodal Coordinate Kinematic Boundary Conditions Static Boundary Conditions
𝑆1 𝑥 =0m
𝜉𝑧 (𝑥) = 𝜉𝑦 (𝑥) = 𝜙(𝑥) = 0 𝜉𝑧′′ (𝑥) = 𝜉𝑦′′ (𝑥) = 𝜙 ′′ (𝑥) = 0
𝑆2 𝑥 = 40 m

5.1.1 UNDAMPED FREE VIBRATION ANALYSIS

The undamped natural frequencies of vibration calculated by the two developed beam models are
compared in Table 5.3 and Table 5.4 with the results given by the shell finite element model implemented
in SAP2000. The values are presented either including or neglecting the warping effect in the structural
model. The inclusion of the cross-sectional rotary inertia in the models is also evaluated being listed the
corresponding results in terms of natural frequencies.

In what concerns the vertical bending motion in the plane 𝑥𝑧 it is verified that the natural frequencies
obtained with the shell model vary when the warping of the cross sections is considered or neglected,
which would not be expected to occur. This fact is due to shear lag problems, motivated by shear
distortions of the shell elements.

In Table 5.3 and Table 5.4 the so-called exact frequency values correspond to the classical analytical
solution to the elementary case of a uniform simple beam, [Clough and Penzien, 1995]:

𝐸𝐼𝑦𝑦
𝑝𝑛 = (𝑛𝜋)2 √ (5.1)
𝜌𝐴𝐿4

59
In the case of coupled lateral bending-torsional vibrations, it is observed that when warping is constrained
the natural frequencies result higher, in special the vibration modes in which the torsional component is
dominant.

The frequency values given by the numerical model are in excellent agreement with the analytical values.
When the cross section’s rotary inertia is neglected in the numerical model the results coincide with the
analytical ones, since this inertial property constitutes the main difference between both beam models.

The in-plane rigid cross sections were simulated in the shell model by assigning body constraints to the
model joints. For the case of free warping, only the relative rotation about the 𝑥 axis was constrained.
When no allowance is made for warping displacements the rotations about 𝑦 and 𝑧 were also constrained.

Table 5.3 – natural frequencies of vibration of the double-T cross section girder.
Natural Frequencies of Vibration in the plane 𝑥𝑧 (Hz)
Model of Analysis
Numerical Analytical Shell FEM Exact
Inc. Inc. Negl. Negl.
Inc. Warping Negl. Warping
Warping Warping Warping Warping
Inc. Negl. Inc. Negl. Negl. Inc. Inc. Negl.
Frequency
Rotary Rotary Rotary Rotary Rotary Rotary Rotary Rotary
Number
Inertia Inertia Inertia Inertia Inertia Inertia Inertia Inertia
1 3.4579 3.4682 3.4579 3.4682 3.4682 3.3870 3.3959 3.4682
2 13.7107 13.8727 13.7107 13.8727 13.8727 12.6340 12.7380 13.8727
3 30.4107 31.2136 30.4107 31.2136 31.2136 25.8590 26.2500 31.2136
4 53.0263 55.4909 53.0263 55.4909 55.4909 41.3880 42.3570 55.4909

Natural Frequencies of Vibration in the plane 𝑥𝑦 (Hz)


Model of Analysis
Numerical Analytical Shell FEM
Inc. Inc. Negl.
Inc. Warping Negl. Warping
Warping Warping Warping
Inc. Negl. Negl.
Frequency Inc. Rotary Negl. Rotary Inc. Rotary Inc. Rotary
Rotary Rotary Rotary
Number Inertia Inertia Inertia Inertia
Inertia Inertia Inertia
1 3.9540 3.9629 12.1990 12.6739 3.9629 3.9190 10.0910
2 12.9714 13.0866 44.1593 50.6811 13.0866 12.1590 28.6130
3 13.0936 13.6822 - - 13.6822 10.7230 29.0480
4 27.5861 28.1344 87.0927 113.9780 28.1344 23.7680 48.5770
5 46.6852 54.5991 - - 54.5991 29.6510 57.6180
6 47.5080 49.1742 134.5508 202.4902 49.1742 37.1910 69.0200

The mode shapes of vibration corresponding to the natural frequencies listed in Table 5.3 are plotted
from Figure 5.3 to Figure 5.8, being separately presented the cases where warping effect is considered or
not.

60
(a) – Frequency Number 1 (b) – Frequency Number 2
1 1

0.5 0.5
𝑧-displacement

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1 -1
(c) – Frequency Number 3 (d) – Frequency Number 4
1 1

0.5 0.5
𝑧-displacement

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1 -1
𝑥 (m) 𝑥 (m)
∎ Numerical Model (including rotary inertia) ∎ Numerical Model (neglecting rotary inertia)
∎ Analytical Model ∎ Shell FEM Model
Figure 5.3 – vertical bending vibration mode shapes of the double-T cross section girder (including warping).

(a) – Frequency Number 1 (b) – Frequency Number 2


1 1

0.5 0.5
𝑧-displacement

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1 -1
(c) – Frequency Number 3 (d) – Frequency Number 4
1 1

0.5 0.5
𝑧-displacement

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1 -1
𝑥 (m) 𝑥 (m)
∎ Numerical Model (including rotary inertia) ∎ Numerical Model (neglecting rotary inertia)
∎ Shell FEM Model
Figure 5.4 – vertical bending vibration mode shapes of the double-T cross section girder (neglecting warping).

61
(a) – Frequency Number 1
1 1

0.5 0.5
𝑦-displacement

𝜙-rotation
0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1 -1
(b) – Frequency Number 2
1 1

0.5 0.5
𝑦-displacement

𝜙-rotation
0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1 -1
(c) – Frequency Number 3
1 1

0.5 0.5
𝑦-displacement

𝜙-rotation

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1 -1
(d) – Frequency Number 4
1 1

0.5 0.5
𝑦-displacement

𝜙-rotation

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1 -1
(e) – Frequency Number 5
1 1

0.5 0.5
𝑦-displacement

𝜙-rotation

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1 -1
𝑥 (m) 𝑥 (m)
∎ Numerical Model (including rotary inertia) ∎ Numerical Model (neglecting rotary inertia)
∎ Analytical Model ∎ Shell FEM Model
Figure 5.5 – lateral bending-torsional vibration mode shapes of the double-T
cross section girder (including warping) (Part 1/2).

62
(f) – Frequency Number 6
1 1

0.5 0.5
𝑦-displacement

𝜙-rotation
0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1 -1
𝑥 (m) 𝑥 (m)
∎ Numerical Model (including rotary inertia) ∎ Numerical Model (neglecting rotary inertia)
∎ Analytical Model ∎ Shell FEM Model
Figure 5.6 – lateral bending-torsional vibration mode shapes of the double-T
cross section girder (including warping) (Part 2/2).

(a) – Frequency Number 1


1 1

0.5 0.5
𝑦-displacement

𝜙-rotation

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1 -1
(b) – Frequency Number 2
1 1

0.5 0.5
𝑦-displacement

𝜙-rotation

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1 -1
(c) – Frequency Number 3
1.0 1.0

0.5 0.5
𝑦-displacement

𝜙-rotation

0.0 0.0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1.0 -1.0
𝑥 (m) 𝑥 (m)
∎ Numerical Model (including rotary inertia) ∎ Numerical Model (neglecting rotary inertia)
∎ Shell FEM Model
Figure 5.7 – lateral bending-torsional vibration mode shapes of the double-T
cross section girder (neglecting warping) (Part 1/2).

63
(d) – Frequency Number 4
1 1

0.5 0.5
𝑦-displacement

𝜙-rotation
0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1 -1
(e) – Frequency Number 5
1.0 1.0

0.5 0.5
𝑦-displacement

𝜙-rotation
0.0 0.0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1.0 -1.0
(f) – Frequency Number 6
1 1

0.5 0.5
𝑦-displacement

𝜙-rotation

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1 -1
𝑥 (m) 𝑥 (m)
∎ Numerical Model (including rotary inertia) ∎ Numerical Model (neglecting rotary inertia)
∎ Shell FEM Model
Figure 5.8 – lateral bending-torsional vibration mode shapes of the double-T
cross section girder (neglecting warping) (Part 2/2).

From Table 5.3 one can notice that the natural frequencies of vibrations in the plane 𝑥𝑦 are highly
influenced when the cross-sectional warping is neglected, since the torsional stiffness is greatly increased
when the warping displacements are constrained.

In Figure 5.7 and Figure 5.8 it should be noticed that the frequencies number 3 and number 5 were not
determined by the beam models for the range of frequencies analysed. In these vibration modes the
torsional component is dominant, which may justify this fact. The shell model is able to obtain such
modes due to shear deformations of the shell elements.

64
Table 5.4 – natural frequencies of vibration of the box cross section girder.
Natural Frequencies of Vibration in the plane 𝑥𝑧 (Hz)
Model of Analysis
Numerical Analytical Shell FEM Exact
Inc. Inc. Negl. Negl.
Inc. Warping Negl. Warping
Warping Warping Warping Warping
Inc. Negl. Inc. Negl. Negl. Inc. Inc. Negl.
Frequency
Rotary Rotary Rotary Rotary Rotary Rotary Rotary Rotary
Number
Inertia Inertia Inertia Inertia Inertia Inertia Inertia Inertia
1 3.1593 3.1671 3.1593 3.1671 3.1671 3.0696 3.0798 3.1671
2 12.5446 12.6684 12.5446 12.6684 12.6684 10.9120 11.0310 12.6684
3 27.8885 28.5039 27.8885 28.5039 28.5039 21.2250 21.6560 28.5039
4 48.7760 50.6736 48.7760 50.6736 50.6736 32.5370 33.5630 50.6736

Natural Frequencies of Vibration in the plane 𝑥𝑦 (Hz)


Model of Analysis
Numerical Analytical Shell FEM
Inc. Inc. Negl.
Inc. Warping Negl. Warping
Warping Warping Warping
Inc. Negl. Negl.
Frequency Inc. Rotary Negl. Rotary Inc. Rotary Inc. Rotary
Rotary Rotary Rotary
Number Inertia Inertia Inertia Inertia
Inertia Inertia Inertia
1 10.8876 10.9022 12.0596 12.5179 10.9022 10.6410 10.8480
2 12.1086 12.5599 - - 12.5599 11.1090 29.3200
3 22.4422 22.4950 43.7510 50.0709 22.4950 22.0070 33.4630
4 35.1033 35.2877 86.4820 112.6568 35.2877 33.3260 58.9080
5 43.7880 50.1226 - - 50.1226 33.0170 58.5130
6 49.3274 49.7866 133.8460 200.2717 49.7866 45.0310 84.7370

In Figure 5.9 to Figure 5.15 it is plotted the vibration mode shapes of the box cross section beam which
natural frequencies were presented in Table 5.4.

(a) – Frequency Number 1 (b) – Frequency Number 2


1 1

0.5 0.5
𝑧-displacement

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1 -1
𝑥 (m) 𝑥 (m)
∎ Numerical Model (including rotary inertia) ∎ Numerical Model (neglecting rotary inertia)
∎ Analytical Model ∎ Shell FEM Model
Figure 5.9 – vertical bending vibration mode shapes of the box cross section girder (including warping) (Part 1/2).

65
(c) – Frequency Number 3 (d) – Frequency Number 4
1 1

0.5 0.5
𝑧-displacement

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1 -1
𝑥 (m) 𝑥 (m)
∎ Numerical Model (including rotary inertia) ∎ Numerical Model (neglecting rotary inertia)
∎ Analytical Model ∎ Shell FEM Model
Figure 5.10 – vertical bending vibration mode shapes of the box cross section girder (including warping) (Part 2/2).

(a) – Frequency Number 1 (b) – Frequency Number 2


1 1

0.5 0.5
𝑧-displacement

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1 -1
(c) – Frequency Number 3 (d) – Frequency Number 4
1 1

0.5 0.5
𝑧-displacement

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1 -1
𝑥 (m) 𝑥 (m)
∎ Numerical Model (including rotary inertia) ∎ Numerical Model (neglecting rotary inertia)
∎ Shell FEM Model
Figure 5.11 – vertical bending vibration mode shapes of the box cross section girder (neglecting warping).

66
(a) – Frequency Number 1
1 1

0.5 0.5
𝑦-displacement

𝜙-rotation
0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1 -1
(b) – Frequency Number 2
1 1

0.5 0.5
𝑦-displacement

𝜙-rotation
0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1 -1
(c) – Frequency Number 3
1 1

0.5 0.5
𝑦-displacement

𝜙-rotation

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1 -1
(d) – Frequency Number 4
1 1

0.5 0.5
𝑦-displacement

𝜙-rotation

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1 -1
(e) – Frequency Number 5
1 1

0.5 0.5
𝑦-displacement

𝜙-rotation

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1 -1
𝑥 (m) 𝑥 (m)
∎ Numerical Model (including rotary inertia) ∎ Numerical Model (neglecting rotary inertia)
∎ Analytical Model ∎ Shell FEM Model
Figure 5.12 – lateral bending-torsional vibration mode shapes of the box
cross section girder (including warping) (Part 1/2).

67
(f) – Frequency Number 6
1 1

0.5 0.5
𝑦-displacement

𝜙-rotation
0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1 -1
𝑥 (m) 𝑥 (m)
∎ Numerical Model (including rotary inertia) ∎ Numerical Model (neglecting rotary inertia)
∎ Analytical Model ∎ Shell FEM Model
Figure 5.13 – lateral bending-torsional vibration mode shapes of the box
cross section girder (including warping) (Part 2/2).

(a) – Frequency Number 1


1 1

0.5 0.5
𝑦-displacement

𝜙-rotation

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1 -1
(b) – Frequency Number 2
1.0 1.0

0.5 0.5
𝑦-displacement

𝜙-rotation

0.0 0.0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1.0 -1.0
(c) – Frequency Number 3
1 1

0.5 0.5
𝑦-displacement

𝜙-rotation

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1 -1
𝑥 (m) 𝑥 (m)
∎ Numerical Model (including rotary inertia) ∎ Numerical Model (neglecting rotary inertia)
∎ Shell FEM Model
Figure 5.14 – lateral bending-torsional vibration mode shapes of the box
cross section girder (neglecting warping) (Part 1/2).

68
(d) – Frequency Number 4
1 1

0.5 0.5
𝑦-displacement

𝜙-rotation
0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1 -1
(e) – Frequency Number 5
1.0 1.0

0.5 0.5
𝑦-displacement

𝜙-rotation
0.0 0.0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1.0 -1.0
(f) – Frequency Number 6
1 1

0.5 0.5
𝑦-displacement

𝜙-rotation

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
-0.5 -0.5

-1 -1
𝑥 (m) 𝑥 (m)
∎ Numerical Model (including rotary inertia) ∎ Numerical Model (neglecting rotary inertia)
∎ Shell FEM Model
Figure 5.15 – lateral bending-torsional vibration mode shapes of the box
cross section girder (neglecting warping) (Part 2/2).

Once again one should have noticed that two vibration modes were not obtained by the beam models in
the case of neglected warping, see Figure 5.14(b) and Figure 5.15(e). In these modes the torsional
component is dominant, and shear deformations enable the shell model to obtain these vibration modes.
In this regard the Benscoter beam theory could have been used in the formulation of the beam models in
order to take into account shear distortions of beam’s walls.

69
5.1.2 UNDAMPED FORCED VIBRATION ANALYSIS

As a first approach, the undamped motion of the simply supported beam is analysed. This enables to
compare the dynamic response with the exact results proposed by [Fryba, 1972] for thin walled beams
subjected to a single moving load.

Considering the same governing expressions as stated in (2.50), and applying them to symmetric cross
sections about the vertical axis 𝑧 , Fryba deduces the following analytical solution for the beam’s
displacement field:

𝑗𝜋𝑥
∞ 2𝑃𝑍 sin (
) 𝑝
𝜉𝑧 (𝑥, 𝑡) = ∑ 𝐿 [sin(𝑝𝑡) − sin(𝑝𝑧𝑗 𝑡)] ,
2 2 𝑝
(5.2)
𝜇
𝑗=1 𝑧 𝐿(𝑝𝑧𝑗 − 𝑝 ) 𝑧𝑗

∞ 𝑃𝑍 𝑒𝑐𝑐 sin(𝑘𝜋𝑥/𝐿) 𝑏𝑦 1 ′ 2 ′ 2
𝜉𝑦 (𝑥, 𝑡) = ∑ ′ 2 ′ 2 𝜇
{ ′ 2 ′ 2 [−𝑟2 2 (𝑝𝑦𝑘 − 𝑝𝜙𝑘 ) cos(𝑟2 𝑡)
𝑘=1 𝐿(1 − 𝑏𝑦 𝑏𝜙 )(𝑝𝑦𝑘 − 𝑝𝜙𝑘 ) 𝜙 (𝑝𝜙𝑘 − 𝑟2 2 )(𝑝𝑦𝑘 − 𝑟2 2 )
′ 2
′ 2 ′ ′ ′ ′ 2 2
+ 𝑝𝜙𝑘 (𝑝𝑦𝑘 − 𝑟2 2 ) cos(𝑝𝜙𝑘 𝑡) − 𝑝𝑦𝑘 (𝑝𝜙𝑘 − 𝑟2 2 ) cos(𝑝𝑦𝑘 𝑡)]
1 ′ ′ 2 ′ 2 ′ ′ 2 2
− ′ 2 ′ 2
[−𝑟1 2 (𝑝𝑦𝑘 − 𝑝𝜙𝑘 ) cos(𝑟1 𝑡) + 𝑝𝜙𝑘 (𝑝𝑦𝑘 − 𝑟1 2 ) cos(𝑝𝜙𝑘 𝑡)
(𝑝𝜙𝑘 − 𝑟1 2 )(𝑝𝑦𝑘 − 𝑟1 2 )

′ 2′ 2 ′
− 𝑝𝑦𝑘 (𝑝𝜙𝑘 − 𝑟1 2 ) cos(𝑝𝑦𝑘 𝑡)]} , (5.3)

∞ 𝑃𝑍 𝑒𝑐𝑐 sin(𝑘𝜋𝑥/𝐿) 1 1 ′ 2 ′ 2
𝜙(𝑥, 𝑡) = ∑ ′ 2 ′ 2 𝜇
{ ′ 2 ′ 2
[(𝑝𝑦𝑘 2 − 𝑟2 2 )(𝑝𝑦𝑘 − 𝑝𝜙𝑘 ) cos(𝑟2 𝑡)
𝑘=1 𝐿(1 − 𝑏𝑦 𝑏𝜙 )(𝑝𝑦𝑘 − 𝑝𝜙𝑘 ) 𝜙 (𝑝𝜙𝑘 − 𝑟2 2 )(𝑝𝑦𝑘 − 𝑟2 2 )
′ ′ 2 2 ′ ′ ′ ′ 2 2
− (𝑝𝑦𝑘 2 − 𝑝𝜙𝑘 )(𝑝𝑦𝑘 − 𝑟2 2 ) cos(𝑝𝜙𝑘 𝑡) + (𝑝𝑦𝑘 2 − 𝑝𝑦𝑘 )(𝑝𝜙𝑘 − 𝑟2 2 ) cos(𝑝𝑦𝑘 𝑡)]
1 ′ ′ 2 2
− ′ 2 ′ 2
[(𝑝𝑦𝑘 2 − 𝑟1 2 )(𝑝𝑦𝑘 − 𝑝𝜙𝑘 ) cos(𝑟1 𝑡)
(𝑝𝜙𝑘 − 𝑟1 2 )(𝑝𝑦𝑘 − 𝑟1 2 )

2 2 2 2

− (𝑝𝑦𝑘 2 − 𝑝𝜙𝑘 ′
)(𝑝𝑦𝑘 ′
− 𝑟1 2 ) cos(𝑝𝜙𝑘 ′
𝑡) + (𝑝𝑦𝑘 2 − 𝑝𝑦𝑘 ′
)(𝑝𝜙𝑘 ′
− 𝑟1 2 ) cos(𝑝𝑦𝑘 𝑡)]} , (5.4)

where

𝑗𝜋𝑐 𝑘𝜋𝑐
𝑝= or 𝑝 = ,
𝐿 𝐿

𝑗 2 𝜋 2 𝐼𝑦𝑦 𝑘 2 𝜋 2 𝐼𝑧𝑧 𝑘 2 𝜋 2 𝐼𝜔𝜔


𝜇𝑧 = 𝜌 (𝐴 + ) , 𝜇𝑦 = 𝜌 (𝐴 + ) , 𝜇𝜙 = 𝜌 (𝐴𝑎𝐶𝐶 2 + 𝐼𝑦𝑦 + 𝐼𝑧𝑧 + ) ,
𝐿2 𝐿2 𝐿2

𝜌𝐴 𝜌𝐴
𝑏𝑦 = 𝑎𝐶𝐶 , 𝑏𝜙 = 𝑎𝐶𝐶 ,
𝜇𝑦 𝜇𝜙

𝑗 4 𝜋 4 𝐸𝐼𝑦𝑦 𝑘 4 𝜋 4 𝐸𝐼𝑧𝑧 𝑘 4 𝜋 4 𝐸𝐼𝜔𝜔 𝑘 2 𝜋 2 𝐺𝐽


𝑝𝑧𝑗 2 = , 𝑝𝑦𝑘 2 = , 𝑝𝜙𝑘 2 = + 2 ,
𝐿4 𝜇𝑧 𝐿4 𝜇𝑦 𝐿4 𝜇𝜙 𝐿 𝜇𝜙

2 2
2
𝑝𝑦𝑘 2 + 𝑝𝜙𝑘 2 2
1 𝑝𝑦𝑘 + 𝑝𝜙𝑘
2
𝑝𝑦𝑘 2 𝑝𝜙𝑘 2 2
𝑝𝑦𝑘 2 + 𝑝𝜙𝑘 2 2
1 𝑝𝑦𝑘 + 𝑝𝜙𝑘
2
𝑝𝑦𝑘 2 𝑝𝜙𝑘 2
𝑝′𝑦𝑘 = +√ ( ) − , 𝑝′𝜙𝑘 = −√ ( ) − ,
2(1 − 𝑏𝑦 𝑏𝜙 ) 4 1 − 𝑏𝑦 𝑏𝜙 1 − 𝑏𝑦 𝑏𝜙 2(1 − 𝑏𝑦 𝑏𝜙 ) 4 1 − 𝑏𝑦 𝑏𝜙 1 − 𝑏𝑦 𝑏𝜙

𝜋 𝜋
𝑟1 = +𝑝 and 𝑟2 = −𝑝 ,
2𝑡 2𝑡

70
being 𝑐 is the load velocity and 𝑎𝐶𝐶 the coordinate of the shear centre 𝐶𝐶 along the principal 𝑧 axis.

The undamped structural dynamic response is illustrated in Figure 5.16 by the vertical displacement
influence lines at the midpoint of the span’s lane, for both the double-T and the box cross section girder.

The results obtained from the beam models are successfully compared with the results given by the shell
model implemented in SAP2000. The results given by the Fyba’s analytical formulation are also in good
agreement with the remaining ones.

From Figure 5.16 it is clear that the torsional response is highly influenced when the cross sectional
warping is neglected. This remark was already made concerning the natural frequencies of vibration. This
fact particularly affects open cross sections, since warping stiffness represents the primary source of
torsion resistance and therefore the dynamic displacements are greatly influenced by its consideration.

(a) – Double-T Cross Section


9.0E-03

6.0E-03
(m)
𝑧-displacement

3.0E-03

0.0E+00

-3.0E-03
0 5 10 15 20 25 30 35 40

(b) – Box Cross Section


1.2E-02

8.0E-03
(m)
𝑧-displacement

4.0E-03

0.0E+00

-4.0E-03
0 5 10 15 20 25 30 35 40
𝑥 (m)
Including Warping & Including Rotary Inertia: ∎ Numerical Model ∎ Shell FEM Model ∎ Fryba’s Formulation
Including Warping & Neglecting Rotary Inertia: ∎ Analytical Model
Neglecting Warping & Including Rotary Inertia: ∎ Numerical Model ∎ Shell FEM Model
Figure 5.16 – dynamic influence lines of the 𝑧-acceleration at the track’s midpoint.

71
5.2 CONTINUOUS BEAM

In this section, the damped dynamic response of a continuous beam is investigated through the developed
beam models. The longitudinal model is shown in Figure 5.17 and the prescribed boundary conditions at
the supports are listed in Table 5.5.

𝑆1 𝑆2 𝑆3 𝑆4

30 40 30
𝑥

Figure 5.17 – longitudinal model of the three-span bridge considered in the practical examples (m).

Table 5.5 – prescribed boundary conditions at the supports.


Support Nodal Coordinate Kinematic Boundary Conditions Static Boundary Conditions
𝑆1 𝑥 =0m 𝜉𝑧 (𝑥) = 𝜉𝑦 (𝑥) = 𝜙(𝑥) = 0 𝜉𝑧′′ (𝑥) = 𝜉𝑦′′ (𝑥) = 𝜙 ′′ (𝑥) = 0
𝜉𝑧 (𝑥) = 𝜉𝑦 (𝑥) = 𝜙(𝑥) = 0
𝑆2 𝑥 = 30 m 𝜉𝑧′′ (𝑥 − ) = 𝜉𝑧′′ (𝑥 + )
𝜉𝑧′ (𝑥 − ) = 𝜉𝑧′ (𝑥 + )
𝜉𝑦′′ (𝑥 − ) = 𝜉𝑦′′ (𝑥 + )
𝜉𝑦′ (𝑥 − ) = 𝜉𝑦′ (𝑥 + )
𝑆3 𝑥 = 70 m 𝜙 ′′ (𝑥 − ) = 𝜙 ′′ (𝑥 + )
𝜙 ′ (𝑥 − ) = 𝜙 ′ (𝑥 + )
𝑆4 𝑥 = 100 m 𝜉𝑧 (𝑥) = 𝜉𝑦 (𝑥) = 𝜙(𝑥) = 0 𝜉𝑧′′ (𝑥) = 𝜉𝑦′′ (𝑥) = 𝜙 ′′ (𝑥) = 0

The dynamic responses are obtained by taking into account the high-speed universal trains HSLM-A1 and
HSLM-A10 proposed by the [EN 1991-2]. The general dimensions and load magnitudes are illustrated in
Figure 5.18.

HSLM

D N×D D

4×P 3×P 2×P


d d
3 11 3 3.525 d d d d 3.525 3 11 3

N D (m) d (m) P (kN)


HSLM-A1 18 18 2 170
HSLM-A10 11 27 2 210
Figure 5.18 – general dimensions and load magnitudes of the high-speed universal trains proposed by [EN 1991-2].

72
5.2.1 UNDAMPED FREE VIBRATION ANALYSIS

The undamped free vibration analysis of the two considered structures is performed by using the
numerical and the analytical models developed in Chapter 3 and Chapter 4, respectively. The natural
frequencies calculated by the two beam models are compared in Table 5.6 and Table 5.7 with the values
given by the shell finite element model implemented in SAP2000. From the analysis of the simply
supported girder it is clear the influence of warping stiffness on the dynamic response of open cross
section beam elements. Therefore, the warping effect is taken into account in the sequel.

Table 5.6 – natural frequencies of vibration of the double-T cross section girder.
Natural Frequencies of Vibration in the plane 𝑥𝑧 (Hz)
Model of Analysis
Frequency
Numerical Analytical Shell FEM
Number
Inc. Rotary Inertia Negl. Rotary Inertia Inc. Rotary Inertia
1 4.5777 4.5942 4.4265
2 7.4770 7.5202 7.0536
3 8.9727 9.0216 8.1927
4 17.0665 17.2890 15.0470
5 25.8383 26.4031 22.1980
6 28.1767 28.7827 23.3090

Natural Frequencies of Vibration in the plane 𝑥𝑦 (Hz)


Model of Analysis
Frequency
Numerical Analytical Shell FEM
Number
Inc. Rotary Inertia Negl. Rotary Inertia Inc. Rotary Inertia
1 4.8724 4.8853 4.7617
2 7.4149 7.4469 7.0966
3 8.5349 8.5700 7.9137
4 15.7794 15.9336 14.1110
5 17.1743 18.1063 12.7420
6 23.5509 23.9380 20.4600

The mode shapes of vibration corresponding to the natural frequencies listed in Table 5.6 are plotted
from Figure 5.19 to Figure 5.21. Hereafter allowance is made for warping displacements and the
numerical model is evaluated considering the cross-sectional rotary inertia, whereas the analytical model
does not consider this property.

73
(a) – Frequency Number 1 (b) – Frequency Number 2
1 1

0.5 0.5
𝑧-displacement

0 0
0 20 40 60 80 100 0 20 40 60 80 100
-0.5 -0.5

-1 -1
(c) – Frequency Number 3 (d) – Frequency Number 4
1 1

0.5 0.5
𝑧-displacement

0 0
0 20 40 60 80 100 0 20 40 60 80 100
-0.5 -0.5

-1 -1
(e) – Frequency Number 5 (f) – Frequency Number 6
1 1

0.5 0.5
𝑧-displacement

0 0
0 20 40 60 80 100 0 20 40 60 80 100
-0.5 -0.5

-1 -1
𝑥 (m) 𝑥 (m)
∎ Numerical Model ∎ Analytical Model ∎ Shell FEM Model
Figure 5.19 – vertical bending vibration mode shapes of the double-T cross section girder.

(a) – Frequency Number 1


1 1

0.5 0.5
𝑦-displacement

𝜙-rotation

0 0
0 20 40 60 80 100 0 20 40 60 80 100
-0.5 -0.5

-1 -1
𝑥 (m) 𝑥 (m)
∎ Numerical Model ∎ Analytical Model ∎ Shell FEM Model
Figure 5.20 – lateral bending-torsional vibration mode shapes of the double-T cross section girder (Part 1/2).

74
(b) – Frequency Number 2
1 1

0.5 0.5
𝑦-displacement

𝜙-rotation
0 0
0 20 40 60 80 100 0 20 40 60 80 100
-0.5 -0.5

-1 -1
(c) – Frequency Number 3
1 1

0.5 0.5
𝑦-displacement

𝜙-rotation
0 0
0 20 40 60 80 100 0 20 40 60 80 100
-0.5 -0.5

-1 -1
(d) – Frequency Number 4
1 1

0.5 0.5
𝑦-displacement

𝜙-rotation

0 0
0 20 40 60 80 100 0 20 40 60 80 100
-0.5 -0.5

-1 -1
(e) – Frequency Number 5
1 1

0.5 0.5
𝑦-displacement

𝜙-rotation

0 0
0 20 40 60 80 100 0 20 40 60 80 100
-0.5 -0.5

-1 -1
(f) – Frequency Number 6
1 1

0.5 0.5
𝑦-displacement

𝜙-rotation

0 0
0 20 40 60 80 100 0 20 40 60 80 100
-0.5 -0.5

-1 -1
𝑥 (m) 𝑥 (m)
∎ Numerical Model ∎ Analytical Model ∎ Shell FEM Model
Figure 5.21 – lateral bending-torsional vibration mode shapes of the double-T cross section girder (Part 2/2).

75
Table 5.7 – natural frequencies of vibration of the box cross section girder.
Natural Frequencies of Vibration in the plane 𝑥𝑧 (Hz)
Model of Analysis
Frequency
Numerical Analytical Shell FEM
Number
Inc. Rotary Inertia Negl. Rotary Inertia Inc. Rotary Inertia
1 4.1828 4.1953 3.9576
2 6.8344 6.8673 6.1834
3 8.2011 8.2384 7.0112
4 15.6181 15.7881 12.5610
5 23.6786 24.1110 18.1190
6 25.8200 27.1762 18.7120

Natural Frequencies of Vibration in the plane 𝑥𝑦 (Hz)


Model of Analysis
Frequency
Numerical Analytical Shell FEM
Number
Inc. Rotary Inertia Negl. Rotary Inertia Inc. Rotary Inertia
1 11.3923 11.3998 11.1040
2 15.0977 15.1142 14.6010
3 15.1884 15.2136 14.6620
4 15.8960 16.6161 13.3490
5 23.4978 23.5562 21.9030
6 25.3300 27.1762 19.6110

The mode shapes of vibration that correspond to the natural frequencies listed in Table 5.7 are depicted in
Figure 5.22 to Figure 5.25.

From Figure 5.25(c) and Figure 5.25(e) it is seen that, although their similarity, the mode shapes obtained
by the beam models are not the same as the ones given by the shell model. In Table 5.7 the frequencies
obtained by the shell model were sorted according to the shape of the corresponding modes. Once again,
shear deformations of the shell elements may cause these differences.

(a) – Frequency Number 1 (b) – Frequency Number 2


1 1

0.5 0.5
𝑧-displacement

0 0
0 20 40 60 80 100 0 20 40 60 80 100
-0.5 -0.5

-1 -1
𝑥 (m) 𝑥 (m)
∎ Numerical Model ∎ Analytical Model ∎ Shell FEM Model
Figure 5.22 – vertical bending vibration mode shapes of the box cross section girder (Part 1/2).

76
(c) – Frequency Number 3 (d) – Frequency Number 4
1 1

0.5 0.5
𝑧-displacement

0 0
0 20 40 60 80 100 0 20 40 60 80 100
-0.5 -0.5

-1 -1
(e) – Frequency Number 5 (f) – Frequency Number 6
1 1

0.5 0.5
𝑧-displacement

0 0
0 20 40 60 80 100 0 20 40 60 80 100
-0.5 -0.5

-1 -1
𝑥 (m) 𝑥 (m)
∎ Numerical Model ∎ Analytical Model ∎ Shell FEM Model
Figure 5.23 – vertical bending vibration mode shapes of the box cross section girder (Part 2/2).

(a) – Frequency Number 1


1 1

0.5 0.5
𝑦-displacement

𝜙-rotation

0 0
0 20 40 60 80 100 0 20 40 60 80 100
-0.5 -0.5

-1 -1
(b) – Frequency Number 2
1 1

0.5 0.5
𝑦-displacement

𝜙-rotation

0 0
0 20 40 60 80 100 0 20 40 60 80 100
-0.5 -0.5

-1 -1
𝑥 (m) 𝑥 (m)
∎ Numerical Model ∎ Analytical Model ∎ Shell FEM Model
Figure 5.24 – lateral bending-torsional vibration mode shapes of the box cross section girder (Part 1/2).

77
(c) – Frequency Number 3
1 1

0.5 0.5
𝑦-displacement

𝜙-rotation
0 0
0 20 40 60 80 100 0 20 40 60 80 100
-0.5 -0.5

-1 -1
(d) – Frequency Number 4
1 1

0.5 0.5
𝑦-displacement

𝜙-rotation
0 0
0 20 40 60 80 100 0 20 40 60 80 100
-0.5 -0.5

-1 -1
(e) – Frequency Number 5
1 1

0.5 0.5
𝑦-displacement

𝜙-rotation

0 0
0 20 40 60 80 100 0 20 40 60 80 100
-0.5 -0.5

-1 -1
(f) – Frequency Number 6
1 1

0.5 0.5
𝑦-displacement

𝜙-rotation

0 0
0 20 40 60 80 100 0 20 40 60 80 100
-0.5 -0.5

-1 -1
𝑥 (m) 𝑥 (m)
∎ Numerical Model ∎ Analytical Model ∎ Shell FEM Model
Figure 5.25 – lateral bending-torsional vibration mode shapes of the box cross section girder (Part 2/2).

The developed numerical and analytical models are in excellent agreement. The results given by both
models concerning the natural frequencies are very similar and the corresponding modes shapes are
almost coincident.

78
5.2.2 DAMPED FORCED VIBRATION ANALYSIS

A three-span bridge girder subjected to the high-speed trains represented in Figure 5.18 is analysed
through the developed beam models in Appendix II. The two cross section geometries plotted in Figure
5.1 are considered in order to evaluate the structural dynamic response. The bridge girder is modelled as a
continuous thin walled beam, which dynamic behaviour is investigated for a set of four different velocity
values of the moving loads, namely 60, 80, 100 and 120 ms-1. The load magnitude is also considered to be
constant in time and the modal damping is assumed to be of 0.01.

In the sequel, only the double-T cross section girder traversed by the high-speed train HSLM-A10 is
considered. The dynamic influence lines of the generalized displacements and accelerations at the
midpoint of the central span are presented from Figure 5.26 to Figure 5.31 for two different velocities.
For the sake of completeness, the influence lines of the vertical displacement and vertical acceleration at
the track’s position are also presented.

(a) – Velocity: 80 ms-1


2.4E-03

1.6E-03
(m)
𝑧-displacement

8.0E-04

0.0E+00

-8.0E-04
0 50 100 150 200 250 300 350 400 450 500

(b) – Velocity: 120 ms-1


4.8E-03

3.2E-03
(m)

1.6E-03
𝑧-displacement

0.0E+00

-1.6E-03

-3.2E-03
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ Numerical Model ∎ Analytical Model
Figure 5.26 – dynamic influence lines of the 𝑧-displacement at the midpoint of the central span.

79
(a) – Velocity: 80 ms-1
2.0E-04

0.0E+00
(rad)
𝜙-rotation

-2.0E-04

-4.0E-04
0 50 100 150 200 250 300 350 400 450 500

(b) – Velocity: 120 ms-1


2.0E-04
(rad)

0.0E+00
𝜙-rotation

-2.0E-04

-4.0E-04
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ Numerical Model ∎ Analytical Model
Figure 5.27 – dynamic influence lines of the 𝜙-rotation at the midpoint of the central span.

(a) – Velocity: 80 ms-1


8.0E-01
(ms-2)

4.0E-01
𝑧-acceleration

0.0E+00

-4.0E-01

-8.0E-01
0 50 100 150 200 250 300 350 400 450 500

(b) – Velocity: 120 ms-1


2.8E+00
(ms-2)

1.4E+00
𝑧-acceleration

0.0E+00

-1.4E+00

-2.8E+00
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ Numerical Model ∎ Analytical Model
Figure 5.28 – dynamic influence lines of the 𝑧-acceleration at the midpoint of the central span.

80
(a) – Velocity: 80 ms-1
5.0E-02
(rads-2)

2.5E-02

0.0E+00
𝜙-acceleration

-2.5E-02

-5.0E-02
0 50 100 150 200 250 300 350 400 450 500

(b) – Velocity: 120 ms-1


2.0E-01
(rads-2)

1.0E-01

0.0E+00
𝜙-acceleration

-1.0E-01

-2.0E-01
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ Numerical Model ∎ Analytical Model
Figure 5.29 – dynamic influence lines of the 𝜙-acceleration at the midpoint of the central span.

(a) – Velocity: 80 ms-1


3.0E-03

2.0E-03
(m)
𝑧-displacement

1.0E-03

0.0E+00

-1.0E-03
0 50 100 150 200 250 300 350 400 450 500

(b) – Velocity: 120 ms-1


6.0E-03
(m)

3.0E-03
𝑧-displacement

0.0E+00

-3.0E-03
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ Numerical Model ∎ Analytical Model ∎ Shell FEM Model
Figure 5.30 – dynamic influence lines of the 𝑧-displacement at the track’s midpoint of the central span.

81
(a) – Velocity: 80 ms-1
8.0E-01

4.0E-01
(ms-2)
𝑧-acceleration

0.0E+00

-4.0E-01

-8.0E-01
0 50 100 150 200 250 300 350 400 450 500
(b) – Velocity: 120 ms-1
3.0E+00

1.5E+00
(ms-2)
𝑧-acceleration

0.0E+00

-1.5E+00

-3.0E+00
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ Numerical Model ∎ Analytical Model ∎ Shell FEM Model
Figure 5.31 – dynamic influence lines of the 𝑧-acceleration at the track’s midpoint of the central span.

In the next the mass effect of the moving vehicles is taken into account in the dynamic response of the
three-span bridge. Therefore, the considered loading is written as follows:
𝑛𝑙𝑜𝑎𝑑 𝑑2 𝜈𝑘
𝑓𝑧 (𝑥, 𝑡) = ∑ δ(𝑥 − 𝑥𝑘 (𝑡)) [𝑉𝑧𝑘 (𝑡) − 𝓂𝑘 ] , (5.5a)
𝑘 𝑑𝑡 2

𝑚𝑥 (𝑥, 𝑡) = 𝑓𝑧 (𝑥, 𝑡) 𝑒𝑐𝑐 , (5.5b)

being 𝑥𝑘 (𝑡) the position and 𝓂𝑘 the mass of the 𝑘-th load, 𝜈𝑘 is the vertical displacement of the moving
mass 𝓂𝑘 , δ(𝑡) the Dirac delta function, 𝑛𝑙𝑜𝑎𝑑 the number of loads and 𝑒𝑐𝑐 is the eccentricity of the track’s
position. Since the load is acting on a moving coordinate and therefore, [Ouyang, 2011]:
2 2 2
𝜈𝑘 (𝑡) = 𝜉𝑧 (𝑐𝑡, 𝑡) , 𝑑𝜈𝑘 = 𝜕𝜉𝑧 + 𝑐 𝜕𝜉𝑧 and 𝑑 𝜈2𝑘 = 𝜕 𝜉2𝑧 + 2𝑐 𝜕𝜉𝑧 𝜕𝜉𝑧 + 𝑐 2 𝜕 𝜉2𝑧 (5.6)
𝑑𝑡 𝜕𝑡 𝜕𝑥 𝑑𝑡 𝜕𝑡 𝜕𝑡 𝜕𝑥 𝜕𝑥

where 𝑐 represents the velocity of the moving masses. Notice that it is assumed that the masses do not
separate from the beam during their travel and vertical vibration.

The influence of the inertia of the moving masses on the continuous beam dynamic response is small in
the analysed case, see Figure 5.32 and Figure 5.33. This does not exclude that in other cases the mass
effect does not have greater importance. This behaviour is studied in more detail by [Ichikawa et al., 2000]
for Euler-Bernoulli continuous beams under single moving loads.

82
In Figure 5.32 and Figure 5.33 the influence of constrained warping in the structural dynamic response is
also evaluated by using the numerical model. It is observed that both the displacements and accelerations
in the 𝑧 direction are lower than when the warping effect is considered.

Velocity: 120 ms-1


5.7E-03

3.8E-03
(m)

1.9E-03
𝑧-displacement

0.0E+00

-1.9E-03

-3.8E-03
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
Including the Warping Effect & Neglecting the Mass Effect: ∎ Numerical Model ∎ Analytical Model
Including the Warping Effect & Including the Mass Effect: ∎ Numerical Model ∎ Analytical Model
Neglecting the Warping Effect & Including the Mass Effect: ∎ Numerical Model
Figure 5.32 – influence of the mass effect on the dynamic influence lines of the 𝑧-displacement
at the track’s midpoint of the central span.

Velocity: 120 ms-1


3.0E+00

1.5E+00
(ms-2)
𝑧-acceleration

0.0E+00

-1.5E+00

-3.0E+00
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
Including the Warping Effect & Neglecting the Mass Effect: ∎ Numerical Model ∎ Analytical Model
Including the Warping Effect & Including the Mass Effect: ∎ Numerical Model ∎ Analytical Model
Neglecting the Warping Effect & Including the Mass Effect: ∎ Numerical Model
Figure 5.33 – influence of the mass effect on the dynamic influence lines of the 𝑧-acceleration
at the track’s midpoint of the central span.

83
The results obtained by the numerical and the analytical models are in good agreement. Both deflections
and accelerations given by the analytical model are larger, which is related with the fact of having been
neglected the cross-sectional rotary inertia.

The selection of the time step Δ𝑡 should not be related with the small differences obtained. In general,
unsuitable time steps lead to losses of accuracy in the solutions. The errors involved in the direct
integration may cause a shift in the period and amplitude of responses, which may be increased
throughout time steps, as shown in Figure 5.34. Although differences in amplitudes were obtained, the
periods are coherent.

𝑢(𝑡)
period elongation
exact solution
amplitude amplitude decay

period 𝑡
numerical solution

Figure 5.34 – period elongation and amplitude decay errors.

84
6
Conclusions and
Future Developments

A s it was stated in the introduction to this dissertation, the main objective of the present work was to
provide two beams models capable of performing dynamic analyses of railway bridges, being
considered the warping displacements.

In the previous chapters, a numerical and an analytical beam models were developed and applied to practical
cases. The performed analyses focused on the consideration of the dynamic response of both open and
closed cross section girders. The torsional response was shown to be significantly influenced by the
warping deformations.

The developed beam models gave good results, being in good agreement with the shell finite element
models implemented in commercial software. Thus, the developed models perform efficient dynamic
analyses that consider the warping effect, which may be very useful at early design stages.

6.1 CONCLUSIONS

After having derived the coupled governing differential equations of thin walled beams, a numerical and
an analytical models have been developed in order to solve those equations. In this regard, a modal
analysis along with a direct integration scheme was performed.

The numerical model consists on a beam finite element formulation, where an additional degree of
freedom was introduced to consider the warping displacements. The displacement fields were referred to
the centre of gravity, motivating the presence of coupling terms in the finite element matrices. The

85
Hermite’s polynomials were used as approximation functions and the finite element equations were
deduced without neglecting terms. The natural frequencies and the corresponding mode shapes were
determined by solving the associated eigenvalue problem.

In what concerns the analytical model, the solution of the coupled governing equations is achieved in an
exact sense. By neglecting the cross-sectional rotary inertia terms, the homogeneous form of the
governing differential equations are solved and the natural frequencies met exactly. The vibration mode
shapes are obtained by explicit expressions.

By applying the Newmark’s 𝛽 method, the forced vibrations due to moving vehicles are calculated. A
simply supported girder traversed by a moving vertical force at constant velocity with eccentricity is
analysed and the results compared with the exact solution of [Fryba, 1972]. Afterwards, a three-span
continuous girder acted by arrays of moving loads is investigated. The mass effect of the moving vehicles
is taken into account and several velocities considered.

The effectiveness of the beam models is verified by comparison with a shell finite element model and the
warping effect was proved to be a fundamental aspect to be considered in the structural analysis of open
cross section girders.

6.2 FUTURE DEVELOPMENTS

The presented beam models do not take into account shear distortions. The comparisons made with the
shell finite element model have shown that shear deformations justify some differences between the
determined mode shapes of vibrations. In order to take into account the in-plane shear deformability the
Benscoter’s beam theory may be used.

Further verifications of the developed models may be carried out by using commercial beam finite
elements with the warping degree of freedom. Software such as ANSYS or ABAQUS provide these
elements, but the understanding of their correct use is not so simple at a first approach.

The presented models could serve as a starting point for further in-depth study of the subject and may be
considered as a first step tending to more complete formulations accounting for other aspect that have
not been considered, such as:

consideration of the Benscoter’s beam theory that considers shear distortions, which is an
important aspect in the analysis of closed cross section girders;

86
the variation of the cross section height along the beam axis, specially important for the analysis
of box girders with variable cross section;
the consideration of curved girders in plan, regardless of the cross section type;
the consideration of the track-structure interaction that must be taken into account according to
the [EN 1991-2];
the consideration of the real support conditions, i.e., the eccentricity of the bearings along the
vertical direction, which motivates coupling between the vertical and the tensional motions;
the adoption of more sophisticated vehicle models;
the considerations of energy dissipation devices;
the introduction of the ballast, which increases the complexity of the whole structural model.

87
References

1. Ahmad, M.F., Complex Frequency Analysis of Damped Thin-Walled Beams with Open Cross Sections, Ph.D. Dissertation,
University of New Hampshire, 1984.
2. Banerjee, J.R., Coupled bending-torsional dynamic stiffness matrix for beam elements, International Journal for
Numerical Methods in Engineering, 28, 1283-1298, 1989.
3. Banerjee, J.R., Guo, S. and Howson, W.P., Exact dynamic stiffness matrix of a bending-torsion coupled beam
including warping, Computers & Structures, 59(4), 613-621, 1996.
4. Benscoter, S.U., A theory of torsion bending for multicell beams, Journal of Applied Mechanics, 25-34, 1954.
5. Biggs, J.M., Suer, H.S. and Louw, J.M., Vibration on simple-span highway bridges, Transactions of the American
Society of Civil Engineers, 124, 291-318, 1959.
6. Bishop, R.E.D. and Johnson, D.C., The Mechanics of Vibration, Cambridge University Press, Cambridge, UK,
1979.
7. Bishop, R.E.D., Cannon, S.M. and Miao, S., On coupled bending and torsional vibration of uniform beams,
Journal of Sound and Vibration, 131(3), 457-464, 1989.
8. Cai, Y., Chen, S.S., Rote, D.M. and Coffey, H.T., Vehicle/guideway interaction for high-speed vehicles on a
flexible guideway, Journal of Sound and Vibration, 175, 625-646, 1994.
9. Chan, T.H.T. and Ashebo, D.B., Theoretical study of moving force identification on continuous bridges,
Journal of Sound and Vibration, 295, 870-883, 2006.
10. Clough, R.W. and Penzien, J., Dynamics of Structures, Computers and Structures, Inc., Berkeley, USA, 1995.
11. Dokumaci, E., An exact solution for coupled bending and torsion vibrations of uniform beams having single
cross-sectional symmetry, Journal of Sound and Vibration, 119, 443-449, 1987.
12. Dugush, Y.A. and Eisenberger, M., Vibrations of non-uniform continuous beams under moving loads, Journal
of Sound and Vibration, 254, 911-926, 2002.
13. EN 1991-2: Actions on structures – Part 2: Traffic loads on bridges, CEN – European Committee for
Standardization.
14. ERRI D214 committee, Rail Bridges for Speed > 200 km/h, Final Report, Part A, Synthesis of The Result of D
214 research, Part B, Proposed UIC Leaflet, European Rail Research Institute ERRI, 1999.
15. Fish, J. and Belytschko, T., A First Course in Finite Elements, John Wiley & Sons, England, 2007.
16. Flügge, W., Stresses in Shells, Springer-Verlag, New York, 1966.
17. Foda, M.A. and Abduljabbar, Z., A dynamic Green function formulation for the response of a beam structure
to a moving mass, Journal of Sound and Vibration, 210, 295-306, 1998.
18. Friberg, P.O., Coupled vibrations of beams – an exact dynamic element stiffness matrix, International Journal for
Numerical Methods in Engineering, 19, 479-493, 1983.

89
19. Friberg, P.O., Beam element matrices derived from Vlasov’s theory of open thin-walled elastic beams,
International Journal for Numerical Methods in Engineering, 21, 1205-1228, 1985.
20. Fryba, L., Vibration of Solids and Structures under Moving Loads, Thomas Telford, 1972.
21. Fryba, L., Non-stationary response of a beam to a moving random force, Journal of Sound and Vibration, 46, 323-
338, 1976.
22. Fryba, L., Estimation of fatigue life of railway bridges under traffic loads, Journal of Sound and Vibration, 70, 527-
541, 1980.
23. Fryba, L., Nakagiri, S. and Yoshikawa, N., Stochastic finite elements for a beam on a random foundation with
uncertain damping under moving force, Journal of Sound and Vibration, 163, 31-45, 1993.
24. Fryba, L., Dynamics of Railway Bridges, Thomas Telford Services Ltd., Prague, 1996.
25. Gere, J.M., Torsional vibrations of beams of thin-walled open section, Journal of Applied Mechanics, 21, 381-387,
1954.
26. Gere, J.M. and Lin, Y.K., Coupled vibrations of thin-walled beams of open section, Journal of Applied Mechanics,
373-378, 1958.
27. Gjelsvik, A., The Theory of Thin Walled Bars, John Wiley & Sons, New York, 1981.
28. Graça, A., Modelação da Resposta em Pontes Devido à Passagem de Comboios de Alta Velocidade – formulação e
implementação computacional de um modelo para simulação do efeito da passagem de comboios de alta velocidade em pontes,
Master Dissertation, Técnico Lisboa, 2011.
29. Hallauer, W.L. and Liu, R.Y.L., Beam bending-torsion dynamic stiffness method for calculation of exact
vibration modes, Journal of Sound and Vibration, 85, 105-113, 1982.
30. Hilal, M.A. and Zibdeh, H.S., Vibration analysis of beams with general boundary conditions traversed by a
moving force, Journal of Sound and Vibration, 229, 377-388, 2000.
31. Hillerborg, A., Dynamic Influences of Smoothly Running Loads on Simply Supported Girders, Kungliga Tekniska
Högskolan (KTH), Stockholm, 1951.
32. Ichikawa, M., Matsuda, A. and Miyakawa, Y., Simple analysis of a multi-span beam under moving loads with
variable velocity, Transactions of the Japan Society for Aeronautical and Spaces Sciences, 41, 168-173, 1999.
33. Ichikawa, M., Miyakawa, Y. and Matsuda, A., Vibration analysis of the continuous beam subjected to a moving
mass, Journal of Sound and Vibration, 230(3), 493-506, 2000.
34. Inglis, C.E., A Mathematical Treatise on Vibration in Railway Bridges, The Cambridge University Press, 1934.
35. Kolousek, V., Dynamics of Civil Engineering Structures – Part I: General Problems; Part II: Continuous Beams and Frame
Systems; Part III: Selected Topics, SNTL, Prague, 1956.
36. Krylov, A.N., Über die erzwungenen Schwingungen von gleichformigen elastischen Staben, Mathematische
Annalen, 61, 211-234, Mathematical Collection of Papers of the Academy of Sciences, Peterburg, 1905.
37. Law, S.S., Chan, T.H.T. and Zeng, Q.H., Moving force identification: a time domain method, Journal of Sound
and Vibration, 201, 23-41, 1997.
38. Lee, H.P., Dynamic response of a beam with a moving mass, Journal of Sound and Vibration, 191, 289-294,
1996(1).
39. Lee, H.P., Transverse vibration of a Timoshenko beam acted upon by an accelerating mass, Applied Acoustics,
47, 319-330, 1996(2).
40. Lee, H.P., Dynamic response of a beam on multiple supports with a moving mass, Journal of Structural
Engineering and Mechanics, 4, 303-312, 1996(3).

90
41. Lee, S.Y. and Yhim, S.S., Dynamic behaviour of long-span box girder bridges subjected to moving loads:
numerical analysis and experimental verification, International Journal of Solids and Structures, 42, 5021-5035, 2005.
42. Lisi, Diego, A Beam Finite Element Model Including Warping – Application to the Dynamic and Static Analysis of Bridge
Decks, IST, Universidade Técnica de Lisboa, 2011.
43. Logan, D.L., A First Course in the Finite Element Method, Fourth Edition, Thomson, 2007.
44. Lowan, A.N., On transverse oscillations of breams under the action of moving variable loads, Philosophical
Magazine, Series 7 19 (127), 708-715, 1935.
45. Mei, C., Coupled vibrations of thin walled beams of open section using the finite element method, International
Journal of Mechanical Sciences, 12, 883-891, 1970.
46. Michaltsos, G.T., Sophianopoulos, D. and Kounadis, A.N., The effect of a moving mass and other parameters
on the dynamic response of a simply supported beam, Journal of Sound and Vibration, 191, 357-362, 1996.
47. Michaltsos, G.T., Sarantithou, E. and Sophianopoulos, D.S., Flexural-torsional vibration of simply supported
open cross-section steel beams under moving loads, Journal of Sound and Vibration, 280, 479-494, 2005.
48. Murray, N.W., Introduction to the Theory of Thin-Walled Structures, Clarendon Press, 1984.
49. Newmark, N.M., A Method of Computation for Structural Dynamics, ASCE, Vol. 85(EM3), Part 1, pp.67-94, 1959.
50. Olsson, M., On the fundamental moving mass problem, Journal of Sound and Vibration, 145, 299-307, 1991.
51. Ouyang, H., Moving-load dynamic problems: A tutorial (with a brief overview), Mechanical Systems and Signal
Processing, 25, 2039-2060, 2011.
52. Reddy, J.N., Energy Principles and Variational Methods in Applied Mechanics, 2nd Edition, John Wiley & Sons, Inc.,
August 2002.
53. Reddy, J.N., An Introduction to the Finite Element Method, 3rd Edition, McGraw-Hill, New York, 2006.
54. Sophianopoulos, D.S. and Michaltsos, G.T., Combined torsional-lateral vibrations of beams under vehicular
loading – I: formulation and solution techniques, Automatic Control & Robotics, 2, 877-886, 1999.
55. Stayridis, L.T. and Michaltsos, G.T., Eigenfrequency analysis of thin-walled girders curved in plan, Journal of
Sound and Vibration, 227, 383-396, 1999.
56. Stokes, G., Discussion of a differential equation relating to the breaking of railway bridges, Transactions of the
Cambridge Philosophical Society, 5, 707-735, 1849.
57. Tanaka, M. and Bercin, A.N., Finite element modelling of the coupled bending and torsional free vibration of
uniform beams with an arbitrary cross-section, Applied Mathematical Modelling, 21, 339-344, 1997.
58. Tanaka, M. and Bercin, A.N., Free vibration solution for uniform beams of nonsymmetrical cross section using
Mathematica, Computers and Structures, 71, 1-8, 1999.
59. Thambiratnam, D. and Zhuge, Y., Dynamic analysis of beams on an elastic foundation subjected to moving
loads, Journal of Sound and Vibration, 198, 149-169, 1996.
60. Timoshenko, S.P., History of the Strength of Materials, D. van Nostrand Company, Inc., New York, 1953.
61. Timoshenko, S.P. and Woinowsky-Krieger, S., Theory of Plates and Shells, Second Edition, McGraw-Hill, New
York, 1959.
62. Timoshenko, S.P., Forced Vibration of Prismatic Bars, Izvestiya Kievskogo Politekhnicheskogo Instituta, 1908.
63. Timoshenko, S.P., Young, D.H. and Weaver Jr.,W., Vibration Problems in Engineering, Wiley, New York, 1974.
64. Ting, E.C., Genin, J. and Ginsberg, A general algorithm for moving mass problems, Journal of Sound and
Vibration, 33, 49-58, 1974.

91
65. Tso, W.K., Coupled vibrations of thin-walled elastic bars, Journal of the Engineering Mechanics Division, ASCE,
91(EM3), 33-52, 1965.
66. Vlasov, V.Z., Thin-Walled Elastic Beams, English translation from Russian, published for the U.S. Science
Foundation and the U.S. Department of Commerce by the Israel Program for Scientific Translations,
Jerusalem, 1961.
67. Von Kármán, T. and Christensen, N.B., Method of analysis for torsion with variable twist, Journal of the
Aeronautical Sciences, 11, 110, 1944.
68. Wunderlich, W. and Pilkey, W.D., Mechanics of Structures – Variational and Computational Methods, Second Edition,
CRC Press, 2003.
69. Xia, H., De Roeck, G., Zhang and N., Maeck, J., Experimental analysis of a high-speed railway bridge under
Thalis trains, Journal of Sound and Vibration, 268, 103-113, 2003.
70. Yang, Y.B., Liao, S.S. and Lin, B.H., Impact formulas for vehicles moving over simple and continuous beams,
Journal of Structural Engineering, American Society of Civil Engineers, 121, 1644-1650, 1995.
71. Yang, Y.B., Yau, J.D. and Wu, Y.S., Vehicle-Bridge Interaction Dynamics, World Scientific Publishing Co., 2004.
72. Zimmermann, H., Die Schwingungen eines Tragers mit bewegter Lasts, Centralblatt der Bauverwaltung, 16, 249-
251, 257-260, 264-266, 288, 1896.
73. Zhu, X.Q. and Law, S.S., Precise time-step integration for the dynamic response of a continuous beam under
moving loads, Journal of Sound and Vibration, 240, 962-970, 2001.

92
Appendix I

In order to obtain the ℬ𝑗𝑛𝑠 constants of eqn (4.23b) the following set of equations was solved using the
symbolic algebra software Mathematica:

being the corresponding solution given by:

93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
Appendix II

In the sequel the dynamic response of the two bridge girders presented in Chapter 5 is evaluated for a set
of four different velocity values of the high-spreed trains illustrated in Figure 5.18.

DOUBLE-T CROSS SECTION GIRDER UNDER THE ACTION OF HSLM-A1

(a) – Numerical Model


3.0E-03
(m)

1.5E-03
𝑧-displacement

0.0E+00

-1.5E-03
0 50 100 150 200 250 300 350 400 450 500
(b) – Analytical Model
3.0E-03
(m)

1.5E-03
𝑧-displacement

0.0E+00

-1.5E-03
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.1– dynamic influence lines of the 𝑧-displacement at the midpoint of the central span.

111
(a) – Numerical Model
4.0E-04
(m)

2.0E-04
𝑦-displacement

0.0E+00

-2.0E-04
0 50 100 150 200 250 300 350 400 450 500
(b) – Analytical Model
4.0E-04
(m)

2.0E-04
𝑦-displacement

0.0E+00

-2.0E-04
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.2– dynamic influence lines of the 𝑦-displacement at the midpoint of the central span.

(a) – Numerical Model


1.5E-04

0.0E+00
(rad)
𝜙-rotation

-1.5E-04

-3.0E-04
0 50 100 150 200 250 300 350 400 450 500

(b) – Analytical Model


1.5E-04

0.0E+00
(rad)
𝜙-rotation

-1.5E-04

-3.0E-04
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.3– dynamic influence lines of the 𝜙-rotation at the midpoint of the central span.

112
(a) – Numerical Model
3.0E-03
(m)

1.5E-03
𝑧-displacement

0.0E+00

-1.5E-03
0 50 100 150 200 250 300 350 400 450 500
(b) – Analytical Model
3.0E-03
(m)

1.5E-03
𝑧-displacement

0.0E+00

-1.5E-03
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.4– dynamic influence lines of the 𝑧-displacement at the track’s midpoint of the central span.

(a) – Numerical Model


1.5E+00

7.5E-01
(ms-2)

0.0E+00
𝑧-acceleration

-7.5E-01

-1.5E+00
0 50 100 150 200 250 300 350 400 450 500

(b) – Analytical Model


1.5E+00

7.5E-01
(ms-2)

0.0E+00
𝑧-acceleration

-7.5E-01

-1.5E+00
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.5– dynamic influence lines of the 𝑧-acceleration at the midpoint of the central span.

113
(a) – Numerical Model
1.4E-01
(ms-2)

7.0E-02
𝑦-acceleration

0.0E+00

-7.0E-02

-1.4E-01
0 50 100 150 200 250 300 350 400 450 500
(b) – Analytical Model
1.4E-01
(ms-2)

7.0E-02
𝑦-acceleration

0.0E+00

-7.0E-02

-1.4E-01
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.6– dynamic influence lines of the 𝑦-acceleration at the midpoint of the central span.

(a) – Numerical Model


8.0E-02
(rads-2)

4.0E-02

0.0E+00
𝜙-acceleration

-4.0E-02

-8.0E-02
0 50 100 150 200 250 300 350 400 450 500

(b) – Analytical Model


8.0E-02
(rads-2)

4.0E-02

0.0E+00
𝜙-acceleration

-4.0E-02

-8.0E-02
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.7– dynamic influence lines of the 𝜙-acceleration at the midpoint of the central span.

114
(a) – Numerical Model
1.5E+00
(ms-2)

7.5E-01
𝑧-acceleration

0.0E+00

-7.5E-01

-1.5E+00
0 50 100 150 200 250 300 350 400 450 500
(b) – Analytical Model
1.5E+00
(ms-2)

7.5E-01
𝑧-acceleration

0.0E+00

-7.5E-01

-1.5E+00
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.8– dynamic influence lines of the 𝑧-acceleration at the track’s midpoint of the central span.

DOUBLE-T CROSS SECTION GIRDER UNDER THE ACTION OF HSLM-A10

(a) – Numerical Model


5.0E-03
(m)

2.5E-03
𝑧-displacement

0.0E+00

-2.5E-03

-5.0E-03
0 50 100 150 200 250 300 350 400 450 500

(b) – Analytical Model


5.0E-03
(m)

2.5E-03
𝑧-displacement

0.0E+00

-2.5E-03

-5.0E-03
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.9– dynamic influence lines of the 𝑧-displacement at the midpoint of the central span.

115
(a) – Numerical Model
6.0E-04
(m)

3.0E-04
𝑦-displacement

0.0E+00

-3.0E-04
0 50 100 150 200 250 300 350 400 450 500
(b) – Analytical Model
6.0E-04
(m)

3.0E-04
𝑦-displacement

0.0E+00

-3.0E-04
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.10– dynamic influence lines of the 𝑦-displacement at the midpoint of the central span.

(a) – Numerical Model


2.0E-04

0.0E+00
(rad)
𝜙-rotation

-2.0E-04

-4.0E-04
0 50 100 150 200 250 300 350 400 450 500

(b) – Analytical Model


2.0E-04

0.0E+00
(rad)
𝜙-rotation

-2.0E-04

-4.0E-04
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.11– dynamic influence lines of the 𝜙-rotation at the midpoint of the central span.

116
(a) – Numerical Model
6.0E-03
(m)

3.0E-03
𝑧-displacement

0.0E+00

-3.0E-03
0 50 100 150 200 250 300 350 400 450 500
(b) – Analytical Model
6.0E-03
(m)

3.0E-03
𝑧-displacement

0.0E+00

-3.0E-03
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.12– dynamic influence lines of the 𝑧-displacement at the track’s midpoint of the central span.

(a) – Numerical Model


3.0E+00

1.5E+00
(ms-2)

0.0E+00
𝑧-acceleration

-1.5E+00

-3.0E+00
0 50 100 150 200 250 300 350 400 450 500

(b) – Analytical Model


3.0E+00

1.5E+00
(ms-2)

0.0E+00
𝑧-acceleration

-1.5E+00

-3.0E+00
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.13– dynamic influence lines of the 𝑧-acceleration at the midpoint of the central span.

117
(a) – Numerical Model
3.0E-01
(ms-2)

1.5E-01
𝑦-acceleration

0.0E+00

-1.5E-01

-3.0E-01
0 50 100 150 200 250 300 350 400 450 500
(b) – Analytical Model
3.0E-01
(ms-2)

1.5E-01
𝑦-acceleration

0.0E+00

-1.5E-01

-3.0E-01
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.14– dynamic influence lines of the 𝑦-acceleration at the midpoint of the central span.

(a) – Numerical Model


2.0E-01
(rads-2)

1.0E-01

0.0E+00
𝜙-acceleration

-1.0E-01

-2.0E-01
0 50 100 150 200 250 300 350 400 450 500

(b) – Analytical Model


2.0E-01
(rads-2)

1.0E-01

0.0E+00
𝜙-acceleration

-1.0E-01

-2.0E-01
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.15– dynamic influence lines of the 𝜙-acceleration at the midpoint of the central span.

118
(a) – Numerical Model
3.0E+00
(ms-2)

1.5E+00
𝑧-acceleration

0.0E+00

-1.5E+00

-3.0E+00
0 50 100 150 200 250 300 350 400 450 500
(b) – Analytical Model
3.0E+00
(ms-2)

1.5E+00
𝑧-acceleration

0.0E+00

-1.5E+00

-3.0E+00
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.16– dynamic influence lines of the 𝑧-acceleration at the track’s midpoint of the central span.

BOX CROSS SECTION GIRDER UNDER THE ACTION OF HSLM-A1

(a) – Numerical Model


3.0E-03
(m)

1.5E-03
𝑧-displacement

0.0E+00

-1.5E-03
0 50 100 150 200 250 300 350 400 450 500

(b) – Analytical Model


3.0E-03
(m)

1.5E-03
𝑧-displacement

0.0E+00

-1.5E-03
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.17– dynamic influence lines of the 𝑧-displacement at the midpoint of the central span.

119
(a) – Numerical Model
5.0E-06

0.0E+00
(m)
𝑦-displacement

-5.0E-06

-1.0E-05

-1.5E-05
0 50 100 150 200 250 300 350 400 450 500
(b) – Analytical Model
5.0E-06

0.0E+00
(m)
𝑦-displacement

-5.0E-06

-1.0E-05

-1.5E-05
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.18– dynamic influence lines of the 𝑦-displacement at the midpoint of the central span.

(a) – Numerical Model


3.0E-05

0.0E+00
(rad)

-3.0E-05
𝜙-rotation

-6.0E-05

-9.0E-05
0 50 100 150 200 250 300 350 400 450 500

(b) – Analytical Model


3.0E-05

0.0E+00
(rad)

-3.0E-05
𝜙-rotation

-6.0E-05

-9.0E-05
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.19– dynamic influence lines of the 𝜙-rotation at the midpoint of the central span.

120
(a) – Numerical Model
3.0E-03
(m)

1.5E-03
𝑧-displacement

0.0E+00

-1.5E-03
0 50 100 150 200 250 300 350 400 450 500
(b) – Analytical Model
3.0E-03
(m)

1.5E-03
𝑧-displacement

0.0E+00

-1.5E-03
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.20– dynamic influence lines of the 𝑧-displacement at the track’s midpoint of the central span.

(a) – Numerical Model


8.0E-01

4.0E-01
(ms-2)

0.0E+00
𝑧-acceleration

-4.0E-01

-8.0E-01
0 50 100 150 200 250 300 350 400 450 500

(b) – Analytical Model


8.0E-01

4.0E-01
(ms-2)

0.0E+00
𝑧-acceleration

-4.0E-01

-8.0E-01
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.21– dynamic influence lines of the 𝑧-acceleration at the midpoint of the central span.

121
(a) – Numerical Model
2.0E-02
(ms-2)

1.0E-02
𝑦-acceleration

0.0E+00

-1.0E-02

-2.0E-02
0 50 100 150 200 250 300 350 400 450 500
(b) – Analytical Model
2.0E-02
(ms-2)

1.0E-02
𝑦-acceleration

0.0E+00

-1.0E-02

-2.0E-02
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.22– dynamic influence lines of the 𝑦-acceleration at the midpoint of the central span.

(a) – Numerical Model


3.0E-02
(rads-2)

1.5E-02

0.0E+00
𝜙-acceleration

-1.5E-02

-3.0E-02
0 50 100 150 200 250 300 350 400 450 500

(b) – Analytical Model


3.0E-02
(rads-2)

1.5E-02

0.0E+00
𝜙-acceleration

-1.5E-02

-3.0E-02
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.23– dynamic influence lines of the 𝜙-acceleration at the midpoint of the central span.

122
(a) – Numerical Model
8.0E-01
(ms-2)

4.0E-01
𝑧-acceleration

0.0E+00

-4.0E-01

-8.0E-01
0 50 100 150 200 250 300 350 400 450 500
(b) – Analytical Model
8.0E-01
(ms-2)

4.0E-01
𝑧-acceleration

0.0E+00

-4.0E-01

-8.0E-01
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.24– dynamic influence lines of the 𝑧-acceleration at the track’s midpoint of the central span.

BOX CROSS SECTION GIRDER UNDER THE ACTION OF HSLM-A10

(a) – Numerical Model


3.6E-03
(m)

1.8E-03
𝑧-displacement

0.0E+00

-1.8E-03
0 50 100 150 200 250 300 350 400 450 500

(b) – Analytical Model


3.6E-03
(m)

1.8E-03
𝑧-displacement

0.0E+00

-1.8E-03
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.25– dynamic influence lines of the 𝑧-displacement at the midpoint of the central span.

123
(a) – Numerical Model
5.0E-06

0.0E+00
(m)
𝑦-displacement

-5.0E-06

-1.0E-05

-1.5E-05
0 50 100 150 200 250 300 350 400 450 500
(b) – Analytical Model
5.0E-06

0.0E+00
(m)
𝑦-displacement

-5.0E-06

-1.0E-05

-1.5E-05
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.26– dynamic influence lines of the 𝑦-displacement at the midpoint of the central span.

(a) – Numerical Model


3.0E-05

0.0E+00
(rad)

-3.0E-05
𝜙-rotation

-6.0E-05

-9.0E-05
0 50 100 150 200 250 300 350 400 450 500

(b) – Analytical Model


3.0E-05

0.0E+00
(rad)

-3.0E-05
𝜙-rotation

-6.0E-05

-9.0E-05
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.27– dynamic influence lines of the 𝜙-rotation at the midpoint of the central span.

124
(a) – Numerical Model
3.6E-03
(m)

1.8E-03
𝑧-displacement

0.0E+00

-1.8E-03
0 50 100 150 200 250 300 350 400 450 500
(b) – Analytical Model
3.6E-03
(m)

1.8E-03
𝑧-displacement

0.0E+00

-1.8E-03
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.28– dynamic influence lines of the 𝑧-displacement at the track’s midpoint of the central span.

(a) – Numerical Model


1.2E+00

6.0E-01
(ms-2)

0.0E+00
𝑧-acceleration

-6.0E-01

-1.2E+00
0 50 100 150 200 250 300 350 400 450 500

(b) – Analytical Model


1.2E+00

6.0E-01
(ms-2)

0.0E+00
𝑧-acceleration

-6.0E-01

-1.2E+00
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.29– dynamic influence lines of the 𝑧-acceleration at the midpoint of the central span.

125
(a) – Numerical Model
3.6E-02
(ms-2)

1.8E-02
𝑦-acceleration

0.0E+00

-1.8E-02

-3.6E-02
0 50 100 150 200 250 300 350 400 450 500
(b) – Analytical Model
3.6E-02
(ms-2)

1.8E-02
𝑦-acceleration

0.0E+00

-1.8E-02

-3.6E-02
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.30– dynamic influence lines of the 𝑦-acceleration at the midpoint of the central span.

(a) – Numerical Model


5.0E-02
(rads-2)

2.5E-02

0.0E+00
𝜙-acceleration

-2.5E-02

-5.0E-02
0 50 100 150 200 250 300 350 400 450 500

(b) – Analytical Model


5.0E-02
(rads-2)

2.5E-02

0.0E+00
𝜙-acceleration

-2.5E-02

-5.0E-02
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.31– dynamic influence lines of the 𝜙-acceleration at the midpoint of the central span.

126
(a) – Numerical Model
1.2E+00 MEF (60)
MEF (80)
6.0E-01 MEF (100)
(ms-2)

MEF (120)
𝑧-acceleration

0.0E+00

-6.0E-01

-1.2E+00
0 50 100 150 200 250 300 350 400 450 500
(b) – Analytical Model
1.2E+00 Analytical (60)
Analytical (80)
6.0E-01 Analytical (100)
(ms-2)

Analytical (120)
𝑧-acceleration

0.0E+00

-6.0E-01

-1.2E+00
0 50 100 150 200 250 300 350 400 450 500
𝑥 (m)
∎ 60 (ms-1) ∎ 80 (ms-1) ∎ 100 (ms-1) ∎ 120 (ms-1)
Figure A.32– dynamic influence lines of the 𝑧-acceleration at the track’s midpoint of the central span.

127
129

Você também pode gostar