Engineering Analysis with Boundary Elements 63 (2016) 134–153

Contents lists available at ScienceDirect

Engineering Analysis with Boundary Elements

journal homepage: www.elsevier.com/locate/enganabound

Dynamic analysis of Euler–Bernoulli beams by the time-dependent

boundary element method formulation
R.F. Scuciato a, J.A.M. Carrer a,n, W.J. Mansur b
a
Programa de Pós-Graduação em Métodos Numéricos em Engenharia, Universidade Federal do Paraná, Caixa Postal 19011, CEP 81531-980,
Curitiba, PR, Brasil
b
Programa de Engenharia Civil, Universidade Federal do Rio de Janeiro, Caixa Postal 68506, CEP 21945-970, Rio de Janeiro, RJ, Brasil

art ic l e i nf o a b s t r a c t

Article history: This paper deals with the solution of the Euler–Bernoulli equation for dynamic bending of beams by the
Received 15 June 2015 time-dependent Boundary Element Method formulation. Initially, an overview of the Euler–Bernoulli
Received in revised form beam theory is presented. Next, the time-dependent fundamental solution is introduced and some of its
13 October 2015
properties are discussed. In the sequence, the integral formulation, obtained through a weighted resi-
Accepted 9 November 2015
duals technique, is presented. Three different numerical implementations are proposed. Finally, the
Available online 9 December 2015
numerical results are compared with the available analytical solutions.
Keywords: & 2015 Elsevier Ltd. All rights reserved.
Dynamic analysis
Euler–Bernoulli beams
Time-dependent BEM

1. Introduction
In the context of the classical theory of beams, or the
Euler–Bernoulli beam theory, the reader can find, for instance,
the works by Providakis and Beskos [1], Schanz [2], and
de Langre et al. [3].
The first of these works is concerned with free and forced
flexural vibrations, in which the forced vibration problem is trea-
ted with the aid of the Laplace transform and the response is
obtained by a numerical inversion of the transformed solution.
In the second work a rather interesting method, called Con-
volution Quadrature Method, proposed by Lubich [4,5] to perform
the convolution in the time-dependent integral equation, is
employed. The main characteristics of this method, which can be
applied to problems where time-domain fundamental solutions
are not available, are (i) the use of fundamental solutions in the
Laplace domain and (ii) the numerical approximation of the time
convolution integrals, presented in the time-dependent Boundary
Element Method (TD-BEM) equations, by a quadrature formula
based on a linear method of multiple steps, which provides direct
solution in the time-domain.
The third work presents a TD-BEM formulation, which is very
interesting from the mathematical point of view. The fundamental
solution for this formulation can be found in Graff [6] and
n
Corresponding author.
E-mail addresses: raphael.scuciato@ufpr.br (R.F. Scuciato),
carrer@ufpr.br (J.A.M. Carrer), webe@coc.ufrj.br (W.J. Mansur).
Campbell and Foster [7]. One of the conclusions found in reference
[3] is transcribed here “It is nevertheless clear that the Laplace
transform domain method is the most adequate for the solution of
linear problems of flexural vibrations of beams”.
This conclusion sounded as a challenge to the authors: if others
TD-BEM formulations were successfully developed for other types
of problems, e.g., Wrobel [8], Mansur [9], why the same would not
be possible for the dynamic analysis of Euler–Bernoulli beams?
Bearing this in mind, this paper is concerned with the solution of
the Euler–Bernoulli equation for dynamic bending of beams by the
TD-BEM formulation. In other words, the motivation was to
develop a TD-BEM formulation capable of providing accurate
results and, consequently, of encouraging further developments.
These include beams over elastic basis, continuous beams, and
Timoshenko beams.
Initially an overview of the Euler–Bernoulli beam theory is
presented together with the analytical solutions used for com-
parison with the numerical values provided by the proposed TD-
BEM formulation. The time-dependent fundamental solution is
then introduced and some of its properties are discussed. In the
sequence, the basic TD-BEM equation, obtained by following a
weighted residuals approach is presented. Note that the problem
is one-dimensional; consequently, the boundary is constituted
only by the extreme nodes of the beam. As the domain is the
length of the beam, say L, one has 0 r x r L. As the differential
equation that governs the problem is non-homogeneous, due to
the presence of the loading term, a double integral in space and
time, containing this term, appears in the BEM equations. The
others integrals are time integrals, evaluated from t ¼ t 0 until the

http://dx.doi.org/10.1016/j.enganabound.2015.11.003
0955-7997/& 2015 Elsevier Ltd. All rights reserved.
 R.F. Scuciato et al. / Engineering Analysis with Boundary Elements 63 (2016) 134–153 135

present time. The beams considered are named according to their called, respectively, rotation, bending moment, shear force, and
boundary conditions as pinned–pinned (PP), clamped–pinned velocity.
(CP), clamped–clamped (CC), and clamped–free (CF). Four types of The problem presents four boundary conditions (two at x ¼ 0
loading are considered: the first one is assumed to be linearly and two at x ¼ L) and two initial conditions. The boundary con-
distributed along the domain and to act continuously in time (DD); ditions are dependent on the beam type under consideration and
the second type consists of a concentrated load acting con- are classified as essential boundary conditions, when u or ϕ are
tinuously in time (CD); the third and the fourth types are short prescribed, or natural boundary conditions, when M or Q are
duration loadings; the former is distributed along the domain prescribed. The initial conditions, conversely, are always of the
(DSD) and the later is concentrated (CSD). same type, namely, the values of u and v prescribed at time t ¼ t 0 .
Good agreement is observed between the numerical and ana- The analytical solutions of Eq. (2), for the beams considered in
lytical results, which are presented before the final conclusions. this work, can be found, for instance, in Rao [14]. The reader is
The results for the bending moments and shear forces at internal referred to Sheehan and Debnath [15] for an interesting discussion
points are also presented, as these quantities play a fundamental concerning the theory of the transient Euler–Bernoulli beam on
role in the dimensioning of concrete and steel beams. elastic foundation, taking into account the effects of linear
For problems with non-homogeneous initial conditions, not damping and axial loading. Another interesting work was pre-
taken into account here, all the domain or, at least, part of it, must sented by Chen et al. [16], which deals with Rayleigh damped
be discretized with cells, as a domain integral containing the non- Euler–Bernoulli beams subjected to multi-support motion.
null initial displacements and rotations appear in the BEM integral The fundamental solution for the problem is the function un
equations. Such a discretization can be carried out by employing that satisfies the equation
internal cells, treated as isogeometric boundary elements (see, for
instance, [10]). ∂ 4 un ∂ 2 un
c2 þ 2 ¼ δðx  ξÞδðt  τÞ; ð4Þ
An interesting extension of the present work could be that ∂x4 ∂t
based on the assumption of functionally graded materials (see, for
instance, [11–13]). In this case, the integral equations would dis- where δ is the Dirac delta function, and corresponds to the
play a domain integral containing the longitudinal elasticity deflection of a beam of infinite length measured at a field point x
modulus. The same would occur under the assumption of variable in a time t when an impulsive concentrated load is applied at a
transverse section. In general, for variable elasticity modulus and/ source point ξ in a time τ. Such function is given by
or moment of inertia, domain integrals appear in the formulation.       pffiffiffi   2  2 
1 r r r a r r
un ¼ S pffiffiffiffiffiffiffiffiffi  C pffiffiffiffiffiffiffiffiffi þ pffiffiffiffiffiffi sin þ cos ;
c 2 2π a 2π a 2π 4a 4a
ð5Þ
2. Euler–Bernoulli beam theory and BEM formulation
where r ¼ x  ξ and a ¼ cðt  τÞ.
Dynamic bending problems of uniform slender beams are The solution of Eq. (4) can be found in de Langre et al. [3],
governed by the classical Euler–Bernoulli equation which is, to the best of the authors' knowledge, the first BEM
∂4 u ∂2 u formulation which is based on the use of this fundamental solu-
EI þ ρA 2 ¼ f ðx; tÞ; ð1Þ tion. However, de Langre et al. [3] did not succeed in obtaining
∂x 4
∂t
accurate results and concluded that the Laplace transform was the
where uðx; tÞ is the vertical deflection of the beam, f ðx; tÞ is the most adequate method for the solution of linear problems of
applied external loading, EI is the flexural rigidity of the beam, ρ is flexural vibrations of beams. The fundamental solution can also be
the density of the material, A is the cross-sectional area, L is the found in Graff [6], Campbell and Foster [7], and Kythe [17].
length of the beam, x is the spatial variable, and t denotes time. The functions S and C, called Fresnel integrals, are defined by
Fig. 1 shows an schematic layout of the beam under consideration.
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Z Z
Dividing Eq. (1) by ρA and defining c ¼ ðEIÞ=ðρAÞ one gets the z
π z
π
SðzÞ ¼ sin ζ 2 dζ ; and CðzÞ ¼ cos ζ 2 dζ : ð6Þ
expression 0 2 0 2
∂4 u ∂2 u f ðx; tÞ Such integrals have no analytical solution and are evaluated
c2 þ ¼ ; ð2Þ
∂x4 ∂t 2 ρA numerically using the algorithm presented by Boersma [18].
which is the governing equation adopted in this work. The function un has the following properties:
From uðx; tÞ the following quantities are defined:
 reciprocity: un ðx; t; ξ; τÞ ¼ un ðξ;  τ; x;  tÞ;
∂u ∂2 u  time translation: un ðx; t; ξ; τÞ ¼ un ðx; t þ Δt; ξ; τ þ ΔtÞ.
ϕðx; tÞ ¼ ; Mðx; tÞ ¼  EI 2 ;
∂x ∂x
∂3 u ∂u Regarding the characteristics of un , it is worth mentioning that
Q ðx; tÞ ¼  EI ; vðx; tÞ ¼ ð3Þ
∂x3 ∂t un does not present singularity when r ¼ 0, though this kind of
singularity always occurs in the fundamental solutions of 2D and
3D problems. In a similar fashion, the fundamental solutions for
Timoshenko beams listed in Carrer et al. [19] are also not singular
when r ¼ 0.
From Eq. (5) one gets
      
1 1 r r ∂r
ϕn ¼ þ S pffiffiffiffiffiffiffiffiffi  C pffiffiffiffiffiffiffiffiffi ; ð7Þ
c 2 2π a 2π a ∂x

   2  2  2
EI 1 r r ∂r
Mn ¼  pffiffiffiffiffiffiffiffiffi sin  cos ; ð8Þ
Fig. 1. Schematic layout of the beam under consideration. c 2 2π a 4a 4a ∂x
136 R.F. Scuciato et al. / Engineering Analysis with Boundary Elements 63 (2016) 134–153

   2  2  3
EI r r r ∂r potential function; in the case of Euller–Bernoulli beams, the
Qn ¼  pffiffiffiffiffiffiffiffiffiffiffi sin þ cos ; ð9Þ bending moment and shear force. The assumption of linear time
c 4 2π a3 4a 4a ∂x
variation enforces continuity of the variables and, therefore, the
∂r
where ¼ 71 as x≷ξ, respectively. discontinuity represented by the jumps is not taken into account
∂x
Using a weighted residuals approach, the integral equation for appropriately, which tends to deteriorate the numerical response.
the deflections, in the absence of initial conditions, is written as The assumption of linear time variation of the variables associated
Z th i with the essential, or Dirichlet, boundary conditions, by its turn,
1
uðξ; tÞ ¼  un Q  ϕ M þ M n ϕ Q n u
n
dτ emphasizes the mentioned common sense. Following this rea-
ρA 0 x¼0
soning, in the third formulation u and ϕ are assumed to vary
Z th i
1 linearly inside each time interval, whereas for M and Q a constant
un Q  ϕ M þ M n ϕ Q n u
n
þ dτ
ρA 0 x¼L time variation is imposed. The loading f also present a linear time
Z tZ L
1 variation, both in time and space. Note that all the integrals pre-
þ un f dx dτ : ð10Þ
ρA 0 0 sent in these formulations can be analytically evaluated.
It is important to mention the alternative formulation devel-
Note that the solution of the problem involves four boundary
oped by Yu et al. [21], in which all the variables are assumed to
unknowns and requires two distinct integral equations to turn the
vary linearly in the modified time interval θΔt, instead of in the
solution of the problem possible. The required additional equation
habitual time interval Δt. In this formulation, the boundary
can be obtained by taking the derivatives of Eq. (10) with respect to
unknowns are initially computed at a fictitious time t n þ θ and
the source point, thus obtaining an integral equation for rotations,
brought back to the actual time according to the formula
Z t h i 
1 ∂
ϕðξ; tÞ ¼  un Q  ϕ M þ M n ϕ Q n u
n
dτ 1
V n þ 1 ¼ V n þ ðV n þ θ  V n Þ; ð14Þ
ρA ∂ ξ 0 x¼0
θ
Z t h i 
1 ∂ n
where the generic variable V can represent u, ϕ, M, or Q .
þ un Q  ϕ M þ M n ϕ  Q n u dτ
ρA ∂ ξ 0 x¼L
As mentioned earlier, Eqs. (10) and (11) are used for the solu-
Z t Z L 
1 ∂ tion of the problem, i.e., are simultaneously applied at the
þ un f dx dτ ; ð11Þ
ρA ∂ ξ 0 0 boundary nodes of the beam. Such procedure leads to a matrix
equation of the form
bending moments,
Z t h i  q
½Hqq fugq ¼ ½Gqq fqgq þ ½Fqq ffg þ fbg; ð15Þ
∂2
Mðξ; tÞ ¼ þ c2 2 un Q  ϕ M þ M n ϕ  Q n u
n
dτ qq qq q q
∂ξ 0 x¼0 where ½H and ½G are 4  4 matrices. The vectors fug , fqg , and
Z t h i  q
ffg contain, respectively, the values of the deflections and rota-
∂ 2
un Q  ϕ M þ M n ϕ  Q n u
n
 c2 2 dτ tions, bending moments and shear forces, and loads at the present
∂xi 0 x¼L
time t ¼ t q . The diagonal elements of ½Hqq are equal to 1/2. The
Z t Z L 
∂ 2
vector fbg stores the contributions of the previous times, either
 c2 2 un f dx dτ ; ð12Þ
∂ξ 0 0 boundary variables and load values, and is given by
and shear forces, qX
1
j
Z t h i  fbg ¼ ½Gqj fqgj  ½Hqj fugj þ ½Fqj ffg : ð16Þ
∂3 n
Q ðξ; tÞ ¼ þ c2 3 un Q  ϕ M þ M n ϕ  Q n u dτ j¼1

∂ξ 0 x¼0
All the coefficients in the matrices present in Eqs. (15) and (16)
Z t h i 
∂ 3
are computed considering the present time t ¼ t q but, due to time
un Q  ϕ M þ M n ϕ  Q u
n
 c2 3 n
dτ
∂ξ 0 x¼L translation properties, can be evaluated by taking into account the
Z t Z L 
∂ 3 values obtained in the previous time-step (i.e., for t ¼ t q  1 ) plus
 c2 3 un f dx dτ : ð13Þ the values of the respective integrals over de first time-step only
∂ξ 0 0
(the reader can see [9] for more details about this procedure). This
This work employs Eqs. (10) and (11) for the solution of the results in great computer time savings, as the analysis advances,
problem; Eqs. (12) and (13) are only used in the post- since most of the coefficients are reused. In addition, after the
processing stage.
For the evaluation of the time integrals in Eqs. (10)–(13), the Table 1
overall time of analysis is initially divided in intervals Δt. Similarly, Geometrical and material properties.
when necessary, the length of the beam can be divided in cells of
Quantity Value Unit
length Δx for the evaluation of the space integrals. According to
the assumed behavior of the variables u, ϕ, M, and Q , and also of E 2:0000E þ 11 N=m2
the external loading f , inside each time interval, three BEM for- I 1:7393E  04 m4
mulations arise. The authors’ experience is discussed in detail in ρ 7:8500E þ 03 kg=m3
Scuciato [20]. In what follows, a brief discussion is carried out. In A 6:2000E  03 m2
the first formulation, a constant time variation was assumed for all
the variables. The second formulation, developed with the aim of
improving the results produced by the first one, assumed a linear
time variation for all the variables. Contrary to the common sense Table 2
Load value for each type of loading.
(it must be pointed out that the following discussion is typical for
TD-BEM formulations, which assumes that the higher degrees of Loading Value Unit Position Duration
approximation are responsible for better accuracy), not reliable
results were produced under this assumption. Variables associated DD 10 kN/m — —
with natural, or Newmann, boundary conditions can present CD 10  L kN 3L=4 —
DSD 1000 kN/m — 100 μs
jumps along the time history: in the case of the scalar wave CSD 1000  L kN 3L=4 100 μs
equation, studied by Mansur [9], the normal derivative of the
 R.F. Scuciato et al. / Engineering Analysis with Boundary Elements 63 (2016) 134–153 137

Fig. 2. Load type: DD/beam type: PP.

introduction of the prescribed boundary conditions, Eq. (15) can be The final aspect discussed is related with the behavior of the
rewritten simply as ½Afxg ¼ fyg. The matrix ½A arises from con- developed TD-BEM formulation when are considered specialized
tributions of matrices ½Hqq and ½Gqq and its inverse is evaluated loadings. Loadings of the type CD are modeled with the aid of the
only once for the whole analysis. Dirac delta function as f ðx; tÞ ¼ δðx  xk ÞFðtÞ, where xk is the point of
138 R.F. Scuciato et al. / Engineering Analysis with Boundary Elements 63 (2016) 134–153

Fig. 3. Load type: DD/beam type: CP.

application of the loading. Similarly, loadings of the type DSD are definitions H is the Heaviside function and Hðt D  tÞ ¼ 1 for 0 r t
defined as f ðx; tÞ ¼ FðxÞHðt D tÞ, and, by extension, loadings of the rt D and Hðt D  tÞ ¼ 0 otherwise. For the CSD loading, F is the
type CSD are given by f ðx; tÞ ¼ δðx  xk ÞHðt D  tÞF. In the previous concentrated load intensity.
 R.F. Scuciato et al. / Engineering Analysis with Boundary Elements 63 (2016) 134–153 139

Fig. 4. Load type: DD/beam type: CC.

3. Numerical examples
In the examples of this section, the TD-BEM results are always
compared with the available analytical solutions (see, for instance,
[14]). The CP beam has L ¼ 2 m; for the other three types of beams
one has L ¼ 4 m. The geometrical and material properties are
presented in Table 1; the load values for each type of loading are
listed in Table 2. The prescribed boundary conditions and initial
conditions are null in all examples. The results presented here
were obtained after performing several analysis, ranging the time-
140 R.F. Scuciato et al. / Engineering Analysis with Boundary Elements 63 (2016) 134–153

Fig. 5. Load type: DD/beam type: CF.

step value from 100 μs to 5 μs (see [20] for a more detailed dis- For each type of loading, the results for each type of beam are
cussion concerning this matter). By choosing Δt ¼ 5 μs, good presented: first, those related to the boundary unknowns; then, those
results were obtained and this value was adopted for all the gra- related to u, ϕ, M, and Q at x ¼ L=4. In this way, the boundary results
phics included here. are: for the PP beam, ϕ and Q ; for the CP beam, M and Q at x ¼ 0, and
 R.F. Scuciato et al. / Engineering Analysis with Boundary Elements 63 (2016) 134–153 141

Fig. 6. Load type: CD/beam type: PP.

ϕ and Q at x ¼ L; for the CC beam, M and Q ; and for CF beam, M and Q An overview of Figs. 2–5 enables one to conclude that, for the
at x ¼ 0, and u and ϕ at x ¼ L. Figs. 2–5 depict results for the DD loading DD, the BEM results are always in good agreement with the
loading; Figs. 6–9, the results for CD loading; Figs. 10–13, the results for analytical solutions, both at the boundary nodes of the problem and at
the DSD loading, and, finally, Figs. 14–17, the results for the CSD loading. the selected internal point. Even the results for shear forces can be
142 R.F. Scuciato et al. / Engineering Analysis with Boundary Elements 63 (2016) 134–153

Fig. 7. Load type: CD/beam type: CP.

considered very good, in spite of the oscillations that appear durations loading also present good agreement with the analytical
in the respective analytical solution. For the CD loading, the results in solutions, see Figs. 10–13 for the DSD loading, and Figs. 14–17 for the
Figs. 6–9 allow the same conclusion. The results concerning short CSD loading.
 R.F. Scuciato et al. / Engineering Analysis with Boundary Elements 63 (2016) 134–153 143

Fig. 8. Load type: CD/beam type: CC.

The algorithm presented by Boersma [18] gives errors of 1:6  errors can be considered small, an improved algorithm, based on
10  9 and 0:5  10  9 for 0 r z r 4 and z 4 4, respectively, if z is the τ-method, see Lanczos [22], can be developed in such way that
taken as the argument of the Fresnel Integrals. Although these (it is the authors' hope), with minor errors occurring, an improved
144 R.F. Scuciato et al. / Engineering Analysis with Boundary Elements 63 (2016) 134–153

Fig. 9. Load type: CD/beam type: CF.

accuracy for the TD-BEM results is achieved. This suggestion Finally, it is important to mention that the selection of the time-
remains open for future developments, as it is far from the scope of step does not suffer the limitations found, for instance, in the TD-BEM
this work. Note that Boersma's algorithm is based on the τ-method. formulation for 2D scalar wave problem (see [9] for more details).
R.F. Scuciato et al. / Engineering Analysis with Boundary Elements 63 (2016) 134–153 145

Fig. 10. Load type: DSD/beam type: PP.

146 R.F. Scuciato et al. / Engineering Analysis with Boundary Elements 63 (2016) 134–153

Fig. 11. Load type: DSD/beam type: CP.

R.F. Scuciato et al. / Engineering Analysis with Boundary Elements 63 (2016) 134–153 147

Fig. 12. Load type: DSD/beam type: CC.

148 R.F. Scuciato et al. / Engineering Analysis with Boundary Elements 63 (2016) 134–153

Fig. 13. Load type: DSD/beam type: CF.

R.F. Scuciato et al. / Engineering Analysis with Boundary Elements 63 (2016) 134–153 149

Fig. 14. Load type: CSD/beam type: PP.

150 R.F. Scuciato et al. / Engineering Analysis with Boundary Elements 63 (2016) 134–153

Fig. 15. Load type: CSD/beam type: CP.

R.F. Scuciato et al. / Engineering Analysis with Boundary Elements 63 (2016) 134–153 151

Fig. 16. Load type: CSD /beam type: CC.

152 R.F. Scuciato et al. / Engineering Analysis with Boundary Elements 63 (2016) 134–153

Fig. 17. Load type: CSD/beam type: CF.

 R.F. Scuciato et al. / Engineering Analysis with Boundary Elements 63 (2016) 134–153
4. Conclusions
The Euler–Bernoulli beam theory, also called classic beam
theory, is relatively simple and gives, from the practical point of
view, very acceptable results, which justifies the attention it has
received for quite some time and the development of the TD-
BEM formulation presented here. For this reason this classical
problem is revisited and solved here by employing a very elegant
and challenging numerical tool, the TD-BEM formulation. Yet,
analyzing the numerical results presented, the effort was
rewarded: accurate results were obtained in most of the cases
studied, what demonstrates, without questioning, the feasibility
of the formulation. Considering what was discussed, it seems
that the outcome of this work is positive and represents only the
beginning of a promising line of research, which can include
continuous beams, beams over elastic basis, beams with variable
cross section, not to mention similar studies concerning
Timoshenko beams.
References
[1] Providakis CP, Beskos DE. Dynamic analysis of beams by the boundary element
method. Comput Struct 1986;22(6):957–64.
[2] Schanz M. A time domain boundary integral formulation for Euler–Bernoulli
beams. Z Angew, Math Mech 1999;79(S2):S575–6.
[3] de Langre E, Axisa F, Guilbaud D. Forced flexural vibrations of beams using a
time-stepping boundary element method. In: Brebbia CA, Chaudouet-Miranda
A, editors. Boundary elements in mechanical and electrical engineering: pro-
ceedings of the international boundary element symposium (Nice, France,
May 15–17, 1990). Southampton: Computational Mechanics Publications;
1990. p. 139–50.
[4] Lubich C. Convolution quadrature and discretized operational calculus I.
Numer Math 1988;52(2):129–45.
153
[5] Lubich C. Convolution quadrature and discretized operational calculus II.
Numer Math 1988;52(4):413–25.
[6] Graff KF. Wave motion in elastic solids. New York: Dover Publications; 1991.
[7] Campbell GA, Foster RM. Fourier integrals for practical applications. New
Jersey: D. van Nostrand Company, Princeton; 1961.
[8] Wrobel LC. Potential and viscous flow problems using the boundary element
method [Ph.D. thesis]. Southampton, England: University of Southampton;
1981.
[9] Mansur WJ. A time-stepping technique to solve wave propagation problems
using the boundary element method [Ph.D. thesis]. Southampton, England:
University of Southampton; 1983.
[10] Simpson RN, Bordas SPA, Trevelyan J, Rabczuk T. A two-dimensional isogeo-
metric boundary element method for elastostatic analysis. Comput Methods
Appl Mech Eng 2012;209–212:87–100.
[11] Chakraborty A, Gopalakrishnan S, Reddy JN. A new beam finite element for the
analysis of functionally graded materials. Int J Mech Sci 2003;45(3):519–39.
[12] Miers LS, Telles JCF. Two-dimensional elastostatic analysis of FGMs via BEFM.
Eng Anal Bound Elem 2008;32(12):1006–11.
[13] Sankar BV. An elasticity solution for functionally graded beams. Compos Sci
Technol 2001;61(5):689–96.
[14] Rao SS. Mechanical vibrations. 5 ed.. Upper Saddle River: Prentice-Hall; 2010.
[15] Sheehan J, Debnath L. On the dynamic response of an infinite Bernoulli–Euler
beam. Pure Appl Geophys 1972;97(1):100–10.
[16] Chen JT, Hong H-K, Yeh CS, Chyuan SW. Integral representations and reg-
ularizations for a divergent series solution of a beam subjected to support
motions. Earthq Eng Struct Dyn 1996;25(9):909–25.
[17] Kythe PK. Fundamental solutions for differential operators and applications.
Boston: Birkhäuser; 1996.
[18] Boersma J. Computation of Fresnel integrals. Math Comput 1960;14(72):380.
[19] Carrer JAM, Fleischfresser SA, Garcia LFT, Mansur WJ. Dynamic analysis of
Timoshenko beams by the boundary element method. Eng Anal Bound Elem
2013;37(12):1602–16.
[20] Scuciato RF. Dynamic analysis of Euler–Bernoulli beams by the boundary
element method with time-dependent fundamental solutions (in Portuguese)
[D.Sc. thesis]. Curitiba, Brasil: Universidade Federal do Paraná; 2014.
[21] Yu G, Mansur WJ, Carrer JAM, Gong L. A linear θ method applied to 2D time-
domain BEM. Commun Numer Methods Eng 1998;14(12):1171–9.
[22] Lanczos C. Applied analysis. New York: Dover Publications, Inc.; 1988.