You are on page 1of 12

acta mechanica solida sinica 30 (2017) 596–607

Available online at www.sciencedirect.com

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

Closed-form solution of beam on Pasternak


foundation under inclined dynamic load

Yu Miao a, Yang Shi a, Guobo Wang b,∗, Yi Zhong a


a School
of Civil Engineering and Mechanics, Huazhong University of Science and Technology, Wuhan 430074, China
b HubeiKey Laboratory of Roadway Bridge & Structure Engineering, Wuhan University of Technology, Wuhan, Hubei
430070, China

a r t i c l e i n f o a b s t r a c t

Article history: The dynamic response of an infinite Euler–Bernoulli beam resting on Pasternak foundation
Received 20 June 2017 under inclined harmonic line loads is developed in this study in a closed-form solution. The
Revised 25 October 2017 conventional Pasternak foundation is modeled by two parameters wherein the second pa-
Accepted 27 October 2017 rameter can account for the actual shearing effect of soils in the vertical direction. Thus,
Available online 5 December 2017 it is more realistic than the Winkler model, which only represents compressive soil resis-
tance. However, the Pasternak model does not consider the tangential interaction between
Keywords: the bottom of the beam and the foundation; hence, the beam under inclined loads cannot
Beam be considered in the model. In this study, a series of horizontal springs is diverted to the
Harmonic line load face between the bottom of the beam and the foundation to address the limitation of the
Pasternak foundation Pasternak model, which tends to disregard the tangential interaction between the beam and
Tangential interaction between the the foundation. The horizontal spring reaction is assumed to be proportional to the relative
beam and the foundation tangential displacement. The governing equation can be deduced by theory of elasticity and
Fourier transform Newton’s laws, combined with the linearly elastic constitutive relation and the geometric
equation of the beam body under small deformation condition. Double Fourier transforma-
tion is used to simplify the geometric equation into an algebraic equation, thereby conve-
niently obtaining the analytical solution in the frequency domain for the dynamic response
of the beam. Double Fourier inverse transform and residue theorem are also adopted to de-
rive the closed-form solution. The proposed solution is verified by comparing the degraded
solution with the known results and comparing the analytical results with numerical results
using ANSYS. Numerical computations of distinct cases are provided to investigate the ef-
fects of the angle of incidence and shear stiffness on the dynamic response of the beam.
Results are realistic and can be used as reference for future engineering designs.
© 2017 Published by Elsevier Ltd on behalf of Chinese Society of Theoretical and Applied
Mechanics.

applications in several fields of engineering, such as geotech-


1. Introduction nical, transportation, and tunnel engineering [1]; thus, it
has gained interest from researchers and engineers over
The classical problem of beams on elastic or viscoelastic
the past few decades. The two basic methods, namely the
foundations under static and dynamic loads has numerous
analytical and numerical methods, are generally employed
in contemporary dynamic design and in the analysis of

Corresponding author. beams resting on either elastic or viscoelastic foundations.
E-mail address: wgb16790604@126.com (G. Wang). Numerical methods, such as the finite element method, are

https://doi.org/10.1016/j.camss.2017.10.006
0894-9166/© 2017 Published by Elsevier Ltd on behalf of Chinese Society of Theoretical and Applied Mechanics.
acta mechanica solida sinica 30 (2017) 596–607 597

effective in analyzing complex structures [2–6]. However, time force existing in practical engineering, thereby changing the
and spatial increments in a numerical algorithm must be internal force distribution. However, previous studies have yet
considered in dynamic problems, such as wave propagation to propose a closed-form solution of the dynamic response
problems. Thus, numerical methods are time-consuming, of beam on Pasternak foundation considering the tangential
and their computational accuracy is related to the integrated interaction between the beam and the foundation under
algorithm. In response, some researchers have proposed dynamic loads that are not perpendicular to the beam.
analytical solutions to improve computational efficiency. In this study, the governing equation considering the
Considerable research has been conducted on improv- tangential interaction between the beam and the foundation
ing the computational efficiency of numerical methods is deduced by conducting a mechanical analysis of a beam
[7]. Kenney [8] first proposed the closed-form solutions of element. Double Fourier transformation is used to simplify
an infinite Euler–Bernoulli beam on a Winkler foundation the governing equation into an algebraic equation, thereby
[9] under a concentrated load moving at a constant velocity. allowing the equation in the frequency domain to be solved
Mathews [10,11] extended Kenney’s solution for an oscil- easily. Residue theorem is employed to calculate the im-
lating moving force. Achenbach and Sun [12] obtained the proper integral, and the closed-form solution of the dynamic
steady-state response of a beam resting on Winkler founda- response of an infinite beam resting on Pasternak foundation
tions with damping under a moving concentrated load and can be obtained while considering the tangential interaction
discussed the effects of linear damping in the analysis. Sun between the beam and the foundation under harmonic line
[13–15] employed double Fourier transformation to obtain the loads. Several numerical examples are calculated to probe
closed-form solutions of an infinite Euler–Bernoulli beam on the effects of the angle of incidence, shear stiffness, and hor-
elastic or viscoelastic foundations under harmonic, moving, izontal spring stiffness on the beam response. The proposed
and moving harmonic line loads. Kim and Roesset [16] used solution can be an effective tool for engineering practitioners.
double Fourier transformation to obtain the dynamic dis-
placement response of an infinite beam on elastic foundation
with frequency-independent linear hysteretic damping sub- 2. Governing equation
jected to constant amplitude or a harmonic moving load.
Shamalta and Metrikine [17] obtained the dynamic response Fig. 1 depicts an infinite Euler–Bernoulli beam on an elastic
of an embedded railway subjected to moving load using Pasternak foundation under a harmonic line load that is not
the Winkler foundation model. Yu and Yuan [18] recently perpendicular to the beam. The harmonic line load F(x, t) is
obtained the integral representation of the solution for an decomposed into an normal load F(x, t)cos θ and a tangential
infinite beam resting on a viscoelastic foundation subjected load F(x, t)sin θ . Herein, E, A, I, and ρ represent the elastic
to arbitrary dynamic loads and applied the solutions to modulus, cross-sectional area, moment of inertia around the
multipoint shaking table test for long tunnels subjected to y-axis, and beam density, respectively. qz , qx , and qs are the
non-uniform seismic loadings [19,20]. These previous studies normal foundation reaction, tangential foundation reaction,
have idealized the ground as the Winkler model; however, the and shearing effect of soils, respectively. The vertical and
shear interaction between disconnected soil springs has not horizontal displacements of the beam are defined as
been considered. Given that the mechanical resistance in soil ⎧
⎨w(x, z, t) = w0 (x, t)
arises from compressive and shear interactions, a realistic
⎩u(x, z, t) = u0 (x, t) − z ∂ w0 (x, t)
(1)
analysis using the two-parameter model (i.e., the Pasternak
∂x
model) has been proposed to examine the dynamic behavior
of beams under moving loads [21]. This model can account where w0 (x, t) and u0 (x, t) are the vertical and horizontal
for the shear effect of soil in vertical directions by introducing displacements of the beam axis, respectively.
a shear parameter in the analysis. Meanwhile, Karganovin A series of horizontal springs is diverted to the face be-
and Younesian [22] investigated the dynamic response of an tween the bottom of the beam and the foundation to consider
infinite Timoshenko beam on Pasternak foundation under a their tangential interaction. The horizontal spring reaction is
moving load by applying complex Fourier transform. Uzzal assumed to be proportional to the relative tangential displace-
et al. [23] obtained the dynamic response of a beam resting ment. In addition, we assume that the Euler–Bernoulli beam
on Pasternak foundation under moving load and moving is isotropic with linear elastic, small deformation. Only the
mass. Basu and Rao [24] presented analytical solutions for in-plane deformation of the beam is considered in this study.
an infinite beam resting on Pasternak foundation subjected Based on the above assumptions, normal foundation reac-
to a concentrated moving load. Yu and Yuan [25] presented tion qz , shearing effect of soils qs , and tangential foundation
analytical solutions for an Euler–Bernoulli beam on Pasternak reaction qx are defined respectively as
foundation subjected to arbitrary dynamic loads. ⎧

⎪ qz = −Kz wb (x, t)
The abovementioned studies only considered cases ⎪
⎨q = −K u (x, t)
x x b
wherein loads are perpendicular to the beam. Hence, the tra- (2)


ditional Winkler and Pasternak foundations do not consider ⎪ 2
⎩q = G ∂ wb (x, t)
s p
the effect of horizontal force. However, in actual situations, ∂ x2
cases wherein loads are not perpendicular to the beam, such where Kz and Kx are the vertical and horizontal spring stiff-
as the oblique incidence of seismic waves, are relatively ness, respectively; Gp is the shear stiffness, which is different
common. In this situation, the beam is in a state of eccentric from the shear modulus of the soil material; and wb (x, t) and
compression instead of a bending state due to the horizontal ub (x, t) are the relative displacements between the foundation
598 acta mechanica solida sinica 30 (2017) 596–607

Fig. 1 – Calculation model for a beam on an elastic foundation.

and the bottom of the beam, respectively. The negative sign


on the right-hand side of Eq. (2) indicates the directions of the
foundation reactions and the opposite relative displacements
between the foundation and the bottom of the beam.
Based on Eq. (1), we can obtain the relative displacement
between the foundation and the bottom of the beam as


⎨wb (x, t) = w0 (x, t)
(3)
⎩u (x, t) = u0 (x, t) − h ∂ w0 (x, t)

b
2 ∂x

where h denotes the depth of beam.


Strain can be obtained according to the geometric equation
of the beam under small deformation condition, as shown as
follows:

∂u(x, z, t) ∂ u0 (x, t) ∂ 2 w0 (x, t)


εx = = −z (4)
∂x ∂x ∂ x2

The linearly elastic constitutive equation of the beam


(Eq. (5)) can be used to determine stress.
Fig. 2 – Force analysis of the beam element on an elastic
σx = E ε x (5) foundation.

The relationship between the internal force and stress can


be expressed as
⎧  ⎧
⎪ ∂ 2 w0 (x, t) ∂V(x, t)

⎪ ⎪
⎪m = − qz + F (x, t) cos θ + qs
⎨N(x, t) = σx d A ⎪
⎪ ∂t2 ∂x


A . (6) ∂ 2 u0 (x, t) ∂N(x, t)

⎪ = − qx + F (x, t) sin θ (9)
⎩M(x, t) = σx z d A ⎪m
⎪ ∂ t 2 ∂x
A ⎪



⎩0 = ∂M(x, t) − V(x, t) − h (q + F (x, t) sin θ )
x
Combining Eqs. (4), (5), and (6) yields ∂x 2


⎪ ∂ u0 ∂ 2 w0 where m = ρA represents mass per unit length. F(x, t) is a
⎨N(x, t) = EA − ES
∂x ∂ x2 harmonic line load expressed by
(7)

⎪ ∂ u ∂ 2w
⎩M(x, t) = ES 0
− EI
0
∂x ∂ x2 H(r2 − x2 )
F (x, t) = P exp(it) (10)
where S is static moment around the y-axis, and S = 0 because 2r
the x-axis is the centroidal axis. Thus, Eq. (7) can be simplified
where r, , and P are the half-width, frequency, and ampli-
as
tude of the line load, respectively. H( • ) is the Heaviside step

⎪ ∂ u0 function, which can be defined as

⎨N(x, t) = EA
∂x
(8) ⎧

⎩M(x, t) = −EI ∂ w0
2
⎪ ⎪
⎨0 x < x0
∂ x2 H(x − x0 ) = 1
x = x0 (11)
⎪2

As shown in Fig. 2, the harmonic line load F(x, t) is divided 1 x > x0
into a normal component F(x, t)cos θ and a tangential compo-
∂ (F(x,t) sin θ )
nent F(x, t)sin θ . Newton’s laws and the equilibrium equation ∂x = 0; thus, by combining Eqs. (2), (3), (8), and (9), the
of bending moment can be used to obtain governing equation becomes
acta mechanica solida sinica 30 (2017) 596–607 599

⎧ 4 ⎧
⎪ ∂ w (x,t) ∂ 2 w (x,t) ∂ u (x,t) hKx

⎪EI ∂ 0x4 + m ∂0t 2 + Kz w0 (x, t) + hK2 x 0∂x ⎪
⎪ B1 = B2 = iξ
⎨ ⎪

2 2
∂ w (x,t) ⎪
⎨ 2

−( h 4Kx + G p ) ∂ 0x2 = F (x, t) cos θ (12)



⎪ h2 Kx (19)

⎩E A ∂ u0 (x,t) − K u (x, t ) + hKx ∂ w0 (x,t) − m ∂ 2 u0 (x,t) = −F (x, t) sin θ
2 ⎪

4
D1 = EIξ − mω + Kz + 2
+ Gp ξ 2
x 0 ∂x ⎪
⎪ 4
∂ x2 2 ∂t2 ⎪

D2 = EAξ 2 + Kx − mω2
The initial conditions of the beam displacement and
velocity are defined as zero, i.e., By solving Eq. (18), we can easily obtain

w0 (x, t)|t=0 = 0 , u0 (x, t) |t=0 = 0 ⎪
⎪ (D2 cos θ − B1 sin θ )F˜ (ξ , ω)

⎨w˜ (ξ , ω) =
∂ w0 (x, t) B1 B2 + D1 D2
∂ u0 (x, t) (20)
|t=0 = 0 , |t=0 = 0 (13) ⎪
⎪ ( B cos θ + D1 sin θ )F˜ (ξ , ω)
∂t ∂t ⎪
⎩u˜ (ξ , ω) =
2
B1 B2 + D1 D2
The beam in the problem is infinite. Thus, the boundary
conditions are given by Eq. (20) is the dynamic response of the beam resting on the
  elastic Pasternak foundation considering horizontal tangen-
∂ n w0 (x, t)  ∂ n u0 (x, t)  tial effect under harmonic line loads in the frequency domain.
lim t=0 = 0, lim t=0 = 0(n = 0, 1, 2, 3)
x→±∞ ∂tn  x→±∞ ∂tn 
3.2. Double Fourier inverse transform
(14)

Eqs. (10)–(14) constitute the complete mathematical de- To solve the problem further, we convert the frequency do-
scription of the problem. The dynamic response of the beam main solution to the time domain by double Fourier inverse
under harmonic line loads can be obtained by solving these transform, as shown as follows:
linear partial differential equations in the following section. ⎧  +∞

⎪ 1
⎪w0 (x, t) =



2
(2π ) −∞

⎪  +∞

⎪ ( D cos θ − B1 sin θ )F˜ (ξ , ω) i(ξ x+ωt)

⎪ 2
e dξ dω
3. Integral representation of the solution ⎨
−∞ B1 B2 + D1 D2
 +∞ (21)

⎪ 1

⎪ u (x, t) =
Eq. (10) is a fourth-order linear partial-differential equation ⎪

0 2
(2π ) −∞


that is difficult to solve directly. Hence, we first simplify the ⎪  +∞ (B2 cos θ + D1 sin θ )F˜ (ξ , ω) i(ξ x+ωt)



⎩ e dξ dω
equation into an algebraic equation by integral transfor- −∞ B1 B2 + D1 D2
mation. The double Fourier transform and its inversion are
defined respectively as where D1 , D2 , B1 , andB2 are defined by Eq. (19); F˜ (ξ , ω) is the
double Fourier transform of F(x, t), which can be obtained by
 +∞  +∞
f˜ (ξ , ω) = Four[ f (x, t)] = f (x, t)e[−i(ξ x+ωt)] dxdt (15) substituting Eq. (15) into Eq. (11).
−∞ −∞  
+∞ +∞ H(r2 − x2 )
 F˜ (ξ , ω) = P exp(it)e[−i(ξ x+ωt)] dxdt (22)
1 +∞  +∞ −∞ −∞ 2r
−1
f (x, t) = Four [ f˜ (ξ , ω)] = 2
f˜ (ξ , ω)ei(ξ x+ωt) dξ dω
(2π ) −∞ −∞ Hence, we can easily obtain
(16)  
+∞ H(r2 − x2 ) (−iξ x ) +r e−iξ x sin(ξ r )
e dx = dx = (23)
where Four[ • ] and Four − 1 [
• ] denote double Fourier transform −∞ 2r −r 2r ξr
and its inversion, respectively.  +∞
e[i(−ω)t] dt = 2π δ( − ω) (24)
3.1. Double Fourier transform −∞

By substituting Eqs. (23) and (24) into Eq. (22), we can


By applying the double Fourier transform to both sides of obtain
Eq. (12) and considering the boundary conditions shown in
sin(ξ r )
Eq. (14), we can obtain F˜ (ξ , ω) = 2π P δ( − ω) (25)
ξr
⎧


⎪ h2 Kx hKx Substituting Eq. (25) into Eq. (21) yields
⎪ 4
⎪ EIξ − mω + Kz +
2
+ Gp ξ 2 w ˜ (ξ , ω) + iξ u˜ (ξ , ω)

⎪ 4 2
⎨ ⎧ 
= F˜ (ξ , ω) cos θ (17) ⎪
⎪ P exp(it) +∞ sin(ξ r )(D2 cos θ − B1 sin θ )ω= iξ x
⎪ ⎪w0 (x, t) =
⎨ e dξ

⎪ 2π −∞ ξ r(D1 D2 + B1 B2 )ω=

⎪  +∞

⎩(EAξ 2 + Kx − mω2 )u˜ (ξ , ω) − hKx iξ w ⎪ P exp(it) sin(ξ r )(B2 cos θ + D1 sin θ )ω= iξ x
˜ (ξ , ω) = F˜ (ξ , ω) sin θ ⎪

2 ⎩u0 (x, t) = e dξ
2π −∞ ξ r(D1 D2 + B1 B2 )ω=
where F (x, t) sin θ. For convenience, Eq. (17) can be abbreviated (26)
as
⎧ where (D2 cos θ − B1 sin θ )ω =  and (D1 D2 + B1 B2 )ω =  represent
⎨D1 w˜ (ξ , ω) + B1 u˜ (ξ , ω) = F˜ (ξ , ω) cos θ the replacement of ω with  in these formulas. Next, we
. (18)
⎩D u˜ (ξ , ω) − B w ˜ determine the integral representation of the solution in the
2 2 ˜ (ξ , ω) = F (ξ , ω) sin θ
time domain. When the harmonic line load becomes a con-
where centrated point load Fpoint (x, t) = Pδ(x)exp (it), the dynamic
600 acta mechanica solida sinica 30 (2017) 596–607

response in this case can be obtained by taking the limit on The root distribution of Eq. (32) is affected by discriminant
both sides of Eq. (26) as , which can be expressed as follows:
⎧ 

⎪ P exp(it) +∞ (D2 cos θ − B1 sin θ )ω= iξ x = (q/2)2 + (p/3)3 (34)
⎨wpoint (x, t) =

2π (D1 D2 + B1 B2 )ω=
e dξ
−∞
 +∞

⎪ P exp(it) (B2 cos θ + D1 sin θ )ω= iξ x We divide our discussion into three cases, namely, > 0,

⎩upoint (x, t) = e dξ
2π −∞ (D1 D2 + B1 B2 )ω= = 0, and < 0.
(27)
(1) > 0
In deriving Eq. (27), the following limit is used as
In the case where > 0, one real root and two conjugate
sin(rξ ) symmetric complex roots exist. The roots of Eq. (32) can be
lim =1 (28)
r→0 rξ obtained by
Eqs. (26) and (27) can be further calculated by employing ⎧ √ √

⎪ y1 = 3 −q/2 + + 3 −q/2 −
the theorem of residue in complex function theory, which ⎪

3
√ 2 3

⎪y2 = w −q/2 + + w −q/2 −
is addressed in the following section. Only the real part of (35)

the integrand must be considered if Eqs. (26) and (27) are the ⎩y = w2 3 −q/2 + √ + w 3 −q/2 − √

3
complex functions.
√ √
where w = (−1 + i 3 )/2, and w2 = (−1 − i 3 )/2. Here, y1 is the
real root.
4. Closed-form representation of the solution
(2) = 0
4.1. Roots of the characteristic equation (a) p = q = 0

In this case, all three roots of Eq. (32) are identical to zero.
Investigating the distribution of the roots of the characteristic
Therefore, we can obtain
equation of the type ξ (D1 D2 + B1 B2 )ω =  = 0 is necessary before
further calculating the integration of Eq. (26). Here, ξ = 0 is a y1 = y2 = y3 = 0 (36)
real root of this algebraic equation. Thus, we only discuss the
roots of the equation (b) (q/2)2 = −(p/3)3 = 0

(D1 D2 + B1 B2 )ω= = 0 (29) In this case, three real roots exist, and two of these are
identical. These roots can be expressed as
Substituting Eq. (19) into Eq. (29) with simplification yields

y1 = 2(−q/2)1/3
a1 X 3 + a2 X 2 + a3 X + a4 = 0 (30) (37)
y2 = y3 = (q/2)1/3
where
⎧ where y2 and y3 are both double roots.

⎪ X = ξ2



⎪ (3) < 0

⎪a1 = E 2 AI




⎨ In the case where < 0, three real roots exist, which differ
h2 Kx
a2 = (Kx − m2 )EI + + G p EA (31) from one another. By means of the triangular functions, the

⎪ 4


roots are given respectively as

⎪ 2

⎪ 2 2 h Kx
⎪ 3
⎪ a = ( Kz − m  ) EA − m  + G p + Kx G p ⎧

⎪ 4 ⎪ 1/3

⎩ ⎨y1 = 2R cos α
a4 = (Kx − m2 )(Kz − m2 ) y2 = 2R1/3 cos(α + 2π /3) (38)


y3 = 2R1/3 cos(α + 4π /3)
Here, we assume that none of the parameters Kx , Kz , m, and
 is of zero value. Characteristic equation (30) is a third-order
where R = ( − p/3)3/2 and α = 13 arccos(−q/2R ).
algebraic equation with parameters depending on the load,
Thus far, we obtain the roots of Eq. (32) for all three cases.
the foundation, and the beam. The distribution of the roots of
By using Eqs. (31) and (33), the distribution of the roots of the
this equation depends on the distribution and combination of
characteristic equation ξ (D1 D2 + B1 B2 )ω =  = 0 can be obtained.
these parameters. According to the algebraic equation theory,
Eq. (30) can be written as
4.2. Closed-form representation of the solution
y3 + py + q = 0 (32)
According to the symmetric property of the problem, the
where deflection at the position of x > 0 is identical to those of x < 0.
⎧ a2 Without loss of generality, we only consider the case where

⎪y = X + 3 a

⎪ x ≥ 0. For each case, a real root at ξ = 0 exists. Thus, the integral


1
⎨ a3 a2 2
p= − of Eq. (26) is in the sense of the Cauchy principal value (p.v.) of
(33)

⎪ a1 3a1 2 the integration. According to the theorem of residue, proving



⎪ 2 a 3 a2 a3 a4
⎩q = 2
− + that the contribution of the indicated semi-circular path
27a1 3 3a1 2 a1 vanishes as its radius approaches infinity is a straightforward
acta mechanica solida sinica 30 (2017) 596–607 601

process. In the case where x ≥ 0, the improper integral of Eq. contribute to the integral. Based on the above conclusion, Eq.
(26) can be rewritten as follows: (40) in this scenario can be expressed as
 +∞
sin(ξ r )(D2 cos θ − B1 sin θ )ω= iξ x ⎧ ⎧⎧ ⎫ ⎧ ⎫⎫
p.v. e dξ ⎪ ⎨⎨  T (ξ ) ⎬ ⎨  T (ξ ) ⎬⎬
−∞ ξ r(D1 D2 + B1 B2 )ω= ⎪
⎪ iP exp (i t) w w
  ⎪
⎪ w (x, t) = 2 +
 sin(ξ r )(D2 cos θ − B1 sin θ )ω= iξ x ⎨ 0 2r ⎩⎩ Q  (ξ ) ⎭ ⎩ Q  (ξ ) ⎭⎭
= 2π i Res e ⎧⎧ Imξ >0 ⎫ ⎧ Imξ =0 ⎫⎫ (42)
ξ r(D1 D2 + B1 B2 )ω= ⎪
⎪ iP exp(it) ⎨⎨  Tu (ξ ) ⎬ ⎨  Tu (ξ ) ⎬⎬
Imξ >0 ⎪

   ⎩u0 (x, t) =

2r ⎩⎩
2 +
Q  (ξ ) ⎭ ⎩ Q  (ξ ) ⎭⎭
sin(ξ r )(D2 cos θ − B1 sin θ )ω= iξ x Imξ >0 Imξ =0
+π i Res e
ξ r(D1 D2 + B1 B2 )ω=
Imξ =0
 +∞ Eq. (42) is the closed-form solution of an infinite Euler–
sin(ξ r )(B2 cos θ + D1 sin θ )ω= iξ x
p.v. e dξ Bernoulli beam resting on Pasternak foundation considering
−∞ ξ r(D1 D2 + B1 B2 )ω=
   horizontal tangential effect under harmonic line loads in the
sin(ξ r )(B2 cos θ + D1 sin θ )ω= iξ x
= 2π i Res e cases where > 0 and < 0.
ξ r(D1 D2 + B1 B2 )ω=
Imξ >0 In the case where = 0, the first- and second-order
  
sin(ξ r )(B2 cos θ + D1 sin θ )ω= iξ x poles are located in the complex plane. The residue of the
+π i Res e , (39)
ξ r(D1 D2 + B1 B2 )ω= first-order poles can be obtained by Eq. (41), and that of the
Imξ =0
second-order poles of Eq. (40) can be evaluated as
where p.v. pertains to the principal value of the integration;
Res[·] denotes the residue of the function inside the paren- ⎧  
sin(ξ r ) (D2 cos θ−B1 sin θ )ω= iξ x

⎪ Res e , ξi
thesis; and Im[ξ ] represents the imaginary parts of complex ⎪
⎪ ξ ( D D
1 2 + B B )
1 2 ω= 


variable ξ , which are the isolated singularities of the complex ⎨= limξ →ξ d (ξ − ξi )2 sin(ξ r )(D2 cos θ−B1 sin θ )ω= eiξ x
i dξ ξ (D1 D2 +B1 B2 )ω=
sin(ξ r ) (D2 cos θ−B1 sin θ )ω= iξ x   (43)
function z(ξ ) = ξ r(D1 D2 +B1 B2 )ω=
e . In the case where ⎪
⎪ Res
sin(ξ r ) (D2 cos θ−B1 sin θ )ω= iξ x
e , ξi

⎪ ξ ( D D
1 2 + B B )
x > 0, the first summation term of the right-hand side of ⎪

1 2 ω= 
⎩= lim d
(ξ − ξ )
2 sin(ξ r ) (D2 cos θ−B1 sin θ )ω= iξ x
e
Eq. (39) pertains to the sum of residues of complex function ξ → ξi d ξ i ξ (D1 D2 +B1 B2 )ω=

z(ξ ) in isolated singularities, which is in the upper complex


ξ -plane. The second summation term denotes the sum of the where ξ i represents the second-order poles. As previously
residues, the isolated singularities of which are located on the mentioned, the residue of ξ = 0 is zero. Substituting Eqs.
(41) and (43) into Eq. (40) thus yields
real axis. By substituting Eq. (39) into Eq. (26), we can obtain
the generic form of Eq. (26). ⎧ ⎧
⎪ ⎪

⎧     ⎪
⎪ ⎪
⎪ ⎧ ⎫
iP exp(it) sin(ξ r ) (D2 cos θ−B1 sin θ )ω= iξ x ⎪
⎪ ⎪

⎪ w0 (x, t) = 2 Imξ >0 Res e ⎪
⎪ ⎨⎨  T (ξ )  Tw (ξ ) ⎬

⎪ 2 r ξ ( D D + B B ) ⎪
⎪ iP exp (i t) w

⎪   1 2 1 2 ω=  ⎪
⎪ w 0 (x, t) =
⎪ ⎩
2 +

⎨ sin(ξ r ) (D2 cos θ−B1 sin θ )ω= iξ x ⎪

2r
⎪ Q  (ξ ) Q  (ξ ) ⎭
+ Imξ =0 Res ξ (D1 D2 +B1 B2 )ω=
e ⎪





Imξ >0 Imξ =0
    ⎪
⎪ ⎪
⎩   
⎪ iP exp(it) sin(ξ r ) (B2 cos θ+D1 sin θ )ω= iξ x ⎪

⎪ u (x, t) = 2 Res e ⎪
⎪ first order poles ⎫

⎪ 0 2 r Imξ > 0 ξ ( D D + B B ) ⎪
⎪ ⎪

⎪   1 2 1 2 ω=  ⎪
⎪ ⎪

⎩ sin(ξ r ) (B2 cos θ+D1 sin θ )ω= iξ x ⎪
⎪ ⎧ ⎫ ⎪
+ Imξ =0 Res ξ ( D D +B B )
e ⎪
⎪ ⎨ 
⎪     ⎬⎪⎪

1 2 1 2 ω= ⎪
⎪ d 2 Tw (ξ ) d 2 Tw (ξ )

⎪ + 2 lim (ξ − ξi ) + lim (ξ − ξi )
(40) ⎪
⎪ ⎩ ξ → ξi d ξ Q(ξ ) ξ → ξi d ξ Q(ξ ) ⎭ ⎪


⎪ Imξ >0 Imξ =0 ⎪


⎪   ⎪


⎨ ⎭
To further calculate Eq. (40), the residues in the equation at ⎧ second order poles
.

⎪ ⎪

different locations of the complex plane must be represented. ⎪
⎪ ⎪⎧
⎪ ⎫

⎪ ⎪


⎪ ⎨⎨  T (ξ )  Tu (ξ ) ⎬
The isolated singularities of complex function z(ξ ) should be ⎪
⎪ iP exp(it) u
⎪u0 (x, t) =

⎪ 2r ⎪ ⎩
2  (ξ )
+
Q  (ξ ) ⎭
determined. In the case where Kx = 0, the numerator of z(ξ ) ⎪




⎪ Imξ >0
Q
Imξ =0

⎪ ⎪
⎪  
is not equal to zero, whereas (D1 D2 + B1 B2 )ω =  = 0. The roots ⎪
⎪ ⎩

⎪ first order poles ⎫


of the characteristic equation are the poles of z(ξ ). When ⎪

⎪ ⎪


⎪ ⎧ ⎫ ⎪

ξ = 0, the numerator of z(ξ ) becomes zero. ξ = 0 is the moving ⎪

⎪ ⎨      ⎬⎪⎪


⎪ d 2 Tu (ξ ) d 2 Tu (ξ )
singularity. The closed-form representation of Eq. (40) can be ⎪
⎪ + 2 lim (ξ − ξ ) + lim (ξ − ξ )

⎪ ⎩ ξ → ξi d ξ
i
Q(ξ ) ξ → ξi d ξ
i
Q(ξ ) ⎭ ⎪


⎪ Imξ >0 Imξ =0 ⎪

obtained using the theory of complex function. ⎪
⎪ ⎪
⎩   ⎪

In the cases where > 0 and < 0, all poles in the complex second order poles

plane are of the first order. The residue of the first-order poles (44)
of Eq. (40) can be evaluated as
⎧   Eq. (44) is the closed-form solution of the infinite Euler–

⎪ sin(ξ r )(D2 cos θ − B1 sin θ )ω= iξ x Tw (ξ )

⎨Res e , ξi =  Bernoulli beam resting on Pasternak foundation considering
ξ (D1 D2 + B1 B2 )ω= Q (ξ )
  (41) the tangential interaction between the beam and the founda-

⎪ sin(ξ r )(B2 cos θ + D1 sin θ )ω= iξ x Tu (ξ )
⎪Res
⎩ e , ξi =  tion under harmonic line loads in the case where = 0.
ξ (D1 D2 + B1 B2 )ω= Q (ξ )
We can obtain the closed-form solutions corresponding
where Tw (ξ ) = sin (ξ r)(D2 cos θ − B1 sin θ )ω =  exp (iξ x), to all the different cases. The velocity response v(x, t) and
Q(ξ ) = ξ (D1 D2 + B1 B2 )ω =  , Tu (ξ ) = sin (ξ r)(B2 cos θ + acceleration response a(x, t) of the beam can be obtained by
D1 sin θ )ω =  exp (iξ x), and Q (ξ ) is the first-order derivative of using the first and second derivatives of the deflection re-
Q(ξ ). We determine the residue of the first-order poles. sponse with respect to time domain t. The dynamic response
Given that ξ = 0 is the moving singularity, we derive the of a concentrated point load can be obtained by applying the
residue of ξ = 0 as zero. This singularity actually does not same method to Eq. (27).
602 acta mechanica solida sinica 30 (2017) 596–607

Table 1 – Closed-form deflection of the degraded solution (C = 0, x ≥ 0).

Frequency Deflection y(x, t)


2 
4

4
 < 0 (4rEIKˆ )−1 iP exp(it) exp(iξ j x ) sinrξ j where ξ1 = Kˆ eiπ/4 , ξ2 = Kˆ ei3π/4
j=1
 = 0 D oes not exist

2 
4

4
 > 0 (4rEIKˆ )−1 iP exp(it) exp(iξ j x ) sinrξ j where ξ1 = i Kˆ , ξ2 = −i Kˆ
j=1

(3)  > 0
5. Degraded solution
Eq. (48) becomes ξ 4− Kˆ = 0. The fours roots of this equation
4
To verify the validation of the proposed solution, a special are expressed as ξ = Kˆ exp[i( jπ /2)] ( j = 0, 1, 2, 3) .
case where Kx = 0 and θ = 0 is considered. In this scenario, Based on the above discussion, all poles in the complex
Eq. (19) becomes plane are of the first order. Eq. (44) can be expressed as
⎧ ⎧ ⎫

⎨B 1 = B 2 = 0 iP exp(it) ⎨  Ts (ξs )  Ts (ξs ) ⎬
D1 = EIξ 4 + Kz − mω2 + G p ξ 2 (45) ws (x, t) = 2 + (49)
⎪ 2r ⎩ Q  s (ξs ) Q  s (ξs ) ⎭
⎩ Imξs >0 Imξs =0
D2 = EAξ 2 − mω2
where Ts (ξ ) = sin (ξ r)(EAξ 2
− m2 )exp (iξ x), Qs (ξ ) =
By substituting Eq. (19) into Eq. (20) and assuming θ = 0,
4 2 2 2
ξ (ξ + K̄ − m̄ )(EAξ − m ). Here, ξ s are the roots of
we obtain
ξ 4 + K̄ − m̄2 = 0, and EAξ 2 − m2 does not become zero.
(EAξ 2 − mω2 )F˜ (ξ , ω) Thus, Eq. (49) can be simplified as
˜ s (ξ , ω) =
w (46)
(EIξ 4 + Kz − mω2 + G p ξ 2 )(EAξ 2 − mω2 ) Eq. (47) can be simplified as

Similarly, Eq. (26) can be expressed as
iP exp(it) ⎨  sin(ξs r ) exp(iξs x )
ws (x, t) = 2
P exp(it) 2r ⎩ 5ξs 4 + K̄ − m̄2
ws (x, t) = Imξs >0
2π EI ⎫
 +∞  sin(ξs r ) exp(iξs x ) ⎬
sin(ξ r )(EAξ 2 − m2 )
× eiξ x dξ (47) + (50)
−∞ ξ r (ξ 4 + K̄ − m̄2 + Ḡ p ξ 2 )(EAξ 2 − m2 ) 5ξs 4 + K̄ − m̄2 ⎭
Imξs =0
 +∞
where K̄ = Kz /E I, m̄ = m/E I and Ḡ p = Ḡ p /EI. The roots P exp(it) sin(ξ r )
ws (x, t) = eiξ x dξ (51)
of ξ 4 + K̄ − m̄2 + Ḡ p ξ 2 = 0, EAξ 2 − m2 = 0, and ξ 2π EI −∞ ξ r(ξ 4 + K̄ − m̄2 )
= 0 are the isolated singularities of complex function
sin(ξ r )(EAξ 2 −m2 ) Eq. (51) is consistent with Eq. (17), where the foundation is
zs (ξ ) = eiξ x . Similar to the case
ξ r(ξ 4 +K̄−m̄2 +Ḡ p ξ 2 )(EAξ 2 −m2 ) considered a Winkler foundation in the case where damp C =
where ξ = 0, as previously mentioned, the numerator of zs (ξ ) 0. On the basis of the aforementioned analysis, we obtain the
2
becomes zero when ξ 2 = mEA . Hence, the roots of EAξ 2 − m2 closed-form solution for this special case. The final results
= 0 are also a moving singularity, the residue of which be- are provided in Table 1. The results are consistent with those
come zero. The singularity actually does not contribute to the in a previous study [13]. Therefore, the proposed closed-form
integral as ξ = 0. Moreover, the roots of ξ 4 + K̄ − m̄2 + Ḡ p ξ 2 = 0 solution of a beam resting on Pasternak foundation consid-
are the poles of zs (ξ ). ering the tangential interaction between the beam and the
As mentioned in Section 4.1, the distribution of the roots foundation under harmonic line loads is validated.
of ξ 4 + K̄ − m̄2 + Ḡ p ξ 2 = 0 should be investigated. Assuming
that Gp = 0, ξ 4 + K̄ − m̄2 + Ḡ p ξ 2 = 0 can be simplified as
6. Parameter study
ξ 4 + K̄ − m̄2 = 0 (48)

We define the equivalent stiffness Kˆ = |K̄ − m̄2 | and res- Based on the analytical solutions presented in the previous
onance frequency 0 = K/m. Next, we divide our discussion sections, several numerical examples are calculated to con-
into three cases, as described as follows. sider the effects of the angle of incidence θ and shear stiffness
Gp on beam response. A rectangular infinite beam with 3-m
(1)  < 0 width and 0.2-m height is considered. The midpoint of the
harmonic line load is at the origin of the coordinate system.
Eq. (48) becomes ξ 4+ Kˆ = 0. The fours roots of this equation
4 The parameters of the harmonic line load are as follows:
are expressed as ξ = Kˆ exp[i(1 + 2 j)π /4] ( j = 0, 1, 2, 3) . amplitude P= 21 kN, circular frequency  = 4π rad/s, and
(2)  = 0 the half length of the load r= 1 m. The beam is considered
linear elastic, homogeneous, and isotropic with density ρ =
Eq. (48) becomes ξ 4 = 0. All fours roots of this equation are 7850 kg/m3 and Young’s modulus E= 2.1 × 1011 Pa. The
zero. foundation is considered linear elastic with spring stiffness
acta mechanica solida sinica 30 (2017) 596–607 603

Fig. 3 – Comparison diagram of the vertical displacement responses at the position of x = 0 of the beam under harmonic line
load.

Kz = 3.5 × 107 N/m2 . To consider the effects of the angle crease of the angle of incidence and ultimately becomes
of incidence θ and shear stiffness Gp on beam response, zero when θ = π /2, which is similar to the vertical velocity
the controlling variable method is employed in all three response.
cases, as described in the next sections. To compare the
numerical solution with the original analytical solution for 6.2. Shear stiffness
validation, we let angle of incidence θ , shear stiffness Gp ,
and horizontal spring stiffness Kx as π /4, 107 N/m2 , and 3.5 In this case, the angle of incidence θ = π /4, and horizontal
× 106 N/m2 , respectively. Fig. 3 indicates the responses of the spring stiffness Kx = 3.5 × 106 N/m2 . Figs. 7, 8 and 9 show
vertical velocity of the beam. The figure indicates that the the steady-state responses of the beam’s vertical velocity,
numerical solution is consistent with the original analytic horizontal velocity, and bending moment, respectively, with
solution, which further validates the proposed closed-form six different shear stiffness values (i.e., Gp = 0, 106 , 107 , 108 ,
solution. and 109 N/m2 ). The observation point is set at the origin of the
coordinate system x = 0, as shown above. The shear stiffness
6.1. Angle of incidence has a dramatic effect on the vertical velocity and the bending
moment responses of the beam, as indicated in Figs. 7 and 9.
In this case, shear stiffness Gp = 106 N/m2 , and horizontal In the case where Gp = 0, 106 , 107 N/m2 , the vertical velocity
spring stiffness Kx = 3.5 × 106 N/m2 . Figs. 4, 5, and 6 indicate responses of the beam are similar to one another, which is the
the steady-state responses of the beam’s vertical velocity, hor- same as the bending moment response. However, when shear
izontal velocity, and bending moment, respectively, with five stiffness Gp is no less than 108 N/m2 , the vertical velocity
different angles of incidence (i.e., θ = 0, π /6, π /4, π /3, π /2). The and the bending moment of the beam significantly decrease
observation point is set at the origin of the coordinate system with the increase of shear stiffness. By contrast, Fig. 8 shows
x = 0. Figs. 4 and 5 show that the angle of incidence sig- that the horizontal velocity responses are irrelevant to shear
nificantly influences the vertical and horizontal velocity stiffness Gp .
responses of the beam. The vertical velocity response
of the beam decreases with the increase of the angle
of incidence and ultimately becomes zero when θ = π /2. 7. Conclusions
The horizontal velocity response is opposite to the ver-
tical velocity response. This result is consistent with the The closed-form solution of the dynamic response of an
above operation, which divides the harmonic line load infinite Euler–Bernoulli beam on Pasternak foundation con-
F(x, t) into a normal component F(x, t)cos θ and a tangen- sidering the tangential interaction between the bottom of
tial component F(x, t)sin θ . Fig. 6 shows that the bending the beam and the foundation under harmonic line loads is
moment response of the beam also decreases with the in- presented in this study. Double Fourier transform is used
604 acta mechanica solida sinica 30 (2017) 596–607

Fig. 4 – Horizontal velocity responses at the position of x = 0 of the beam under a harmonic line load (with θ changed).

Fig. 5 – Vertical velocity responses at the position of x = 0 of the beam under a harmonic line load (with θ changed).
acta mechanica solida sinica 30 (2017) 596–607 605

Fig. 6 – Bending moment responses at the position of x = 0 of the beam under a harmonic line load (with θ changed).

Fig. 7 – Vertical velocity responses at the position of x = 0 of the beam under a harmonic line load (with Gp changed).
606 acta mechanica solida sinica 30 (2017) 596–607

Fig. 8 – Horizontal velocity responses at the position of x = 0 of the beam under a harmonic line load (with Gp changed).

Fig. 9 – Bending moment responses at the position of x = 0 of the beam under a harmonic line load (with Gp changed).
acta mechanica solida sinica 30 (2017) 596–607 607

to simplify the governing equation, thereby obtaining the [4] H.T. Yu, Y. Yuan, A. Bobet, Multiscale method for long tunnels
dynamic response of the beam easily in the frequency subjected to seismic loading, Int. J. Numer. Anal. Methods 37
domain. Double Fourier inverse transform is employed to (4) (2013) 374–398.
[5] G.S. Payette, J.N. Reddy, Nonlinear quasi-static finite element
convert the frequency domain solution into the time do-
formulations for viscoelastic Euler-Bernoulli and
main, and theorem of residue is employed to calculate the
Timoshenko beams, Int. J. Numer. Methods Biomed. 26 (12)
improper integral. Hence, the closed-form solution is deter- (2010) 1736–1755.
mined. The presented solution is validated by comparing [6] Y.X. Wu, Y.F. Gao, Analytical solutions for simply supported
the degraded solution with the known results reported in viscously damped double-beam system under moving
literature. harmonic loads, J. Eng. Mech.–ASCE 141 (7) (2015) 04015004.
Several numerical examples are used to probe the effects [7] H.T. Yu, Y. Yuan, Z.Z. Qiao, Y. Gu, Z.H. Yang, X.D. Li, Seismic
analysis of a long tunnel based on multi-scale method, Eng.
of the angle of incidence θ and shear stiffness Gp on beam
Struct. 49 (2013) 572–587.
response. The controlling variable method is employed for [8] J.T. Kenney, Steady-state vibrations of beam on elastic
this purpose. The principal conclusions of these numerical foundation for moving load, J. Appl. Mech. 21 (4) (1954)
computations are summarized as follows. 359–364.
[9] M.A. Biot, Bending of an infinite beam on an elastic
(1) The angle of incidence significantly influences the ver- foundation, J. Appl. Mech. 4 (1) (1937) A1–A7.
tical velocity, horizontal velocity, and the bending mo- [10] P.M. Mathews, Vibration of a beam on elastic foundation, J.
ment responses of the beam. The vertical velocity and Appl. Math. Mech. 38 (3–4) (1958) 105–115.
the bending moment responses of the beam decrease [11] P.M. Mathews, Vibration of a beam on elastic foundation II, J.
Appl. Math. Mech. 39 (1–2) (1959) 13–19.
with the increase of the angle of incidence and ulti-
[12] J.D. Achenbach, C. Sun, Dynamic response of beam on
mately become zero when θ = π /2. The horizontal ve- viscoelastic subgrade, J. Eng. Mech. Div. 91 (5) (1965) 61–76.
locity response shows the opposite effect of the vertical [13] L. Sun, A closed-form solution of a Bernoulli-Euler beam on a
velocity and the bending moment responses. viscoelastic foundation under harmonic line loads, J. Sound
(2) Shear stiffness dramatically affects the vertical veloc- Vib. 242 (4) (2001) 619–627.
ity and bending moment responses of the beam. In the [14] L. Sun, A closed-form solution of beam on viscoelastic
subgrade subjected to moving loads, Comput. Struct. 80 (1)
cases where Gp = 0, 106 , 107 N/m2 , the vertical veloc-
(2002) 1–8.
ity responses of the beam are similar to one another,
[15] L. Sun, An explicit representation of steady state response of
which is the same as the bending moment response. a beam on an elastic foundation to moving harmonic line
When shear stiffness Gp is no less than 108 N/m2 , the loads, Int. J. Numer. Anal. Methods 27 (1) (2003) 69–84.
responses significantly decrease with the increase of [16] S.M. Kim, J.M. Roesset, Dynamic response of a beam on a
shear stiffness. By contrast, the horizontal velocity re- frequency-independent damped elastic foundation to
sponse is irrelevant to shear stiffness Gp . moving load, Can. J. Civil Eng. 30 (2) (2003) 460–467.
[17] M. Shamalta, A.V. Metrikin, Analytical study of the dynamic
The presented closed-form solution can be an effective response of an embedded railway track to a moving load,
Arch. Appl. Mech. 73 (2003) 131–146.
tool for engineering practitioners who aim to conduct prefea-
[18] H.T. Yu, Y. Yuan, Analytical solution for an infinite
sibility studies and check the validity of the results obtained
Euler-Bernoulli beam on a viscoelastic foundation subjected
from numerical computations. to arbitrary dynamic loads, J. Eng. Mech.–ASCE 140 (3) (2014)
542–551.
[19] Y. Yuan, H.T. Yu, C. Li, X. Yan, J.Y. Yuan, Multi-point shaking
Acknowledgments table test for long tunnels subjected to non-uniform seismic
loadings – Part I: theory and validation, Soil Dyn. Earthquake
Eng. (2017), doi:10.1016/j.soildyn.2016.08.017.
This study was financially supported by the National
[20] H.T. Yu, Y. Yuan, G.P. Xu, Q.K. Su, X. Yan, C. Li, Multi-point
Key Research and Development Program of China (no. shaking table test for long tunnels subjected to non-uniform
2016YFC0800206) and the National Natural Science Founda- seismic loadings – Part II: application to the HZM immersed
tion of China (nos. 51778260, 51378234, 51678465). tunnel, Soil Dyn. Earthquake Eng. (2017),
doi:10.1016/j.soildyn.2016.08.018.
[21] D. Kerr A, Elastic and viscoelastic foundation models, Int. J.
references Appl. Mech. 31 (3) (1964) 23–32.
[22] M.H. Karganovin, D. Younesian, Dynamics of Timoshenko
beams on Pasternak foundation under moving load, Mech.
Res. Commun. 31 (2004) 713–723.
[1] H.T. Yu, C. Cai, X.F. Guan, Y. Yuan, Analytical solution for long [23] R.U.A. Uzzal, R.B. Bhat, W. Ahmed, Dynamic response of a
lined tunnels subjected to travelling loads, Tunn. Undergr. beam subjected to moving load and moving mass supported
Sp. Tech. 58 (2016) 209–215. by Pasternak foundation, Shock Vib. 19 (2012) 205–220.
[2] L. Andersen, S.R.K. Nielsen, P.H. Kirkegaard, Finite element [24] D. Basu, N.S.V.K. Rao, Analytical solutions for Euler-Bernoulli
modeling of infinite Euler beams on Kelvin foundations beam on visco-elastic foundation subjected to moving load,
exposed to moving loads in convected coordinates, J. Sound. Int. J. Numer. Anal. Methods 37 (8) (2012) 945–960.
Vib. 241 (4) (2001) 587–604. [25] H.T. Yu, C. Cai, Y. Yuan, M.C. Jia, Analytical solutions for
[3] T.P. Chang, Y.N. Liu, Dynamic finite element analysis of a Euler–Bernoulli beam on Pasternak foundation subjected to
nonlinear beam subjected to a moving load, Int. J. Solids arbitrary dynamic loads, Int. J. Numer. Anal. Methods 41 (8)
Struct. 33 (12) (1996) 1673–1688. (2017) 1125–1137.

You might also like