Professional Documents
Culture Documents
Vertical Elastic Dynamic Impedance of A Large Diameter and Thin-Walled Cylindrical Shell Type Foundation
Vertical Elastic Dynamic Impedance of A Large Diameter and Thin-Walled Cylindrical Shell Type Foundation
A R T I C L E I N F O A BS T RAC T
Keywords: This paper studies the vertical vibration of a large diameter and thin-walled cylindrical shell type foundation
Offshore wind turbine embedded in a fully saturated porous seabed in contact with a seawater half-space. The solution of the coupled
Dynamic impedance fluid-shell foundation-soil vibration problem is obtained using the ring-load Green's functions for both the shell
Elastic shell and the layered fluid-seabed half-space. By considering the fully coupled boundary conditions at the shell-soil
Poroelastic seabed
interface, the shell vibration problem is reduced to Fredholm integral equations. Through an analysis of the
Vertical vibration
corresponding Cauchy singular equations, the intrinsic singular characteristics of the problem are rendered
explicit. With the singularities clear, an effective numerical method involving the Gauss-Chebyshev method is
developed to solve the governing Fredholm equations. Selected numerical results for the dynamic contact load
distributions, displacements, and dynamic impedance functions are examined based on different shell lengths,
soil materials, shell properties, and frequencies of excitation. Moreover, the results are analysed for cases in
which there is and is no fluid overlying seabed to examine the effect of fluid.
⁎
Corresponding author.
E-mail address: nehzoug@163.com (Z. Guo).
http://dx.doi.org/10.1016/j.soildyn.2017.01.034
Received 27 August 2015; Received in revised form 13 November 2016; Accepted 26 January 2017
Available online 06 February 2017
0267-7261/ © 2017 Elsevier Ltd. All rights reserved.
R. He et al. Soil Dynamics and Earthquake Engineering 95 (2017) 138–152
Fig. 1. (a) The image of a wind turbine in shallow water; (b) The image of a wind turbine in deep sea.
boundary finite element method. However, it is still not clear about the equations may not applicable and saturated soils that consist of fine to
elastic dynamic impedance of shell type foundations, which has an medium sands may be described by Biot's theory, so we use a sandy
important effect on the natural frequencies of OWTs. For these new seabed in the numerical analysis part. For the shell problem, the
types of foundations for OWTs that engineers can provide estimates coupled water-shell-porous seabed system is treated as a superposition
using the static stiffness of piles [11], the p-y and t-z curves developed of an intact water-seabed half-space and a reduced and constrained
for slender piles [12–16], Novak's simplified method [17,18], experi- shell, whose reduced elastic modulus and mass density are defined in
ments [19,20], or numerical methods [21–25] to calculate the stiffness Section 4. The Green's functions for the thin shell and the layered
of OWT foundations. However, Versteijlen et al. [26] noted that “Over seawater-seabed half-space are used to formulate the governing
the recent years, measured natural frequencies of installed OWTs have Fredholm equations by considering appropriate boundary conditions
been found to be higher than designed for.” This inconsistency is on the contact interface of the shell and the seabed. By analysing the
mainly caused by the underestimation of the stiffness of OWT founda- corresponding Cauchy kernels, the fundamental singular characteris-
tions. To save costs and make the natural frequency computation more tics of the interfacial reactions acting on the shell are given explicitly.
accurate, a new and more realistic model involving dynamic interaction The Fredholm equations are numerically solved by a piecewise linear
of seawater-shell type foundation-soil is needed. interpolation of the functions of the interfacial reactions using the
Therefore, to guarantee the safety of OWTs, it is critical that Gauss-Chebyshev method. The governing equations for the fluid and
researchers understand how shell type foundations respond to vertical the soil are presented in Section 2. In Section 3, the ring load Green's
dynamic forces. Although there are many papers related to the dynamic functions for the shell and layered fluid-soil half-space are presented.
interaction of traditional slender piles within elastic soil strata due to In Section 4, the coupled dynamic vibration problem is considered. The
vertical dynamic forces [27–32], there are few studies on the vertical numerical results are presented and discussed in Section 5.
vibration of cylindrical shell type foundations. Liingaard et al. [33] Conclusions are presented in Section 6.
studied the impedance of bucket foundations in a viscoelastic medium
using the coupled BEM-FEM method. For a free open-ended cylind- 2. Governing equations
rical shell embedded in soil, Pak and Ji [34] obtained an analytical
result by considering the soil as a single-phase elastic solid. Ji [35] 2.1. Governing equations for the fluid
studied the vertical dynamic vibration of an impermeable shell in soil
using the boundary element method (BEM). However, the shell in a The governing equations for the compressible inviscid fluid can be
real shell type foundation is not free and open-ended but constrained written as [36].
by a top disc or the tower of OWT, and the seafloor is a natural two-
phase medium composed of water and soil skeletons. Thus, the ∂ 2Pw ∂ 2Uw
cw2 ∇2 Pw = , ρw + ∇Pw = 0.
∂t 2 ∂t 2 (1)
constraint of the top ending and the poroelastic effect of the seabed
should be considered. Additionally, He et al. [36] noted that the Considering time-harmonic motion with angular frequency ω , and
seawater overlying the seabed might have a strong effect on the the relation Pw = pw e−iωt , Uw = uw e−iωt , we have [37].
dynamic response of foundations embedded near the water-soil inter-
1
face due to interface waves. It would be reasonable to take the interface ∇2 pw + k w2 pw = 0, uw = ∇pw ,
ρw ω 2 (2)
effect into consideration because shell-type foundations are assembled
very close to the seawater-seabed interface. where pw and uw are the amplitude of the pressure and the displace-
In this study, to extend the rigorous analysis of the dynamic ment of the fluid, respectively. k w = ω / cw is the wave number of the
problem to offshore applications, a coupled seawater-shell-seabed fluid, cw = (λ w / ρw )1/2 is the speed of sound in the fluid, and λ w is the
model is used. In this model (refer to Fig. 2), the seawater half-space compressional modulus of the fluid. Besides, the term e−iωt will be
is described by the Euler equations for a compressible inviscid fluid omitted from all quantities below for simplicity.
[37]; The seabed is modelled as a Biot poroelastic medium [38],
whereas the shell type foundation is modelled using an elastic thin shell 2.2. Governing equations for the shell
theory [39]. As noted by Lin et al. [40], when permeability of soil is less
than 10−6 m/s, relative pore fluid flow is negligible, in which Biot's wave To model the shell type foundation, elastic thin shell theory will be
139
R. He et al. Soil Dynamics and Earthquake Engineering 95 (2017) 138–152
used, and the governing equations for the shell under time-harmonic f1 (z ) = akp e kp z [ cos(kp z ) − sin(kp z )],
axisymmetric loads may be written as [39].
f2 (z ) = akp e−kp z [ cos(kp z ) + sin(kp z )],
d4wr dw f d2wz νp dwr f f3 (z ) = akp e kp z [ cos(kp z ) + sin(kp z )],
+ γp wr + νp γp a z = − r , + = z,
dz 4 dz G dz 2 a dz G1 (3) f4 (z ) = akp e−kp z [ cos(kp z ) − sin(kp z )],
f5 (z ) = 0, f6 (z ) = −aνp, g1 (z ) = −νp e kp z cos(kp z ),
where
g2 (z ) = νp e−kp z cos(kp z ), g3 (z ) = −νp e kp z sin(kp z ),
γp = 12/ a2h2 , fr = pr − ρp hω 2wr , fz = pz − ρp hω 2wz , 3(1 − νp2)
g4 (z ) = −νp e−kp z sin(kp z ), g5 (z ) = 1, g6 (z ) = z, kp4 = . (7)
G= μp h3 /6(1−νp ), G1 = 2μp h /(1−νp ), a2h2
(4)
and wr and wz are the displacements of the shell, μp and νp are the shear
2.3. Governing equations for soil
modulus and Poisson's ratio of the shell, respectively; pr and pz are the
resultants of the distributed radial and vertical contact stresses acting
When time-harmonic motion with an angular frequency ω is
on the surface of the shell, respectively; a is the mid-surface radius, l is
considered, using Biot's theory for saturated soil, and the theory of
the length, and h is the thickness of the shell with h < < a ; and ρp is the
potential, we can write the displacements of the solid part and pore
density of the shell.
pressure of the seabed as [37]
The stress-resultants per unit length in the vertical, angular,
bending, and shear directions acting on an infinitesimal shell element 1
ur = φf , r + φs, r + ψ, rz, uz = φf , z + φs, z − r ψ, r − ψ, rr ,
are expressed as [39]
ρ∼22 ω 1 1
pf = [Af (φf , rr + r φf , r + φf , zz ) + As (φs, rr + r φs, r + φs, zz )],
φ (8)
∂ 2w ∂ 3w
Nz = G1 (εz + νp εθ ), Nθ = G1 (εθ + νp εz ), Mz = G 2r , Qz = G 3r
∂z ∂z (5)
Table 1
Mechanical properties of the shell and soil.
The solution to (3) can be written as [34]
Material E (Pa) ρ (kg/m3) ν νu φ B k (m/s)
6 6
wr (z; c ) = ∑i =1 Ci fi (z ), wz (z; c ) = ∑i =1 Ci gi (z ), 0 ≤ z < c,
6 6 Steel 2e11 7800 0.3
wr (z; c ) = ∑i =1 Pi fi (z ), wz (z; c ) = ∑i =1 Pi gi (z ), c < z ≤ l , (6) Concrete 3e10 2500 0.2
Loose sand 1.5e7 1600 0.25 0.495 0.5 1.0 e-3
where, Dense sand 6e7 1900 0.35 0.495 0.4 1.0 e-4
140
R. He et al. Soil Dynamics and Earthquake Engineering 95 (2017) 138–152
Fig. 3. (a) Comparison of the vertical dynamic contact load distribution given in Pak and Ji [34] versus that obtained in this study; (b) Comparison of the vertical displacement of the
shell given in Pak and Ji [34] versus that obtained in this study; (c) Comparison of the vertical dynamic impedance given in Rajapakse and shah [29] versus that obtained in this study;
(d) Comparison of the vertical dynamic compliance given in Ji [35] versus that obtained in this study.
where ur and uz are the radial and vertical displacements of the solid represented as
part of the seabed, respectively. pf is the pore pressure of the seabed. δ (r − a )
ϕf , ϕs and ψi are three potentials for the fast, slow P wave and S wave, fz (z ) = δ (z − c ), fr (z ) = 0,
2πr (10)
respectively. Af and As are coefficients defined in [37]. The mass
coefficients ρ∼mn is defined as ρ∼mn = ρmn + (−1)m + n ω , and b = ηφ 2 / κ
ib for the vertical ring-load and
represents the resistive damping due to the relative motion between δ (r − a )
fz (z ) = 0, fr (z ) = δ (z − c ),
the solid and the fluid [38], η is the fluid viscosity, κ is Darcy's 2πr (11)
coefficient of permeability. The stress of the soil can be represented as
for the radial ring-load, where δ () is the Dirac's delta function.
[37]
For the shell of a bucket foundation, there are two limiting
σrz = μ (ur , z + uz, r ), boundary conditions for the top ending of the shell
1 R+Q
σz = 2μuz, z + λ (uz, z + r ur + ur , r ) − ϕ R
pf ,
1 R+Q (a) Free top: totally no constraint on the top ending of the shell
σr = 2μur , r + λ (uz, z + r ur + ur , r ) − ϕ R pf , (9)
wz (0) = 0, Mz (0) = 0, Qz (0) = 0, z = 0, (12)
where λ and μ are the Lame constants of the solid part, and φ is the
which means the top of the shell can deform freely.
porosity of the porous medium. σr , σz and σrz are total stresses of the
(b) Fixed top: perfect constraint on the top ending of the shell
seabed. The coupling between the solid and the fluid is characterized by
the two parameters Q and R, which can be calculated from the bulk wz (0) = 0, wr (0) = 0, dwr (0)/dz = 0, z = 0, (13)
moduli of the solid and the fluid [41].
which means the top of the shell cannot deform. The real boundary
conditions will be within the two limiting conditions, e.g. hinge top:
3. Ring load Green's function
wz (0) = 0, wr (0) = 0, Mz (l ) = 0, z = 0, (14)
In this section, Green's function for the shell and the layered fluid-
which means the weld between the shell and the top disc is not so
soil half-space under ring-load will be studied.
strong or a plastic hinge is formed here. We will compare the numerical
results obtained by the three boundary conditions in Section 5.
3.1. Green's functions for a shell The appropriate boundary conditions for the bottom end of the
shell may be written as [34]
When a shell is subjected to an vertical or radial ring-load of unit
total weight at the plane z = c, the boundary conditions can be Nz (l ) = 0, Mz (l ) = 0, Qz (l ) = 0, z = l , (15)
141
R. He et al. Soil Dynamics and Earthquake Engineering 95 (2017) 138–152
Fig. 4. (a) Comparison of the vertical dynamic contact load distribution obtained with three different top ending boundary conditions of the shell; (b) Comparison of the radial dynamic
contact load distribution obtained with three different top ending boundary conditions of the shell; (c) Comparison of the real part of vertical displacement of the shell obtained with
three different top ending boundary conditions; (d) Comparison of the imaginary part of the vertical displacement of the shell obtained with three different top ending boundary
conditions; (e) Comparison of the radial displacement of the shell obtained with three different top ending boundary conditions; (f) Comparison of the real part of the internal axial force
of the shell obtained with three different top ending boundary conditions; (g) Comparison of the imaginary part of the internal axial force of the shell obtained with three different top
ending boundary conditions; (h) Comparison of the real part of the internal shear of the shell obtained with three different top ending boundary conditions; (i) Comparison of the
imaginary part of the internal shear of the shell obtained with three different top ending boundary conditions; (j) Comparison of the K vs of a steel shell embedded in dense sand obtained
with three different top ending boundary conditions.
which are appropriate for friction pile foundations. for the vertical ring-load case, and
From the load conditions and the required continuous conditions at
dwr (c−) dwr (c+)
the plane z = c, the boundary conditions at the plane z = c can be wr (c−) = wr (c + ), wz (c−) = wz (c + ), = ,
dz dz
written as d2wr (c−) d2wr (c+) d3wr (c−) d3wr (c+) 1
= , G −G = , Nz (c + ) − Nz (c−) = 0,
dz 2 dz 2 dz 3 dz 3 2πa
142
R. He et al. Soil Dynamics and Earthquake Engineering 95 (2017) 138–152
Fig. 4. (continued)
shell, the corresponding Green's functions can be obtained easily, and η1 = −Af k f2 ρ∼22 ω / ϕ, η2 = −As ks2 ρ∼22 ω / ϕ,
only the results for the fixed top will be given in Appendix A for
α 2 = ξ 2 − k f2, β 2 = ξ 2 − ks2, γ 2 = ξ 2 − kt2, ϑ 2 = ξ 2 − k w2, (21)
simplicity.
z = c is the plane where the loads act, and k f , ks and kt are complex
wave numbers for the fast, slow P wave and S wave, respectively [37].
3.2. Green's functions for layered fluid-soil half-space The way to obtain the Green's functions for layered fluid-soil half-
space due to vertical and radial ring load is almost the same as that
Through the theory of potential and the governing equations for the given in [37], except that the discontinuous conditions at the plane z
porous medium, the Green's function for the seabed can be written as = c should be replaced by
∞
uz +(r , z ) = ∫ (−αA1 e−α z − βA2 e−β z + ξA3 e−γ z) J0 (ξr ) ξ dξ, δ (r − a )
0 σz (c + )−σz (c − ) = − .
∞ 2πr (22)
ur +(r , z ) = ∫ (−(A1 e−α z + A2 e−β z) ξ + A3 γ e−γ z) J1 (ξr ) ξ dξ,
0
∞ for the vertical ring-load, and
pf + (ξ, z ) = ∫ (η1 A1 e−α z + η2 A2 e−β z) J0 (ξr ) ξ dξ, (18)
0
δ (r − a )
σzr (c + )−σzr (c − ) = − .
for c ≤ z ≤ ∞, and 2πr (23)
143
R. He et al. Soil Dynamics and Earthquake Engineering 95 (2017) 138–152
Fig. 5. (a) The vertical dynamic contact load distribution for a steel shell with l = 2a in various type of soils; (b) The radial dynamic contact load distribution for a steel shell with l = 2a
in various type of soils; (c) Real part of the vertical displacement of the shell for a steel shell with l = 2a in various type of soils; (d) Imaginary part of the vertical displacement of the shell
for a steel shell with l = 2a in various type of soils; (e) The radial displacement of the shell for a steel shell with l = 2a in various type of soils; (f) Real part of the internal axial force of the
shell for a steel shell with l = 2a in various type of soils; (g) Imaginary part of the internal axial force of the shell for a steel shell with l = 2a in various type of soils.
where the variables ki , Δi , etc., appearing in (24) are defined in the have the displacements of the soil under the interfacial forces pr and pz
Appendix B, and the coefficients for the radial ring load of radius a
l l
ur (r , z ) = ∫ uˆrR (r , z; s ) pr (s )ds + ∫ uˆrZ (r , z; s ) pz (s )ds,
0 0
A1 = A5 + A4 e2α c , A2 = A7 + A6 e2β c , A3 = A9 − A8 e2γ c ,
l l
A4 = η2 e−αcJ1 (ξa )/4παμN2 , uz (r , z ) = ∫ uˆzR (r , z; s ) pr (s )ds + ∫ uˆzZ (r , z; s ) pz (s )ds,
0 0
l l
A5 = (Δ1 η2 βξ e−αc−2Δ2 η1 αξ e−βc−2Δ3 k1 k3 α e−γc) J1 (ξa )/(4παβμN2 R1), pf (r , z ) = ∫ pˆ fR (r , z; s ) pr (s )ds + ∫ pˆ fz (r , z; s ) pz (s )ds. (27)
0 0
A6 = η1 ξ e−βcJ1 (ξa )/4πβμN2 , A8 = −k1 e−γcJ1 (ξa )/4πμN2 ,
For the shell, considering Eqs. (3), (4) and (6), the displacements
A7 = (2Δ4 η2 ξβ e−αc−Δ5 η1 αξ e−βc−2Δ6 αk1 k3 e−γc) J1 (ξa )/(−4παβμN2 R1),
caused by the interfacial forces pr and pz can be represented as
A9 = (4Δ7 βη2 ξ 2 e−αc − 4Δ8 αη1 ξ 2 e−βc− Δ9 k1 e−γc) J1 (ξa )/(−4παβμN2 R1),
l l
B1 = (Δ10 βξη2 e−αc + Δ11 η1 αξ e−βc + Δ12 αk1 k3 e−γc) J1 (ξa )/(−2παβμN2 R1). wr (a, z ) = ∫ wˆrR (z; s ) fr (s )ds + ∫ wˆrZ (z; s ) fz (s )ds,
0 0
(25) l l
wz (a, z ) = ∫ wˆzR (z; s ) fr (s )ds + ∫ wˆzZ (z; s ) fz (s )ds + Δz , (28)
0 0
4. The coupled dynamic vibration problem where Δz is the vertical displacement of the top ending of the shell.
In this study, the foundation embedded in the seabed is in full
We now consider the dynamic interaction of the shell and the soil. contact with the seabed, and no slippage and separation between the
For the reduced shell, its elastic modulus and mass density are defined shell and the soil are allowed. For the hydraulic boundary condition
as [42] alongside the shell in detail, the contact shell-soil interface can be
Ep* = Ep − E , ρp* = ρp − ρ , considered fully permeable or impermeable, Halpern and Christiano
(26)
[43] stated that there are negligible differences between the load
where Ep and E are elastic modulus of the shell and the soil, transfer mechanism and the vertical compliances of impermeable and
respectively. In the coupled dynamic vibration problem, the reduced permeable rigid plates on a poroelastic half-space in the low frequen-
modulus and density are used for the shell in the following. The Green's cies (ω < 1.0 ). Similar results can also be found in [37]. So, it is
function obtained in Section 3 will be used, and will be written with a reasonable to assume that the hydraulic boundary condition is not
roof, for example, ŵrR means the Green's function of wr of the shell important in the vertical vibration model. This assumption will be
caused by a radial ring load, other Green's functions used below has a confirmed in Section 5.1 by comparing this study with Ji's work, who
similar definition. For this coupled problem, assume the interfacial considered a totally impermeable shell using BEM [35]. In this paper,
forces on the soil half-space and the reduced shell are pr and pz , we can we do not impose a hydraulic boundary condition on the contact
144
R. He et al. Soil Dynamics and Earthquake Engineering 95 (2017) 138–152
Fig. 5. (continued)
surface following Zeng and Rajapakse [30]. The perfectly bonded From the point view of engineering, we can also define
contact conditions can be written as
Kvs = Fv / Δz , (34)
wz (a, z ) = limr → a ± uz (r , z ), 0 ≤ z ≤ l ,
wr (a, z ) = limr → a ± ur (r , z ), 0 ≤ z ≤ l , as the vertical dynamic impedance for the shell in saturated seabed,
(29)
where Fv is the external force which sustains the shell's vibration, and
Combining the above equations, we can have the following bound- the dimensionless vertical dynamic impedance coefficient
ary conditions for the shell and the soil
K vs = Kvs / μa, (35)
wz (a, z ) = uz (a, z ), 0 ≤ z ≤ l ,
wr (a, z ) = ur (a, z ), 0 ≤ z ≤ l , (30) which can be divided into stiffness and damping ratio
which are Fredholm integral equations for pr and pz , and can be solved K vs = Re(K vs ) − iIm(K vs ) = k − iωc , (36)
numerically.
Following the way of [34], it will be useful to study the correspond-
ing Cauchy singular equations
dwz (a, z ) ∂uz (r , z )
dz
= limr → a ± ∂z
, 0 < z < l,
dwr (a, z ) ∂ur (r , z )
dz
= limr → a ± ∂z , 0 < z < l, (31)
and by the theory of singular integral equations, the singularity of pr
and pz , can be represented as
gz (z )
pz (z ) = , 0 < Re(αz ) < 1, 0 < Re(βz ) < 1,
z αz (l − z ) βz
gr (z )
pr (z ) = , 0 < Re(αr ) < 1, 0 < Re(βr ) < 1,
z αr (l − z ) βr (32)
where αz , βz , αr , and βr satisfy the characteristic equations
145
R. He et al. Soil Dynamics and Earthquake Engineering 95 (2017) 138–152
Fig. 7. (a) The influence of the seawater on Re(K vs ) for a steel shell with various length/radius ratio embedded in dense sand; (b) The influence of the seawater on Im(K vs ) for a steel shell
with various length/radius ratio embedded in dense sand.
where k and c mean stiffness and damping coefficients the foundation linear interpolation of gr and gz with the Gauss-Chebyshev method was
can provide, respectively. written to calculate the value of gr and gz . Then, we can obtain the full
responses of the foundation and K vs by using the corresponding Green's
functions.
5. Numerical results Interpolate functions Ni (z ) satisfy Ni (zi ) = 1 and Ni (zj ) = 0 if j ≠ i ,
which are defined as
When the singularities are known, the Fredholm integral Eq. (30)
can be solved numerically. A computer program using a piecewise
Fig. 8. (a) The difference of Re(K vs ) between a steel shell and a concrete shell embedded in loose sand; (b) The difference of Im(K vs ) between a steel shell and a concrete shell embedded in
loose sand; (c) The difference of Re(K vs ) between a steel shell and a concrete shell embedded in dense sand; (d) The difference of Im(K vs ) between a steel shell and a concrete shell
embedded in dense sand.
146
R. He et al. Soil Dynamics and Earthquake Engineering 95 (2017) 138–152
Fig. 9. (a) K vs for a steel shell with l = a embedded in various types of soils; (b) K vs for a steel shell with l = 2a embedded in various types of soils; (c) K vs for a steel shell with l = 4a
embedded in various types of soils; (d) K vs for a steel shell with l = 10a embedded in various types of soils; (e) K vs for a concrete shell with l = a embedded in various types of soils; (f) K vs
for a concrete shell with l = 2a embedded in various types of soils; (g) K vs for a concrete shell with l = 4a embedded in various types of soils.
⎧ (z − z )/ h1, z1 ≤ z ≤ z2 l
Hi, j = ∫ uˆzz (zi , s ) αz
j N (s ) l
ds, Hi,(n +1)+ j = ∫ uˆzr (zi , s ) αr
jN (s )
ds,
N1 (z ) = ⎨ 2 , 0 s (l − s ) βz 0 s (l − s ) βr
⎩ 0, other
l j N (s )
⎧ (z − zi −1)/ hi −1, zi −1 ≤ z ≤ zi H(n +1)+ i, j = ∫ uˆrz (zi , s ) αz ds ,
s (l − s ) βz
⎪ 0
Ni (z ) = ⎨ (zi +1 − z )/ hi , zi ≤ z ≤ zi +1 , l j N (s )
⎪ 0, H(n +1)+ i,(n +1)+ j = ∫ uˆrr (zi , s ) αr ds ,
⎩ other 0 s (l − s ) βr
l N (s ) l N (s )
⎧ (z − zn )/ hn , zn ≤ z ≤ zn +1 Wi, j = ∫ wˆzz (zi , s ) αz
j
ds, Wi,(n +1)+ j = ∫ wˆzr (zi , s ) αr
j
ds ,
Nn +1 (z ) = ⎨ , 0 s (l − s ) βz 0 s (l − s ) βr
⎩ 0, other (37) l j N (s )
W(n +1)+ i, j = ∫ wˆrz (zi , s ) αz ds,
0 s (l − s ) βz
where,hi = zi +1 − zi , i = 1, 2.... n. l j N (s )
Using functions Ni (z ), continuous functions gz ,gr ,wz and wr can be W(n +1)+ i,(n +1)+ j = ∫ wˆrr (zi , s ) αr ds ,
0 s (l − s ) βr
represented as l l
Wˆi, j = ∫ wˆzz (zi , s ) Nj (s )ds, Wˆi,(n +1)+ j = ∫ wˆzr (zi , s ) Nj (s )ds,
0 0
n +1 n +1
gz (z ) = ∑i =1 Ni (z ) gzi , gr (z ) = ∑i =1 Ni (z ) gri , l l
Wˆ(n +1)+ i, j = ∫ wˆrz (zi , s ) Nj (s )ds, Wˆ(n +1)+ i,(n +1)+ j = ∫ wˆrr (zi , s ) Nj (s )ds.
0 0
n +1 n +1
wz (z ) = ∑i =1 Ni (z ) wzi , wr (z ) = ∑i =1 Ni (z ) wri . (38) (41)
where, gzi and gri are the regular part of the vertical and radial The unknown forces are
distributing forces on the shell-soil interface in the i-direction, respec-
tively. wzi and wri are the vertical and radial displacement of the shell in g(2(n + 1) × 1) = {gz1 ~gzn +1, gr1 ~grn +1}. (42)
the i-direction, respectively.
The initial displacement of the shell u 0 is
The Gauss-Chebyshev integral formula can be written as
1
u 0 (2(n + 1) × 1) = {Δz ... Δz , 0...0}T , (43)
f (x ) n
∫−1 dx = ∑i =0 Ai f (xi ),
1 − x2 and (40) can be solved numerically.
π ⎛ (2i + 1) π ⎞ In Sections 5.2 and 5.3, some numerical results are given. For
Ai = , xi = cos ⎜ 2(n + 1) ⎟ .
n+1 ⎝ ⎠ (39) better practical engineering use, the dynamic responses of two types of
shells, steel shells and concrete shells embedded in two types of soils,
By substituting (37) and (38) into (27), (28) and (30), we can obtain loose sand and dense sand, are fully studied. The material properties of
j − W) g = u 0,
(H + ρp hω 2 WH (40)
the shell and the soil can be found in Table 1, where B is Skempton's
pore pressure coefficient and k is the hydraulic conductivity of the soil.
where, νu is the undrained Poisson's ratios of the soil. The radius of the shell is
147
R. He et al. Soil Dynamics and Earthquake Engineering 95 (2017) 138–152
Fig. 9. (continued)
a = 5 m. The thickness of the steel shell and the concrete shell are h elastic shell vibrating in an elastic half-space given by Pak and Ji [34],
= 0.01a and h = 0.03a, respectively. Length of the shells of l = a, l = 2a, and Rajapakse and shah [29]. The solution for a shell embedded in an
l = 4a, and l = 10a are studied. The parameters R, Q, and b can be elastic half-space can be obtained by setting cw , b , R , Q , and ρf to
obtained from the given parameters [41]. For seawater, we have equally small values. Fig. 3(a)-(c) show that our solution corresponds
ρw = 1.0 × 10 3 kg/m3 and vw = 1.414 × 10 3 m/s. well with those provided in [34] and [29], which validates the Green's
functions and the numerical implementation in this study.
To validate the statement “the hydraulic boundary conditions of the
5.1. Comparison with existing solutions shell are unimportant for vertical impedance”, we compare our result
with Ji's work, who considered a totally impermeable shell using BEM
To verify the correctness of this method and the computer program, [35]. From the comparison in Fig. 3(d), the correctness of the
we compare the results obtained here with the existing solutions for an
Fig. 10. (a) The influence of the radius a of the shell on K vs for a steel shell with l = 2a embedded in dense sand; (b) The influence of the thickness h of the shell on K vs for a steel shell
with l = 2a embedded in dense sand.
148
R. He et al. Soil Dynamics and Earthquake Engineering 95 (2017) 138–152
Fig. 11. (a) The influence of the permeability k of the seabed on K vs for a steel shell with l = 2a embedded in dense sand; (b) The influence of the drained Poisson's ratio ν of the seabed
on K vs for a steel shell with l = 2a embedded in dense sand; (c) The influence of the undrained Poisson's ratio νu of the seabed on K vs for a steel shell with l = 2a embedded in dense sand;
(d) The influence of the Elastic modulus E of the seabed on K vs for a steel shell with l = 2a embedded in dense sand; (e) The influence of the elastic modulus E of the seabed on K vs for a
steel shell with l = 10a embedded in dense sand.
statement is validated. interaction forces pz and pr and wz , wr , Nz and Qz of the shell at a non-
dimensional frequency ω = 0.25 (ω = ωa / μ / ρ ) with l = 2a, and K vs for
5.2. Dynamic responses of the shell the shell with l = 2a and l = 10a between the three boundary
conditions. From Fig. 4(a)-(j), we find that the differences of the
The influence of the boundary conditions of the top ending of the dynamic responses between the three cases are localised for the shell
shell on the dynamic responses of the shell-soil vibration problem must (refer to pz ,wr and Qz at the vicinity of the top ending of the shell), and
first be made clear. A steel shell embedded in dense sand case is the impedance coefficient K vs is minimally influenced by the boundary
illustrated here for an example. We choose to compare the dynamic
149
R. He et al. Soil Dynamics and Earthquake Engineering 95 (2017) 138–152
effect. Similar results are found for shells with different lengths in other l / a ≤ 4 and a concrete shell with a length/radius ratio l / a ≤ 2 .
types of soils. Therefore, the following parameter analysis can be based
on any one of the three boundary conditions, and we use the fixed top 5.3.2. The influence of a and h on K vs
ending boundary conditions (Eq. (13)) here. To know whether the results obtained above can be used when the
To understand the coupled dynamic vibration problem, it is radius and the thickness of the shell are changed, the effect of changing
necessary to know the resultant dynamic interaction forces pz and pr the radius a and the thickness h on K vs is studied below. From Fig. 10(a),
on the shell and the soil. Fig. 5(a) and (b) show both pz and pr for the we can see that for the steel shell with a small length ratio (l = 2a) in dense
case of a steel shell with l = 2a embedded in various types of soils at sand, when changing the radius a from 0.025 to 5 m (from a small model
ω = 0.25, respectively. From Fig. 5(a) and (b), we find that both pz and bucket foundation to a real bucket foundation), K vs varies minimally,
pr are singular at both the top and bottom of the shell, and the forces in except when a is smaller than 0.5 m at a very small frequency, which
the lower portion are larger than the forces on the upper half portion, means K vs obtained by a model test can be used directly for a real case
which means most of the load transfer occurs at the lower portion of when the radius of the model bucket is properly chosen (larger than
the shell. pz and pr in the two cases are very similar, which means the 0.5 m). From Fig. 10(b), we also find that varying the shell thickness h
mechanism of the load transfer in dense sand and loose sand are very from 0.01a to 0.1a (in the range of h < < a , if h is comparable to a, the
similar. The displacements of the shell are plotted in Fig. 5(c)–(e). We bottom of the shell should be treated as an annulus disc) also has very
can see that shell has larger displacements when the soil has a larger small influence on K vs , however, when h is 0.001a, one will obtain a much
stiffness; however, for the steel shell with l/a=2, both the vertical and smaller K vs . Similar results are obtained for other types of soils and a large
the radial displacements are very small, which means the steel shell length/radius ratio (l/a = 10) shell. From the above results, we find that
with a small length/radius ratio acts like a rigid one. The axial force K vs is not sensitive to the radius of the shell, and not sensitive to h/a of the
Nz (s ) of the shell can be obtained by integrating the inertial force and shell, except when h/a is very small.
the vertical contacting force from z=s to the end of the shell, and Nz (0)
is equal to Fv defined above. The results of Nz (s ) for a steel shell in dense 5.3.3. The influence of k, ν , νu and E of the soil on K vs
sand are given in Fig. 5(f) and (g), from which we can see that the real To determine whether the above obtained results can be used when
part of Nz (s ) for the two types of soils are almost the same, whereas the the properties of the soils are not the same as we have used, we studied
imaginary part of Nz (s ) has some difference, however, the values are all how the parameters of the soil affect K vs , and take the steel shell
very small. embedded in dense sand covered by seawater as an example. For dense
sand, we assume the hydraulic conductivity k is 10−3 m/s~10−5 m/s, the
5.3. Numerical results for the dynamic impedance drained Poisson's ratio is 0.25–0.45, the undrained Poisson's ratio is
0.49–0.4995, and the elastic modulus E is 50 MPa~120 MPa. From
To obtain a first sight of the dynamic response of a shell, a comparison Fig. 11, we find that varying ν , κ , νu and E has a very small effect on the
of the dynamic impedance between a rigid disc with r = a and a shell with K vs , which means that the results obtained above for dense sand can be
l = a in dense sand is studied. From Fig. 6, we find that the real part of the used for dense sand even if their mechanic parameters are not the
impedance of the shell is about 20% larger than that of the disc, while the same. In contrast to the small length/radius ratio case, E has a
imaginary part of the shell is about 60% larger than that of the disc. pronounced effect on K vs in the large length/radius ratio case, as seen
in Fig. 11(e). Similar results can also be obtained for loose sand. From
5.3.1. Numerical results for K vs the above results, we reach the conclusion that for the dynamic
The numerical results for the vertical dynamic impedance coeffi- impedance coefficient of a steel or concrete shell embedded in common
cient K vs as a function of the shell length/radius ratio (l/a), shell type, sands at a low frequency, the most dominant factors are the relative
soil type, and the non-dimensional frequency ω are presented below. rigidity of the shell and the soil. For the shell, its rigidity increases with
From Fig. 7, we find that for a steel shell in loose sand, the existence of decreasing length/radius ratio l/a, increasing thickness/radius ratio h/
a fluid half-space only has some influence on Im(K vs ), which means a and increasing elastic modulus Ep , whereas for the soil, its rigidity is
damping increases. The length/radius ratio l/a has a very large most affected by its elastic modulus E. That means K vs of a shell can be
influence on K vs . From the cases l = a to l = 10a, K vs increases gradually. written as a function of l/a, h/a,Ep / E and ω . For the shell with a small
For the case of a steel shell embedded in other types of soils or a length/radius ratio, its rigidity is large enough compared to any type of
concrete shell, very similar results can be obtained. common sands, so varying the soil's rigidity does not influence the
To consider the difference between a steel shell and a concrete shell, results significantly. For the shell with a large ratio, its rigidity is
we find from Fig. 8 that the difference of K vs between them increases comparable with the soil's rigidity, and K vs is affected by both the shell
with the length/radius ratio for the case embedded in dense sand, and and the soil's rigidity. The threshold value for small length/radius ratio
when l / a = 10 ,the difference is considerable, whereas the differences and large length/radius ratio is dependent on the elastic modulus of the
of K vs for the loose sand are much smaller. These results mean when the shell. For a steel shell, we suggest l = 4a and for the concreted shell l
soil type is given, whether a steel shell or concrete shell is used, similar = 2a.
dynamic responses are obtained, except in the case when the shell is
embedded in dense sand with a large length/radius ratio. 6. Conclusions
Fig. 9 shows the results for both a steel shell and a concrete shell
embedded in various types of soils with different length/radius ratios. This study investigates the problem of an offshore shell- type
From Fig. 9, we find an interesting phenomenon: when the length/ foundation vibrating vertically in a sandy seabed, potential theory
radius ratio of the steel shell is l = a, l = 2a, and l = 4a, and when the and the integral equation method are used. The problem is shown to be
length/radius ratio of the concrete shell is l = a and l = 2a, the obtained reducible to a couple of Cauchy equations and Fredholm equations.
K vs is not related to the soil types and is almost only dependent on l/a From the theoretical analysis, we find that both the top and bottom of
and ω , which means K vs in these cases can be written as a simple the shell have stress concentrations of Cauchy type, relates to the
equation of ω . The fitted equations are given in Fig. 9. The error of the drained Poisson's ratio of the soil and keeps the same for all soils for
fitted equation and the real value is acceptable: the maximum errors the top and bottom, respectively. By solving the Fredholm equations
are within 5%. Considering the case a rigid shell embedded in soil, K vs is numerically, we find the non-dimensional impedance coefficient K vs is
independent on the soil's rigidity, this phenomenon can be explained highly related to four non-dimensional quantities: length to radius
by the fact that the rigidity of the shell embedded in any type of the two ratio l/a, non-dimensional frequency ω , thickness to radius ratio h/a
sands is large enough with a steel shell with a length/radius ratio and elastic modulus ratio Ep / E . The detailed results are:
150
R. He et al. Soil Dynamics and Earthquake Engineering 95 (2017) 138–152
1) Seawater only has some effect on Im(K vs ). 6) When l/a and h/a are chosen, varying the permeability κ , drained
2) When h/a is properly chosen, the dynamic properties of a steel shell Poisson's ratio ν , and undrained Poisson's ratio νu of the sand has a
and a concrete shell are similar for shells with a small l/a. The small effect on K vs .
difference between them increases with increasing l/a. 7) When h/a is properly chosen, K vs of the shell with small l/a is not
3) When h/a is properly chosen, for the shell with a small l/a (steel sensitive to E of the soil, whereas K vs of the shell with a large l/a is
shell with l ≤ 4a , and concrete shell with l ≤ 2a ), K vs is mostly sensitive to E.
related to l/a and ω and can be treated independent of the soil
types. Acknowledgements
4) When l/a is chosen, K vs is not sensitive to varying h/a of the shell,
except when h/a is very small. A very small h/a will induce a quite The first author would like to acknowledge the support of the Grant
small K vs . No. 51509082 from the National Natural Science Foundation of China,
5) When l/a and h/a are chosen, K vs is not sensitive to varying radius the Grant No. BK20150804 from the Natural Science Foundation of
of the shell, except when a is very small, so K vs obtained from a Jiangsu province, and to acknowledge China Scholarship Council
small scale model test can be used directly in the real case when a is supporting him to study in University of Colorado, Boulder. The fourth
properly chosen (mechanic properties of the soil in the model test author would like to acknowledge support grant 51209183 from the
and real case should be the same). National Natural Science Foundation of China.
P1 = (e−kp cνp (4e kp c + 4e kp (2l + c) + e 2kp (l + c) cos(kp (2l − c )) − (1 + 2e kp l + 3e 2kp c + 2e 2kp (l + c) )cos(kp c ) − e 2kp l cos(kp (2l + c ))
− 4e kp (2l + c) sin(2kp l ) + 2e 2kp l sin(kp (2l − c )) + e 2kp (l + c) sin(kp (2l − c )) + (1 + e 2kp c )sin(kp c ) + e 2kp l sin(kp (2l + c )))/Sv,
P2 = −(e 2kp l − kp cνp (−4e kp c − 4e kp (2l + c) − cos(kp (2l − c )) + (2 + 3e 2kp l + 2e 2kp c + e 2kp (l + c) )cos(kp c ) + e 2kp c cos(kp (2l + c ))
− 4e kp c sin(2kp l ) + (1 + 2e 2kp l )sin(kp (2l − c )) + (e 2kp l + e 2kp (l + c) )sin(kp c ) + e 2kp c sin(kp (2l + c )))/ Sv,
P3 = −(e−kp cνp (−4e kp (2l + c) − 4e kp (2l + c) cos(2kp l ) + e 2kp l (2 + e 2kp c )cos(kp (2l − c )) + (1 + 4e 2kp l − e 2kp c )cos(kp c ) + e 2kp l cos(kp (2l + c ))
− e 2kp (l + c) sin(kp (2l − c )) + (1 + 2e 2kp l + e 2kp c + 2e 2kp (l + c) )sin(kp c ) + e 2kp l sin(kp (2l + c )))/ Sv,
P4 = −(e 2kp l − kp cνp (−4e kp c − 4e kp c cos(2kp l ) + (1 + 2e 2kp c )cos(kp (2l − c )) − (e 2kp l − 4e 2kp c − e 2kp (l + c) )cos(kp c ) + e 2kp c cos(kp (2l + c ))
+ sin(kp (2l − c )) − (2 + e 2kp l + 2e 2kp c + e 2kp (l + c) )sin(kp c ) − e 2kp c sin(kp (2l − c )))/ Sv,
P5 = 4(e kp c ((e kp c + 4e 2kp (l + c) + e kp (4l + c) + 2e kp (2l + c) cos(2kp l )) kp c + (e kp c − e kp (4l + c) ) νp2 + (1 + e 2kp l )(e 2kp l − e 2kp c ) νp2 cos(kp c )
− 2e kp (2l + c) νp2 sin(2kp l ) + (e 2kp l + e 2kp (l + c) ) νp2 sin(kp (2l − c )) − (e 2kp l + e 2kp (l + c) ) νp2 sin(kp c ))/Sv,
P6 = 0.
where,
Sv = 4G1 kp (−1 + νp2 )(1 + 4e 2kp l + e 4kp l + 2e 2kp l cos(2kp l )).
2) For radial ring load on z = c
C1 = (e−kp c (−2e 2kp l cos(k p (2l − c )) − (3e 2kp l + e 4kp l + e 2kp c + e 2kp (l + c) )cos(k p c ) − (e 2kp l − e 2kp (l + c) )sin(k p (2l − c )) + (e 2kp l + 2e 2kp c + e 2kp (l + c) )sin(k p c ))/ Sr ,
C2 = (e−kp c (2e 2kp (l + c) cos(k p (2l − c )) + (e 2kp l + e 4kp l + e 2kp c + 3e 2kp (l + c) )cos(k p c ) + (e 2kp l − e 2kp (l + c) )sin(k p (2l − c )) + (e 2kp l + 2e 4kp l + e 2kp (l + c) )sin(k p c ))/ Sr ,
C3 = C4 = (e−kp c ((e 2kp l − e 2kp (l + c) )cos(k p (2l − c )) − e 2kp l (−1 + e 2kp c )cos(k p c ) − (1 + e 2kp l )(e 2kp l + e 2kp c )sin(k p c )))/ Sr ,
C5 = −2(e−kp cνp ((e 2kp l + e 2kp (l + c) )cos(k p (2l − c )) + (2e 2kp l + e 4kp l + e 2kp c + 2e 2kp (l + c) )cos(k p c ) + + (e 2kp l − e 2kp (l + c) )sin(k p (2l − c ))
+ (e 4kp l − e 2kp c )sin(k p c ))/ Sr ,
C6 = 0,
(A-2)
151
R. He et al. Soil Dynamics and Earthquake Engineering 95 (2017) 138–152
P1 = (e−kp c (−e 2kp l cos(kp (2l − c )) − (1 + e 2kp l )(−1 + e 2kp c )cos(kp c ) − (e 2kp l − e 2kp (l + c) )sin(kp (2l − c )) + (e 2kp l + 2e 2kp c + e 2kp (l + c) )sin(kp c ))/Sr ,
P2 = (e−kp c (−e 2kp l cos(kp (2l − c )) − (1 + e 2kp l )(−1 + e 2kp c )cos(kp c ) − (e 2kp l − e 2kp (l + c) )sin(kp (2l − c )) + (e 2kp l + 2e 2kp c + e 2kp (l + c) )sin(kp c ))/Sr ,
P3 = (e−kp c ((e 2kp l − e 2kp (l + c) )cos(kp (2l − c )) − e 2kp l (−1 + e 2kp c )cos(kp c ) − (−1 − 3e 2kp l + e 2kp c + e 2kp (l + c) − 2e 2kp l cos(2kp l ))sin(kp c ))/ Sr ,
P4 = (e 2kp l − kp c (−(−1 + e 2kp c )cos(kp (2l − c )) − (−1 + e 2kp c )cos(kp c ) + (−1 − e 2kp l + 3e 2kp c + e 2kp (l + c) + 2e 2kp l cos(2kp l ))sin(kp c ))/ Sr ,
P5 = −2(e−kp cνp (−e kp c − 4e 2kp (l + c) − e kp (4l + c) − 2e kp (2l + c) cos(2kp l ) + (e 2kp l + e 2kp (l + c) )cos(kp (2l − c )) + (2e 2kp l + e 4kp l + e 2kp c + 2e kp (2l + c) )cos(kp c )
+ (e 2kp l − e 2kp (l + c) )sin(kp (2l − c )) + (e 4kp l − e 2kp c )sin(kp c ))/ Sr , P6 = 0,
where,
Sr = 8aGkp4 (1 + 4e 2kp l + e 4kp l + 2e 2kp l cos(2kp l )).
k1 = η1−η2 , k2 = η1 β2−η2 β1, k3 = 2βγμξ, k 4 = 2αγμξ, Ω0 = β1 η2−β2 η1, Ω1 = β1 Γ2−β2 Γ1, Ω2 = η1 Γ2−η2 Γ1, Ω3 = β1 + η1, Ω4 = β2 + η2 , Ω5 = Γ1−η1 Ω ,
Ω = Γ −η Ω , Ω = η Γ + η Γ , Ω = Γ + η Ω , Ω = Γ + η Ω , s = γ 2 + ξ 2, β = 2μα 2−λk 2 + A k 2 ρ∼ ω (R + Q)/ R,
6 2 2 7 1 2 2 1 8 1 1 9 2 2 1 1 f f f 22
152