Soil Dynamics and Earthquake Engineering 75 (2015) 220–233

Contents lists available at ScienceDirect

Soil Dynamics and Earthquake Engineering

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

Rotational components in structural loading

M.R. Falamarz-Sheikhabadi a,n, M. Ghafory-Ashtiany b
a
Civil, Architectural, and Environmental Engineering Department, Drexel University, Philadelphia, PA, USA
b
International Institute of Earthquake Engineering and Seismology, IIEES, Tehran, Iran

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

Article history: In this paper, the rotational loading pattern of multi-storey buildings supported on the spread and
Received 20 January 2014 continuous interconnected single foundations is discussed. To achieve this, simplified relations for the
Received in revised form estimation of (1) point rotations; (2) spatial variation of strong ground motions, and (3) foundation input
9 April 2015
motions are derived. The height-wise variation of the earthquake rotational loading of multi-storey
Accepted 16 April 2015
buildings is parametrically evaluated by considering the location of the first rigid floor diaphragm and
Available online 15 May 2015
foundation type. In addition, the effect of the kinematic soil–structure interaction on the response
Keywords: spectrum of the rotational and translational components is studied. The numerical results provide a
Point, surface, and foundation rotation deeper insight into the rotational loading of structures in the middle-field zone, and show how the
Phase delay
rotational components may detrimentally affect the structural response of multi-storey buildings
Coherency
depending on their kinematic characteristics.
Spatial variation of strong ground motion
Kinematic soil–structure interaction & 2015 Elsevier Ltd. All rights reserved.
Response spectrum
Random vibration analysis

1. Introduction earthquake shaking can be assigned to a single point, the rotational

As seismic waves propagate from the source to the ground
surface along a certain wave passage, their characteristics, such as
their amplitude and frequency content, change as they pass through
the soil layers. Propagation of the seismic waves along different
wave paths in the earth crust causes Spatial Variation of Strong
Ground Motion (SVSGM). The propagation character of the seismic
waves also induces the rotational components at any points on the
ground surface, which may be estimated in terms of spatial
derivatives of the corresponding translational motions [1–5].
A uniform loading pattern including six earthquake compo-
nents corresponding to a point on the ground surface may be
considered as the simplest form of the seismic excitation for the
ordinary multi-storey buildings. However, such a loading pattern
may result in the underestimation of the structural response
considering the kinematic characteristics of the multi-storey
buildings such as the geometrical configuration of their structural
components and foundation properties [6,7,8,47]. Therefore, it is
necessary that the seismic loading of the multi-storey buildings is
determined considering their kinematic characteristics.
From the engineering aspect, three approaches have been used
for the evaluation of the rotational motions and consequently, the
rotational loading of structures subjected to the seismic waves,
namely: point, surface, and foundation rotations [1–15]. When the
n
Corresponding author.
E-mail address: mf585@drexel.edu (M.R. Falamarz-Sheikhabadi).
loading of structures may be performed by the point rotation [1–5],
which corresponds to the gradient of the translational displacements
at that point on the ground surface. The rotational excitation of the
structures supported on large mat foundations is usually applied
using surface rotation [9,10] or foundation rotation [6,7]. The surface
rotation is defined as a mean slope of the area of the ground surface
where the foundation is supported on. Approximate value of the
input rotational motions at foundation level may be evaluated based
on the surface rotation method while the seismic ground motions of
at least two points under foundation surface are known [9,10].
However, when the ground motion pattern at all points under the
foundation surface is mathematically determined; a better approx-
imation of the input rotational motions at the foundation level can be
obtained using the foundation rotation method [6,7].
The rotational loading of structures may be notably intensified
due to the foundation effects on the input excitations [16]. For the
structures supported on rigid mat foundations, considering the
effects of the kinematic soil–structure interaction leads to filtering
high-frequency components of the translational response and to the
amplification of the rotational motions. Therefore, the rotational
contribution to the seismic response of structures may be amplified
if the effect of the soil–structure interaction is considered in the
structural loading and modeling [13,17]. Asymmetric permanent
strains at the foundation-soil level due to the soil nonlinearity [18]
and foundation damage [19] during Strong Ground Motions (SGMs)
can also cause the rotational excitation of structures.
Recent seismological data indicated that the ratio of the ampli-
tude of the rotational components to the translational components

http://dx.doi.org/10.1016/j.soildyn.2015.04.012
0267-7261/& 2015 Elsevier Ltd. All rights reserved.
 M.R. Falamarz-Sheikhabadi, M. Ghafory-Ashtiany / Soil Dynamics and Earthquake Engineering 75 (2015) 220–233
in the near field can be significantly larger than that was expected
in this zone [20,21]. In the past decade, this observation leaded to
the attraction of the theoretical studies toward the near-field effects
of the rotational excitations on the seismic behavior of structures
[22–25], which showed that the rotational components may result
in significant damages in multi-storey buildings and bridges.
Despite these research, the influence of the rotational components
on the structural response is still neglected by most of the seismic
design codes probably due to: (1) lack of knowledge on the
characteristics of the rotational motions because of insufficient
amount of the recorded data and difficulty in presenting a quanti-
tative assessment of the rotational components for given transla-
tional components, and (2) complexity in the derivation of simple
loading patterns for structures subjected to the rotational excita-
tions. The only exception is a very simplified approach to rotational
effects presented in Eurocode 8 part 6 [26]. To overcome these
difficulties, extensive investigations on the rotational components
for generating synthetic rotational accelerograms, which have most
compatibility with the existing recorded data as well as the
cognition of the types of the rotational motions and their influences
on the behavior of structures, are necessary.
Considering the wave passage effect, simplified approaches
have been proposed for the application of the rotational loading
in the seismic analysis of multi-storey buildings, such as (1) gen-
erating rotational accelerograms [3,5,13,32,33]; (2) introducing
rotational response spectrum [8,14,15,26]; (3) deriving accidental
eccentricities due to the torsional earthquake component
[1,14,15,27–29], and (4) modifying base shear due to the combined
action of the rocking and horizontal earthquake components
[14,15]. The main objective of the present research is to provide
more accurate relations for the rotational excitation of structures
in comparison to the previous studies. In this case, after a brief
review of the characteristics of the rotational components due to
the wave passage effects, the combined action of the coherency
and wave passage effects in deriving the rotational components in
the middle-field zone is discussed. The rotational input motions
corresponding to the spread and continuous interconnected single
foundations are estimated and the foundation effect on the
rotational response spectrum is studied. Finally, the height-wise
Fig. 1. Schematic diagram of the considered wave propagation system: (a) seismic wave propagation from the hypocenter to the site; (b) geometric interpretation of the
system considered for SVSGM.
221
variation of the earthquake rotational loading of multi-storey
buildings is addressed by considering the location of the first rigid
floor diaphragm and foundation type. The generated input
motions in this study are derived in the linear frequency domain
using random vibration approach and assuming ground motions
as stationary Gaussian processes.
2. Mathematical representation of rotational excitation
Herein, three different approaches for the estimation of the
rotational components are discussed in Sections 2.1 and 2.2, and
the rotational loading of the multi-storey buildings is parametri-
cally studied in Section 2.3.
2.1. Point rotation
The displacement components of the SGMs always accompany
the rotational components induced by the spatial variation of
seismic waves. The rotation may be considered as the gradient of
the translational displacement at a point on the ground surface. In
this subsection, two approaches for deriving point rotations are
discussed by considering (1) wave passage effect, and (2) combined
action of wave passage and coherency effects.
2.1.1. Wave passage effects
The rotational components are commonly estimated by con-
sidering the wave passage effects [1–3,13–15]. Herein, such an
approach is briefly examined by comparing actual and synthesized
Spectral Density Function (SDF) of rotational components. n g The
o
rotational acceleration components of ground motions, θ€ ðt Þ ,
induced by the spatial variation of the seismic waves, in terms of
g g g
the translational components, fu€ x ðt Þ; u€ y ðt Þ; u€ z ðt Þg, along the Carte-
sian coordinate axes (Fig. 1(a)), for small deformations, may be
expressed as:
" g #
n g o ∂u€ z ðt Þ! ∂u€ z ðt Þ! 1 ∂u€ y ðt Þ ∂u€ x ðt Þ !
g g g
θ€ ðt Þ ¼ i  j þ  k ð1Þ
∂y ∂x 2 ∂x ∂y
222 M.R. Falamarz-Sheikhabadi, M. Ghafory-Ashtiany / Soil Dynamics and Earthquake Engineering 75 (2015) 220–233

The first two terms in the right-hand side of Eq. (1) are known ground motions then V y -1, and Eq. (4) simplifies to:
as the rocking components, which are correlated to the vertical 8 9
> 0 0 0 >
ground motion. Considering the fact that the amplitude of the >
> >
>
n o < =
0 ω2 Sz ðωÞ
2 g
0
high-frequency motions of the vertical earthquake component can S € ð ωÞ ¼
g
Vx ð5Þ
θ >
> >
be considerably larger than the horizontal ones in the near-field >
:0 0 ω2 Sg ðωÞ >
>
;
4V 2 y
zone in comparison to the far-field zone; these two rotational x

components may lead to the severe structural damages in the

The stochastic characteristics of the rotational SDF matrix were
near-field zone for the SGMs, particularly during thrust-slip
parametrically discussed in Ref. [15]. However, because of lacking
events. The third term in Eq. (1) entitled the torsional component
the recorded data on the rotational components, it is still difficult
is correlated to the horizontal motions and for sites near the end of
to establish a mathematical/physical platform for a parametric
faults may be specifically notable during strike-slip events due to
discussion on the cross-correlation characteristics of the rotational
the rupture directivity effects.
and translational components, i.e. it is not feasiblento study o the
The SDF matrix of six earthquake components can be
stochastic characteristics of the entries of matrices Sg€ € ðωÞ and
written as: uθ
fSg€ € ðωÞg. At the present time, any argument on the cross-
θu
8  n o 9 correlation characteristics of the rotational and translational
< Su€ u€ ðωÞ 33
> Sg€ € ðωÞ >
g
n o uθ
= components may lead to the fallacious mathematical interpreta-
S6EC ðωÞ ¼ n g o n o33 ð2Þ tions, which are not in agreement with the seismological observa-
66 >
: S € € ðωÞ Sg€ € ðωÞ >
;
θu 33 θθ 33 tions [20,21]. For example, it can be mathematically shown that
the rocking and vertical components are not correlated as follows:
When six earthquake components are decomposed along the h g i Z
1 T=2 € g
E θ€ y ðt Þu€ z ðt Þ ¼ limT- þ 1 θ ðt Þu€ gz ðt Þdt
g
principal axes of the translational components, one can obtain the
T  T=2 y
translational SDF matrix as:
0 2 1
Z þ1 €g  Z þ1
8 g 9 U ð ωÞ 
1 ω B z C ω g
> Sx ðωÞ
> 0 0 >
> ¼ limT- þ 1 @ Ad ω ¼ Sz ðωÞdω ¼ 0
 g  < = 2π  1 V x T 1 Vx
Su€ u€ ðωÞ ¼ 0 Sgy ðωÞ 0 ð3Þ
>
> >
>
: 0 0 Sgz ðωÞ ; ð6Þ

where E[…] gives the expected value of the quantity under the
where ω is the circular frequency, and Sgx ðωÞ, Sgy ðωÞ and Sgz ðωÞ are bracket. This relation (Eq. (6)) may not hold if the dispersion
the SDF of translational accelerations. Herein, the principal axes characteristics of surface waves [32,33] or the coherency effects
are considered as the axes in which the cross correlation of the (Section 2.1.2) are considered in deriving the SDFs corresponding
ground motions along them are minimized. It has been shown that to the rotational components.
the principal axes of the translational components are along the Considering the diagonal entries of the rotational SDF matrix
radial, tangential, and vertical directions of seismic waves in the (Eq. (4)), the validity of the following relation:
far-field zone [48]. However, this is not generally the case for the (  )
n o
ω2 Sx ðωÞ þ Sy ðωÞ
g
Sg€ ðωÞ Sg€ ðωÞ Sg€ ðωÞ ω2 SzVðω2 Þ ; ω2 SzVðω2 Þ ;
g g g
middle- and near-field zones because the seismic source cannot be ; ; ¼
θx θy θz y x
4 V 2y V 2x
defined as an immovable point source. In spite of this, throughout
this paper it is assumed that the principal axes exist at least as ð7Þ
obtained in an approximate way for any analyzed points on the
is examined for an event occurred at the HACC station of the HGS-
surface.
Taiwan array due to a moderate-magnitude earthquake (2007)
The rotational SDFs can be estimated in different forms [15].
recorded in the middle-field zone. In this evaluation, the following
From engineering aspect, it is usually difficult to determine the
spectrum is fitted to the translational acceleration of the consid-
relative contribution of different types of seismic waves and it is
ered event [15,34–38]:
normally assumed that the contribution of the shear waves is
h  2  2 i
primary to seismic motions in the middle- and near-field zones 2 2
exp  ω=ωh =2 1 þ ξh
XN 1 þ4ξf ω=ωf Sf
[30]. In addition, the shear waves are non-dispersive except in S ðωÞ ¼
g
h  2 i2 2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2  4
2
highly attenuated media [31] and thus it can be roughly assumed f ¼ 1 1  ω=ωf þ4ξf ω=ωf 1 þ 1  2ξl ωl =ω
that the phase velocity of these waves is frequency independent. ð8Þ
In this case, the phase velocity along jth-direction (y- or x-axis) of
the coordinate axis may be assumed as V j ¼ V S = sin αj in which VS where index f relates to the prevailing frequencies of the site; Sf is
is the shear wave velocity in the surface layer and αj is incident an intensity factor, and h and l indicate the highest and lowest
angle with the vertical axis and jth-direction. Assuming a constant prevailing frequency of the site, respectively. Table 1 shows the
phase velocity for the seismic wave propagation and considering respective parameters for the best fit of the considered spectrum
the direction of the coordinate axis (Fig. 1(a)) along the principal to the SDFs of the translational accelerations. The constant phase
axes of the translational components, the rotational SDF matrix velocity is estimated as V x ¼ V y ffi0:6 km=s by fitting the actual
can be estimated as [15]: SDF of the rotational acceleration components to their correspond-
ing synthetic SDF ones. A fairly good agreement between the
8 ω2 Sg ðωÞ 9
 Vωx V y Sgz ðωÞ frequency content of the synthetic and real rotational SDFs can be
2
>
> V 2y z
0 >
>
>
> >
> seen in Fig. 2. From the practical viewpoint, this observation may
n o >
< >
=
 Vωx V y Sgz ðωÞ ω2 Sg ðωÞ
2
0 verify current assumption for the estimation of the rotational
Sg€ ðωÞ ¼ V 2x z ð4Þ
θ >
>  >
> components in most research studies [1,2,3,13–15], i.e. assuming a
>
> ω2 Sgx ðωÞ Sgy ðωÞ >
>
>
: 0 0 þ >
; reasonable phase velocity may result in a fairly good estimation of
4 V 2y V 2x
the diagonal elements of the rotational SDF matrix. In the next
subsection, a detailed discussion on the combined action of the
For the point source case or far-field conditions, when the
wave passage and coherency effects on the rotational components
x-axis is considered along the radial direction of the seismic
is presented.
 M.R. Falamarz-Sheikhabadi, M. Ghafory-Ashtiany / Soil Dynamics and Earthquake Engineering 75 (2015) 220–233 223

Table 1
The spectral density function parameters compatible with the selected SGM
records.
 
Sf ξf ωf rad=s

Horizontal acceleration in x-direction

5.3 0.3 17
1.2 0.15 27
0.4 0.15 49
Horizontal acceleration in y-direction
4 0.3 17
0.8 0.15 26
0.4 0.15 36
Vertical acceleration
0.53 0.15 26
0.1 0.1 33
0.02 0.08 45

2.1.2. Wave passage and coherency effects

In this part, a simple method for the estimation of the spatial
variation of the translational and rotational motions is presented
for a homogeneous random field. The random field for SGMs refers
to a multivariate random process where information is available
for all locations on the ground surface and homogeneity indicates
that the SDF of the SGMs is independent of their location and does
not vary significantly at different locations [39]. Homogeneity
assumption may be only applied to the uniform site conditions
and therefore, the proposed relations in this section cannot be
used for non-uniform site conditions. For the seismic waves that
propagate along the x-direction with phase velocity Vx as shown in
Fig. 1(b), the rotation with respect to the displacement at points A
and B in the Cartesian coordinate can be written as:

U B ðωÞ  U A ðωÞ U g ðx; y; ωÞ  U g ðωÞ

θAB ðx; y; ωÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
x2 þ y2 x2 þ y2
 
    

U g ðωÞexp iω t  x=V x þ iϕ ðωÞ  U g ðωÞexp iωt þ iϕ ðωÞ
B A
¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
x2 þ y2
 
U g ðωÞ



¼ p ffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp½iωt  exp  iωx=V x þ iϕB ðωÞ  exp iϕA ðωÞ ð9Þ
x2 þ y2

where x and y are the separation distances between points A and

B, Ug is the recorded ground displacement at point A, and φðωÞ is
the initial phase. The absolute value (norm) of θAB can be
expressed as:
 
   g ðωÞ 

θAB ðx; y; ωÞ ¼ pUffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1  exp iϕ ðωÞ  iϕ ðωÞ þ iωx=V x 
A B
x2 þ y2
 
U g ðωÞ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

ffi
¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2  2 cos ϕA  ϕB þ ωx=V x ð10Þ
x2 þ y2

Eq. (10) may be rewritten as:

  pffiffiffiffiffiffiffiffiffiffiffiffiffiffi Fig. 2. Real and synthetic SDF of rotational acceleration along Cartesian
θAB ðx; y; ωÞ x2 þ y2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

ffi coordinate axes.

U g ðωÞ
 pffiffiffi ¼ 1  cos ϕ A  ϕ B þ ω x=V x ð11Þ
2

On the other hand, the SDF of θAB can also be obtained as:
Eq. (12) can be simplified:
1
SθAB ðx; y; ωÞ ¼ 2 ½SAA ðωÞ þ SBB ðωÞ  SAB ðωÞ  SBA ðωÞ ð12Þ
x þ y2 SθAB ðx; y; ωÞ x2 þy2    
¼ 1  γ AB ðx; y; ωÞ cos ωx=V x ð14Þ
Su ðωÞ 2
Considering SAA ¼ SBB ¼Su and defining the coherency function as:
Combining Eqs. (11) and (14), one can approximately get:
     

SAB ðx; y; ωÞ   iωx γ ðx; y; ωÞ cos ωx=V x ¼ cos ϕ  ϕ þ ωx=V x ð15Þ
γ AB ðx; y; ωÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ γ AB ðx; y; ωÞexp  ð13Þ AB A B
SAA ðωÞSBB ðωÞ Vx
224 M.R. Falamarz-Sheikhabadi, M. Ghafory-Ashtiany / Soil Dynamics and Earthquake Engineering 75 (2015) 220–233

1 0.14
VS = 100 m/s
0.998 VS = 200 m/s
0.12
VS = 300 m/s
0.996
VS = 400 m/s
Coherency Loss Function

0.1
0.994

ϕAB (rad)
0.992 0.08

0.99 0.06

0.988
VS = 100 m/s
0.04
0.986
VS = 200 m/s
VS = 300 m/s 0.02
0.984
VS = 400 m/s
0.982 0
0 5 10 15 0 5 10 15
Frequency (Hz) Frequency (Hz)

Fig. 3. Variations of coherency loss functions considered for different shear wave Fig. 4. Variations of ϕAB versus frequency for different values of shear wave
velocity and x¼ y¼ 0.5 m. velocity, x¼ y¼ 0.5 m and incident angle of 60 deg.

with ϕAB ¼ ϕA  ϕB , Eq. (15) can be rewritten as: where β is ω=V x , α is  2λ ω2 x=V 2S , and λ is a dimensionless
2

 


 parameter [40]; Eq. (20) can be expressed as:
γ ðx; y; ωÞ ¼ cos ϕ  sin ϕ
AB tan ωx=V x ð16Þ  2 
AB AB  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2CD α  2Dβ 1 þD2  C þ 2jDjβ 1 þ D2  C  αCDjDj

 dϕAB
If x-ð2n þ 1Þπ V x =2ω ðn A naturalnumbersÞ then sin ϕAB -0, ¼ h  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffii  

 dx 2D 1 þ D2 1 þ D2  C sin ϕAB
and
cos ϕAB - 7 1. For a positive coherency loss function,


cos ϕAB -1 can only be accepted. Solving Eq. (16) for cos ϕAB ð22Þ
will lead to:
The SDF corresponding to the rotation at point B can also be
2  
q
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

  3 obtained as:
γ ðx; y; ωÞ þ  tan ωx=V x  1þ tan 2 ωx=V x  γ ðx; y; ωÞ
ϕB ¼ ϕA  cos  1 4
AB


AB
5  
1þ tan 2 ωx=V x ω dϕAB 2
SθB ðωÞ ¼ Su ðωÞ  ð23Þ
Vx dx
ð17Þ
In the case of the maximum or minimum initial phase differ-
For the coherency loss function shown in Fig. 3, the typical ence, Eq. (23) may be simplified to:
variations of ϕAB versus frequency for different shear wave
ω2 Su ðωÞ
velocities is shown in Fig. 4. In a special case when x-0, SθB ðωÞ ¼ ð24Þ
Eq. (17) may be simplified to ϕB ¼ ϕA  cos  1 γ AB ð0; y; ωÞ. The V 2x
Comparison of this with Eq. (17) clearly shows that the coupling Eqs. (14) and (23) show the application of two different
effects of the wave passage and coherency cannot generally be concepts in the estimation of the rotational components. Eq. (14)
ignored. Using Eq. (17), the displacement at point B in terms of is derived based on the surface rotation (average rotation) concept,
displacement at point A (reference point) can be defined as: and Eq. (23) is estimated on the basis of the point rotation concept.
 
  
 Eq. (14) can also be expressed based on the point rotation concept.
U B ðωÞ ¼ U g ðx; y; ωÞ ¼ U g ðωÞexp iω t x=V x þ iϕB ðωÞ ¼ U A ðωÞexp  iωx=V x  iϕAB ðωÞ
For example, to find the SDF of the rotational component along
ð18Þ y-axis (by setting y in the denominator to zero), one can get:
" 2 #
2Su ðωÞ
γ ðx; 0; ωÞ cos ωx=V x ¼ 2ω2 Su ðωÞ λ þ 1
   
With this one can estimate the rotation at point B as: limx-0 SθAB ðx; 0; ωÞ ¼ limx-0 1  AB
x2 V 2S 2V 2x
 
 ð25Þ
θB ðx; y; ωÞ ¼  iω=V x  idϕAB ðωÞ=dx U A ðωÞexp  iωx=V x iϕAB ðωÞ
ð19Þ Eqs. (24) and (25) become the same if the effect of the loss of
coherency is ignored (λ ¼0). Using Eqs. (23) and (25), the typical
where effects of the excitation frequency, shear wave velocity, angle of
incidence and loss of coherency on the rotational components
2  
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi


  3
dϕAB 1
  
d 4 γ AB ðx; y; ωÞ þ tan ωx=V x
 1 þ tan 2 ωx=V x  γ ðx; y; ωÞ
AB
when V x ¼ V S = sin ðαÞ, are shown in Figs. 5–7. As it can be seen in
¼  
 5
dx sin ϕAB dx 1 þ tan 2 ωx=V x these figures, the frequency amplitude of the rotational compo-
nents in comparison to the corresponding translational ones
ð20Þ
increases with the increase of the excitation frequency, incident
angle, λ, and decreases with the increase of the shear wave
assuming:
velocity. In addition, Eq. (25) estimates the smaller amplitudes
  dC for the rotational components in comparison to Eq. (23). However,
C ¼ γ AB ðx; y; ωÞ - ¼ αC
dx for a constant loss of coherency, the difference between the
  dD  outputs of two methods decreases for medium and hard soil
D ¼ tan ωx=V x - ¼ β 1 þ D2 conditions as it is expected from Eq. (25). The material presented
dx
" in Section 2.1 may be directly used to find a rotational loading
2 2 2 #
γ ðx; y; ωÞ ¼ exp  λ ω x þ y
  2 pattern for structures with spread foundation system; however, it
AB ð21Þ
V 2S is not the case for structures with mat or continuous intercon-
nected single foundation systems as discussed in the next section.
 M.R. Falamarz-Sheikhabadi, M. Ghafory-Ashtiany / Soil Dynamics and Earthquake Engineering 75 (2015) 220–233 225

1.2 1
α = 90o VS = 100 m/s
α = 60o 0.9
VS = 200 m/s
1 α = 30o
0.8 VS = 300 m/s
α= 0o
VS = 400 m/s
0.7
0.8

Sθ/Su (rad/m)2
Sθ/Su (rad/m)2

0.6

0.6 0.5

0.4
0.4
0.3

0.2
0.2
0.1

0 0
0 5 10 15 0 5 10 15
Frequency (Hz) Frequency (Hz)

1 0.8
α = 90o VS = 100 m/s
0.9 α = 60o 0.7 VS = 200 m/s
α = 30o VS = 300 m/s
0.8
α = 0o 0.6 VS = 400 m/s
0.7
0.5
Sθ/Su (rad/m)2

Sθ/Su (rad/m)2
0.6

0.5 0.4

0.4
0.3
0.3
0.2
0.2
0.1
0.1

0 0
0 5 10 15 0 5 10 15
Frequency (Hz) Frequency (Hz)

Fig. 5. Variation of spectral density function of rotational to translational compo- Fig. 6. Variation of spectral density function of rotational to translational compo-
nents given in Eq. (23) (top figure) and Eq. (25) (bottom figure) for VS ¼ 100 m/s and nents given in Eq. (23) (top figure) and Eq. (25) (bottom figure) for α ¼60 deg and
λ ¼ 0:2. λ ¼ 0:2.

2.2. Foundation rotation where UF and Dg are respectively the Fourier transforms of uF and
dg, ½K dyn  is the foundation’s dynamic stiffness matrix, Ω is the
As seismic waves pass through the foundation, the high- coordinate of the foundation surface, and G defines the stress
frequency waves of the free-field motions are filtered by the under the foundation and is given in Appendix A [41]. Different
foundation. To obtain these filtered motions, the effects of the forms of Eq. (26) in the previous studies have been used to
kinematic soil–structure interaction between the foundation and estimate the input rotational and translational motions of the
surrounding medium due to the SVSGM should be considered rigid mat footings [6,7]. The FIMs corresponding to other types of
[6,7,40,41]. Herein, at the first step, the current method for the foundations, such as a continuous interconnected single founda-
estimation of the Foundation Input Motions (FIMs) for rigid mat tion, can be similarly estimated using a weighted average method.
footings, for which the embedment effects [45,46] can be A continuous interconnected single foundation is a surface
neglected, is briefly reviewed. The FIM indicates the kinematic foundation that consists of a grid of single foundations, which
response of the foundation to the seismic excitations and is are interconnected to each other by grade beams (Fig. 8(b)).
considered as input ground motion for the structural loading. Herein, to evaluate the FIMs of such a foundation,
  grade beams
Next, a simple relation is proposed to estimate the input rotational are assumed as rigid tie beams. Describing uF as the FIMs at the
and translational motions for a special type of the continuous center O (Fig. 8(b)), the translational displacements at pth single
interconnected single foundations. foundation or pth mesh/element of the foundation can generally be
 FThe
T n FIMs at the stiffness o center of a rigid mat footing, expressed as:
¼ uFx ; uFy ; uFz ; θx ; θy ; θz , corresponding to the free-field
F F F
u
 g T  g g g  8 F 9 8 9
¼ ux ; uy ; uz , can numerically be calcu- uFx  yp θz þ zp θy >
F F
ground motions, d >
> dxp >> >
> >
>
< > >
= < >
=
lated using the weighted average method. The FIMs are evaluated
dyp ¼ uy þ xp θz  zp θx
F F F F
by finding the ratio of the external force imposed to the founda- ð27Þ
>
> > >
> > >
>
: dFzp >
> : uF  xp θF þ y θF >
; > ;
tion, due to the free-field ground motions, to its dynamic stiffness. z y p x
In this case, considering the pattern proposed in Eq. (18) for the
free-field ground motions, the FIMs can be obtained, in frequency where xp, yp and zp denote the coordinates of the geometrical center of
domain, by: the pth mesh (zp ¼0 for a surface foundation). Now, by allocating
n o h i1 six-degrees-of-freedom to each mesh and assuming that the interface
UF ¼ K dyn between the foundation and soil is a welded interface, the dynamic
61 66
  energy equilibrium of the system under seismic loading, for while the
∬Ω ½G63 Dg 31 exp½  iωx=V x  iϕAB ðωÞdxdy ð26Þ
226 M.R. Falamarz-Sheikhabadi, M. Ghafory-Ashtiany / Soil Dynamics and Earthquake Engineering 75 (2015) 220–233

mass of the foundation is excluded [44], will be given as: XN   2 1  2 1  

1 g 2
kθxp θx  θxp þ kθyp θy  θyp þ kθzp θz  θzp
F g F g F
þ
2 2 2
N  2  p¼1
X 1  F 2 1 
F
2 1 
F
E¼ kxp dxp  ugxp þ kyp dyp ugyp þ kzp dzp ugzp ð28Þ
p¼1
2 2 2
where superscript g indicates the ground motion along the principal
 
axes of the structure and kxp ; kyp ; kzp ; kθxp ; kθyp ; kθzp is the impe-
1.4 dance matrix of pth mesh. Minimizing the energy with respect to the
γ = 0.4
displacement will result in:
γ = 0.3
1.2 γ = 0.2
st   F  N h
X i n o
γ = 0.1 K U u ¼ K^ p U ugp ð29Þ
1 p¼1

 h i
where K st is the foundation’s static stiffness matrix, and K^ P is
Sθ/Su (rad/m)2

0.8

the participation matrix of the pth single foundation in the FIMs

0.6
(given in Appendix A). For the free-field motion uf due to the
propagation of the seismic waves at angle ϕ to the x-axis of
0.4  
the structure with T ϕ (directional cosine matrix), the ground
motion along the principal axes of the structure and its SDF
0.2
 g 
  n f o n o
can be obtained by up ¼ T ϕ U up and Sgpq ðωÞ ¼
0
  n f o
 
T ϕ U Spq ðωÞ U T ϕ . With this definition, the SDF of the
T
0 5 10 15
Frequency (Hz)
FIMs becomes:
( )
1 n o
1 N h
N X
X i n o h iT 
0.9
γ = 0.4 SFϕ ðωÞ ¼ K st U K^ p U Sgpq ðωÞ U K^ q
γ = 0.3
p¼1q¼1
γ = 0.2
h
iT
st   1
0.8 γ = 0.1
 K ð30Þ
0.7

Certainly, as shown in Fig. 8(c), this method may also be used to

Sθ/Su (rad/m)2

0.6
estimate the FIMs of the rigid mat foundations when the separa-
0.5
tion distances between single foundations, l, set to zero. To
0.4 increase the accuracy of the method in the estimation of the FIMs
0.3 of the mat footings, the largest
 dimension of each element/mesh
should be smaller than min 0:1Dx ; 0:1Dy ; V x Δt where Δt is the
0.2
time step of the free-field record, and Dx and Dy are the foundation
0.1 dimensions as shown in Fig. 8(a). In the following, the foundation
0
transfer functions, TF, corresponding to a special case of Eq. (26)
0 5 10 15 are derived in Appendix B and are used to evaluate the influences
Frequency (Hz) of the foundation size on the FIMs.
Fig. 7. Variation of spectral density function of rotational to translational compo-
A set of continuous interconnected single foundations with
nents given in Eq. (23) (top figure) and Eq. (25) (bottom figure) for VS ¼ 100 m/s and given characteristic in Table 2 are chosen. Foundations A, C, E and F
α ¼60 deg. are selected in order to evaluate the effects of the foundation

Fig. 8. Description of the soil, foundation and structure model, and coordinate systems: (a) considered system; (b) continuous interconnected single foundation; (c) mat
foundation; (d) spread foundation.
 M.R. Falamarz-Sheikhabadi, M. Ghafory-Ashtiany / Soil Dynamics and Earthquake Engineering 75 (2015) 220–233 227

Table 2
Dimension and characteristics of considered foundations.

Type of foundation A B C D E F

Dx  m 20 40 40 20 60 80
Dy  m 20 40 20 40 20 20
Nx 2 4 4 2 6 8
Ny 2 4 2 4 2 2

dimension ratio on the input acceleration, and effects of the

foundation length are studied by comparing the FIMs of founda-
tions A and B. Also, the influences of the propagation direction of
the seismic waves on the FIMs are investigated by comparing the
FIMs of foundations D and C. Other assumptions which are used
for this investigation are as follows: (1) the main direction of the
seismic waves is considered along the x-axis; (2) the Hao’s
coherency function [10], is used and constants aj ¼ 5  10  6 and
bj ¼ 2 are considered to model the coherency loss function; (3) the
ratio r xy ðωÞ is considered constant and equal to one, and (4) the
foundations are supported on a site with soft soil condition with
shear modulus of 74 MPa and Poisson’s ratio of 0.38.

2.2.1. Evaluation of input torsional acceleration

Fig. 9 shows the variations of the normalized torsional transfer

1=2
function, T Fθz U D2x þ D2y , as a function of the frequency and
apparent velocity. It can be seen that the foundations which have
larger dimension along x-axis quickly attain their maximum value

1=2
of T Fθz U D2x þ D2y for a constant value of V x with increase of the
frequency (see foundation F). Further, amongst foundations with
the same length along the x-axis, the foundations which have
shorter length along the y-axis obtain larger peak value of

1=2
T Fθz U D2x þ D2y , see the comparison of the foundations B and C,

1=2
or A and D. An important point in this figure is the variations of the
peak values of the normalized torsional transfer functions with
respect to the apparent velocity. Indeed, the decrease in value of
apparent velocity causes that peak values of the normalized
torsional transfer function shift toward lower frequencies. There-
fore, it can be inferred that the behavior of the normalized torsional
transfer function cannot be predicted without considering the real
value of the apparent velocity corresponding to seismic waves, even
if the foundation shape and its dimensions are known.
2.2.2. Evaluation of input rocking acceleration
The variations of the normalized rocking transfer functions
along the x- and y-axis, T Fθx ðωÞ U H and T Fθy ðωÞ U H, are shown in
Figs. 10 and 11, respectively. Here, H is described as the effective
height of structure (  0:7 total height) which is assumed to be
20 m in this study. As shown in Fig. 10, the foundations with
respectively smaller and larger length along the x- and y-axis have
larger values of T Fθx ðωÞ. T Fθx ðωÞ is more sensitive to the variation of
the foundation dimension along the x-axis than the y-axis, since
the SVSGM effects along the x-axis due to the propagation of
seismic waves along this direction is more than the y-axis. Thus,
foundations D and F have largest and smallest values of T Fθx ðωÞ,
respectively. It should also be mentioned that although the
prediction of the behavior of T Fθy ðωÞ is to some extent complicated
relative to the behavior of T Fθx ðωÞ; it is still obvious that the
decrease of the apparent velocity leads to the increase of T Fθy ðωÞ.
In addition, the behavior of T Fθy ðωÞ is mostly insensitive to the
variation of Dy due to two reasons: (1) the SVSGM along the x-axis
is larger than the SVSGM along the y-axis; and (2) Ax in Eq. (B3)
(given in Appendix B) is more dependent on Dx than Dy.
Fig. 9. Variation of foundation transfer function, T Fθz U D2x þ D2y , for the six
different foundation: (a) V x ¼ 1 km=s; (b) V x ¼ 2 km=s.
Finally, it is worth noting that considering the rigid foundation
behavior for short wave-lengths due to the flexibility of actual
foundations may lead to non-conservative estimates of relative
deformations in the structure [42,43]. Therefore, the above-
mentioned results might be different when the foundations are
not assumed to be rigid. In addition, the foundation embedment
may lead to increasing (decreasing) amplitude of the rotational
(translational) components of the FIMs at high excitation frequen-
cies due to the greater scattering of the short seismic waves [45,46].
2.2.3. Acceleration response spectrum
In this part, the acceleration response spectrum corresponding to
the free-field motions and the FIMs are calculated and the relationship
between the torsional and translational response spectrum is studied.
The following steps are performed to obtain the response spectra:
 Choosing the acceleration time-history series from HACC sta-
tion as the free-field ground motions recorded in the middle-
filed zone.
 Modifying the time history free-field record for the FIMs
corresponding to the foundations A, C and E with Eq. (B4)
given in Appendix B.
 Calculating the torsional and translational acceleration
response spectrum of the FIM and free-field ground motion.
228 M.R. Falamarz-Sheikhabadi, M. Ghafory-Ashtiany / Soil Dynamics and Earthquake Engineering 75 (2015) 220–233

Fig. 10. Variation of the normalized foundation transfer function, T Fθx ðωÞ U H, for the Fig. 11. Variation of the normalized foundation transfer function, T Fθy ðωÞ U H, for the
six different foundation: (a) V x ¼ 1 km=s; (b) V x ¼ 2 km=s. six different foundation: (a) V x ¼ 1 km=s; (b) V x ¼ 2 km=s.

 Scaling the torsional and translational response spectrum to response spectrum from engineering viewpoint. Certainly,
the corresponding Peak Ground Acceleration (PGA) of the more rotational SGM data is needed to verify this statement.
rotational and translational ground motions.
2.3. Rotational loading of multi-storey buildings

Fig. 12 shows the acceleration response spectrum of torsional, Herein, a simple method is proposed to include the rotational
SAθ , and translational, SAu , motions. The following points may be components in the structural loading of typical multi-storey build-
drawn from these spectra: ings. Consider a shear-type building model with spread foundation
system (see Fig. 8(a) and (d)), rigid floor diaphragm, and three-
 Despite decrease of the PGA corresponding to the FIM relative degrees-of-freedom at the floor level. The equations of motion of
to the PGA of the free-field ground motions, the response such a system under earthquake excitations can be written as:
spectrum of the FIM is more dominant in low-frequency part "
 # 8 n to 9 "
 # 8 n to 9
which this is because of kinematic soil–structure interaction ½M ss  M sg < u€ = ½C ss  C sg < u_ =


 U   þ

 U  
and filtering the high-frequency motions. M gs M gg : u€ g ; C gs C gg : u_ g ;
 The influence of the kinematic soil–structure interaction on the " #

 ( ) ( )
rotational components is more pronounced than translational com- ½K ss  K sg ut fP s g
þ

 U ¼   ð31Þ
ponents, because the rotational motions have more high-frequency K gs K gg fug g Pg
contents in comparison to the corresponding translational motions.
 For the torsional component as the period of the structural where ½M , ½C  and ½K  are the mass, damping, and stiffness
vibration increases, the rate of the reduction of the acceleration matrices; subscripts s and g, respectively, relates to structure
response spectrum is higher than the translational ones. This and supports (ground); sg and gs refers to coupling terms
 
point implies that the torsional component cannot be detri- between the structure and supports; ut and fug g are the total
mental in overall responses of torsionally-soft structures. horizontal displacement vector of structure and supports, fP s g
 
 Because at larger distances the surface waves have low- and P g are the force vectors of the structure and supports.
frequency contents, their contributions to the acceleration Introducing following matrices:
response spectrum of the rotational components are expected



½rK  ¼  ½K ss   1 K sg ; ½r C  ¼  ½C ss   1 C sg ; ½rM  ¼  ½Mss   1 M sg ð32Þ
to be small. Thus, using a constant apparent velocity may be a
reasonable approximation in order to evaluate the acceleration and assuming that: (1) there is no mass coupling between the structure
 M.R. Falamarz-Sheikhabadi, M. Ghafory-Ashtiany / Soil Dynamics and Earthquake Engineering 75 (2015) 220–233 229

0 0
coordinates of the pth column with respect to the floor’s CS, kxp and kyp
are the stiffness of pth column along the x- and y-axis, and h is the
height of the floor level from the ground surface. In a special case, when
0 0 0
kxp ¼ kyp ¼ k , the quasi-static accelerations of this structure become:
8 N  9
> X >
>
> u€ g
þ h €g
θ >
>
>
> xp yp >
>
>
> >
>
>
>
p¼1 >
>
>
> N >
>
>
> >
>
8 qs 9 > > X N  >
>
€ >
> uyp þ hθxp
€ g € g >
>
> ux >
> > > > >
>
 qs  < qs = < =
u€ ¼ uy € ¼
p¼1
N
> > > >
> ; >
: θ€ qs > >X
> N h i h i >
>
z >
>
> u€ g
þ hθ€ g Ux þ u€ g þ hθ€ g U y > >
>
>
>
> yp xp p xp yp p >
>
>
> p¼1 >
>
>
> >
>
>
> X N >
>
>
> 2 2 >
>
>
> x þy >
>
: p p ;
p¼1

ð35Þ
In frequency domain, using Eqs. (18), (19), and (35), the quasi-
static displacements for a regular m-storey building which its first
floor diaphragm located in fth storey can be expressed as:
8 N 
9
> X g >
> U€ xp þ hj θ€ yp >
g
>
> >
>
>
> >
>
>
> p ¼ 1 >
>
>
> >
>
>
>
N >
>
>
> X  >
>
>
>
N
€ g € g >
>
>
>
> U yp þ h θ
j xp
>
>
>
n qs o < p¼1
=
€
Uj ¼ N
>
> >
>
>X
> N h
g i h g g i >
>
>
> € g € € € >
>
> U yp þ hf θxp Uxp þ U xp þ hf θyp U yp > >
>
>
> >
>
>
> p¼1 >
>
>
> >
>
>
> XN >
>
>
> x 2
þ y 2 >
>
>
: p p >
;
Fig. 12. Acceleration response spectra of torsional and translational components p¼1
with damping ratio of 0.05 for the free-field ground motions and the FIM. " !#
X
j
hi
exp iω t  ð36Þ

 i¼1
vsi
and the supports (mass idealization), and the coupling term M sg is
zero: ½r M  ¼ ½0; (2) system is linear with stiffness-proportional damp- where index j relates to stories f to m, hj is the height of jth storey,
ing: ½r  ¼ ½r K  ¼ ½r C , and (3) excitation of the system is only due to the and vs is the velocity of the shear waves along the height of the
SGMs at its foundation: fP s g ¼ f0g; Eq. (31) can be simplified to: structure. The seismic loading of such a structure at stories lower
n o n o n o  qs  than fth storey may approximately be applied similar to a multiple
d
½M ss  u€ þ ½C ss  u_ d þ ½K ss  ud ¼  ½M ss  u€ ð33Þ
support structure and in terms of the quasi-static acceleration of
 qs   g n o n o  
where u€ ¼ ½r  u€
d
and u€ ¼ u€  ½r  u€
t g
are the quasi-static each storey’s columns. In this case, quasi-static translational
and dynamic acceleration vector of the structure, respectively. This acceleration of pth column connected to the jth storey becomes:
equation shows that to obtain the dynamic acceleration of structures, 8 qs 9 8 g g 9 " !#
n qs o < U€ xpj = < U€ xp þ hj θ€ yp = X j
hi
the quasi-static acceleration of the structure need to be determined. €
U pj ¼ ¼ exp  iω t  ð37Þ
Describing dynamic accelerations with respect to the floor’s Center of : U€ qs ; : U€ g þ h θ€ g ; vs
i ¼ 1 pi
ypj j yp xp
Stiffness (CS), the quasi-static accelerations become:
8 9 where j¼ 1, 2, …, f  1. To evaluate the effects of the rotational
> X N  g >
>
> k
0
€ gxp þhθ€ yp
u >
> components on the seismic loading of structures with rigid floor
>
> xp >
>
>
> >
> diaphragm, using Eq. (7) and assuming V x ¼ V y as the worst-case
>
>
p¼1 >
>
>
> X N >
> scenario [15], the transfer function corresponding to Eq. (36),
>
> 0 >
>
> >
j ðωÞ, can be obtained as follows:
>
> k xp >
> T qs
>
> >
>
>
> p ¼ 1 >
> 8 2 9 8 Sqs ðωÞ 9
>
> >
> > > > >
8 qs 9 > >
> X N  >
> > qs
ð ω Þ > > xjg >
0 € g > >
> T >
> > > Sx ðωÞ > >
>
> u€x > >
> >
> k yp
€
u
g
yp þ h θ xp
>
>
> >
>
<
xj
>
2 >
= >
>
< Sqs ðωÞ >
>
=
 qs  < u€ qs = < p¼1
=
T j ð ωÞ ¼
qs
T yj ðωÞ
qs yj
u€ ¼ y ¼ X
 Sy ðωÞ
g
> > > N
> >
> > > >
; >
: θ€ qs >
> >
>
0
kyp >
>
> >
>  qs 2 >
>
> >
> qs >
> Sθzj ðωÞ >
>
> > >
> > > >
z >
>
> p¼1
>
>
> : T θzj ðωÞ > ; > : Sg ðωÞ > ;
>
> >
> y
>
>
> XN   >>
> 8 9
> 0 € g 0 € g > 2
ω ωÞ=V 2x
> kxp yp u€ xp þ hθyp þ kyp xp u€ yp þ hθxp > 1 þ hj r zx ð
g g 2 2
>
> >
> >
> AN >
>
>
>p¼1 >
> >
> 2
N >
>
>
> >
> >
> >
>
>
> >
> >
> 1 þ hj ω2 r 2zy ðωÞ=V 2x
2
>
>
>
> XN >
> >
< AN >
=
>
> k
0
y 2
þ k
0
x 2 >
>
 N2


>
: xp p yp p >
; ¼ ð38Þ
> 1 þ h2f ω 2 r2 ð ωÞ=V x Ax þ
2
ωÞ þ h2f ω2 r2zy ðωÞ=V 2x Ay >
r2xy ð
p¼1 >
>
zy
!2 >
>
>
> X
N >
>
ð34Þ >
> >
>
>
> x2i þ y2i >
>
: ;
i¼1
where N is the number of structure columns, xp and yp denote the
230 M.R. Falamarz-Sheikhabadi, M. Ghafory-Ashtiany / Soil Dynamics and Earthquake Engineering 75 (2015) 220–233

fourth floor is the first level with rigid floor diaphragm of structure
 1=2
(f ¼4) and in this case, the variation of the T qs 2 2
θz U Dx þDy for
this storey and all stories above it are the same and plotted as
 1=2
storey 4. Similarly, the variations of the T qs 2 2
θz U Dx þ Dy for three
other cases (storey 6, 8 and 10) are calculated and plotted.
The numerical results imply that: (1) when first rigid floor
diaphragm of the structure is located in the upper stories of
structures, the seismic loading of the torsional component is more
dominant in comparison to the case which first rigid floor diaphragm
is located in the lower stories of the structure, and (2) the SGM
rotational components can severely change seismic loading of upper
stories of tall buildings and their effects on the seismic loading of
structures can never be neglected. This conclusion can be extended,
of course, to the seismic loading of the structures with rigid mat
foundation or continuous interconnected single foundation. It should
also be mentioned that the effects of the rotational components on
the seismic behavior of structures can even be amplified if the effects
of the dynamic soil–structure interaction is considered.

3. Limitations

The approach developed herein is based on a few fundamental

assumptions that may limit the framework of the proposed
relations and hence engineering judgment is necessary in order
to use them in practice. Although the authors are aware of the
limitations of these assumptions, it is believed that their use may
be justified as a first step in the development of simple quantita-
tive criteria for engineering applications. Here are the major
limitations of the present study:

 The effect of the spatial variation of the body waves is solely

considered to present formulas for the rotational loading of
Fig. 13. Variation of the normalized transfer functions of quasi-static accelerations
structures. In this case, a single apparent velocity is assumed to

1=2
versus frequency for a ten-storey building (a) T qs qs 2 2
x ðωÞ; (b) T θz U Dx þ Dy
 g   g   g   g
      
where r zy ðωÞ ¼ U€ z ðiωÞ=U€ y ðiωÞ and r zx ðωÞ ¼ U€ z ðiωÞ=U€ x ðiωÞ.
Consider a uniform ten-storey building with 40 m height, which
is supported on a spread foundation with the special assumptions:
(1) the center of stiffness of all floors lie on a common vertical axes
and height of each storey is 4 m; (2) the floor systems have
characteristics: Dx ¼ Dy ¼ 40 m and N x ¼ Ny ¼ 4; and (3) the Hao’s
coherency function [10], is used and constants aj ¼ 5  10  6 and
bj ¼ 2 are considered to model the coherency loss function, and
(4) the ratios-r xy ðωÞ, r zy ðωÞ and r zx ðωÞ set to one.
The variations of the transfer function of the translational
quasi-static acceleration T qs x ðωÞ versus frequency for V x ¼ 1 km=s
and even stories of this structure is shown in Fig. 13, for which the
main direction of the seismic waves is along the x-axis and f ¼2.
From Fig. 13, it can be observed that the T qs x ðωÞ increases in the
height of building, and its maximum occurs at tenth storey of the
structure. This increase implies that the effects of the rocking
component can be destructive on the behavior of tall buildings and
parts of structural damages of such structures during SGMs can be
attributed to the rocking components. The variations of the
. take into account the wave-passage effects and therefore the
dispersion of the seismic waves is ignored. It should be noted that
there are other phenomena that may lead to the rotational
 excitations of structures, such as surface waves, special rotational

waves, block rotation, topographic effects, foundation damage,
soil–structure interaction, and asymmetric permanent strains or
settlement in the soil. Among these phenomena, the soil–struc-
ture interaction may be considered as the most important one.
 The contribution of the rotational components to the seismic
response of structures may be detrimental in the middle- and
near-field zones. In these zones, the nature of seismic wave
propagation and soil response is complicated and more recorded
data on the rotational components is required to justify the
considered assumptions or to examine their application. For the
far-field zone, Penzien and Watabe [48] showed that three
translational components are not correlated if they are decom-
posed along the radial, transverse and vertical directions of
seismic wave propagation. However, it is not generally the case
for the near- and middle-field zones and therefore, there is not a
platform, criterion or specified axis for an in-depth discussion on
the characteristics of six earthquake components.
 The numerical modeling of the foundation as a rigid system
relative to its surrounding soil may unrealistically amplify scatter-

1=2 ing of the seismic waves and lead to the increase of the rotational
normalized torsional quasi-static acceleration, T qs 2
θz U Dx þ Dy
2
, loading of structures. In this case, the soil nonlinearity, foundation
of the second floor (f¼ 2) versus frequency shown in Fig. 13 embedment, and the possibility of sliding, separation (uplift), and
indicate the value for this storey and all stories above it are the damage of the foundation during the SGMs may significantly
same and shown as storey 2. At the second step, it is assumed that affect scattering of the seismic waves as well.
 M.R. Falamarz-Sheikhabadi, M. Ghafory-Ashtiany / Soil Dynamics and Earthquake Engineering 75 (2015) 220–233 231

 The superposition of the effects of the kinematic and dynamic Appendix A

soil–structure interaction is only feasible assuming a linear
behavior for the foundation and its surrounding soil. Consider- The matrix of the stress distribution corresponding to Eq. (22)
ing the fact that soil becomes nonlinear at very low strains, the is defined as following matrix:
interaction between these two phenomena during the SGMs 2 3
may appreciably affect the FIMs and the structure response. τx ðx; y; ωÞ 0 0
6 τy ðx; y; ωÞ 7
6 0 0 7
6 7
6 0 0 σ z ðx; y; ωÞ 7
6 7
½G ¼ 6 7 ðA1Þ
6 0 0 yσ z ðx; y; ωÞ 7
6 7
6 0 0  xσ z ðx; y; ωÞ 7
4 5
 yτ x ðx; y; ωÞ xτy ðx; y; ωÞ 0
4. Conclusions

This study tries to provide a deeper insight into the rotational in which τx and τy are the shear stresses under the foundation
loading of structures in the middle-field zone and presents (1) simple along the x- and y-direction over the area of the foundation, and
methods for the estimation of the earthquake rotational components σ z is the normal stress under the foundation obtained using the
and their SDFs; (2) a procedure to include effects of kinematic soil– reciprocal theorem of elastodynamics [41].
structure interaction in evaluating the FIMs of the continuous The participation matrix of the pth single foundation to the
interconnected single foundation systems; (3) a comparison between FIMs corresponding to Eq. (29) is considered as:
the torsional and translational acceleration response spectrum with 2 3
kxp 0 0 0 0 0
considering foundation effects, and (4) a discussion on the effects of 6
the rotational components on the seismic loading of multi-storey 6 0 kyp 0 0 0 0 7
7
6 7
h i 6 0 0 kzp 0 0 0 7
buildings. The following conclusions may be drawn based on the 6 7
K^ p ¼ 6 ðA2Þ
numerical results of the present study: 6 0  kyp zp kzp yp kθxp 0 0 7
7
6 7
6 kxp zp 0  kzp xp 0 kθyp 0 7
4 5
 The combined action of time delay and loss of coherency may  kxp yp kyp xp 0 0 0 kθzp
significantly increase the amplitude of rotational components.
 Because the rotational motions have more high-frequency
content than the translational ones, the foundation input
rotational components are more dependent on the effects of
the kinematic soil–structure interaction and due to the founda-
tion presence, their PGA considerably reduce in comparison to Appendix B
the PGA of the translational ones.
 Rotational acceleration components can be detrimental in the Herein, as an example for practical application, the foundation
seismic loading of stiff structures; however, their influences on transfer functions corresponding to a simplified form of Eq. (30)
the dynamic responses of structures decrease rapidly as are derived step-by-step. In a special case of a flat foundation with
increasing the periods of structural vibration. In addition, the equal stiffness for all single foundations, co-location of the
contribution of the surface waves to the acceleration response coordinate system at the stiffness center of the foundation system,
spectrum of the rotational components may be negligible. and propagation of the seismic waves along the x-axis, ϕ ¼ 0; the
 Even though the FIMs may have smaller values of PGA and less six-component accelerations of the FIMs become:
high frequency contents than the free-field motions, they have 8 9
larger amplitudes in low-frequency parts of the acceleration >
> X
N >
>
>
> 1 g
u€ xp >
>
response spectrum in comparison to the corresponding free-field >
> N >
>
>
> p¼1 >
>
>
> >
>
motions. Thus, peak ground velocity or peak ground displacement >
> >
>
>
> X
N >
>
of the FIMs may be larger than corresponding free-field motions. >
> 1 g
u€ yp >
>
>
> N >
>
 For the multi-storey or tall buildings, the influences of the seismic >
>
> p¼1 >
>
>
>
> >
>
loading of the rotational components on the upper parts of the >
> X
N >
>
>
> g >
>
structures may be larger than on the lower parts of structures. >
>
1
u€ zp >
>
n o >
< N >
=
 For a safe seismic design of the ordinary multi-storey buildings, u€
F
¼
p¼1
ðB1Þ
adding extra rigidity to the floor diaphragm of lower stories, >
> X N
g k X N >
>
>
>
>
>
kθx
θ€ xp þ z g
yp u€ zp >
>
>
>
preferably first storey, can lead to the reduction of the dama- > K θx K θx p ¼ 1 >
>
> p ¼ 1 >
>
ging effects of the accidental torsion due to the SVSGM. >
> >
>
>
> X N XN >
>
>
> € g kz >
>
>
>
>
kθy
K θy θ yp  x €
u
p zp
g
>
>
>
>
> K θy p ¼ 1 >
>
>
>
p ¼ 1 >
>
>
> >
>
>
> X
N
k X N
k X N >
>
> θ
g
€  x g y g >
> € p yp >
€
kθz
>
> K θz zp y u
p xp þ x u >
>
: p¼1 K θz p ¼ 1 K θz p ¼ 1 ;
Acknowledgments
N 
P
where K θx ¼ kθxp þ y2p kzp , and K θy and K θy can similarly be
The research conducted by the authors has been funded by the p¼1
International Institute of Earthquake Engineering and Seismology obtained. For the case with the not closely correlated translational-
(IIEES) under Grant Number 7722. This support is gratefully acknowl- rotational motions, the diagonal elements of the SDF of the
n o
edged. The authors wish to thank Prof. M. D. Trifunac of the University
foundation input acceleration, SF0 ðωÞ , can be expressed by
of Southern California for his valuable suggestions and comments on
the first version of the paper. We also would like to thank the two defining the coherency between single foundations with
pffiffiffiffiffiffiffiffiffiffiffiffiffi
anonymous reviewers for their constructive comments. γ pq ¼ Spq = Spp Sqq , and the correlation between translational
232 M.R. Falamarz-Sheikhabadi, M. Ghafory-Ashtiany / Soil Dynamics and Earthquake Engineering 75 (2015) 220–233

pffiffiffiffiffiffiffiffiffiffiffiffiffi
components with ρxy ¼ Sxy = Sxx Syy as follows: (27) and (B4) as follows:
8 9 8 F9 8 g g 9
> Sgx ðωÞ
UAN > > € > > T Fx U€ x  yT Fθz U€ y >
>
>
> N2
>
>
> > Dx >
>
< >
=
>
>
<
>
>
=
>
> >
> F € g
€ g
>
> S g
ðω Þ >
> D€ ¼ T F
U þ xT F
U ðB5Þ
>
>
y
UA >
> > > y y θz y
> > >
N y
>
> N 2
>
> >
> > F g >
>
> g
ω >
> : €F > ; > : T U€ xT F U€ g þ yT F U€ g > ;
>
>
S z ð Þ
UA >
> Dz z z θy z θx z
>
> N2 N >
>
>
> >
>
>
> k
2 g
S ðω Þ kz Sz ðωÞ
2 g >
>
n o <> θx θx
2 U AN þ 2 UAy
>
= in which x and y denote the point coordinates from the geometrical
ðK θx jz ¼ 0 Þ ðK θx jz ¼ 0 Þ
diag S0 ðωÞ ¼
F
ðB2Þ center of the foundation system. In actual conditions, the phase of
>
> kθy Sθy ðωÞ
2 g
kz Sz ðωÞ
2 g >
>
>
> 2 UAN þ 2 U Ax
>
> the FIMs is different with the phase of the free-field ground motions
>
> ð θy jz ¼ 0 Þ ð θy jz ¼ 0 Þ >
>
>
>
K K >
> but its application in practice may be justified considering uncertain-
>
> 2 3 >
>
>
> k 2 g
S ðω Þ kx Sx ðωÞ
2 g k
2 g
S ðωÞ >
> ties associated with the (1) the seismic wave propagation; (2) site
>
>
θz θ z
U A þ UA þ y y
UA >
>
>
> 6 2
K θz N 2
K θz y 2
K θz x
7 >
>
>
> 6 p ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 7 >
> conditions; (3) the soil nonlinearity, and (4) the possibility of sliding,
>4
> Sy ðωÞSx ðωÞ
g g 5 >
>
>
>  ρxy
kx ky
UA >
> separation (uplift), and damage of the foundation during the SGMs.
>
: 2
K θz xy >
;

where References
!
X
N 1 X
N 

AN ¼ Nþ2 Real γ pq [1] NM Newmark, Torsion in symmetrical building. In: Proc. fourth world conf.
p ¼ 1 q ¼ pþ1 earthquake eng. Santiago, Chile 2, A.3; 1969. p. 19–32.
[2] Trifunac MD. A note on rotational components of earthquake motions on
!
X
N X
N 1 X
N 
ground surface for incident body waves. Int J Soil Dyn Earthquake Eng
Ax ¼ x2p þ2 xp xq Real γ pq 1982;1:11–9.
p¼1 p ¼ 1 q ¼ pþ1
[3] Ghafory-Ashtiany M, Singh MP. Structural response for six correlated earth-
quake components. Earthquake Eng Struct Dyn 1986;14:103–19.
!
X
N X
N 1 X
N 
[4] Basu D, Whittaker AS, Constantinou MC. Estimating rotational components of
Ay ¼ y2p þ 2 yp yq Real γ pq ground motion using data recorded at a single station. J Eng Mech
2012;138:1141–56.
p¼1 p ¼ 1 q ¼ pþ1
[5] Basu D, Whittaker AS, Constantinou MC. Extracting rotational components of
! earthquake ground motion using data recorded at multiple stations. Earth-
N X
X N
quake Eng Struct Dyn 2013;42:451–63.
Axy ¼ xp yq γ pq ðB3Þ [6] Li B, Al-Bermani FGA, Kitipornchai S. Maximum response of asymmetric
p¼1q¼1 structures subject to a multicomponent earthquake. Earthquake Eng Struct
Dyn 1993;22:1047–66.
in which Realð Þ indicates the real part of the quantity under [7] Veletsos AS, Prasad AM, Wu WH. Transfer functions for rigid rectangular
parentheses. For the case where there is no recorded seismic rotational foundations. Earthquake Eng Struct Dyn 1997;26:5–17.
[8] Rutenberg A, Heidebrecht AC. Response spectra for torsion, rocking and rigid
acceleration, each one of methods discussed in Section 2.1 may be foundations. Earthquake Eng Struct Dyn 1985;13:543–57.
used to obtain the approximate SDF of the rotational accelerations by [9] Zembaty Z, Boffi G. Effect of rotational seismic ground motion on dynamic
assuming that the propagation direction of seismic waves is along response of slender towers. Eur Earthquake Eng 1994;1:3–11.
[10] Hao H. Characteristics of torsional ground motions. Earthquake Eng Struct Dyn
principal axes. For example, substituting Eq. (7) into Eq. (B2) and
1996;25:599–610.
considering V x ¼ V y as the worst-case scenario [15], the foundation [11] Ghayamghamian MR, Nouri GR. On the characteristics of ground motion
transfer function, T F ðωÞ, can be expressed as: rotational components using Chiba dense array data. Earthquake Eng Struct
Dyn 2007;36:1407–29.
8
2 9 8 F 9 [12] M Ghafory-Ashtiany, MP Singh, Seismic response for multi-component earth-
> > > Sx ðωÞ >
> T x ð ωÞ >
F
> > > > quakes, Report no. VPI-E-84, VPI & SU, USA; 1984.
> > > g
ω >
>
2 > > >
S ð Þ
> > > > [13] M Ghafory-Ashtiany, MR Falamarz-Sheikhabadi, Evaluation influence of rota-
> F
x
> >
> >
> SFy ðωÞ >
>
>
> T ð ω Þ >
> >
> >
> tional components on the behavior of structures, Report International Institute
>
> y >
> >
> g
ðω Þ >
>
>
> > >
> >
S y >
>
of Earthquake Engineering and Seismology, IIEES, in Farsi; 2010.
> 
2 >> > >
>
> > >
> Sz ðωÞ >
F
> [14] Falamarz-Sheikhabadi MR, Ghafory-Ashtiany M. Approximate formulas for
> >
< T z ðωÞ = < Sgz ðωÞ >
>
F
= rotational effects in earthquake engineering. J Seismolog 2012;16:815–27.
T ð ωÞ ¼ 
F

2  SF ðωÞ [15] Falamarz-Sheikhabadi MR. Simplified relations for the application of rota-
> > > >
> T θx ðωÞ >
> > > >
F θx tional components to seismic design codes. Eng Struct 2014;59:141–52.
> > >
> Sgz ðωÞ > >
>
> >
> >
> >
> [16] Luco JE, Sotiropouls DA. Local characterization of free-field ground motion and
> 
> F
>
> >
2> > Sθy ðωÞ >
> F
>
> > effects of wave passage. Bull Seismol Soc Am 1980;70:2229–44.
>
>
>
>
T ðωÞ > > Sz ðωÞ >
> >
> g >
>
> θy
> > >
> > >
>
[17] VK Gupta, MD Trifunac. Investigation of building response to translational and
>
> 
>
>
2>
>
> F
ω >
> rotational earthquake excitations, University of Southern California, Depart-
>
> > >
:
S θz ð Þ >
;
: T ð ωÞ ;
F
θz Sg ðωÞ y
ment of Civil Engineering, Report no. 89-02.
[18] Trifunac MD, Ivanivic SS, Todorovska MI. Apparent periods of a building.
8 AN
9 II: Time–frequency analysis. J Struct Eng, 127; 2001. p. 527–37.
>
> >
> [19] Trifunac MD, Ivanivic SS, Todorovska MI, Novikova EI, Gladkov AA. Experi-
>
> N2 >
> mental evidence for flexibility of a building foundation supported by concrete
>
> >
>
>
>
AN
>
> friction piles. Int J Soil Dyn Earthquake Eng 1999;18:169–87.
>
> N2 >
>
>
> >
> [20] Takeo M. Ground rotational motions recorded in near-source region of
>
>
AN >
>
>
> N2 >
> earthquakes. Geophys Res Lett 1998;25:789–92.
< = [21] Huang BS. Ground rotational motions of the 1999 Chi–Chi, Taiwan earthquake
k ω2
2
k2
¼ 2 U AN þ 2 UAy ðB4Þ
θx z
as inferred from dense array observations. Geophys Res Lett 2003;30(40):1–4.
>
> V 2x ðK θx jz ¼ 0 Þ ðK θx jz ¼ 0 Þ >
>
>
> >
> [22] Kalkan E, Grazer V. Coupled tilt and translational ground motion response
>
> kθy ω
2 2 2 >
>
>
> U A þ
kz
UA >
>
spectra. J Earthquake Eng 2007;133:609–19.
>
> 2
ð j Þ
2 N
ð j Þ
2 x >
> [23] Trifunac MD. The role of strong motion rotations in the response of structures
>
>
V x K θy z ¼ 0 K θy z ¼ 0 >
>
>
> >
> near earthquake faults. Int J Soil Dyn Earthquake Eng 2009;29:382–93.
>
> k2
θ ω ð
2 1 þ r 2 ðωÞ
Þ k2
r 2 ðωÞ k 2
>
>
: z
4V 2 K 2
xy
UA N þ x
K2
xy
U A y þ K2
y
UA x;
[24] Jalali RS, Trifunac MD. Response spectra for near-source, differential and
x θz θz θz rotational strong ground motion. Bull Seismol Soc Am 2009;99:1404–15.
[25] Jalali RS, Nouripour-Azgomi M, Trifunac MD. In-plane response of two-story
 g   g  pffiffiffiffiffiffiffiffi
    structures to near-fault ground motion. Int J Soil Dyn Earthquake Eng
in which r xy ðωÞ is U€ ðiωÞ=U€ ðiωÞ, i ¼  1. Assuming the phase of
x y 2013;55:263–74.
the FIMs is the same as the phase of the free-field ground motions; the [26] Zembaty Z. Rotational seismic load definition in Eurocode 8, Part 6, for slender
tower-shaped structures. Bull Seismol Soc Am 2009;99:1483–5.
acceleration components at any point on the continuous intercon- [27] De La Llera JC, Chopra AK. Accidental torsion in buildings due to base
nected single foundation can approximately be obtained using Eqs. rotational excitation. Earthquake Eng Struct Dyn 1994;23:1003–21.
 M.R. Falamarz-Sheikhabadi, M. Ghafory-Ashtiany / Soil Dynamics and Earthquake Engineering 75 (2015) 220–233
[28] Shakib H, Tohidi RZ. Evaluation of accidental eccentricity in buildings due to
rotational component of earthquake. J Earthquake Eng 2002;6:431–45.
[29] Ghayamghamian MR, Nouri GR, Heiner Igel T. Tobita, The effects of torsional
ground motion on structural responses: code recommendation for accidental
eccentricity. Bull Seismol Soc Am 2009;99:1261–70.
[30] Li HN, Sun LY, Wang SY. Improved approach for obtaining rotational compo-
nents of seismic motion. Nucl Eng Des 2004;232:131–7.
[31] Zerva A, Zervas V. Spatial variation of seismic ground motions: an overview.
J Appl Mech Rev—ASME, 55; 2002. p. 271–97.
[32] Lee VW, Trifunac MD. Torsional accelerograms. Int J Soil Dyn Earthquake Eng
1985;4:132–9.
[33] Lee VW, Trifunac MD. Rocking strong earthquake accelerations. Int J Soil Dyn
Earthquake Eng 1987;6:75–89.
[34] Kanai K. Semi-empirical formula for the seismic characteristic of the ground.
Bull Earthquake Res Inst (Tokyo) 1957;35:309–25.
[35] H Tajimi, A statistical method of determining the maximum response of a
building structure during an earthquake. In: Proceedings of second world
conference on earthquake engineering, Tokyo, 2; 1960. p. 781–98.
[36] Zembaty Z, Castellani A, Boffi G. Spectral analysis of the rotational component
of seismic ground motion. Probab Eng Mech 1993;8:5–14.
[37] R. Ruiz, J. Penzien, Probabilistic study of the behaviour of structures during
earthquakes, Report EERC-69-03, Earthquake Engineering Research Center,
University of California at Berkeley, CA; 1969.
[38] Hasseliuan TK, Hart GC. Modal analysis of random structural system. J Eng
Mech Devision—ASCE 1972;115:561–79.
233
[39] Zerva A. Spatial variation of seismic ground motions: modeling and engineer-
ing applications, advances in engineering (A series of reference books,
monographs and textbooks). CRC Press, Taylor & Francis Group; 2009.
[40] Luco JE, Wong HL. Response of a rigid foundation to a spatially random ground
motion. Earthquake Eng Struct Dyn 1986;14:891–908.
[41] Bycroft GN. Soil-foundation interaction and differential ground motions.
Earthquake Eng Struct Dyn 1980;8:397–404.
[42] Lee VW, Trifunac MD, Feng CC. Effects of foundation size on Fourier spectrum
amplitudes of earthquake accelerations recorded in buildings. Int J Soil Dyn
Earthquake Eng 1982;1:52–8.
[43] Trifunac MD. Differential earthquake motion of building foundation. J Struct
Eng—ASCE 1997;123:414–22.
[44] Luco JE. On the relation between radiation and scattering problems for
foundations embedded in an elastics half-space. Int J Soil Dyn Earthquake
Eng 1986;5:97–101.
[45] Todorovska MI, Trifunac MD. In-plane foundation interaction for embedded
circular foundations. Int J Soil Dyn Earthquake Eng 1993;12:283–97.
[46] Stewart JP, Seed RB, Fenves GL. Seismic soil–structure interaction in buildings.
II: Empirical findings. J Geotech Geoenviron Eng 1999;125:38–48.
[47] Heredia-Zavoni E, Leyva A. Torsional response of symmetric buildings to
incoherent and phase-delayed earthquake ground motions. Earthquake Eng
Struct Dyn 2003;32:1021–38.
[48] Penzien J, Watabe M. Characteristics of 3-dimensional earthquake ground
motions. Earthquake Eng Struct Dyn 1975;3:365–73.