Beruflich Dokumente
Kultur Dokumente
Earthq Eng & Eng Vib (2013) 12: 307-317 DOI: 10.1007/s11803-013-0173-1
1. Department of Civil Engineering, Chonbuk National University, Jeonju, Jeollabuk-do 561-756, South Korea
2. Department of Civil and Environmental Engineering, Hanyang University, Ansan, Gyeonggi-do 426-791, South Korea
Abstract: In this study, a new lumped-mass-stick model (LMSM) is developed based on the modal characteristics of
a structure such as eigenvalues and eigenvectors. The simplified model, named the “frequency adaptive lumped-massstick
model,” hasonly a small number of stick elements and nodes to provide the same natural frequencies of the structure and is
applied to a nuclear containment building. To investigate the numerical performance of the LMSM, a time history analysis is
carried out on both the LMSM and the finite element model (FEM) for a nuclear containment building. A comparison of the
results shows that the dynamic responses of the LMSM in terms of displacement and acceleration are almost identical to those
of the FEM. In addition, the results in terms of floor response spectra at certain elevations are also in good agreement.
Keywords: new lumped-mass-stick model; frequency adaptive; nuclear containment buildings; seismic response; floor
response spectrum
natural frequencies and eigenvectors. The new LMSM two points provides a new Point “j”. Moreover, Point “c”
developed in the present study provides the same is close to Point “f” and linear interpolation provides a
natural frequencies of the actual structure and a high new Point “i”. Finally, the four points (“a”, “i”, “j”, and
accuracy in dynamic responses. Detailed procedures “h”) are the new lumped-mass locations for the first to
and examples are presented in the next sections. Note fourth modes. This procedure leads to the same number
that the development of the study is limited to an axial- of lumped-masses as the number of modes considered.
symmetric structure considering their lateral modes and
its elastic behavior. 2.2 Stiffness matrix of LMSM
In the LMSM, equivalent bending and shear
2 Frequency adaptive LMSM stiffnesses are considered in order to represent the
stiffness variation of the actual structure. These
2.1 Determination of lumped-mass locations stiffnesses are defined by the equivalent flexural and
shear rigidities, EIeq and GAs,eq respectively. Herein,
The locations of the lumped-masses of the stick the parameter As is an effective shear area. If the
model are determined in this new approach by sectional shape is uniform or a regular shape, then the
investigating the mode-shapes of the structure. The stiffnesses are calculated by closed-form formulas. As
number of lumped-masses is taken to be equal to the an alternative way to calculate the equivalent flexural
number of target modes or frequencies of the structure, and shear rigidities, a pushover analysis is performed by
based on either a modal mass participation ratio or the applying an arbitrary lateral load (force or displacement)
maximum frequency required by the design response to the FEM, as shown in Fig. 2.
spectrum. A modal mass participation ratio above The total lateral stiffness, kl, evaluated from the
about 90% is generally acceptable (Varma et al., 2002; pushover analysis is related to the bending and shear
Chopra, 2006), but values over 90% can be considered if stiffnesses as follows:
more accurate responses are desired. Figure 1 shows the
procedure to determine the lumped-mass locations. The k b ⋅ ks
kl = (1)
first four mode shapes are presented as an example. The k b + ks
mode shapes or eigenvectors are obtained through an
eigenvalue analysis on the FEM of the actual structure where kb and ks are the equivalent bending and shear
of concern. If there are several nodes on each layer stiffnesses, respectively. For the separation of the total
of the FEM, a representative average deigen vector is stiffness into the flexural and shear stiffnesses, one
considered for each layer. The dots in Fig. 1 indicate of the stiffnesses needs to be defined by a theoretical
the vertexes of each mode shape. As shown in Fig. 1(a), formulation. Normally, the shear stiffness is selected
Point “b” of the second mode is close to Point “d” of since it has a simpler form than the flexural stiffness. The
the third mode. Interpolating these two points linearly, remaining stiffness is then calculated by using Eq. (1).
the new Point “e” on the right hand side of Fig. 1(a) is The equivalent rigidities are calculated by using the
obtained. Point “c” remains where it is since there is separate stiffnesses and the length of the stick element.
no adjacent point. Note that the first mode is omitted The pushover analysis (static method) may be more
because its vertex is always at the top of the structure accurate in the evaluation of each stiffness and rigidity
(Point “a”). The three points (“a”, “c”, and “e”) are if the actual structure has a non-uniform cross-sectional
the lumped-mass locations for the first to third modes. shape. The equivalent flexural and shear rigidities are
Interpolation is then made with the fourth mode, as input into the LMSM implemented in a computational
shown in Fig. 1(b). Since Point “e” is close to Point “g” frame structural analysis platform (Reinhorn et al.,
of the fourth mode, linear interpolation between these 2009; Computers and Structures, 2011).
2nd mode 3rd mode Three nodes Three nodes 4th mode Four nodes
Fig. 1 Interpolation procedure for lumped-mass locations: for example (a) between 2nd and 3rd modes and (b) combined with 4th mode
No.2 Hwasung Roh et al.: New lumped-mass-stick model based on modal characteristicsof structures 309
If the lateral stiffness of the stick elements is required KL is the static condensation stiffness matrix of the stick
in matrix form, theoretical formulations can be used. It model defined in Section 2.2. The parameter ωt is the
is well known that the unit force method provides the matrix diagonally assembled with the eigenvalue of
flexibility matrix, which by inversion gives the stiffness the actual structure. In this study, the matrix is named
matrix. On the other hand, the unit displacement method a target eigenvalue matrix for the LMSM. The matrix
yields the stiffness matrix directly. The lateral stiffness d is the corresponding maxim normalized eigenvector
matrix can also be obtained from the LMSM by using matrix for the same mass locations of the LMSM
a computational frame structural analysis program. defined in Section 2.1 (vertexes of each mode shape).
Arbitrary lumped-masses are input at the locations or The preliminary mass matrix M L0 obtained from Eq. (2)
nodes determined in Section 2.1 and the eigenvalue is neither symmetric nor diagonal since the eigenvector
analysis is performed to obtain the eigenvector and matrix d is not orthogonal. In order to obtain a
eigenvalue matrices. Using the eigen-results and the diagonal mass matrix, the summation of the rows of
arbitrary lumped-mass matrix, the stiffness matrix is the mass matrix is performed as shown in Fig. 3. This
simply evaluated by the relationship k = ( T ) −1 2 T m , consideration provides a mutual mass distribution for
which holds since the stiffness matrix of the stick model each node to cover all mass influence coefficients for u i
does not depend on the distribution of the lumped- as far as possible. In the present study, the diagonal mass
masses. This procedure yields a static condensation matrix obtained by summing the rows of the preliminary
stiffness matrix due to the consideration of the lateral mass matrix ( M L0 ) is named the “initial shooting mass
modes. matrix” and indicated by M L* .
Performing an eigenvalue analysis using the
2.3 Frequency adaptive lumped-mass matrix stiffness matrix (KL) and the initial shooting mass
matrix ( M L* ) provides a new eigenvalue matrix (ωL) and
In the conventional LMSM, the nodal masses are corresponding eigenvector matrix ( L). These matrices
Mass influence co- Mass influence coefficients Mass influence coefficients ML*
efficients for u1 = 1 for u2 = 1 for un = 1
Fig. 3 Preparation for the initial shooting mass matrix
310 EARTHQUAKE ENGINEERING AND ENGINEERING VIBRATION Vol.12
are related as shown below: Repeating Eqs. (6) to (9) leads to a zero value of the
error matrix (ΔML=0) since the eigenvectors of Eq. (6)
K L = ( LT ) ( )
−1 −1
( L )
2
L
(3) become adaptive to the target eigenvalues during the
iterations. Once the mass matrix is diagonal from Eq. (7),
Herein, the matrix L is a mass normalized eigenvector the eigenvalues are equal to the target eigenvalues.
matrix. However, the obtained eigenvalues are A flow chart of the iterative numerical procedure is
required to be equal to the target eigenvalues (ωt). For shown in Fig. 4. In the present study, the procedure
this purpose, Eq. (3) is rearranged, adding the target is implemented in MATLAB (MATLAB, 2010) and
eigenvalues (ωt) as follows: the diagonal mass matrix is the “frequency adaptive
lumped-mass matrix”.
K L = ⎡( LT ) ( )( ) ⎤⎥⎦ ⎡⎣( ) ( )⎤⎦ ⎡⎢⎣( ) ( ) ( )
⎤
−1 T T −1 T −1 −1
⎢⎣ L t ⎥⎦
t t t L L
M L0 (Eq.(2))
Diameter of inside:
∗
45.72 m
Initial shooting mass matrix, M ( Fig.3)
L
Thickness of wall:
24.08 m
1.22 m
Eigenvalue analysis, ( M L∗ , K L )
L , L
77.27 m
new = (L )( L ) −1 ( T )(Eq.(6))
L , L
53.19 m
M L = (new
T
) −1 (new ) −1 (Eq.(7))
Eigenvalue analysis (ML, KL) (Eq.(9))
No
ΔML=0
Yes
Frequency adaptive lumped-mass matrix, ML
Fig. 4 Flowchart for the frequency adaptive lumped-mass matrix Fig. 5 Three-dimensional-FEM of the nuclear containment building
No.2 Hwasung Roh et al.: New lumped-mass-stick model based on modal characteristicsof structures 311
3.2 Frequency adaptive LMSM case special techniques are needed such as the energy
equivalence and average deflection methods (Varma
Using the linear interpolating procedure, as shown et al., 2002). In the use of Eq. (1), the shear stiffness of
in Fig. 6(a), the LMSM of the NC building is obtained, the dome part is assumed to be infinite since the shear
as shown in Fig. 6(b). The number of mass locations deformation of such a dome typed structure is very small
is four since the first to fourth modes are considered. while the flexural deformation of the section and the
Node number 3 is at the location where the dome starts. axial action through the dome thickness are relatively
The second moment of inertia (I) and the effective apparent. Table 2 summarizes the flexural and shear
shear area (As) are calculated by using a simple closed rigidities of the each stick element and the corresponding
formula (geometric method). These values are used to static condensation stiffness matrix is shown in Eq. (10).
define the equivalent flexural and shear rigidities of the Applying 1.0 m of the lateral displacement to the top of
cross-section. Herein, the effective shear coefficient ( ) the building, Fig. 7 shows the comparison of the overall
is 0.5 for the circular-hollow cross-section. Regarding lateral stiffness between the FEM and LMSM. The
the upper part of the NC building, as shown in Fig. LMSM provides 6,066,371 kN/m of the lateral stiffness
6(c), a pushover analysis is conducted to obtain the while the FEM provides 6,280,373 kN/m, the difference
total lateral stiffness. Arbitrary displacement loading is is 3.41%.
applied to all nodes positioned at the same elevation of Using the initial shooting mass matrixas shown in
the top thickness. Displacement loading is considered, Eq. (11) and following the procedure described in Section
rather than force loading, since it makes it easier to 2.3, Fig. 8 shows the variation of the mass matrix (4 by
calculate the total stiffness of the dome part. If arbitrary 4 matrix) during the iterations. All off-diagonal terms
force is applied, the displacement of each node can be of the matrix converge to zero. The total mass obtained
different due to local concentration effects, and thus it from the iterations is 27,776,894 kg which is about 84%
is not easy to determine the representative displacement of the total mass of the actual structure. The variation of
corresponding to the arbitrary force applied. In such the eigenvalues during the iterations is shown in Fig. 9.
24.08 m
41.61 m
19.65 m
21.15 m 20.85 m
14.51 m 12.69 m
2nd mode 3rd mode Three nodes Three nodes 4th mode Four nodes
77.27 m
53.19 m
33.54 m
12.69 m
Fig. 6 LMSM of the nuclear containment building: (a) interpolation for lumped-mass locations; (b) present model and (c)
pushover analysis for the dome part
312 EARTHQUAKE ENGINEERING AND ENGINEERING VIBRATION Vol.12
The eigenvalues of the first and third modes converge to One massless node is added, as shown in Fig. 11,
the target values earlier than those of the other modes. in order to get a floor response spectrum, which will be
Table 3 compares the final or converged eigenvalues and discussed in Section 3.4. The addition of the massless
natural frequencies obtained from the iterations, which node does not affect the dynamic response of the
are almost identical. lumped-mass nodes, and does not change the natural
frequencies of the LMSM. It just provides additional
⎡158672308 −62247472 2156575 4847706 ⎤ → u1 response information at the location of the massless
⎢ −62247472 119464044 −671922002 8296482 ⎥ → u
⎥ kN node. In the present study, such massless nodes are
KL = ⎢ 2
named “optional nodes” while the other nodes (lumped-
⎢ 2156575 −67192002 108882416 −44563245⎥ m → u3
⎢ ⎥ mass nodes) are called “essential nodes”.
⎣ 4847706 8296482 −44563245 29809005 ⎦ → u4
The time history analysis is performed with a
(10) time increment of 0.005 s, considering 5% modal
constant damping ratio. Figures 12 and 13 compare the
displacement, acceleration, and shear force responses
⎡8878535 0 0 0 ⎤ of the LMSM to those of the FEM. Note that in both
⎢ 0 10421351 0 0 ⎥
⎥ kg cases, the results are almost identical. Figures 12(b),
M L0 = ⎢ 13(b) and 13(d) capture the responses up to 10 s in order
⎢ 0 0 6542418 0 ⎥
⎢ ⎥ to show the similarity of the response phase. Figure 14
⎣ 0 0 0 55410503⎦
compares the peak responses at each floor. As shown
(11) in Fig. 14(a), the maximum difference of the peak
displacement between the FEM and LMSM occurs at
3.3 Time history analysis the third floor, while the responses at the other floors
are very close. Figure 14(b) gives a comparison between
The ground motion shown in Fig. 10 is generated the story drift ratios of the two models. Note that the
based on the US NRC RG 1.60 design spectrum,which story drift ratio at the top story in the LMSM is bigger
is normally applied for the seismic design of nuclear than that of the FEM while the drift ratio at the other
containment (NC) buildings (Regulatory Guide, 1973). stories in the LMSM is smaller than that of the FEM.
The total duration is 30 s as per the Standard Review Plan However, the differences are very small. Figure 14(c)
recommendations (Standard Review Plan – Section 3.7.1, shows the peak floor acceleration response. Except for
2007), which recommends that the total duration of the the third and top floors, the peak acceleration response
artificial ground motion be long enough and more than 20 s in both models is almost identical. The differences in
(Standard Review Plan – Section 3.7.1, 2007). The peak peak acceleration at the third and top floors are 0.08 g
ground acceleration (PGA) of the horizontal excitation is and 0.04 g, respectively. Overall, the responses of both
0.3 g for a Safe Shutdown Earthquake (SSE, Regulatory models are very close since the natural frequencies of the
Guide, 1974). LMSM are the same as those of the full FEM within the
5
Base shear (106kN)
2
FE model
1 Stick model
(2.5-scale factor) 0
z
z
0 0.25 0.50 0.75 1.00 1.25
x x
Lateral displacement at top (m)
Fig. 7 Static displacement profile in the FEM and LMSM
No.2 Hwasung Roh et al.: New lumped-mass-stick model based on modal characteristicsof structures 313
8
0 0 0
7458324 kg
4 -0.5 -0.5 -0.5
0 8813656 kg 0 0
6
-0.5 -0.5 -0.5
3
-1.0 -1.0 -1.0
1 10 20 30 1 10 20 30 1 10 20 30 1 10 20 30
1.0 1.0 12 1.0
m31 m32 m33 m34
0.5
Mass (106 kg)
0.5 9 0.5
0 0 0
6
4011830 kg
-0.5 -0.5 -0.5
3
-1.0 -1.0 -1.0
1 10 20 30 1 10 20 30 1 10 20 30 1 10 20 30
1.0 1.0 1.0 12
m41 m42 m43 m44
0.5 0.5 0.5
Mass (106 kg)
9
0 0 0
6 7493083 kg
-0.5 -0.5 -0.5
3
-1.0 -1.0 -1.0
1 10 20 30 1 10 20 30 1 10 20 30 1 10 20 30
Number of iteration Number of iteration Number of iteration Number of iteration
Target
Eigenvalue (2nd mode)
Target
Iteration Iteration Iteration
Iteration
25 75 150 200
22 65 120 150
1 10 20 30 1 10 20 30 1 10 20 30 1 10 20 30
Number of iteration Number of iteration Number of iteration Number of iteration
(a) eigenvlaues of the LMSM (unit of eigenvalue: rad/s)
1.2 1.2 Target 1.2 1.2 Target
Target Target
ωt /ωL (1st mode)
Iteration Iteration
Iteration Iteration
1.5 0.4
Spectra acceleration (g)
Acceleration (g)
Achieved response spectrum
0.2
1.0
0
0.5
-0.2 5 10 15 20 25 30
Time (s)
0 -0.4
0 10 20 30 40
Frequency (Hz)
Fig. 10 Design spectrum RG 1.60 (5% damping) and artificial ground motion (PGA = 0.3g)
design spectrum range (below 33 Hz, see Fig. 10). Also, Node # 77.27 m Top floor
in the system equation (mass-damping-stiffness) of the 4
LMSM, the resisting and input inertia forces ( M L u and
M L 1ug , respectively) are simultaneously reduced due 53.19 m 3rd floor
to the reduced mass amount. Herein, ML is a frequency 3
adaptive mass matrix. Figure14(d) shows the peak shear
response for each story and shows a good agreement in 2
33.54 m 2nd floor
the time domain. From the LMSM, the ratio of the peak Optional node
20 m Optional floor
(massless
base shear corresponding to the total weight is 0.686. It node) 12.69 m 1st floor
is larger than that of the FEM, which is 0.652, but the 1
difference is only 0.035, resulting in a 5% error. Base
0
3.4 Floor response spectrum Fig.11 LMSM of the nuclear containment building: adding
a massless node (left side) and first mode shape in SAP
Floor response spectra (FRS) are required for 2000 (right side)
the seismic design of NC buildings since extensive
mechanical equipment, such as nuclear reactors and a damping ratio of 2% and 5% to the SDOF system.
electric cabinets, are installed there (Leonardo et al., The LMSM provides very similar floor response
2007; Curreri et al., 1983). The LMSM developed in spectrums, compared to the FEM. Due to the small
the present study can provide the acceleration response damping ratio applied, the damped frequencies, which
at any floor level by adding “optional nodes”, as shown correspond to the spikes of the spectral acceleration, are
in Fig. 11. The optional nodes are massless nodes. Since close to the first and second natural frequencies of the
most of the equipment is usually placed on the first and building.
second floors (Ahn et al., 2006; Leonardo et al., 2007),
the location of the optional node is selected at a height
of 20 m from the base. To generate a floor response 4 Remarks and conclusions
spectrum, the acceleration response at a given floor is
obtained and used as ground motion for a single degree A frequency adaptive lumped-mass-stick-model
of freedom (SDOF) system. Figures 15 and 16 show the (LMSM) has been developed, based on the variation
FRS results at the second and optional floors, providing of the frequencies and eigenvectors of a structure. A
(a)
Time (s)
(b)
Time (s)
Fig. 12 Displacement time history response: (a) entire duration and (b) up to 10 s
No.2 Hwasung Roh et al.: New lumped-mass-stick model based on modal characteristicsof structures 315
(a) 1.0
0.5
Acceleration at
2nd floor (g)
0
-0.5
-1.0
Time (s)
(b) 1.0
Acceleration at
2nd floor (g)
0.5
0
-0.5
-1.0
Time (s)
(c) 0.30
2nd floor (106kN)
Shear force at
0.15
0
-0.15
-0.30
Time (s)
(d) 0.30
2nd floor (106kN)
Shear force at
0.15
0
-0.15 10
-0.30
Time (s)
Fig. 13 Time history response: (a) entire duration of the acceleration; (b) up to 10 second; (c) entire duration of the shear force
and (d) up to 10-second
(a) (b)
Optional Optional
(b) (d)
Optional
. . .
Fig. 14 Peak floor responses: (a) displacement; (b) drift ratio; (c) acceleration and (d) shear force
316 EARTHQUAKE ENGINEERING AND ENGINEERING VIBRATION Vol.12
(a) (b)
Fig. 15 Floor response spectrum at the second floor: (a) 2% damping and (b) 5% damping
(a) (b)
Fig. 16 Floor response spectrum at the optional floor: (a) 2% damping and (b) 5% damping
and Structures, Inc., Berkeley, CA. 19th International Conference on Structural Mechanics
Curreri J, Costantino C, Subudhi M and Reich M (1983), in Reactor Technology (SMIRT-19), Toronto, Canada.
“Seismic and Dynamic Qualification of Safety-related Regulatory Guide (1973), “Design Response Spectra for
Electrical and Mechanical Equipment in Operating Seismic Design of Nuclear Power Plants,” Regulatory
Nuclear Power Plants: Development of a Method to Guide 1.60-revision 1.0, U.S. NRC, Washington, DC.
Generate Generic Floor-response Spectra,” NUREG/ Regulatory Guide (1974), “Instrumentation for
CR-3266, Brookhaven National Lab., Upton, NY. Earthquakes,” Regulatory Guide 1.12, Revision 1.0,
Halabian AM, El Naggar MH and Vickery BJ (2002), U.S. NRC, Washington, DC.
“Nonlinear Seismic Response of Reinforced-concrete Reinhorn AM, Oikonomou K, Roh H, Schiff A and
Free-standing Towers with Application to TV Towers Kempner L (2011), “Modeling and Seismic Performance
on Flexible Foundations,” The Structural Design of Tall Evaluation of High Voltage Transformers and Bushing,”
Buildings, 11(1): 51–72. MCEER-11-0006, Multidisciplinary Center of
Halabian AM and Kabiri S (2004), “Soil Structure Earthquake Engineering Research, University at Buffalo,
Interaction Effects on Inelastic Response of R/C Stack- the State University of New York, Buffalo, NY.
Like Structures,” Proceeding of 13WCEE, Vancouver, Reinhorn AM, Roh H, Sivaselvan M, Kunnath SK,
Canada. Valles RE, Madan A, Li C, Ozer C and Park YJ (2009),
Huang YN, Whittaker AS and Luco N (2010), “Seismic “IDARC2D Ver. 7.0: A Program for the Inelastic
Performance Assessment of Base-isolated Safety-related Damage Analysis of Structures,” MCEER-09-0006,
Nuclear Structures,” Earthquake Engineering and Multidisciplinary Center of Earthquake Engineering
Structural Dynamics, 39(13): 1421–1442. Research, University at Buffalo, The State University of
Jadhav MB and Jangid RS (2004), “Response of Base- New York, Buffalo, NY.
isolated Liquid Storage Tanks,” Shock and Vibration, Roh H, Oliveto ND and Reinhorn AM (2012),
11(1): 33–45. “Experimental Test and Modeling of Hollow-core
Leonardo TS, Richard SO, Sener T and Diego PR Composite Insulators,” Nonlinear Dynamics, 69(4):
(2007), “Finite Element Modeling of the AP1000 1651–1663. .
Nuclear Island for Seismic Analyses at Generic Soil and Standard Review Plan-section 3.7.1 (2007), Seismic
Rock Sites,” Nuclear Engineering and Design, 237(12- Design Parameters, NUREG-0800, U.S. NRC,
13): 1474–1485. Washington, DC.
Malhotra PK (1997), “Method for Seismic Base Takada T and Yamaguchi K (2002), “Two-step Seismic
Isolation of Liquid-storage Tanks,” Journal of Structural Limit State Design Procedure Based on Non-linear
Engineering, 123(1): 113–115. LRFD and Dynamic Response Analyses,” Structural
MATLAB (2010), The Math Works, Inc. Natick, Safety, 24(2-4): 397–415.
Massachusetts. Varma V, Reddy GR, Vaze KK and Kushwaha HS
Moussa L and Ali Z (2010), “Approximate Earthquake (2002), “Simplified Approach for Seismic Analysis of
Analysis for Regular Base Isolated Buildings Subjected Structures,” International Journal of Structural Stability
to Near Fault Ground Motions,” Proceedings of the and Dynamics, 2(2): 207–225.
World Congress on Engineering 2010, London, U.K. Wilson JL (2003), “Earthquake Response of Tall
Paskalov A, Saudy A and Elgohary M (2007), “Response Reinforced Concrete Chimneys,” Engineering
of ACR-1000TM Nuclear Power Plant to Eastern North Structures, 25(1): 11–24.
America High Frequency Input Motions,” Proceeding