You are on page 1of 12

Seismic analysis of internal

equipment and components


in structures
J e r o m e L. S a c k m a n and James M. Kelly

Department of Civil Engineering, University of Cali[ornia, Berkeley, USA


(Received 10 January 1979)

An analytical method is developed whereby a simple estimate can be


obtained of the maximum dynamic response of light equipment attached
to a structure subjected to ground motion. The natural frequency of the
equipment, modelled as a single-degree-of-freedom system, is considered
to be close or equal to one of the natural frequencies of the N-degree-of-
freedom structure. This estimate provides a convenient, rational basis for
the structural design of the equipment and its installation.
The approach is based on the transient analysis of lightly damped
tuned or slightly detuned equipment-structure systems in which the mass
of the equipment is much smaller than that of the structure. It is assumed
that the information available to the designer is a design spectrum for the
ground motion, fixed-base modal properties of the structure, and fixed-
base properties of the equipment. The results obtained are simple estimates
of the maximum acceleration and displacement of the equipment. The
method can also be used to treat closely spaced modes in structural sys-
tems, where the square root of the sum of squares procedure is known
to be invalid.
This analytical method is also applied to untuned equipment-structure
systems for which the conventional floor spectrum method is mathe-
matically valid. A closed-form solution is obtained which permits an
estimate of the maximum equipment response to be obtained without the
necessity of computing time histories, as required by the conventional
floor spectrum method.

Introduction is tuned to a structural frequency, we found that for the


combined system there are two closely spaced frequencies
In this paper we present a rational approach to the design on either side of the tuning frequency around which a
of lightly damped relatively light equipment in structures band of high amplification appears, offering a substantial
subjected to seismic loading or other forms of ground target for sympathetic oscillation. The significant inter-
motion. The analysis is carried out in the context of a action of equipment and structure in this case means that
model that consists of an N-degree of freedom structure to the conventional floor spectrum method, which ignores
which is attached a single-degree-of.freedom component. that interaction, will not be valid for the transient analysis
In previous work, l we have described the steady-state problem.
response of such a system to ground motion. Significant This previous research has been extended to transient
interaction effects were shown to occur when the equip- analysis of the undamped equipment.structure inter'action
ment frequency is close or equal to one of the natural problem for tuned systems2 and slightly untuned systems)
frequencies of the structure, a situation often referred to A design ground response spectrum together with fixed-
as tuning. If the equipment frequency is not tuned to a base dynamic properties of the structure alone and of the
structural frequency, the response is roughly the super- equipment alone are used to estimate the peak response of
position of the structural response and the equipment the equipment. By taking advantage of the mathematical
response with little interaction, in which case the conven- structure of the equations and of asymptotic methods made:
tional floor spectrum method should be valid for transient possible by the smallness of the equipment mass in com-
problems. If, on the other hand, the equipment frequency parison with the mass of the structure, simple results were

0141-0296/79/040179-12$02.00
© 1979IPC BusinessPress Eng. Struct., 1979,Vol. 1, July 179
Seismic analysis of equipment and components in structures: d. L. Sackman and d. M. Kelly
obtained that are valid for tuned and nearly tuned systems. -~'~.~mEquipment S dof
In this paper we extend these results to damped tuned,
nearly tuned, and completely untuned systems.
The rationale for using design spectrum methods is that
they are inexpensive and to a certain extent incorporate
the probabilistic nature of the problem, i.e. the uncertainty
involved in specifying the structural parameters and the
earthquake or other input. These uncertainties are accounted
for in the way that the design spectrum is developed from
historical records of earthquakes and from artificial records,
and also in the way that the maximum values in each mode
are combined to predict the maximum for the entire system. '= ug It)
For light equipment and small damping, the results
Ground motion
obtained can be implemented easily and efficiently by a
Figure I Equipment-structuresystem
designer. The most important aspect of the analysis is its
extreme simplicity; namely, if the response spectrum for
the ground motion is available, the response spectrum for
the equipment can be calculated merely by multiplying the The equations of motion of the N-degree-of-freedom
former by an amplification factor. structural system take the form:
This approach is in contrast to several earlier analyses N
of equipment response, such as the floor spectrum Z (Mlj + +xii@
method 4-zz in which the equipment is treated as a single- j=l
degree-of-freedom system subject to a base motion that is N
taken to be that which the structure would experience at = T (cijRiat + Fei
the attachment point in the absence of equipment. Not i=i
0nly does this method neglect interaction, it has the further i= 1,2 . . . . . N (1)
disadvantage of requiring that an expensive time history
analysis of the structure be conducted in order to determine where MU, Cq, and Kit are the mass, damping, and stiffness
a base motion. Approximate techniques 12-is that bypass matrices, respectively, and Ul is the absolute displacement
associated computational problems have been proposed of the ith degree of freedom. The vector R i is a vector of
whereby floor response spectra are developed from ground influence coefficients introduced to couple the actual
spectra, but these are ad hoc methods whose accuracy ground motion ug(t) to the structure, and e i a vector whose
cannot be evaluated. components are zero at every degree of freedom except
It is also possible to account for the interaction by the one to which the equipment is attached, denoted by
considering the system to be an N + 1-degree-of-freedom the index r, where it takes unit value. The term F is the
model z6 which may then be treated by a modal approach interaction force between the equipment and the structure.
and spectral methods or by direct time history analysis The natural frequency ~n and mode shape ~7 of the
using an ensemble of spectrum-consistent earthquake nth mode (n = 1, 2 . . . . , N) are obtained from the
records. Neither approach is, however, ideal. If equipment- equations:
structure interaction is important two closely spaced N N
modes appear, the contributions of which must be summed n2 Z M#@7 = • Xe@7 i = 1,2 . . . . . N (2)
when a modal approach is used, but there exists no con- i=z /=z
sensus as to an appropriate summation procedure. A dis-
If the damping is assumed small enough not to introduce
advantage of the second approach, time history analysis, is
coupling between the modes, equation (1) becomes in
that it is very expensive and for a given item of equipment modal coordinates:
must be carried out for a wide range of earthquake motion.
If the equipment is light the use of the N + 1-degree-off N
freedom model with standard structural dynamics computer
codes may mask significant response. ill
The method developed here allows the design engineer k=l,2 ..... N (3)
to use directly the given ground motion design spectrum to
determine the response of the equipment and thus to avoid where:
the expense of time history analysis and the problems of N N
closely spaced modes. In fact the method proposed here Mk=ZZ *i *Tm;j
can be used to determine the correct form of summation /=1 j-I
for closely spaced modes. This will be presented in a N N
forthcoming paper. 2Bkn k = ~ ~ o,k k Cdmk
i=l j=l

and
M o d a l analysis o f e q u i p m e n t - s t r u c t u r e systems N
In this section we formulate the equations of motion Fi = ~, (C#Ri~g + KliRlut) + Fet
j=l
governing the response of a general N-degree-of-freedom
structure to which is attached equipment modelled as a with Mk the generalized mass and B k the fraction of
single-degree-of-freedom oscillator (Figure 1). critical damping for the kth mode. The Laplace transform

180 Eng. Struct., 1979, Vol. 1, July


Seismic analysis of equipment and components in structures: d. L. Sackman and J. M. Kelly

of the structural response Uj(t) is given by: The zeroes of the term in brackets on the left-hand side
of the equation must be determined to invert the Laplace
N ~¢ k--
transform by residue theory. These zeroes are the poles of
~. dPi c}i Fi
N i=! the transfer function for the equipment response. The case
vj= (4) considered here is that illustrated in Figure 2, where the
r-kX,=mk(P
l, 2 + 2BknkP + n~) equipment frequency is close to a structural frequency,
with: say Q , . The two expressions in the brackets on the left-
hand side of equation (9) have been plotted separately;
N p was replaced by iQ and the graph of the second function
?i = ~ (CitRtP + KitRt) fig + ?ei and the negative of the first function in the bracketed
I=l
expression were then drawn. For simplicity, the undamped
wh.ere p is the Laplace transform parameter and a bar case (fl = B l = . . . = BN = 0) has been plotted. The plot for
above a function denotes its Laplace transform. The corre- the first function is a simple quadratic in Q, zero when
sponding equation of motion for the equipment displace- ~2 = co, the natural frequency of the equipment. The plot
ment u is: for the summation is a complicated curve that reaches
+**when Q = ~k and k = 1,2 . . . . . N, the natural fre-
--mii = F = c(ti - ~J,) + k(u - Ur) (5) quencies of the structure. Two such curves have been
or, in Laplace transforms: plotted, one for equipment of small mass, and another for
equipment of larger mass.
P The values of 91 at the intersections of these two curves
-pXa = - - = (2~cop + co2)(fi - Ur) (6)
m locate the zeroes in the bracketed expression, where equip-
ment-structure interaction is considered. When the equip.
where m, c and k are the mass, damping and stiffness of ment mass is small, these poles, all of which are simple,
the equipment, respectively, with/3 the fraction of critical appear near the natural frequencies of the structure. Two
damping. A relationship between u and U, is obtained closely spaced poles, referred to herein as tuning poles, are
from equation (6) in the form: located near the equipment frequency and the frequency
(pX + 2Ocop + ¢ox)a = ( 2 / ~ p + cos)Or (7) of the structure to which the equipment is nearly tuned,
one below these frequencies and one above them as shown
which, from equation (4) can be written as: in the figure. These two poles coalesce into a double pole
when co = Qn and m--}0. Thus, the contribution to the
t](P 2 + 23cop + cos) = (23cop + cos) sum of the residues at all poles is dominated by the residues
associated with the two tuning poles. The contribution
~Nk k [~e N gtg] of the summation term to the residues at these two poles
~i cbr i + ~ (CnRtP + KuRD
N i=l i=l is dominated by the term k = n since the denominator
of that term is nearly zero. Hence, in the region o f p = ice,
k =! Mk(P x + 2Bk~kP + 91~) equation (9) can be approximated by:
Since P = -mpX~,P can be eliminated. The final trans- m (23cop + co2) ]
formed equation for the equipment response is then: B
[ p2 + 2flcop+ cos) + px
Mn/(~Pnr) 2 p2
[( N mpX(2~Op + coX)c}kre#kr]
pX + 2flcop+co2) + k=tZmk(pX+Xfjk91kp+Q~) j = (2~cop + cos ) ( ~ . ~ . p + ~)

N N
p2 + XBn~nP + 912n
dpkr ~. dp~ ~ (CitRtp +KitRt)
N t'--1 I=1 x E *~eP~MaRdMn fig (10)
= (2flcop + 6os) ~ ag
1 i=l
k=i Mk(P 2 + 2Bk~kP + ~ )
(s) This expression is identical to that for a two-degree-of-
N N freedom system, as shown in Figure 3. The equivalent
The expression E Kn*~ can be written as g2~ X Mn4~ expression for the system shown in that figure is:
i=1 1=1
N
L (~cop +cox) ]
and, given the assumption of small damping, ~ C u ~ can [tp' + + co2)+
i=1 d fi;J
N
['(23coP + cox)(2BftP + 912)1
be represented as 2Bk~ k ~ Mit~p~. Thus, the solution for =[ ; ~ ~ ]~r (11)
i=1
for the multidegree-of-freedom system takes the form:
where 7 = m/Mis the mass ratio. When we compare this
[ - . ~ m@~=(2flcoP+co 2) ] expression to equation (lO), we see that the effective mass
ratio is:
N N
,),eft _ m
dP~rZ O~ Z MuRt(XBkn~P + Q~)(2flcop + co~) Mn/(~)2 (12)
N i= 1 != 1
=X
k=l Mk(p x + 2B k [Xkp + [2~) and the effectiveground motion:
eft _ ~n
(9) ui - c,u 8 (13)

Eng. Struct., 1979, Vol. 1, July 181


Seismic analysis o f equipment and components in structures: J. L. Sackman and J. M. Kelly

iI, ', , 'y \


\
\
\
\

a I
Figure 2 Location of poles of equipment response transfer function with slightly detuned N-degree-of-freedom structure. (el,
location of poles when equipment mass is small. (z~), location of poles when equipment mass is not small

C Analysis o f transfer f u n c t i o n for nearly tuned


-E c
two-degree-of-freedom system
In this section we develop the transfer function for the
K
equipment acceleration; the result for the equipment
U u displacement can be similarly developed. The transformed
F i g u r e 3 Two-degree-of-freedom system giving system parameters.
equipment acceleration fi(p) for the equivalent two.degree-
12 = = K / M ; to = = K / M ; 2~oJ = c / m ; 2 B f l = C / M of-freedom system given by equation (I I) may be written
in the form:
fi = [N(p )/D(p ) ] fig (15)
wh~re:
N N where:
c", =,z,'~ Z T+ "z'~MoRi/M. (14) N(p) = (2/3¢0p + co2) [2B(I + ~)cop + ( 1 + ~)2w2] (16)
i=1 ]=1

In the subsequent development, the contribution of the and


residues from the tuning poles (which are near p = ice) will D(p) = p4 + wp3 [2/3(1 + 3,) + 2B(1 +/;)l
be obtained from an analysis of the equivalent two-degree- +~o2p 2 [2 +-r + 2f+~ 2 +4/3B(1 +~)l
of-freedom system def'med by the above equations. The
+ w3p[2~l +~)2 +2B(I +~)] + ~4(1 +~)2
contributions at the other ( N - 1) poles are straightforward
and will be considered after the two-degree-of.freedom (17)
analysis has been completed. Although it is not essential
In the above, ~ = (I2 - co)[~ is the detuning parameter.
that an equivalent two-degree.of-freedom system be con-
The nature of the solution strongly depends on the
sidered, since it is only conceptual and introduces no location of the zeroes of the denominator D(p). The roots
further approximations beyond those made in passing from
of D(p) will be close to those of the system where 7,/3, B
equation (9) to equation (10), the following development
and ~ are taken to be zero because these parameters are
will be for such a system in order to simplify notation.
small; namely, p = :t i6o. To locate the poles of D(p), we
We will use the notation of expression (11). Thus, B and
replace p in equation (17) by:
will refer to the structural quantities Bn and ~n, and 3' and
ug to 7 err and u~rr, as defined in equations (12) and (13). p = ico(l + <5) (18)

182 Eng. Struct., 1979, Vol. 1, July


Seismic analysis of equipment and components in structures: J. L. Sackman and J. M. Kelly

where 6 is a small quantity. We retain only the plus sign Green's function will be obtained by residue theory, since
since the roots will appear as complex conjugates. Substi- there are no branch cuts in the p plane. The inversion of
tution of equation (18) into equation (17) yields the the transformed Green's function for the general case
expression: (equation (20)) is obtained by writing the denominator
D ( p ) in the form:
84 + {4 --i [2fl(l +3') + 2B(I +/~)]}83
+ 14 - 3 ' - 2~ - ~2 _ 4fiB(1 + ~) - i [6fl(1 +3') D(p) = ( p -- Pl)(P --P I)(P--P2)( rJ --1~2)
+ 6B(1 + ~)1}82 + 1-23' - 4~ - 2~2 - 8flB(1 + ~) where:
- i [2fl(2 + 33' - 2~ - ~2) + 4B(1 + ~)]}8
+ { - 7 - 4flB(l + ~ ) - i[2fl(3'- 2 ~ - ~2)]} = 0 (19) pi=ico(l+~+k~_
When fl = 0, B = 0, and "r * 0, ~ * 0, this equation can be
easily solved, viz.:

6 = [l + ~"+ 27 +~2
7 + ( 7 + ~2 + ~3 +3'~ + 7~22 and Pl and P2 are the complex conjugates of Pl and P2.
Evaluating the residues at each pole and collecting complex
3'2 ~a 1/21/2 conjugate terms in pairs, we obtain the result, correct to
dominant order:
,, _~[~_, (3' +~2yl2]
Also, if 3' = 0, ~ = 0, and fle 0, B * 0, then: //G(t) = •2 +/j2 e -(fl+B)catl2 ;~ sinh -2 cot cos cot sin + cot

8 = ( 1 - f 1 2 ) I / 2 - l+ifl, (1-B2)ll2-1+iB..ifl, iB
- X cosh--cot s i n - cot cos 1 + cot
Throughout the analysis it will thus be assumed that fl, B, 2 2
and 3't/2 are all of the same order, say e, and the various
approximations for 8 will be based on 8 of order e ~ 1. - - cot cos - cot cos I + cot
When the parameters are not of the same order, the modi- 2 2
fications required are obvious.
The solution of equation (19), where terms of order e 2 p cosh p ;~
are retained, is:
,,+, The acceleration response//(t) to a specified imposed
t-+-+/[--i--+2
2 (2o) ground motion iig(t) is obtained by substituting the above
equation in equation (24). We recall that to simplify the
where here and throughout the remainder of the analysis derivation we have considered the equivalent two-degree-
the upper signs are taken together to give one root and the of-freedom system and this result provides the contribution
lower the other. The quantities h and # are given by: of the tuning poles to the response of the equipment-
structure system. These poles dominate response in the
1
h = -'~ {[3' + ~2 _ (lg - B ) 2 + 4~2 (fl -B)2} 1/2 case of light equipment, but contributions from the other
VZ + [3'+~2_(~_~)2]]tl2 (20 poles can be easily computed. To do so, recall that the
nontuning poles of equation (9) are close to their location
I for the structure alone, as indicated in Figure 2. The poles
= ....~.^~ { [3' + ~2 __ (fl _B)212 +4~2(fl_B)2}l/2 for the ruth nontuned mode are taken to be:
Vz
- [3"+~2-(fl-B)2]]U2 (22)
p = - B m ~ 2 m + i[2 m (26)
For fl * 0 and/or B ~ 0, the imaginary part of 8 is always
positive and oscillations are therefore damped. If we evaluate the residues and drop negligible terms, we
obtain, to dominant order, the contribution from the
ruth structure poles as:
Inversion o f t r a n s f o r m s o l u t i o n
1 -
C~
(~,./co)2 S2m e -Bmfzmt sinI2mt, m*
n (27)
The formal inversion of the transform expression (15) is:
1 pN(p),, pt and contributions of the same order from the tuning
if(t) = 21r'--i [ D---~n~ug(p) e dp (23) poles as:
F
N c?
where F is a suitable Bromwich path. If t~g(p) is taken to Z 1 -- (CO/~"2m)2 co e-flwt Sin(or (28)
be 1, then the inversion directly yields Green's function m=l
m:f:n
//G(t) for the solution, the essential component of the
subsequent analysis. The complete solution for the equip- where Crm is defined in equation (14). The complete solu-
ment acceleration for given ground motion//g(t) will take tion for the response of the equipment then takes the
the form: form:
t t
4

//(t) =/ //G (t - r)/2g(r) dr (24) ii(t) ii,(r) =, 1 - ( ~ m / c o ) 2 Qme-BraI2rn( t -- ¢)


O 0 m~n

Eng. Struct., 1979, Vol. 1, July 183


Seismic analysis of equipment and components in structures: d. L. Sackman and d. M. Kefly

x sin~2m(t-r ) +
m = 1 I -- (CO/~"~m) 2
m~n
e-a~o( t - r)
/ Structure
fcurv¢
x sinco(t - r) + C,"iio(t - r ) l dr
!

where//c is given by equation (25) with 7 replaced by


(29)
/i
1'err as given in equation (12).
The character of the two parts of the solution in equa- f~
tion (29) differs. The contributions from the nontuning
poles and the nondominant contributions from the tuning
poles are conventional and would attain their peaks during "A x ~ cEuqrU~epment
the ground excitation or shortly thereafter. The dominant
response from the tuning poles, on the other hand, is con-
trolled by the energy transfer from the structure to the
equipment through beating, which takes a relatively long (~)Structurepote (~ Equipment pete
time 2,~ and the peak response of this contribution will E
occur substantially after the end of the ground motion. Figure 4 Location of poles of equipment responsetransfer func-
The peak response of grossly untuned systems, i.e. where tion for grossly detuned systems. ~), structure pole; (~, equipment
the equipment frequency is far from all structural frequen- pole
cies, can be estimated in the same way. For light equipment,
the structure poles are only slightly shifted from their
location for the structure alone, namely: A p p r o x i m a t e forms using response spectra
p = -Bm~ m + i~"~m (30) The results obtained in the previous section can be applied
to the design of equipment or equipment mounting. The
Additional poles due to the equipment occur close to the least complicated form is that for the undamped system
fixed-base poles: and has been treated elsewhere.2, 3 Methods developed
here for damped and tuned or untuned systems are exten-
p = -/3co + i¢o (31)
sions of the method developed in references 2 and 3.
as shown in Figure 4 where th¢ response of the completely Commonly, a designer begins with a design spectrum
undamped system is illustrated. The residues at the struc- specified by a code or obtained by averaging several possible
ture poles are as before with the contributions from each inputs as, for example, in seismic design. Is, t9 In references
m = 1 to N set of poles given by equation (27). The 2 and 3 we were able to show how the appropriate form
residues at the equipment poles, equation (31), contribute of equation (25) for an undamped system can be used to
a term similar to equation (28) to Green's function in the obtain estimates of maximum acceleration when the
form: information available is the response spectrum of the
ground motion//g.
Three approaches to the design problem were developed.
1 I -- (CO/~'~m) 2 CO e - ~ ' sincot (32) These were the floor spectrum method, the modified
ground motion spectrum method, and the amplified ground
The derivation is standard and similar to the terms from the motion spectrum method. The floor spectrum method was
structure poles. The complete response for the equipment demonstrated to be invalid for undamped tuned systems,
in the grossly untuned case is thus given by: and the modified ground motion spectrum method, while
valid, to be inconvenient. The amplified ground motion
t spectrum was the most convenient for use and led to the
following estimates of maximum acceleration and displace-
f
//(t) = //~,(r)
0
i Cm
1 -- (~'~m/(.JO) 2 am e-Bmam(t--*') ment:
SA o)
M c m I//l=,x = - -
7t/2 (34)
x s i n [ 2 m ( t - r) + ~ co e - a w ( t - r )
m =1 I - - ( ~ 0 / ~ m ) 2
and

x sinc~(t- r)} dr (33) so( , 0)


l u l = , x = - ~t/2
- (35)
This result can be used as an alternative to time history or where SA (co,/3), Sn(¢o, 8) are the acceleration and displace-
modal analysis of the composite N + 1-degree-of-freedom ment design spectra for frequency co and damping factor/3.
system, or as an alternative to the standard floor spectrum The simplicity of the result can be explained on physical
analysis for which the time history of the point of the grounds. In weakly coupled systems with the same fre-
attachment must be computed. Equation (33) is indepen- quency, the response of the system involves a perfect energy
dent of the modal mass ratios, representing, in fact, the exchange between each component at a beat frequency
general closed-form solution of the floor spectrum method. much lower than the natural frequency of each component.
Indeed, the preceding analysis, which led to equation (33), The same phenomenon - a classical beat phenomenon -
mathematically justifies the use of the floor spectrum occurs here. The coupling is weak because the ratio of
method for the grossly untuned system. equipment mass to structure mass is small.

184 Eng. Struct., 1979, Vol. 1, July


Seismic analysis of equipment and components in s~ructures:J. L. Sackman and J. M. Kelly
When a structure is subjected to a ground motion, the
velocity imparted to the structure is mass independent and
determined only by the ground motion. Thus, if the same Peek response
ground motion were applied directly to tuned equipment, \
the same velocity would be transmitted to it. Kinetic _o(t)
~" j EnveLol~ curve
energy, on the other hand, is proportional to the mass of
the system excited; in equipment, that energy would be
much smaller than in a structure. However, if the equip-
ment were attached to a structure and the structure sub-
jected to a ground motion, the kinetic energy imparted to
the latter would be wholly transmitted to the equipment,
lllII 'IHLL:
if tuned, and the velocity imparted would be amplified by ~ 'Peok response
the reciprocal of the square root of the mass ratio.
Damping is clearly important in this process because the 2 n/u: Period of oscittation
energy transfer requires many cycles and much of the
kinetic energy in a damped system could be dissipated Beet period-2rt/~l
before being transmitted. If for a damped system, 3' '~ 1, F~ u m 5 Equipment response history in case of damped beats
then 3' is negligible if4~B ~, 3' (equation (17)), and equip-
ment-structure interaction can be neglected. The floor where:
spectrum method can, in principle, be used to determine
t
the response of the equipment for this case. However, this
method requires that time histories be computed, and the
numerical time-integration algorithms available in structural
dynamics computer programs are not sufficiently reliable 0 = tan -1
f /is(r)
0
cos r/r e -(a +B)w(t-r)/2 cos co(t - r) dr

to calculate the late-occurring peak response of equipment. t


Furthermore, if only the design spectrum and not the
history of the ground motion is available, then the conven-
tional floor spectrum method cannot be used. Amplified
f /is(r)
o
sinr/r e -(a+B)c°(t-T)/2 cosco(t - r)dr

response spectra results equivalent to equations (34) and (38)


(35) for the damped system can be developed using the
dominant contribution from the tuning poles. The develop- We consider r/t i = 2~q/T,~ I, where t t is the duration
ment will be carried out for the equivalent two-degree-of. of ground motion and T is the beat period of the system.
freedom system to simplify the presentation and notation. Then the first integral in equation (37) can be approxi-
In order to emphasize the physical and analytical basis of mated by:
the result we will apply it to the special case of the damped t
tuned system with 3' > (/~-B) 2. Other cases can be
handled in exactly the same way.
When ~ = 0 and 3' > (/3-B) 2 the contribution from
f
0
/is(r) e - ( # * B ) w ( t - r)/2 cosco(t - r) d r

the tuning poles in the general equation (29) becomes, in


and the second neglected since sinr/t will be bounded by
the notation of the equivalent two-degree-of.freedom
r/tl '~ 1 and/i s = 0 for t > q . Thus, we take:
system: t
t
/i(0-- /i (O co sin r/t (
/i(t) = [3' - 7a--E- W/ J u,(,-)
0 %r- '-.-/ j t¢
o
x [e -(0÷B~(t-r)/2 cosco(t - r)sin~7(t - r)] dr
× e -C~+8)~(t-~')/2 c o s ~ ( [ - r ) d r (39)
(36) When the parameters 71/2,/~ and B are small, this result
whe re:
may be interpreted in the following way: for t > q , the
n = [3"- (~-B)21~/2co/2 above expression can be written in the form:
We note that the term in brackets in the integral represents
a damped beat motion as shown in Figure 5 with a damping //(t) = - ¢°2sinr/t e -(t~ +a)totl2R cos(cot- ~)
factor (#+B)/2 and a beat period T given by 21r/r/. When 2rl
the term sinr/(t - r) is expanded: where:
co R = (,4,2
/fit) = [3' - (/3-B)2] ~1~cos(r/t - 0)
with:
t 2
t!
x [( f /is(r)e-(#*a)t°(t-r)12 ~osco(t- r) cos~?rdr) A ~ =f iis(t) e+O~*n)'°t/2 coscot dt
o
o
t 2
+ ( f /i,(r)e-~+B)ta(r-r)12 cosco(t--r)sinrlrdr) ] 112 t I

A 2 = f iis(t) e*O3+a)t°t/2sincot dt
o
(37) 0

Eng. Struct., 1979, Vol. 1, July 185


Seismic analysis of equipment and components in structures: J. L. Sackman and J. M. Kelly

and where Iv(i)l < Iv(~)l. It then follows that:


= tan-~(A2/Al)

The response indicated by the above is illustrated in


Figure 5. In the above, the terms R and ff are constants
independent o f t for t > t~, and R cos(tot- if) is a rapidly = Iv(i)l e-(a*a)t°(:-~)/Zlcosco(t* - i)1
varying function of time. The term:
< I v(F) l e- ( a + s ) ~ : O - e/r*)/z
¢O2 Iv(i') I e-(a +s),,,:/2 (47)
e -(#+a)WU2 sinrlt
217
since i/t*< I. From this we obtain the approximate result:
is a slowly varying envelope curve whose maximum value
must be determined. The maximum value of this envelope Iv(f) l ~ e(#+B)°~r°/2
curve is attained at time t*, expressed by: =

27
tanr/t* = (40)
(~ + a ) ~ ,

The value of t* is thus: (48)


We recognize, however, that to the order of/3 and B Iv(i)l
t* = arctan [2r//a~(# + B)]/r/ (41) is very nearly the pseudo-velocity response spectrum
For lightly damped systems and light equipment, in general Sv(co, (I~+B)/2) for a lightly damped single-degree-of.
t*~, t I. The values of sin r/t and exp [-(/3 +B)tot/2] when freedom oscillator of frequency co and damping factor
the maximum of the envelope is achieved are: (1~+B)/2 subjected to the ground acceleration//g(t). Thus,
an estimate of the maximum equipment acceleration is:
~7
sinrtt* = = c° [sinr/t*[ e-K (6~ ~ 8 )
[,72 + ((3 + B)z~o2/4]l/2
I//Jma~ [3'_(#_B)2]l/Z Sv ,
e -(/~+B)t°:/2 = e -K (42)
where: With the value of sinr/t* from equation (42), we obtain the
final estimate as:
K = (arctan ~')/~" (43)
= [3"- ((3-B)2]U2/([3+B) (44) [ /3+B\
It follows that:
I//Imax (3"+4(3B)V 2 (49)
lalma~ = I//(t*)l = [3'_(I~_B)21j/2 Isinnt*le -~
Recalling that:
t*
cosy = SA = cO2SD
x{e(#+B)(°t°/2lf
i~g(T)e
-(~+B)c°(t*-r)/2
this estimate can be written in the alternative form:
O X COSCO(t*--r)drl} (45)
[//(t) [max (3"+4{3B)u2SA , (50)
In order that this estimate of peak acceleration be
useful for design purposes, it is necessary that the second and, similarly, for the displacement:
factor in braces be interpreted in terms of a ground response
spectrum. To this end, we recognize that the integral is, to
the order of/~ + B, the relative velocity response history [u(t)lmax (3'+4~B)U2 SD , (51)
evaluated at time t* of a lightly damped single-degree-of-
freedom oscillator of frequency co and damping factor Estimates for other cases of damped tuned systems, e.g.
(8 +B)/2 subjected to the ground aeceieration//~(t). At where 3' < (/~-B) 2, 3' = (/3 - B ) 2 and 3' = B 2,/~ = 0, have
some time i'during the ground motion or shortly after it been developed2O; precisely the sameresult was obtained.
ceases (so t h a t / ' < t*), the absolute value of the relative I f # + B is fixed, the maximum value of 4#B and the mini-
velocity will attain its global maximum, denoted as Iv(t')l. mum value o f ( ~ - B ) 2 are achieved when/~ = B, which
The relative velocity response at t*, denoted as v(t*), can yields the smallest value of the amplification factor,
be thought of as that which would occur in a single-degree- (y+4#B)-l/2e -k. Thus, if total damping is fixed, it should
of-freedom system (subjected to the ground acceleration optimally be shared equally by equipment and structure
~g(t)) as a consequence of free vibration beginning at time for the case developed above.
( > t t ) when the absolute value of the relative velocity of For a slightly untuned system, the response'is consider-
the oscillator attains its first local maximum, Iv(t)h after ably modified and is given by:
the end of the earthquake. This instant of time ? is equal
t
to Fiftr occurs after the end of the ground motion; other-
wise, ~>/r. In any event, ~ < t*. Thus, we can write: //(t) = (x 2 + ~:) i i . ( r ) e- < ~ ÷ B ) ° ' < ' - ' / ~
Iv(t*)l = Iv(?)le-(a+a)"~(:-~)/21cos~(t*-i)l (46) 0

186 Eng. Struct., 1979, Vol. 1, July


Seismic analysis of equipment and components in structures: J. L. Sackman and J. M. Kelly,

{ 2. 2
(,, and:
= e-" (co+a /3+B
lulm~,x (7+~2+4/3B)I/2 3D x ~ ' 2 / (55)
+Xcosh2co(t-r)sin~co(t-r)cos(1 + ~ ) c o ( t - r)
For all cases considered a universal result applies: the
appropriate response spectrum evaluated at the average
+l~shrh;co(t-r)cos~co(t-r)cos(l + ~ ) c o ( t - r) damping and average frequency of structure and equipment
is multiplied by the amplification factor:
+ /~cosh2co(t - r)sin ; co(t -r)sin (l + ~)co(t - r)}dr e-g
(3' + ~2 + 4fiB)l/2 (56)
(52)
where:
where X and # are defined in terms of ~, 7,/3 and B in
equations (21) and (22). K = (arctan [')/~" (57)
The reduction of this general form to the required = [3[ + ~2 -- (fl _B)2]l/2/(fl +B) (58)
amplified response spectrum form is illustrated by consider-
ing the case of optimal damping for 3' > (fl - B ) 2, namely Owing to the presence of the four terms in the general
13= B. For this damping, equation (52) takes the form: form, equation (52), a closed-form solution cannot be
t obtained for t*~ However, we have carried out careful and
ii(t) = - (r) e-(#÷B)~°(t-r)12 thorough numerical experiments, having computed the,
peak value of equation (52) for a variety of choices of the
0 parameters leading to expressions in which all four terms
appear. In every case the formulae (54), (57) and (58)
x sin co(t - r) COS \ yield highly accurate estimates of the peak response.
The contribution to the equipment acceleration from
(53) the other modes for which no interaction need be con-
where ~ = (7 + ~2)1/2. When the term sin(h/2)co(t- r) is sidered is quite standard. Using equation (29) in conjunc.
expanded and we recall that for t ~ t l, the duration of the tion with the square root of the sum of squares procedure
ground motion, the integral containing sin(),/2)coT can be the estimate of the maximum acceleration in the early
neglected, we obtain: peak, which will occur during the ground motion, is
given by:
ii(t) = - co e-(fl+B)Wt/2 sin - cotR cos(cot - ~b)
~, 2 cy s. (a,,, B,,,)
where:
I//Imax = L
m=l
"I" (a,,,lw)~ '

m~n
R = (A 2 +A2) 1/2 and ~b = tan-l(A2/At)
+
with: 1 1 - (~/a,,,) ~ S,~(co,~)
tt m ,kn

Al =f iig(t)e+(#+B)~°t/2cos(l +~)cotdt (s9)


where n is the structural mode to which the equipment is
o tuned or nearly tuned. In the notation of the general
tl system the result equation (54) for the later peak, from
A2 = f iig(t) e+(#+B)wt/2sin (l + ~) col dt the tuning poles, is:
0 = Cn e -K {co+[2n .fl+Bn I
I~l,..x I "l(?ert+~2+4/3Bn)l/2Sa\ 2 2 /
The slowly varying envelope function exp [-(/3 +B) cot~2]
sin(h/2)cot reaches maximum at a time t* such that: (60)
X h where:
sin- tot* =
2 [X2 + (fl +B)2]1/2 K = (arctan~')/~"
and: and
e-(fl+B)oJt*12= e -K ~- = [.),eft+ ~2 _ (/3_Bn)2]l/2/(/3+Bn)
where: The detuning parameter is//= (~n - co)/~o, and 7 en is given
K = (arctan ~')/~" by equation (12). For fight equipment and lightly damped
closely tuned systems, the second peak is likely to be more
= Xl(/3 +B) important. However, the early peak could be the larger and
From this result and the argument used for the damped both should therefore be evaluated.
tuned cases,the estimated maximum response value is: The response in the case of a completely untuned system
where the equipment frequency is not close to any struc-
e -K ( . ~ + ~ fl+B~
tural frequency can be estimated in the same way. Using
I//Imax = (7+~2+4/3B)1/2 SA 2 ' 2 '/ (54) equation (33) and the square root of the sum of squares

Eng. Struct., 1979, Vol. 1, July 187


Seismic analysis o f equipment and components in structures: d. L. Sackman and J. 114.Kelly

procedure, the estimate of the peak response is: ~o a


20
I/~l,.ax = C."' 2 0 ~, . . . . . -
..... .....
12 15 18

+ 1' s](~o, ~)j,,,:


1 I - (~-~',.Qm)2J "°° !b
200
(61) o
This result can be used as an alternative to time history -zoo
or modal analysis of the composite N + 1-degree-of- -too( . . . . . 3 . . . . . 6 . . . . . 9 . . . . . 12 15 18
freedom system, or as an alternative to the standard floor
spectrum analysis for which time histories must be com- 1-00
C
puted. 200
C ,^.AAAA IAAAAA AA4AAr ^^ . . . .

_~-a°°~ -- ~vvvvt VVVVVVJVVVVV ,vv .......


Application to equipment mounting design
_,.o.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3 6 9 12 15
Although the results developed above can be used dkectly
in the form given we have recently completed extensive
numerical and physical experimentation which have
suggested ways in which the results can be conveniently
used in practical application to equipment mounting
design.
The physical experiments were carried out at the Earth-
quake Simulator Laboratory of the Earthquake Engineering
~'OO
Research Center of the University of California, Berkeley, 2ooi e
and involved the use of a three-storey steel frame model (31 ._~ . . . . . . . . . ,_
structure around one-third full scale. The response of the
-~00_1
o, . , . . . . . . . . . .
steel frame model is roughly that of a shear building with
0~0 3 6 9 12 15 18
natural frequencies at around 2 Hz, 8 Hz and 15 Hz. Time,(sec)
The tests were conducted on the twenty-by-twenty foot Figure 6 Time history of acceleration for shaking table, third floor
shaking table of that facility. Each floor of the model and oscillators, a, table acceleration; b, third floor absolute accelera-
frame carried concrete blocks weighing a total of 8000 lb. tion; c, first mode oscillator acceleration; d, second mode oscillator
The total weight of the model was 39 500 lbs and its height acceleration; e, third mode oscillator acceleration
was about 20 feet. Three single-degree-of-freedom oscil-
lators were attached to the concrete blocks at the second magnification of accelerations experienced by the oscillators
and third floor levels to simulate equipment in a primary is immediately obvious. The beat phenomenon is clear. The
structure. The test structure was instrumented to measure peak acceleration in the first two oscillators is achieved
displacements and accelerations at each floor and accelera- considerably after the peak acceleration in the input. These
tion of the oscillators. The mechanical oscillators were two oscillators are obviously responding at the coupled
constructed to correspond to the first three natural frequen- frequencies governed by structure-equipment interaction.
cies of the model structure. Each oscillator consisted of a The response of the third oscillator is quite different;
vertical cantileve beam fixed at its base with a mass on top. a highly irregular pattern appears and local maxima occur
Each mass was a 6" x 6" x 2" piece of steel weighing about during as well as after excitation. To clarify this response,
20 lb. The arms of the cantilevers were made of ¼" thick cold Fourier transforms have been taken of the acceleration
rolled steel plate and were slotted so that the position of time histories of the three oscillators shown in Figure 6.
the masses could be adjusted to change the frequency. These transforms are shown in Figure 7. It is clear from
The first three natural frequencies of the frame were these transforms that the first mode oscillator responded
determined accurately using a Rockland spectrum analyser. only at the first mode frequency around 2 Hz; the second
The oscillators were then each damped in turn to a rigid mode oscillator responded predominantly at the second
fixture and tuned with the spectrum analyser to the struc- mode frequency around 8 Hz with a small contribution
tural frequencies. After they had been tuned, the oscillators from the first mode. The dominant contribution to the
were bolted to the concrete weights on the model structure response of the third mode oscillator was from frequencies
using concrete anchors. The first and second mode oscil- around the third mode frequency of 15 Hz, but significant
lators were attached to the third floor and the third mode contributions also appeared from the lower modes. Thus
oscillator was placed on the second floor. These positions the pure damped beat response predicted by the theory
were chosen so that each oscillator would be subjected to was obscured by the contributions from lower modes.
the greatest possible amplitude input in its mode. The These observations are supported by the numerical
Pacoima Dam (1971) and Taft (1950)earthquake signals experiments (to be described in detail elsewhere) and have
as well as time-scaled Paeioma Dam, Taft, and E1 Centre implications for the application of the theoretical results
(1940) signals and some approximate square waves were to design. First, only the structure modes up to and around
used as input to the shaking table. Figure 6 shows the that of the equipment under consideration need be con-
time history of acceleration in a typical run. The top trace sidered and, second, the late peak acceleration given by
shows the input table acceleration, the second the response equation (60) should be summed with the early peak
of the third floor of the structure, and the next three the acceleration given by equation (59) by an appropriate
accelerations of the three tuned oscillators. The very large summation rule. For instance, if the conventional square

188 Eng. S t r u c t . , 1979, V o l . 1, J u l y


Seismic analysis of equipment and components in structures: J. L. Sackman and J. M. Kelly
['OO. N
Crm = ~ ~ (¢gT~MiiRl)/Mm - participation factors
o.7~ il=l
RI vector of influence coefficients that couples
0,.50
ground motion to structural degrees of freedom
~n /th component of ruth mode shape
025
Mi] structural mass matrix
.J N
5 I0 1,5 20
Mm = ~ Mtlrbrfl~Tz generalized mass for mth mode
l,l
tOO
K = (arctan ~')[~"
0.75 = + - +B.)

0.50
For systems that are completely untuned, we recall
equation (61):
0'25

k_, SA(~..,~m ' B m


q i ,¢ •
5 I0
i i i i
15
. . . .
ZO
I//Imax = m=l "1 -- (~-~m/CO) 2

0.10 ]2 s](co, t3)}'n


1 l--(~-Tam) 2]
0.(38
In the above, S.4 is the acceleration design spectrum;
other responses, such as maximum velocity or displacement,
005
can be obtained by using the appropriate design spectra.
The advantages of this approach are its simplicity snd
0.03 ~ ~ .~---..~
adaptability for practical application. A great deal of com-
putational effort is avoided since time history analyses
OI .... IO. . . . '
need not be performed. The equipment and structure need
Frequency,Hertz H~
Figure 7 Fourier transforms o f oscillator time history accelerations.
not be analysed as an N + 1-degree-of-freedom system
a, first mode oscillator; Io, second mode oscillator; e, third mode either by modal or matrix-time.marching methods, and
oscillator errors in estimates of peak response due to the possible
root of the sum of squares rule is used, the estimate is: unreliability of numerical time-integration schemes, or to
uncertainty as to the appropriate procedure for summing
the contributions of the two closely spaced modes, are
I//Ima x = ~. '1 - - ( a m / c o ) 2 thereby avoided. For tuned and nearly tuned systems it
m=l takes into account the important effect, completely neg-
men
lected in the floor spectrum method, of equipment-
structure interaction.
+ s (co, 0) The method advanced here does not require that new
,1 -
m~n
information be generated. Data available from the building
design alone (Mii , Ri, S2m, cI~t , Bin), the equipment alone
r ICrnle -K [CO'I'~'~n ~ "1"~n'~']2~ 1/2

+ 7 '
.)jj (m, co,/3), and the ground shock spectra ( S o or S y or Sa)
are used. The estimates of peak response have been obtained
by rational analysis and are easily evaluated and conveni-
If the late peak occurs well after the peak of the excitation ently used during the design process.
(as in the first mode oscillators) it will be the dominant
term and the result will be nearly the same asif it alone were
considered. On the other hand, if the late peak occurs Acknowledgements
during the excitation (as in the third mode oscillator) This research was partially supported by the National
then it should be superposed with the other modal contri- Science Foundation under Grant No. ENG77-05197, and
butions which may occur around the same time and this by the Defense Nuclear Agency of the United States
will be effected by the formula. For summary purposes, the Department of Defense under Contract DNA001-78-C-0388
terms in the above equation are as follows: with Weidlinger Associates, Menlo Park, California, which
~, co, B m and I2m equipment and structural damping support is gratefully acknowledged.
and frequencies, respectively
n structural mode to which equipment is tuned or References
nearly tuned l Kelly, J. M. and Sackman, J. L. 'Equipment-structure inter-
r structural degree of freedom to which equipment is action at high frequencies', Rep. No. DNA 4298T, Defense
attached Nuclear Agency, Washington, D.C., 1977
=--(I2 n - co)/co - detuning parameter 2 Kelly, J. M. and Sackman, J. L. 'Response spectra design
methods for tuned equipment-structure systems', d. Sound
m mass of equipment Vibration, 1978,59 (2), 171
7 =-M n / ( ~ n ) 2 = effective mass of structure in nth mode 3 Kelly, J. M. and Sackman, J. L. 'Shock spectra design
methods for equipment-structure systems', Bull Shock Vib.
= mass ratio (In press)

Eng. Struct., 1979, Vol. 1, July 189


Seismic analysis of equipment and components in structures: J. L. Sackman and J. M. Kelly
4 Biggs, J. M. and Roesset, J. M. In 'Seismic Design for Nuclear 13 Hadjian, A. H. 'Some problems with the calculation of seismic
Power Plants" (ed. R. J. Hansen), MIT Press, Cambridge, forces on equipment', Proc. Struct. Des. Nucl. Plant Facilities,
Massachusetts, 1970 Chicago, 1 7 - 1 8 December 1973, American Society of Civil
5 Kapur, K. K. and Shao, L. C. 'Generation of seismic tloor Engineers, New York, 1973
response spectra for equipment design'. In Proc. Struct. Des. 14 Peters, K. et al. 'The problem of resonance in the evaluation
Nucl. Plant Facilities, Chicago, 1 7 - 1 8 December 1973, of floor response spectra', Trans. Fourth Int. Con]~ SMIR T,
American Society of Civil Engineers, New York, 1973 Vol. K(b), paper K 4 / l l , 1977
6 Amin, M. et al. 'Earthquake response of multiply connected 15 Penzien, J. and Chopra, A. K. 'Earthquake response of an
light secondary systems by spectrum methods', Symposium on appendage in multi-storey building', Proc. Third World Conf.
Seismic Analysis of Pressure Vessel and Piping Components, Earthquake Eng.. Vol. 2, New Zealand, 1965
ASME, New York, 1971 16 Schmitz, D. and Peters, K. 'Direct evaluation of floor response
7 Newmark, N. M. 'Earthquake response analysis of reactor spectra from a given ground response spectrum', Trans.
structures', Nucl. Eng. Des., 1972, 20 (2), 303 Fourth Int. Conf. SMIRT, Vol. K(b), paper K4/10, 1977
8 Newmark, N. M. et al. 'Seismic design spectra for nuclear 17 'Combining modal responses and spatial components in
power plants', Proc. ASCE Power Div., 1973, 99 (PO2), 287 seismic response analysis', U.S. Nuclear Regulatory Commis-
9 Sato, H. 'The response spectrum analysis of building-appendage sion, Regulatorl' Guide 1.9.2, 1976
system to an artificial earthquake with two ground predominant 18 Housner, G. W. In 'Earthquake Engineering' (Robert L. Wiegel,
periods', Proc. Second Int. Conf. SMIR T, paper K 6/2, 1973 ed.), Chapter 5, Prentice-Hall, Englewood Cliffs, New Jersey,
I0 Tsai, N-C. 'Spectrum-compatible motions for design purpose', 1970
J. Eng. Mech. Div., Proc. ASCE, 1972, 98, EM2,345 19 Newmark, N. M. and Rosenblueth, E. 'Fundamentals of
11 Scanlan, R. L. and Sachs, K. 'Earthquake time histories and Earthquake Engineering', Prentice-Hall, Englewood Cliffs,
response spectra', J. Eng. Mech. Div., Proc. ASCE, 1974, 100, New Jersey, 1971
EM4,635 20 Sackman, J. L. and Kelly, J. M. 'Rational design methods for
12 Hadjian, A. H. 'Earthquake forces on equipment in nuclear light equipment in structures subjected to ground motion',
power plants',./. PowerDiv., Proc. ASCE, 1971, 97, (PO3), Rep. UCB/EERC- 78/19, Earthquake Engineering Research
649 Center, University of California, Berkeley, 1978

190 Eng. Struct., 1979, Vol. 1, July

You might also like