You are on page 1of 96

l

/Wh

ISSN 1018-5593

Commission of the European Communities

SeJJIKKftEESSP^EflSi^S

Application of the random vibration


approach in the seismic analysis of LMFBR
structures - Benchmark calculations

Report
EUR 14153 EN

Commission of the European Communities

echnology
Application of the random vibration
approach in the seismic analysis of LMFBR
structures - Be nchmark calculations
A. P reumont,1 S. Shihab,1 L. Cornaggia,2
M. Reale,3 P . Labbe,4 H. N o e r
1

3
4

Belgonuclaire
2
Ansaldo
University of Firenze
Electricit de France

Contract No RA1-0108

Final re port
This work was performed under the Commission of the European Communities
for the Working Group 'Codes and standards'
Activity Group 2 'Structural analysis' within the Fast Reactor Coordinating Committee

Publication of this report has been supported by the Value programme for the dissemination
and utilization of the results of EC research and technological development activities
(Dissemination of Scientific and Technical Knowledge Unit, Commission of the European Communities,
Directorate-General for Telecommunications, Information Industries and Innovation, Luxembourg)

Directorate-General
Science, Research and Development

1992

PARL EUROP. Biblioth.


N.C. EUR 14153
C1.

Published by the
COMMISSION OF THE EUROPEAN COMMUNITIES
Directorate-General
Telecommunications, Information Industries and Innovation
L-2920 Luxembourg

LEGAL NOTICE
Neither the Commission of the European Communities nor any person
acting on behalf of the Commission is responsible for the use which might
be made of the following information

Cataloguing data can be found at the end of this publication

Luxembourg: Office for Official Publications of the European Communities, 1992


ISBN 92-826-4287-9
ECSC-EEC-EAEC, Brussels Luxembourg, 1992
Printed in Luxembourg

FOREWORD AND EXECUTIVE SUMMARY


The Commission of the European Communities is assisted in its actions regarding fast
breeder reactors by the Fast Reactor Coordinating Committee which has set up the
Safety Working Group and the Working Group Codes and Standards (WGCS). The
latter's mandate is to harmonize the codes, standards and regulations used in the EC
member countries for the design, material selection, fabrication and inspection of
LMFBR components.
The present report is the revised final report of CEC Study Contract RA1-0108-B
performed under WGCS/Activity Group 2 : Structural Analysis. It was executed by
BELGONUCLEAIRE as prime contractor and Ansaldo and EdF as sub-contractors.
The seismic analysis of structures has been one of the important areas addressed by
Activity Group 2 over the years. In previous studies it had been shown that the
random vibration approach could be an effective and attractive tool in seismic analysis
of nuclear power plants. It was therefore decided to set up a benchmark exercise to
compare the random vibration approach with the two traditional methods : the time
history analysis and the response spectrum analysis.
In this benchmark exercise BELGONUCLEAIRE performed the random vibration
analysis, Ansaldo the time history analysis and EdF the response spectrum analysis.
Two cases were treated : a two degree of freedom system and a simplified multidegree
of freedom system. The latter contained a primary structure for which floor response
spectra were calculated and a secondary structure (piping) for which support reactions
and bending moments at selected locations were the required output.
The following conclusions can be drawn from this study :
The statistical variability between different time history analysis corresponding
to different accelerograms generated from the same imposed ground
spectrum is large (of the order of 2). This means that several time history
analysis need to be performed in order to reduce the variability.
The response spectrum method can lead to substantial errors for
multisupported structures.
In the cases treated in this study, the average random vibration results were
in good agreement with the average time history results.
The random vibration method is therefore concluded to be a reliable method
of analysis. Its use is recommended, particularly for preliminary design,
owing to its computational advantage on multiple time history analysis.
L.H. Larsson
CEC, DG.XII/D/1

- T A B L E

O F

CO

N T E N T S

Pages

1. INTRODUCTION

1.1. Background
1.2. Scope

1
2
3

2. DEFINITION OF THE EXCITATION

3. TWO DEGREES OF FREEDOM SYSTEM

10
3.1. Effect of crosscorrelations
3.2. Decoupling, nonclassical damping
3.3. Benchmark calculation
4. MULTIDEGREE OF FREEDOM SYSTEM
4.1. Structural model
4.2. Floor response spectra
4.3. Secondary structure
4.4. Remark

11
14
19
19
22
24
31

5. CONCLUSIONS, RECOMMENDATIONS

33

6 . REFERENCES

35

APPENDIX 1

1.
2.
3.
4.
5.

6.
7.
8.
9.
10.

SEISME A COMPUTER PROGRAMME FOR THE RANDOM VIBRATION


ANALYSIS OF MULTISUPPORTED STRUCTURES UNDER SEISMIC EXCI
TATION
37
Introduction.
37
Equation of motion
39
Effective model mass.
^
Excitation
47
Inputoutput relationships in the frequency domain.
49
5.1. Modal amplitudes^
49
5.2. Absolute accelerations
49
5.3. Support reactions
50
Calculation of the PSD matrices
53
Stress calculations.
55
Postprocessing of the PSD functions.
57
Flow chart of the programme.
59
References
61

APPENDIX II
1.
2.
3.
4.

SOME RANDOM VIBRATION RESULTS ON DECOUPLING AND NONCLAS


SICAL DAMPING.
63

Introduction
Asymptotic response of the two D.O.F. system as >0
Numerical results
References

63
65
69
71

1.

I N T R O D U C T I O N

1.1. BACKGROUND.
This benchmark exercise was decided following a state of the art
review [1] which concluded that the Random Vibration (RV) approach could be
an effective tool in seismic analysis of Nuclear Power Plants (NPP), with
potential advantages on Time History (TH) and Response Spectrum (RS) techniques.

As compared to the latter, RV provides an accurate treatment of multi-

support excitations, non-classical damping as well as the combination of high


frequency modal components.

With respect to the former, RV offers a direct

information on the statistical variability


cheaper computations.

(probability distribution) and

From the point of view of its drawbacks, RV is based

on stationary results, and requires a Power Spectral Density (PSD) input


instead of a RS.
It was therefore decided to set up a benchmark exercise to compare
the three methods from the various aspects mentioned above, on one or several
simple structures.
The share of work was agreed as follows :
- BELGONUCLEAIRE would perform the RV analysis with the programme SEISME ;
- EdF would perform the RS analysis with SISPOU ;
- Ansaldo would perform the TH analysis with PAS.
The programme SEISME is a general software for the stationary analysis of linear structures subjected to multisupport random excitations with
arbitrary correlation.

It is described in Appendix I.

It has been written

to be used as a post-processor of a general purpose Finite Element (FE) programme and is currently connected to SAMCEF.

1 -

1.2. SCOPE.
It was decided that the following aspects would be covered with the
simplest possible models :
(1) statistical variability ;
(2) multisupport excitation ;
(3) non-classical damping.
To cover the statistical variability, it was decided to perform a
set of ten TH analyses.
Section 2).

(The generation of the excitation is discussed in

The effect of neglecting off-diagonal terms in a composite model

formed of two parts having widely different damping ratios has been analysed
with a tuned two d.o.f. oscillator.

Some results on the decoupling of tuned

primary and secondary structure have also be obtained.


Section 3 and Appendix II.

They are discussed in

The effect of multisupport excitation has been

studied on a primary-secondary structure involving two detuned beams.


results are presented in Section 4.

The

2. DEFINITION OF THE EXCITATION

The basic idea in defining the excitation was to use the NRC res
ponse spectra. However, since there is an inconsistency in its high frequen
cy asymptote (this was extensively discussed in [1]), it was decided to defi
ne the excitation as follows :
As in [2], we start from the evolutionary spectrum (Fig. 2.1)
Sx(,t) =4>Mt2

a(
e-o

j)t

(2.1)

where () is the PSD of a stationary gaussian process and

() = a Q + a.^o + a ^

a are given the same value as in [2].


are generated as follows :

(a. ^ 0)

(2.2)

From Sx(o>,t), the various inputs

(a) Generate ten statistically independent accelerograms with the programme


THGE.

A sample is shown in Fig. 2.2.

The maximum acceleration has been

left free, to avoid inconsistency.


(b) Calculate the statistics of the response spectra from the ten accelero
grams (min, mean, max). They are shown in Fig. 2.3. Note that the aver
age spectrum is close to the NRC spectrum, except for the high frequency
asymptote, as expected.

(c) As in [2], the PSD used in the RV method is defined by the energy equali
ty

V/

() = /

Sx(w,t) dt

(2.3)

where the equivalent tationary duration is 14.5 s.


in Fig. 2.4.

() is represented

S(),t)

,u.vX*&

Figure 2.1 : Evolutionary spectrum of the nonseparable NRC compatible


accelerograms.

line ui

6 8

IO

2
4 6 8
FREQUENCY

IO' 2
(HZ)

6 8 IO"

Figure 2.2 : Sample accelerogram for the Evolutionary Spectrum of


figure 2.1 .
Figure 2.3 : Comparison of the response spectra of 10 sample accelerograms
with the NRC horizontal spectrum (= 5%) .

-6

10

,
50
f Crierez)

Figure 2.4 : Power spectral density corresponding to the NRC 5%


horizontal spectrum (equivalent stationary duration = 14.5 s)

-7

3. TWO DEGREE OF FREEDOM SYSTEM

We consider the two d.o.f. oscillator of Fig. 3.1 with the follow
ing notations :

. =

Vffii

Vm2

2 = C2/2 m2 2

>

2/ml

= c

i / 2 mi
(3.1)

= <*>2/

We are interested in the case = 1, where the two oscillators are tuned and
the two modes have participation factors of the same order of magnitude.

Figure 3.1 2. d.o.f. oscillator

3.1. EFFECT OF CROSSCORRELATIONS.

It is well known that, when . 0, the two natural frequencies


become very close and the crosscorrelation between the modal responses con
tributes significantly to the total response (e.g. see [3]). In fact, assum
ing a white noise excitation and a uniform damping ratio ,

it can be shown

that

crosscorrelation
autocorrelation

._ ..
2

where e is the reduced modal spacing :

2 ~1
(^ + >2)

(3.3)

From this, it is expected that the SRSS rule which neglects the crosscorrel
ations will lead to large errors when O .
simulations reported in Section 3.3.

-10

This is confirmed by numerical

3 2. DECOUPLING, NONCLASSICAL DAMPING.

In this section, we analyse the effect of decoupling, and


Appendix II ; the findings

can be summarized as follows :

(a) When the systems are assumed decoupled, the error is on the conservative
side (this is well known) and may become extremely large, even for mass
ratio smaller than = 10" 2 (as recommended in the SRP [4]), when the
secondary structure is lightly damped.
on the ratio

The error is strongly dependent

I2/I1

(b) When ^ and 2 a r e very different (which is typical of primary concrete


secondary steel structures), neglecting the offdiagonal modal damping
terms in a coupled analysis leads to unconservative errors which may
become very large for small .

Surprisingly, when 2 " j, the error of a coupled analysis with


classical damping may be larger (and unconservative) than that of a decoupled
analysis.

This is illustrated in Fig.

3.2, which shows a chart of the va

riance ratios R and Rc defined as follows

a2(decoupled)
(exact)

Rc

"'(classical)

(3>5)

(exact)

For a given error (say 20 % ) , the diagram reveals essentially three regions :

the decoupled region, where a decoupled analysis (neglecting feedback) is


acceptable (for the given error level) ;

11

- the classical damping region, where a coupled analysis is required, but it


can be performed with the assumption of classical damping ;
- finally, the non-classical damping region, where a coupled analysis is
required and the full modal damping matrix must be considered unless unacceptable errors are made.
Charts for different damping ratios are given in Appendix II. Note
that the SRP decoupling criteria, ,u<10~ 2 , can lead to substantial errors, on
the conservative side though.

12-

10

= 10

-4

-3

DECOUPLED

10

-2

re
u
M
UI

re

lo-'j-

o.o

1 .0

Figure 3.2 : Chart of tue variance r a t i o s R, and R as functions of the


j j

c
mass and darning r a t i o s .
1
^ - 0 . 1 and
'
2^

- 13-

3.3. BENCHMARK CALCULATIONS.

In the comparison, the following example has been adopted :


m 1 1 kg

fj_ = f 2 2 Hz

il = 5 7.

2 = 2 %

Four mass ratios are considered, ranging from = 0.1 to 0.

For each one,

the time history integration is performed ten times, with the accelerograms
defined in Section 2.

Three damping models have also been considered, namely

nonclassical (exact), classical and Biggs'.

The RV analysis is carried out

with SEISME, using the exact damping distribution.

B oth the SRSS and the CQC

combination rules are considered in the RS analysis.

The results are as follows :

(a) As expected, no marked differences have been observed between the three
damping models in the TH analysis.
(b) Except in the very low frequency range, the RV results compare extremely
well with the TH analysis for direct floor spectra generation (Fig. 3.3).
(c) For low mass ratio, the statistical variability between different time
history analyses can be very large (more than 2, see Fig. 3.4.c for the
maximum relative displacement between the two masses).
(d) The probability of exceedance curves as predicted by the RV with the
Gumbel distribution and the peak factor formulae of [5] agree well with
TH results (Fig. 3.4).
Table I

compares the average peak value of the relative displace

ment |x2 X]J (m) for the various methods of analysis.

(Only those TH re

sults obtained with the exact damping distribution are reported).

It can be

seen that
(e) SRSS and CQC results diverge when

+. 0.

This is because of the

crosscorrelation problem mentioned in Section 3.1. above.


(f) RV, TH and CQC results agree reasonably well for all mass ratios.
agreement between TH and RV is particularly good.

14

The

Table I
AVERAGE PEAK VALUE OF THE RELATIVE DISPLACEMENT | x 2 - | (m)
FOR THE TWO d.o.f. OSCILLATOR

Mass ratio

Exact damping
distribution

RV

TH

RS

SRSS

CQC

0.1

0.417

0.442

0.434

0.424

Coupled

0.01

0.944

1.046

1.393

1.144

analysis

0.001

1.425

1.563

4.385

1.811

0.

1.565

1.70

0.

1.565

1.70

Decoupled

15

1.97

1.70

max J
min >Time h i s t o r y
mean )
Spectral analysis

10

10

10'
2
FREQUENCY HZ

Figure 3.3: Comparison of response spectra at node 1


for the 2 D.O.F. oscillator.
= 0.1

16

10

= 0.1

Esrom .

= 0.01

AESPOMSE . 1

'J;
.

= ./~

\ \

' '

KES'OHSC . 1

Figure 3.4 : Tuned two d..f. oscillator probability of exceedance curves


of the relative displacement |xx. for various nass ratios.
The discontinuous lines correspond to T.H. results, for various
damoin^ assunDtions.

17

4. MULTIDEGREE OF FREEDOM SYSTEM

4.1. STRUCTURAL MODEL.

The structure considered is represented in Fig. 4.1.


jected to a onedimensional excitation, along the y axis.

It is sub

The primary struc

ture consists of two sticks with lumped masses of 500 t.

The natural fre

quencies and effective modal masses are reported in Table IV/1. Mode 1 is a
local mode of the shorter stick and modes 2 and 3 correspond to the longer
one.

Note that modes 1 and 2 belong to the frequency range where the excit

ation has most of its energy (see Fig. 2.4). Also, we notice that the total
mass (1500 t) of the structure is accounted for by the three modes.

Table IV/1
PRIMARY STRUCTURE

Mode number

(Hz)

" (t)

1.99

500

2.3

808.7

10.2

191.3

1500

19

The secondary structure consists of a 800 kg L shaped pipe (elbow


flexibility factor 1.856) with three supports (x, y and displacements fixed
at nodes 3 and 22 ; y displacement fixed at node 2).

The natural frequencies

2
and effective modal masses (1 ).//". are given in Table IV/2. We observe
that mode 2 (which is tuned with mode 3 of the primary structure) accounts
for more than 60 of the total mass and that the contribution from modes 3
and 4 is expected to be small (the energy content of the excitation is small
in that frequency range).

Note that 13 7. of the total mass is not accounted

for by the model truncated after the 4th mode. One can therefore expect some
discrepancy in the support reactions, if no missing mass correction is appli
ed.

Finally, note also that the secondary structure satisfies, by far, the

usual decoupling criteria [4].

Table IV/2
SECONDARY STRUCTURE

Mode number

(Hz)

(riy)J/^1 (kg)

4.48

67.6

9.66

507.5

18

26.44

Crly)^^

0.10
14.6

589.8

108.7

00 y

Missing mass

20

101.5

Modal damping of respectively 5 and 2 7. have been assumed for the


primary and secondary structures.

The following response quantities are compared :

- floor response spectra at nodes 2, 3 and 22 ;


- support reactions along y at nodes 2, 3 and 22 ;
- bending moment M x at node 7 and M z at node 18.
*

node 3 : 500

node 2 : 500

Figure 4.1 : System geometry.

-21 -

4.2. FLOOR RESPONSE SPECTRA.


Floor response spectra obtained from TH analysis are compared to
those directly generated from RV in Fig. 4.2. For the TH analysis, ten accelerograms have been used.
ported on each diagram.

The maximum, minimum and average spectra are reFor the RV analysis, average floor spectra are ob-

tained from the corresponding PSD using average peak factor formulae [5].
From the scatter between the maximum and minimum curves in the TH
analysis, one can see that the spectral ordinates obtained by direct integration range in a ratio of about 2.
frequency range.

This ratio is uniform over the entire

It is also observed that, except for very low frequencies

(below 0.5 Hz), the mean floor spectrum obtained by RV is very close to the
average of the ten samples in the TH analysis.

(The discrepancy in low fre-

quency is due to the high inaccuracy of the peak factor formulae in that
frequency range).

22

nax
mean
nin

^ cine histo ry

I\3

Figure (\.2 : Floor response spectra.


Comparison of R.V. and T.II. results.

4.3. SECONDARY STRUCTURE.


For the piping, three methods of analysis are compared :
- Direct integration (TH) of the complete structure (primary + secondary).
- RV analysis of the secondary system, using as input the acceleration PSD
matrix at nodes 3, 2 and 22 (obtained as output of the analysis of the
primary structure).
- RS analysis according to the Standard Review Plan [4]. A single average
spectrum is obtained by enveloping the average TH spectra at nodes 3, 2 and
22 (Fig. 4.3).

The modal contributions are combined according to the CQC

rule.
Table IV/3 compiles the main results of the comparison.

It can be

seen that, for all response quantities, the RV results are close to the average value of the TH results, while the RS method leads to large excess errors
(which, likely, are related to the enveloping process).
Fig. 4.4 and 4.5 compare the probability of exceedance curves of
the TH and RV analyses (a Gumbel distribution is assumed in this case).

It

can be observed that the RV predictions are close to the TH simulations, and
that, most of the time, the ratio between the maximum and minimum results in
the TH simulations is close to 2.
Finally, Figs. 4.6 and 4.7 show the PSD of the various random responses in the structure.

24

Table IV/3
COMPARISON OF THE AVERAGE MAXIMUM RESPONSE OF THE SECONDARY STRUCTURE

Response (unit)

Time History

Random Vibration

Envelope spectrum
(CQC)

B.M.7

(N.m)

18 020

18 253

20 400

B.M.18

(N.m)

17 790

16 671

28 000

S.R.3

(N)

9 255

9 484

9 570

S.R.2

(N)

23 314

21 519

32 500

S.R.22

(N)

9 452

9 257

14 420

(B.M. = bending moment ; S.R. = support reaction)

25

10

10

10'

FREQUENCY ( HZ )

Figure 4.3 : Envelope of the average floor response spectra (T.H.)


at nodes 2,3 and 22.

26

10

IO

Figure t\.t\ : Support reactions. Comparison of the probability of excecdanc.e curves


obtained by U.V. and .. analysis.

nods 3 : SOO

M V M M .

nod 2 : SOO

00

MVMM

Figure 4.5 : Rending moments. Comparison of the probability of exceedance curves


obtained by R.V. and T.H. analysis.

ro
co

II

Figure 4.6 : Support reactions. Ff3D functions.

node 3 : SOO

CO
O

Vi " u r o

\. 7 :

Ucndi.nr, mnmcnf.s.

>) f u n c t i o n s ,

4.4. REMARK.
The RV results for BM 18, in Table IV/3, appear as non-conservative as compared to the TH.

This, in fact, is because the artificial TH

are too energetic, as can be assessed from Fig. 2.3.

(The mean spectrum is

generally above the NRC spectrum).


If the RV calculations are reproduced with a PSD consistent
with the mean spectrum of the artificial TH (instead of the NRC spectrum),
the results of Table IV/4 are obtained, which are conservative.
Table IV/4
COMPARISON OF THE AVERAGE MAXIMUM RESPONSE OF THE SECONDARY STRUCTURE
Response (unit) Time history

Random
vibration

Envelope
spectrum (CQC)

B.M.7

(N.m)

18 020

20 439

20 400

B.M.18

(N.m)

17 790

17 922

28 000

S.R.3

(N)

9 255

10 620

9 570

S.R.2

(N)

23 314

24 064

32 500

S.R.22

(N)

9 452

9 913

14 420

(B.M. = bending moment ; S.R. - support reaction)

31 -

5. CONCLUSIONS, RECOMMENDATIONS

A benchmark study has been carried out, of various numerical methods for seismic analysis of nuclear structures, namely the Response Spectrum (RS) method, the Time History analysis (TH), and the Random Vibration
method

(RV). The various analyses were performed by different teams, res-

pectively EdF, ANSALDO and BELGONUCLEAIRE.

The conclusions are as follows :

(1) The statistical variability between different time history analyses can
be large (more than 2 in some cases).
(2) For multisupported structures, using the RS method with envelope spectra
can lead to substantial errors (by excess, in this study).
(3) Throughout this study, the average RV results have been found in good
agreement with the average TH results.

This is true for floor response

spectra, as well as bending moments and support reactions.

Besides, the

probability of exceedance curves obtained in the RV analysis fit well


with the corresponding curves of the ten TH analyses.
Based on this study, it is concluded that the RV method is a reliable alternative method of analysis.

Its use is recommended, particularly

for preliminary design, owing to its computational advantage on multiple TH


analyses.

33

- R E F E R E N C E S -

[1]

A. PREUMONT (1988).
Application of the Random Vibration Approach in the Seismic Analysis of LMFBR Structures.
EUR Report 11369 EN.

[2]

A. PREUMONT (1985).
The Generation of Non-Separable Artificial Earthquake Accelerograms
for the Design of Nuclear Power Plants.
Nucl.Eng.Des., 88, 59-67.

[3]

I. ELISHAKOFF (1982).
Probabilistic Methods in the Theory of Structures.
Wiley & Sons.

[4]

US NRC, Standard Review Plan, Sections 3.7.2. and 3.7.3.

[5]

A. PREUMONT (1985).
On the Peak Factor of Stationary Gaussian Processes.
Journal of Sound and Vibration, 100(1), 15 - 34.

35-

A P P E N D I X

SEISME A COMPUTER PROGRAMME FOR THE RANDOM VIBRATION ANALYSIS


OF MULTISUPPORTED STRUCTURES UNDER SEISMIC EXCITATION

Andr Preumont & Sammy Shihab

1. INTRODUCTION.
The programme SEISME has been developed to perform the random vi
bration analysis of a linear structure subjected to a stationary seismic
excitation.

B y seismic, it is meant that the excitation results from the

support accelerations.

The damping is assumed viscous, but no particular

assumption is made concerning the decoupling of the modal equations (non


classical damping).

The structure is assumed to be multisupported, with an

arbitrary correlation between the acceleration of the various supports.

The

stationary random excitation is given in the form of the power spectral den
sity (PSD) matrix of the support accelerations.

The output of the programme consists of

. The PSD matrix of the acceleration at a set of degrees of freedom (d.o.f.)


for subsequent use in the analysis of secondary systems (cascade analysis).
. The PSD of the support reactions (for every support d.o.f.).
. The PSD of a set of user specified response quantities which are linear
functions of the modal amplitudes (e.g. displacements, stresses).
In addition to the PSD functions, the programme provides

. the spectral moments IDQ, m^, m ;


. the variance ^, central frequency u>c and bandwith .
. the peak factor corresponding to a specified observation period.

37

2. EQUATIONS OF MOTION.

Consider a multisupported structure excited by the motion (possibly


differential) of its supports.

Partitioning the restrained and unrestrained

d.o.f., the equation of motion reads

11

\ /
10 \ / 1

01

/\

oo/ xo<

where the subscript 1 refers to the unrestrained d.o.f., while the subcript 0
refers to the support d.o.f.

f.. represents the excitation force at the sup

ports, that is the support reactions.

The part of equation (1) relative to the unrestrained d.o.f. can be


rewritten

n i +

n i +

n xi

Mio o cio o Kio o

(2)

In this equation, x.. represents the absolute displacements of the structure


and x Q the absolute displacements of the supports. To reduce the size of the
system, it is appropriate to decompose the motion into the normal modes of
the structure fixed at its supports.

B efore doing so, it is necessary to

split x. in its dynamic and quasistatic contributions [1] :

i = xL + j s

(3)

where } are the quasistatic displacements of the structure resulting from


GS

the support displacements, and y

is the dynamic response, x.? can be ob

tained from equation (2) by cancelling out all the time derivatives :
K
K
ll 1
10 0
or
(4)
qs
1
x

x_ =

qs 0

11

10 0

39

Tqs is the quasistatic transmission matrix

; its ith

the static displacements at the unrestrained d.o.f.

column contains

resulting from a unit

displacement at the ith support d.o.f.. For an isostatic structure, TqS


results directly from rigid body kinematics ; for hyperstatic structures,
its columns are obtained from static analyses.

Since x!? is linearly rela

ted to .., the following change of variables can be performed :


' x. \
1 \

/I
/

T \ /y.
qs\ / 1
(5)

0/

\-o

where the dynamic displacements, y,, satisfy homogeneous (zero) boundary con
ditions at the supports, like the mode shapes of the fixed base structure.
Combining (5) and (2) gives
M

n i + c n i + K n Zi = ^IIV'VO

Substituting

TqS

from equation

<C11 T qs

+ C

<K11 T qs

+ K

10) *0

(6)

10> 5 0

(4), it can be seen that the stiffness

contribution to the excitation vanishes.

So does the damping contribution if

the damping matrix is proportional to the stiffness matrix ; this term is


usually small and is frequently neglected.

This will be done io what fol

lows.
Deleting the subscript 1, equation (6) can be rewritten
M + C + Ky_ = ( M T q s + M 1 0 ) xQ

(7)

Because of the essentially lowpass character of most of the physi


cal excitations, most of the response is usually concentrated in the first
few modes.

It is therefore appropriate to operate a change of variables and

decompose the dynamic displacements into their modal components (this is


licit since y_ and the mode shapes d. have the same boundary conditions) accor
ding to :
1

(8)

40

where -

(d..,...,d ) is the m matrix ( = number of d.o.f., m = number


1
m

of modes considered in the analysis), whose columns are the normal modes
of the fixed base structure, satisfying the orthogonality conditions

Mdj

"f*ij

(9)

Sj

or, in matrix form,

= 2 = diagC/^

M = = diagO^)

(9')

Substituting (8) in equation (7), multiplying both sides of the equation by


E

and taking into account the orthogonality conditions (9), it is readily

obtained that
2

* + C E + = (

where

+ M1(J) x Q

(10)

= (
+ , )
qs
10

(11)

is the m n s modal participation matrix.

Each column of gives the work

done on each mode by the inertia forces associated to the quasistatic acce
lerations induced by a unit acceleration applied on the
It is worth noting that in many cases, the term M ^ Q *

ns

support d.o.f..

neglected in equation

(7) (this term vanishes for a lumped mass matrix and is usually small).

41

3. EFFECTIVE MODAL MASS.

When going from equation (7) to (10), a drastic reduction of the


size of the system of equations can be achieved ().

The question ari

ses, then, of how many modes should be considered in the analysis (how large
should be m).

Obviously, the first criterion must be related to the frequen

cy content of the excitation :

All the modes belonging to the bandwidth of the excitation should


be included in the analysis.
Even so, this may not be enough to achieve a good accuracy for the
support reactions, as first pointed out by Powell [2].

(To understand that,

one must think of the extreme case of a rigid structure excited in low fre
quency ; none of the modes react dynamically and still there are support
reactions associated with the "quasistatic inertia").

The modal participation matrix provides a guide as to how well the


structural mass is accounted for in the truncated modal basis.
one forms the matrix

In fact, if

including all the modes (m n), one gets,

after some algebra

r T ' 1 = TJ S M

T qs + ij, M 1 0 + M 01 T q s + M 01 M"} M 10

*oo + M oi

io M oo

<12>

where MQO is the socalled Guyan mass matrix, obtained by static condensation
of the unrestrained d.o.f. according to
X

l
(13)

o.

MQO is obtained by expressing the conservation of the kinetic energy

43

1 .T A
2 0 00 0

which leads to
= T T ..
+TT +
+ M
00
qs 11 qs
qs 10
01 qs
00

(14)

MQO represents the inertia of the structure seen from the supports, when it
responds statically.

If one neglects M

as second order term, equa

tion (12) becomes


rT

^"lr

=S

oo-Moo

(15)

where all the modes are included in the left side.

M Q Q is the mass matrix

directly associated to the supports.


Now, if 1

stands for the unit rigid body translation of supports

along a global axis i, a velocity v 0 along this axis corresponds to support


velocities

1..

The corresponding total kinetic energy is

"" " Vo =

o h M00 h

which means that the total mass of the structure, mj, is related to M Q O by

(16)

" - i Ko h
From equation (15), we see that

I M
1
ii 00 -i
or

-l ^
- 1
1 - 1 / " 1 = 0
-
00 -i
-
-i

m T - m s - 1* '1

1^ = 0

(17)

where mg is the total mass associated to the supports.

Equation (17) applies providing all the modes are included in the
third term.

If only m modes are included, equation (17) will no longer be

satisfied and there will be a residue, which will represent the missing mass,
that is the mass which is not accounted for by the truncated modal expansion
(8).

The single support form of equation (17) is

44

_m

Jl
-I.

^2

"

(18a)
J

is called the effective modal mass of mode i.

It represents the part

of the total mass of the structure which is associated to mode i.


In the multisupported case, the component form the equation (17) is

T~ m s = 2^T^~

(18b)

It is apparent that, in the truncated case, the missing mass depends on the
direction i of the excitation.

From the foregoing discussion, the second criterion for mode selec
tion is
Any mode whose effective mass is a significant part of the total
mass should be included in the analysis.

45

4. EXCITATION.

In practical applications, it is important to distinguish between


the number of supports

(ns) and the number of excitations

(n f ).

This

accounts for the fact that the same excitation can be applied to several
supports, with a possible scalar amplification.
piping system of Fig. 1/1.

As an example, consider the

It has six supports (ns = 6), but if one assu

mes that the three components of the acceleration are the same at every sup
port, there are only three independent excitations (nf = 3). The relation
ship between the support accelerations and the excitations is

-~y

X,

In the general case, one has

= A

"+
*0

(19)

where the ns n f matrix A includes the possibility of scalar amplific


ation.

It is the PSD matrix of x Q , which is provided at the input. The re

duced quasistatic transmission matrix and reduced modal participation matrix


are respectively
T+ =
A
qs
qs

(20a)

+ = A

(20b)

47

5. INPUT OUTPUT RELATIONSHIPS IN THE FREQUENCY DOMAIN.


5.1. Modal amplitudes.
Introducing equation (20b) into (10), one gets
"

?
+ "+
+ C + = xj

(21)

Upon Fourier transformation, one gets the transfer matrix between


+

the modal amplitudes Z(a>) and the excitations () :


() =

wnere

() + *()

() J diag /U (o 2 ) 1 + j C

(22)
(23)

5.2. Absolute accelerations.


Let be the m matrix containing the modal amplitudes at
the iu d.o.f.
ed.

where the PSD matrix of the accelerations is to be calculat

One wishes to determine the n p nf transfer matrix between the

acceleration at these nodes and the excitation. It consists of a dynamic and


a quasistatic contribution, which are respectively

()

and

* *()
qs U

Combining with (22), one gets


() = () +()

'

(24)

() = 2 = + + +

qs

(25)

49

5.3. Support reactions.


If one neglects the contribution of the damping to the support
reactions, the part of equation (1) relative to the restrained d.o.f. provi
des
f
-o

M
oi 5l + c )0 o + K oi x i + K oo o

(26)

Substiti j t i n g ,
X

o +

qs
(modal displacement method), one gets
f

+ M

-o M00 k

01(Tqs k

h + (K 0 0 + K 0 1 T q s ) x Q + K Q 1

(27)

Apart from the second term which represents the coupling inertia
between the unrestrained and the restrained d.o.f., the meaning of the other
terms is quite clear : The first represents the support inertia, the third is
related to the differential displacements and the fourth contains the dynamic
modal reactions (the columns of Kgi are the modal reaction vectors). The
above formulation is subject to the missing mass problem mentioned earlier.
Equation (17) suggests that the results can be improved by applying a quasi
static correction (missing mass correction) of the form

l \ "

oo - M oo -

"

(28)

) So

It represents the quasistatic inerial effect of those modes which have not
been included in the analysis.

In the case M ^ Q = 0, the corrected result

reads
f

-o - ( K 00 + K 01 T qs ) 0 + K 01 *-

( 00 " ^ ^

) *0

(29)

The three terms represent respectively the differential displacements, the


dynamic response and the quasistatic inerial contribution (inertia of the
supports plus missing mass).
An alternative form can be obtained as follows. First, K m x1 is
eliminated from equation (26) by using the first part of equation (1) (the
damping is omitted for the sake of simplification) :

50

* = - K (

i i + M i o * o )

K x =Ti

+
01
- 1'
qs (
+)

Combining with equation (26) gives


f = M m + M.. ) . + ( __ + ._ ) .
\ 01
qs 11 / 1 \ 00
qs 10 / 0
+ ( + \
\ 00
01 qs 0
Upon substituting

l = Tqs k

< 30 >

(nodal acceleration method), one gets, after some straightforward algebra

*o

oo + ( K oo+ KoiT qs ) 2o r T

(31)

After Fourier transformation and using equation (22)


U^

"00 0 ( ) + ( K00 + K01 Tqs ) ?0 ( ) + " 2 <

()

(32)

This form is totally equivalent to (29) although the meaning of the various
terms is not as straightforward.

It is under this form that it has been

implemented. The transfer matrix between the support reactions and the exci
tation is

( )

() ^0()

(33)

Y"> = { " 0 0 + " 2 r T H r ^ ( K 0 0 + K 01 T qs) J A <34>


Note that with the foregoing formulation, the computed support
reactions will be correct, even if some modes carrying an important part of
the effective mass are omitted in the analysis, providing these modes react
in a quasistatic manner (i.e. their natural frequencies must be above the
bandwidth of the excitation).

51

6. CALCULATION OF THE PSD MATRICES.

The PSD matrix of the absolute accelerations is obtained from that


of the excitation using standard random vibration relationships [3].

Fron

(24), one gets

..() = () () ()

(35)

where * stands for the conjugate transposed.

0 ( ) is available as input

data provided by the user, or results from previous calculations (primary


structure).

Both ,.() and () are complex and hermitian (only the lower

X
half is calculated).

The diagonal terms characterize the spectral content

of each acceleration, while the off-diagonal terms characterize their crosscorrelation.

Similarly, the PSD matrix of the support reactions can be obtained


from (33) :

() = H R (w) () *()

(36)

However, on the contrary to the acceleration for which the cross-correlation


structure is important, only the diagonal terms of ^() are used in the
subsequent strength assessment.

Notice that all the above calculations must be performed for a set
of

control frequencies whose spacing

0 () and ,().

allows

a smooth representation of

This requires an adaptive frequency spacing.

53-

7. STRESS CALCULATIONS.

In addition to the previous quantities, the users will be interest


ed in calculating the PSD of various response quantities like stresses, or
bending moments, for subsequent use, for example in fatigue damage assess
ment.

Any of such response quantities can be defined by its modal components

b and its quasistatic components c as

r = b T + T x Q

(37)

From equation (22), the (1 n f ) transfer matrix between r and x is

() = b T () + i= T A

(38)

(this is a row vector) and the PSD function is

( ) () () *()

(39)

55

8. POST-PROCESSING OF THE PSD FUNCTIONS.


The PSD functions calculated by the programme can be post-processed
in various ways, depending on the user's wishes :
(1) Sample time histories with the calculated spectral content can be generated and used in simulations (e.g. qualification of components, fatigue
tests).
(2) Various statistics can be evaluated, like the RMS value (variance), the
central frequency, the bandwidth, or the extreme value over a given period of observation (peak factor).
(3) If the output is a stress and if the SN (Wohler) curve is available, the
expectation of the damage (according to Miner's cumulative law) can also
be calculated.
All these aspects are discussed extensively in Chapters X and XI of
[3].
On the other hand, if the problem is related to earthquake engineering, the user will be interested in the so-called response spectra rather
than the PSD functions.

This is also true at the input, which will often be

specified in terms of a response spectrum instead of a PSD.


between response spectra and PSD is discussed in [4],

57

The relationship

9. FLOW CHART OF THE PROGRAMME.

Response
Spectrum

F.E. A n a l y s i s

<D(nf,nf|n) I

<ns'nf>

Structure

Damping

Excitation

^HTCH

1
r(m,n )

A qs

1
I

v( ,

s'

00(ns*ns)

00

01 T qs

(22)
o
c , (35)*
o,
o.
o
o

^()

(m,nf,n)

..()

( iiln)

()

(r

Save

on disk

Response
Spectrum

)
1

Save (o>)

on disk

Save ()

Stress calcu! ation

Read b & c
en

op on stress

(38)
o
e
o J' (39)
o.
o
o

Hr)

()

L,n f )

<

59

on disk

Post
processing

10. REFERENCES.
[1]

R.W. CLOUGH & J. PENZIEN (1975).


Dynamics of Structures.
Mac Graw-Hill.

[2]

G.H. POWELL (1979).


"Missing Mass" Correction in Modal Analysis of Piping Systems.
SMIRT-5, Berlin, paper K10/3.

[3]

A. PREUMONT (1989).
Vibrations Alatoires et Analyse Spectrale.
Presses Polytechniques Romandes, Lausanne, paratre.

[4]

A. PREUMONT (1983).
On the Connection Between the Response Spectrum and the Spectral
Properties of Earthquake Accelerograms.
ICASP-4, Florence.

61

- A P P E N D I X

II -

SOME RANDOM VIBRATION RESULTS ON DECOUPLING AND


NON-CLASSICAL DAMPING
A. Preumont & S. Shihab

1. INTRODUCTION.
In this chapter, the random response of the two d.o.f. oscillator is
analysed for the various cases outlined in Table I of the previous report.
The study is restricted to tuned case (# = (JL^/CL^ = 1), and special attention
is given to the case where the secondary and primary structures have widely
different damping ratios.
The following questions are the subject of investigation :
(1) What is the effect of the modal cross-correlations on the random response ?
(2) What is the effect of off-diagonal modal damping ? (non-classical).
(3) What is the effect of coupling between the secondary and primary structures (feedback of mass 2 on mass 1).
The first question, relative to modal cross-correlations, has been
analysed in a previous report. Since the modal cross-correlations are automatically included in a Random Vibration analysis, the results reported here ara
only for the sake of future comparison with Response Spectrum Calculations
(CQC vs. SRSS).
The second question, concerning the non-classical damping, has been
addressed in the literature [1]. Here, it is pointed out that when ^ / 2
the assumption of classical damping leads to an error which increases as the
mass ratio decreases to 0
The third question has also been addressed extensively in the literature [2], and decoupling criteria have been developed (e.g. [3,4]).

63

_0.

2. ASYMPTOTIC RESPONSE OF THE TWO DOF. SYSTEM AS

We consider the tuned case ((X = CL>2/(*\ = 1). With the notations
(3.1), the damping matrix of the two d.o.f. system reads :

/ 1 + &

- &

C = 2a>mL I

\
)

(2.1)

According (App.1.9) , the modal matrix behaves asymtotically (/LL^0) as

'

(d , d ) 7=^|

V^l

(2.2)

\<fr

-/TT

Combining equations (2.1) and (2.2), the modal damping matrix behaves asympto
tically as

lim

(k*k

i-2\

f = E C H= W i

(2.3)

This shows that the discrepancy between the damping ratios is responsible for
offdiagonal terms in . Their magnitude is proportional to the difference of
the damping ratios, while the diagonal terms are proportional to their avera
ge. It follows directly that if the offdiagonal contributions of are ne
glected, the response will depend only on ^ + S2 anc* n o t o n c ^ e individual
damping ratios in the primary and secondary structures.
From (2.3) and the asymptotic behaviour of the natural frequencies
of the coupled system :
lim

V = V

*~0
the modal transfer matrix can be written as
9

---

+ (

+ ^

- 5(

- 2)

= - + :^+21 * = M
- 2(^(1 - 12)

+ j6L>(^ + 2)^
(2.4)

65

where

= (^ - 2 + 2 ^^

Since

lim

, , =

- 2 + 2 12)

(2.5)

the asymptotic form of the transfer function of the relative displacement of


mass 1 with respect to the support is

Hn - bTH

r\ yfT

^1

(2.6)

<2 - 2 ) + 2

VT/

which is identical to the uncoupled result. If the offdiagonal term had been
neglected in (2.3), the following expressions would have been obtained (the
superscript c refers to "classical damping") :

a -~ + jt(^1 + 2)c^

H =1

^ 2 + jw(JL + 2)o^

o
where

Ac <^ 2 + jw^c^+pj

and

(2.7)

(^ /) + jo; ^ + 2)

Comparing with equation (26), one sees that the average damping ratio would
have been substituted t o ^ ^ , which may lead to substantial error.
Similarly, providing the (offdiagonal terms are considered in equa
tion (2.4), it can be shown (although not done here for the sake of concise
ness) that, as jLL^0, the transfer function between the support acceleration
and the absolute acceleration of mass 1, and the transfer function between the
support acceleration and the relative displacement of mass 2 relative to mass
1, converge towards the decoupled results :

2 + 2 j
H A l 22
2
-4
A

+
2
j
1{?{*>
l
/.

66

(28)

1 +2 3 i ^

* - + 2 j jojjco

2"

(2")

- + 2 j ^ 2

As expected, these results are identical to those obtained assuming that there
is no feedback of mass 2 on nass 1 On the contrary, if the assumption of
classical damping is made (offdiagonal neglected in 2.4)), one gets, respec
tively :

+ j w ^
H

A\11 "[-
- T 22

+ 2)
(2

+ j W (J^1 (J 1 +~
2)

10)

+ , + 2)
and

KZ =
R

_.

(2.11)

|^ - + ] 1 ( 1 + 2) J

When the damping ratios of the two masses are substantially diffe
rent, neglecting the offdiagonal damping contributions in (2.4), in a coupled
analysis, may lead to larger errors than those associated with a decoupled
analysis. This is illustrated in Figs. 1, 2 and 3, where the isolines of the
following ratios are plotted in the particular case of CK = 1 (tuning) and a
white noise excitation :
_

variance of decoupled case


variance of exact solution

variance of classically damped case


variance of exact solution

These isolines are drawn in the plane defined by the mass ratio fJL and the
ratio S201 Figs 1> 2 and 3 refer respectively to i = 0.1, = 0.05 and
(,1 = 0.02. With reference to Fig. 4, the interpretation of these plots is as
follows : For any value of U and ^2^bl> t n e isolines R and R give res
pectively the error associated with a cascade analysis, neglecting the feed
back of mass 2, and a full analysis but assuing classical damping. Note that
neglecting the feedback of mass 2 leads to an error on the conservative side,
while assuming classical damping leads to an error on the unconservative side.
If one accepts a given error (say 20 % ) , the diagram can be separated in three
zones as indicated in Fig. 4 :
the decoupled region, where a decoupled analysis (neglecting feedback) is
acceptable for the given error level ;
the classical damping region, where a coupled analysis is necessary, but i:
can be carried out with the assumption of classical damping ;
the nonclassical damping region corresponds to those values of the parame
ters which require to consider a full modal damping matrix unless an unac
ceptable error is made.

67

All these results correspond to OL = 1 (^ =^2)

an<

white noise excitation.

According to the SRP [3], the decoupling can be done under the conditions
< 102

or

10"2 ^ ^ 101

for ^1.25 ora^0.8

It is clear from Figs. 1 to 4 that, when OL = 1, the decoupling assumption


leads to a substantial error, even for / ^ l O 2 , especially if the damping is
light. This error, however, is on the conservative side.

68

3 NUMERICAL RESULTS.
The stationary response of the two d.o.f. oscillator to the excit
ation of Fig. 5 has been analysed with the computer programme SEISME. The
following data have been adopted :

= 2 = 47

= 0.05

, 2 0 2

The mass ratio


has been taken as a parameter.
been considered :

Four methods of analysis have

(1) Decoupled systems (cascade analysis of two s.d.o.f. systems.


is asymptotically correct as LL ^ 0 .

This method

(2) Coupled system [2 d.O.f.] with nonclassical damping.


This method is
exact ; the results converge asymptotically towards those of the previous
method as JL^0.
(3) Coupled system [2 d.o.f.] with classical damping.
totically wrong, except when ^ = 2*

The results are asymp

(4) Same as (3), but without crosscorrelations between the modal responses.
This amounts to neglecting the offdiagonal terms in the modal response
PSD matrix. This is done only for the sake of comparison with Response
Spectrum calculations using the SRSS and the CQC combination rules.
The results are illustrated in Figs. 6 to 9 :
Fig. 6 shows the PSD of the relative displacement between the two masses,
X2 , as obtained with method (2), for various values of the mass ratio.
Fig. 7 shows the PSD of the absolute acceleration of mass 1 as obtained with
method (2) (full line) and method (1) (dotted line).
Fig. 8 compares, for U L = 10~4, the results of method (2) and method (3). It
is observed that the curves are significantly different near resonance, as a
result of nonclassical damping. This difference depends essentially on the
ratio 2/^1
Fig. 9 shows the PseudoAcceleration [mean] Response Spectra for 2 7. damp
ing, obtained from the PSD of Fig. 7 by the direct method of [5] (page
AII/20).
Relevant numerical results are summarized in Table I.
observed that :

It can be

as Li0, method (2) gives the same results as method (1) ;


methods (2) and (3) give similar results for large values of {t but differ
when [A ^ 0 ;
neglecting the modal crosscorrelations is acceptable for large f, where the
modes are well separated.
It decomes unacceptable when fJL > 0 , as the
modes become closely spaced.

69

REFERENCES.
[1]

IGUSA, T . , DER K IUREGHIAN, . & SACK MAN, J . L . ( 1 9 8 4 ) .


Modal Decomposition Method for S t a t i o n a r y Response of Non-Classicall y Damped Systems.
E a r t h . E n g . S t r u c t . S y n . , Vol 12, 121-136.

[2]

CRANDALL, S.H. & MARK, W.D. ( 1 9 6 3 ) .


Random V i b r a t i o n i n Mechanical Systems.
Academic p r e s s .

[3]

USNRC Standard Review P l a n , S e c t i o n 3 - 7 . 2 .

[4]

AZIZ, T.S. & DUFF, C G. (1978).


Decoupling Criteria for Seismic Analysis of Nuclear Power Plant
Systems.
ASME/CSME Conference, Montreal, June 25/30.

[5]

PREUMONT, A. (1986).
Application of the Random Vibration Approach in Seismic Analysis of
LMFBR Structures.
EEC Contract Report RAP-086-B(S). WG CS/AG 2.

Report

EUR 1 1 3 6 9 EN

71

Table I
NUMERICAL RESULTS
0.
decoupled

IO" 1

10-2

10-3

47

10.74

11.95

12.37

kn

14.71

13.21

12.77

il
11

0.05

0.0319

0.034

0.0347

0.02

0.0382

0.036

0.0353

(2)

0.156

0.406

0.680

(3)

0.155

0.392

0.622

(4)

0-158

0.475

1.487

(l

(rad/s)

(mS\x
/
Methods /
(

- x2
(1)

0.768

1 d.o.f.
(cascade
analysis)

73

u
m
co
CO

<7 ( c l a s s i c a l damping)
2
O (non-classical) =
(exact)
2

- o.i

<T (decoupled)
2
(exact)

frequency ratio (X - 1.

FIG. 1
75

H -10

--

- 0.05

.
FIG. 2
-76-

- 10

--

,.02

FIG. 3
77

10

-4

- 1"._

DECOUPLED

10

-2

en
(3

'V

1.0
R

O ( c l a s s i c a l damping)
2
2
(7 ( n o n - c l a s s i c a l ) (exact)
0.1

(7 (decoupled)
2

frequency ratio (X - !

(exact)

FIG. 4

78

.)
MX

(rel.)

IN

10

FIC.:5

I L

.
50
f (Hertz)

Power spectral density corresponding to the NRC 2% horizontal


spectrum (equivalent stationary duration = 14.5 s ) .

79

FIG. 6 - PSD o f the relative displacement

oo
o

( x 2 - ) fo r vario us mass ratlo s U (exact)

FIG. 7 PSD of the absolute acceleration of masa I for various mass ratlos IL
Full line
= coupled (2 d.o.f.)
Dotted line =decoupled (I d.o.f.)

10

io'

=.

='

4 :

10

00

6
4

<

< ;

4 :

_
=

"4

10

(RflO/Sl

"

10

U (RO/31

10

10

(RR0/SI

(/* = 0.0001) PSD of the absolute acceleration at mass 1


Full line
= coupled [2 d.o.f.] system with modal damping
Dotted line decoupled [1 d.o.f] system (^coupled with
non-classical damping)

The ratio between the


maxima is :

= 2.04

4 6 8 10 2
* 8 10 2
CRAD/SI

4. sa io

FIG. 8

82

FIG. 9 (Directly generated from PSD)


Pseudoacceleration [mean]* Response Spectrum of mass 1 (F

0.02)

Full line
=coupled (2 d.o.t.)
Dotted line =decoupled (1 d.o.f.)

10'

m/s

* t r

m/s'

=.1

10'

10'

/ 2
m/s

=.0

IO2

=.00

oo
CO


4 _

10'

10

10'

10

10"

2_
10

u m i unu

10-}

102

'

'

CRA0/SI

irr 1

10 2

10*

*Usin the formulae for the average of the peak

IH M
4 8

10 2

factoi

I I Mill
4

'

10*2

(RR0/SI

10 2

I I IIIUJ
4

10"

10

nun
io;

4 102

4 ( 8

nini

012
4
(RR0/SI

10 2 2

4 ( 8

10'

For up-to-date information on


European Community research
consult
CORDIS
The Community Research
and Development
Information Service
CORDIS is an on-line service set up under the VALUE programme to give quick and easy
access to information on European Community research programmes.
The CORDIS service is at present offered free-of-charge by the European Commission Host
Organisation (ECHO). A menu-based interface makes CORDIS simple to use even if you are
not familiar with on-line information services. For experienced users, the standard Common
Command Language (CCL) method of extracting data is also available.
CORDIS comprises eight databases:
RTD-News: short announcements of Calls for Proposals, publications and events in the
R&D field
RTD-Programmes: details of all EC programmes in R&D and related areas
RTD-Projects: containing 14,000 entries on individual activities within the programmes
RTD-Publications: bibliographic details and summaries of more than 50,000 scientific
and technical publications arising from EC activities
RTD-Results: provides valuable leads and hot tips on prototypes ready for industrial
exploitation and areas of research ripe for collaboration
RTD-Comdocuments: details of Commission communications to the Council of Ministers
and the European Parliament on research topics
RTD-Acronyms: explains the thousands of acronyms and abbreviations current in the
Community research area
RTD-Partners : helps bring organisations and research centres together for collaboration
on project proposals, exploitation of results, or marketing agreements.
For more information and CORDIS registration forms, contact
ECHO Customer Service
CORDIS Operations
BP2373
L-1023 Luxembourg
Tel.: (+352) 34 98 11 Fax: (+352) 34 98 12 34

If you are already an ECHO user, please indicate your customer number.

European Communities - Commission


EUR 14153 - Application of the random vibration approach in the seismic
analysis of LMFBR structures - Benchmark calculations
A. Preumont, S. Shihab, L Cornaggia, M. Reale, P. Labbe, H. Noe
Luxembourg: Office for Official Publications of the European Communities
1992 - V, 84 pp., num. tab., fig. - 21.0 x 29.7 cm
Nuclear science and technology series
ISBN 92-826-4287-9
Price (excluding VAT) in Luxembourg: ECU 10.50

This benchmark exercise is the continuation of the state-of-the-art review


(EUR 11369 EN) which concluded that the random vibration approach could
be an effective tool in seismic analysis of nuclear power plants, with potential
advantages on time history and response spectrum techniques. As compared to the latter, the random vibration method provides an accurate
treatment of multisupport excitations, non-classical damping as well as the
combination of high-frequency modal components. With respect to the
former, the random vibration method offers direct information on statistical
variability (probability distribution) and cheaper computations. The disadvantages of the random vibration method are that it is based on stationary
results, and requires a power spectral density input instead of a response
spectrum.
A benchmark exercise to compare the three methods from the various
aspects mentioned above, on one or several simple structures has been
made.
The following aspects have been covered with the simplest possible models:
(i) statistical variability,
(ii) multisupport excitation,
(iii) non-classical damping.
The random vibration method is therefore concluded to be a reliable method
of analysis. Its use is recommended, particularly for preliminary design,
owing to its computational advantage on multiple time history analysis.

Venta y suscripciones Salg og abonnement Verkauf und Abonnement


Sales and subscriptions Vente et abonnements Vendita e abbonamenti
Verkoop en abonnementen Venda e assinaturas
3ELGIQUE/BELGI

FRANCE

SUOMI

CANADA

Moniteur belge /
Belgisch Staatsblad
Rue de Louvain 42 / Leuvenseweg 42
B-1000 Bruxelles / B-1000 Brussel
Tl. (02) 512 00 26
Fax (02) 511 01 84

Journal officiel
Service des publications
des C o m m u n a u t s europennes
26, rue Desaix
i - 7 5 7 2 7 Paris Cedex 15
Tl. ( 1 ) 4 0 58 75 00
Fax (1) 40 58 75 74

A k a t e e m i n e n Kirjakauppa
Keskuskatu 1
PO Box 128
SF-00101 Helsinki
Tel. (0) 121 41
Fax (0) 121 44 41

Renouf Publishing C o . Ltd

Autres distributeurs /
Overige verkooppunten
Librairie e u r o p e n n e /
Europese boekhandel
Rue de la Loi 244/
"Wetstraat 244
B-1040 B r u x e l l e s / B-1040 Brussel
Tl. (02) 231 04 35
Fax (02) 735 08 60
Jean De Lannoy
Avenue du Roi 202 /Koningslaan 202
B-1060 Bruxelles / B-1060 Brussel
Tl. (02) 538 51 69
Tlex 63220 U N B O O K
Fax (02) 538 08 41
Document delivery:
Credoc
Rue de la M o n t a g n e 34 / Bergstraat 34
Bte 11 / Bus 11
B-1000 Bruxelles / B-1000 Brussel
Tl. (02) 511 69 41
Fax (02) 513 31 95

NORGE
IRELAND
Government
4-5 Harcourt
Dublin 2
Tel. (1) 61 31
Fax (1) 78 06

Narvesen information center


Bertrand Narvesens vel 2
PO Box 6125 Etterstad
N-0602 Oslo 6
Tel. (2) 57 33 00
Telex 79668 NI C
Fax (2) 68 19 01

Supplies Agency
Road
11
45

Licosa Spa
Via Duca di Calabria, 1/1
Casella postale 552
1-50125 Firenze
Tel. ( 0 5 5 ) 6 4 54 15
Fax 64 12 57
Telex 570466 LI COSA I

BTJ
Tryck Traktorwgen 13
S-222 60 Lund
Tel. (046) 18 00 00
Fax (046) 18 01 25

AUSTRALIA

DANMARK

G R A N D - D U C H DE LUXEMBOURG
Messageries Paul Kraus
1 1 , rue Christophe Plantin
L-2339 Luxembourg
Tl. 499 88 88
Tlex 2515
Fax 499 88 84 44

OSEC
Stampfenbachstrae 85
CH-8035 Zrich
Tel. (01) 365 54 49
Fax (01) 365 54 11

DEUTSCHLAND

NEDERLAND

Bundesanzeiger Verlag
Breite Strae
Postfach 10 80 06
D-W 5000 Kln 1
Tel. (02 21) 20 29-0
Telex ANZEI GER B O N N 8 882 595
Fax 2 02 92 78

S D U Overheidsinformatie
Externe Fondsen
Postbus 20014
2500 EA 's-Gravenhage
Tel. (070) 37 89 911
Fax (070) 34 75 778

Kinokuniya C o m p a n y Ltd
17-7 Shinjuku 3 - C h o m e
Shinjuku-ku
Tokyo 160-91
Tel. (03) 3439-0121

NIS
Havelkova 22
13000 Praha 3
Tel. (02) 235 84 4 6
Fax 42-2-264775

PORTUGAL

G.C. Eleftheroudakis SA
International Bookstore
Nikis Street 4
GR-10563 Athens
Tel. (01) 322 63 23
Telex 219410 ELEF
Fax 323 98 21

Imprensa Nacional
Casa d a M o e d a , EP
Rua D. Francisco Manuel d e Melo, 5
P-1092 Lisboa Codex
Tel. (01) 69 34 14

Sucursal:
Librera Internacional A E D O S
Consejo de Ciento, 391
E-08009 Barcelona
Tel. (93) 488 34 92
Fax (93) 487 76 59
Llibreria de la Generalitt
de Catalunya
Rambla deis Estudis, 118 (Palau Moja)
E-08002 Barcelona
Tel. (93) 302 68 35
302 64 62
Fax (93) 302 12 99

Journal Department
PO Box 55 Chitse
Tokyo 156
Tel. ( 0 3 ) 3 4 3 9 - 0 1 2 4

MAGYARORSZAG

GREECE/

Distribuidora d e Livros
Bertrand, Ld."
Grupo Bertrand, SA
Rua das Terras d o s Vales, 4-A
Apartado 37
P-2700 A m a d o r a Codex
Tel. (01) 49 59 050
Telex 15798 BERD I S
Fax 49 60 255

Hunter Publications
58A Gipps Street
Collingwood
Victoria 3066

JAPAN

CESKOSLOVENSKO

MundiPrensa Libros, SA
Castell, 37
E-28001 Madrid
Tel. (91) 431 33 99 (Libros)
431 32 22 (Suscripciones)
435 36 37 (Direccin)
Tlex 49370-MPL I -E
Fax (91)575 39 98

Toronto Store:
211 Yonge Street
Tel. (416) 363 31 71

UNIPUB
4611 -F Assembly Drive
Lanham, M D 20706-4391
Tel. Toll Free (800) 274 4888
Fax (301) 459 0056

SVERIGE

J. H . Schultz Information A / S
EFPublikationer
Ottiliavej 18
DK-2500 Valby
Tit. 36 44 22 66
Fax 36 44 01 41

Boletn Oficial del Estado


Trafalgar, 29
E-28071 Madrid
Tel. (91)538 22 95
Fax (91) 538 23 4 9

O t t a w a Store:
61 Sparks Street
Tel. ( 6 1 3 ) 2 3 8 89 85

UNITED STATES OF AMERI CA


ITALIA

SCHWEIZ / SUI SSE / SVI ZZERA

ESPANA

Mail Orders Head Office:


1294 Algoma Road
O t t a w a , Ontario K1B 3W8
Tel. (613) 741 43 33
Fax (613) 741 54 39
Telex 0534783

EuroInfoService B.T.

RUSSIA

Rdy u. 2 4 / B
H-1092 Budapest
Tel. (1) 36 1 118
Fax (1) 36 1 72 83

C C E C (Centre for Cooperation with


t h e European Communities)
9. Prospekt 60-let Oktyabria
117312 M o s c o w
Tel. 007 095 135 52 87
Fax 007 095 420 21 44

POLSKA
Business Foundation
ul. Krucza 38/42
00-512 Warszawa
Tel. (22) 21 99 93, 628-28-82
International Fax&Phone
(0-39) 12-00-77

ISRAEL
ROY International
PO Box 13056
41 Mishmar Hayarden Street
Tel Aviv 69865
Tel. 00972 3 496 108
Fax 00972 3 544 60 39

CYPRUS
UNITED KI NGDOM
H M S O Books (PC 16)
H M S O Publications Centre
51 Nine Elms Lane
London SW8 5DR
Tel. (071) 873 2000
Fax GP3 873 8463
Telex 29 71 138

OSTERREICH
M a n z ' s c h e Verlags
und Universittsbuchhandlung
Kohlmarkt 16
A - 1 0 1 4 Wien
Tel. (0222)531 61-0
Telex 112 500 B O X A
Fax (0222) 531 61-39

36642

Cyprus C h a m b e r of C o m m e r c e a n d
Industry
Chamber Building
38 Ghvas Dhigenis Ave
3 Deligiorgis Street
PO Box 1455

Nicosia

Tel. (2)449500/462312
Fax (2) 458630

SINGAPORE
Legal Library Services Ltd
STK A g e n c y
Robinson Road
PO Box 1817
Singapore 9036

TURKIYE

AUTRES PAYS
OTHER COUNTRIES
ANDERE LNDER

Pres Gazete Kitap Dergi


Pazarlama Dagitim Ticaret ve sanavi
A
Narlibahe Sokak . 15
Istanbul-Cagaloglu
Tel. (1) 520 92 96 - 528 55 66
Fax 520 64 57
Telex 23822 DSVO-TR

Office des publications officielles


d e s C o m m u n a u t s europennes
2, rue Mercier
L-2985 Luxembourg
Tl. 499 28 1
Tlex PUBOF LU 1324 b
Fax 48 85 73/48 68 17

42670 ' 41

im I I *

oll '/. mri I /

You might also like