Publisher’s version / Version de l'éditeur:
Vous avez des questions? Nous pouvons vous aider. Pour communiquer directement avec un auteur, consultez la première page de la revue dans laquelle son article a été publié afin de trouver ses coordonnées. Si vous n’arrivez pas à les repérer, communiquez avec nous à PublicationsArchive-ArchivesPublications@nrc-cnrc.gc.ca.
Questions? Contact the NRC Publications Archive team at
PublicationsArchive-ArchivesPublications@nrc-cnrc.gc.ca. If you wish to email the authors directly, please see the first page of the publication for their contact information.
https://publications-cnrc.canada.ca/fra/droits
L’accès à ce site Web et l’utilisation de son contenu sont assujettis aux conditions présentées dans le site LISEZ CES CONDITIONS ATTENTIVEMENT AVANT D’UTILISER CE SITE WEB.
Building Research Note, 1966-10-01
READ THESE TERMS AND CONDITIONS CAREFULLY BEFORE USING THIS WEBSITE. https://nrc-publications.canada.ca/eng/copyright
NRC Publications Archive Record / Notice des Archives des publications du CNRC : https://nrc-publications.canada.ca/eng/view/object/?id=0042641f-eda2-4874-9fad-c62f4a73fad7 https://publications-cnrc.canada.ca/fra/voir/objet/?id=0042641f-eda2-4874-9fad-c62f4a73fad7
NRC Publications Archive
Archives des publications du CNRC
This publication could be one of several versions: author’s original, accepted manuscript or the publisher’s version. / La version de cette publication peut être l’une des suivantes : la version prépublication de l’auteur, la version acceptée du manuscrit ou la version de l’éditeur.
For the publisher’s version, please access the DOI link below./ Pour consulter la version de l’éditeur, utilisez le lien DOI ci-dessous.
https://doi.org/10.4224/40000651
Access and use of this website and the material on it are subject to the Terms and Conditions set forth at
Investigation of methods of determining structural damping from
random vibrations
11
s e r
TH
B92I
Building
~esearch
Note
no.
57
c, 2C
t
DGINVESTIGATION
OF METHODS OF
DETERMINING STRUCTURALDAMPING
FROM
RANDOM VIBRATIONSH.S. W a r d and
L.
HalabiskyINVESTIGATION O F METHODS O F DETERMlNING STRUCTURAL
DAMPING FROM RANDOM VIBRATIONS
by
H.S. W a r d and L, I-Ialabisky
The analysis of random vibrations of buildings, caused by
winds, has been u ~ e d to esti'mate the damping characteristics of these
buildings. A s the true values of the damping w e r e not known, there was no check on the accuracy or validity of the methods u s e d . A n analogue computer simulation of the r a n d o m vibrations of a theoretical
five
-
s t o r e y building was therefore carried out to i n v e s t i g a t e these matters, In both the field and experimental work v i s c o u s damping wasassumed.
So
Ear, practical measurements have supported thisas sumptian,
Band limited white noise excitation ( 0 to 35 cps) w a s used because analysis of wind records shows its power spectrum is
essentially flat in the range of building frequencies. Autocor relation
and power spectrum a n a l y s i s had both been u s e d to determine t h e damping values of actual buildings; therefore, e a c h of them was investigated in the analogue simulation. Cherry and Brady (1) have
recently reported some work involving the use of autacorrelation a n a l y s i s to determine structuraL damping values from random v i b r a t i o n records.
EXPERIMENTAL PROCEDURE
A five - s t o r e y shear-type structure w a s simulated on the
analogue computer. It was assumed to b e one bay wide and supported
on a rigid b a s e . A w e i g h t ,
W ,
was assumed to be concentrated ateach floor level and all t h e columns and s t o r e y heights w e r e assumed
t o be equal, The equation of motion for the nth floor of an N-storey
building when e a c h is subjected to a f o r c e of the form
?!.
2 is g i v e n byg f
In Eq. ( I ) , g is the acceleration due t o gravity,
p
is the damping c o e f f i c i e n t , and the K n i t s are the elements of the J'tiffness matrix of the structure. The stiffness matrix of a f i v e - s t o r e y shear-( S E I / L ~ ~
,
Qo
-4
E i s Young" modulus, 1 is the second moment of area of the columns,
and L is the distance between f l o o r s . For the investigation the value
of W L ~ / ~ E I ~ w a s taken equal to 1.0 x
l o q 2
(secl2. In t h i s case thefive natural frequencies of undamped vibration are 0.907 cps, 2.65 c p s ,
4.18 c p s , 5 . 3 8 c p s and 6.20 c g s .
The analogue structure w a s excited for 15 min, during which time the v e l o c i t y outputs or displacements w e r e recorded f o r analysis
purposes. Each 15 min of simulation generally represented a case when
the damping value on each floor w a s the same. The 15-min recording
period corresponds approximately t o the recording time used i n acixal
buildings,
V e l o c i t i e s f r o m the analogue simulation w e r e measured with
0, 1, 2 , 3 , 4 and 10 per cent c r i t i c a l damping. Displacements were
recorded with 0, 1, 2, 3 , 4 and 5 per c e n t c r i t i c a l damping. T h e s e
damping values for any particular t e s t w e r e equal on each floor except
in ofie instance. 1n this c a s e , d i f f e r e n t values of damping w e r e assigned to each f l o o r to investigate the effect on the measured values.This had
some practical importance as it is likely that the damping in multi-
storey structures will vary throughout the building.
A seven-channel f r e q u e n c y -modulated tape recorder w a s u s e d
t o r e c o r d the output voltages from the analogue computer s a that a
power -spectrum d e n s i t y a n a l y s i s and autocorrelation of these outputs
could be performed. The output voltages w e r e recorded on every floor
f o r both velocities a n d displacements. A11 recordings on the F M tape
recorder were at 1 718 i n . / s e c ,
In o r d e r t.o do a p o w e r - s p e c t r u l n density analysis of each
recording, it was played back at 1 5
in.
/ s e c and r e c o r d e d on an 8-Itloop at 1 7/8 i n . / s e c . T h e loop w a s t h e n played back at 60 i n . / s e c into a spectrum analyser. T h e total speed-up factor was 2 5 6 ; i.e,, 256 cps
(analysis time) = 1 cps (real t i m e ) . T h e p o w e r - s p e c t r u m analysis w a s
performed with a bandwidth of 1 c p s and an averaging time of 1. 5 s e c .
The d a m p i n g w a s f o u n d f r o m the power spectrum by measuring
the bandwidth a t half - p o w e r p o i n t s . W i t h a icnowledge of this bandwidth
i t was p o s s i b l e to c a l c u l a t e the Q value 01 the system f r o m Q = fJ(half- p o w e r b a n d w i d t h ) , w h e r e f, is the r e s o n a n t f r e q u e n c y . The percentage
The autocorrelation analysis was c a r r i e d out on the S . D.S, 9 2 0 computer. Ten min of r e c o r d of each s e l e c t e d track were played
into the cortlputer at a speed of I 7/8 in, /sec. The sampling time w a s
s e t at 6 1 milliseconds, which gave approximately 10,000 sample
readings over which the analysis was done. The autocorrelation function
was calculated for 190 time-delay values. An X - Y recorder w a s used to
record the output with time measured on the x axis and the relative
amplitude on the y axis. The damping w a s found from a chart by
measuring the ratios of different peaks to the f i r s t , i . ~ .
,
%/x, w h e r eXn is the amplitude of the nth peak, with xo the amplitude of the initial
peak. From the ratio %/xo the percentage of critical damping can be
found from a chart similar t o that shown in Figure 1 .
The autocorrelation analysis was done f a r the fundamental
frequency only. T h i s w a s isolated by means of a band-pass filter with the low frequency cut-off at 0 . 1 cps and the high cut-off at 1.7 c p s / s e c . THEORY O F DAMPING
IN
RANDOM VIBRATIONSL n ~ u t -Olttnut Relations for S D ~ ctral Densitie s
Consider t h e case of a linear system that has a complex
frequency response, H(rn). If i t i s excited by a function, f ( k ) , then the
response, x(t), is defined by
w h e r e is the F o u r i e r transform of f ( t ) , If f ( t ) i s a stationary
random p r o c e s s It c a n be shown ( 2 ) that the power spectrum of t h e response. S x ( a ) , is related t o the power spectrum of the input. S f ( w ) ,
by the relation,
If it i s now assumed that the input has a uniform spectral density,
FY
,
0From this result w e can conclude that the response spectral
density is equal to a constant times the absolute value of the square of the complex frequency response of the system in the c a s e of white
n o i s e excitation, This proves that i t i s possible to measure directly
the Q value of a system from t h e power spectrum of the output produced by an input of white noise. A comparison of the half-power bandwidth
obtained from the analysis of displacement and velocity of a high
system is considered in Appendix A , with the results shown in Figure Awl,
It can be seen that there is no significant error in the u s e of velocity measurements t o obtain the Q value of such a system.
Autocar relation Analysis
The autocorrelaticn function, Rf(T), of a function, f ( t ) , is defined by
T
~
~
=
(
Lim~
1
-L
J
f(t) t(t+
71
dt.T j o
2T
Also by definition the autocorrelatian function is related to the power
spectrum, Sf(lu). of f(t) by
Consider the free vibrations of a single -degree -of-f reedom
system that h a s an initial displacement, A, and no initial velocity condition. For a lightly damped system the displacement as a function of time is
given by
f ( t )
=
Ae-at
cos w t ... .
( 8 )w h e r e
a
is directly proportional to the damping for values l e s s than 10 per cent of critical damping, and is the a n g u l a r frequency of free vibrations. F r o m Eq.(6)
the autocarrelation function, R( = I ,
of f(t3 isf given b y h 2 -a(2t
+
T I R f ( d =-
T 0 c o s ~ j t ( c o s ,ut cos (u 7-
sin l o t s i n. .
( 9 )T h e i n t e g r a l is o n l y pcrformccl b c t w e e ~ l the limits 0 to T because for
The i n t e g r a l can b e put into the f o r m ,
A~ n a ( Z t + s )
TI'
J
e { C O B w,+
C O S ~ 1 7 C 0 8 2 l U t 0-
sin sin 2 wt1
dt.The separate terms in the integral can be integrated by parts and when
T = -
and n is l a r g e , the value ofR
( T ) i s given byW f
A~ CI
C -a7
{,
cos 1tI-T+
2 C O S W T R i d
=
2Ie
+ U J )A
sin WT.
r - a 110)-
2 2 i a + w )1
The f i r s t t e r m in Eq. (10) is dominant since
a
is small, and itcan be seen that the autocorrelation function has an exponentially-damped amplitude directly proportional t o the damping value of the s y s t e m .
If
the velocity of vibration i s analysed, then again the autocorrelation
function will have a dominant damped t e r m directly proportional t o the
damping of the system when this damping is small. This c a n be demon-
strated by d i f f e r e n t i a t i n g Eq. ( 8 ) to obtain the velocity of the system, v ( t ) ,
The second t e r m in Eq, (11) c a n be ignored because (11 i s small and the dominant term of the autocorrelation function, R of v(t3 will
v
be given by
The preceding e x p r e s s i o n s for the autocorrelation $unction r e f e r to the case when t h e r e i s n o forcing function, If the system is forced t o vibrate then the expression g i v e n i n
Eq.
(7) must be used to evaluate theautocorrelation function, T h i s h a s been done in Appendix B f o r t h e
power spectrum. S ( " ] ) , of a lightly damped single-degree-of -freedom
system given by
L
tU
2
w h e r e w, is the natural frequency of the system,
C
is the amount of viscous damping andCcritical
is the amount of critical damping, An-
example of the autocorrelation function when a.
=
1 andc
/
c
-
~
~
~
~
~
~
~
~
0.0 1 is plotted in Figure B-I f r o m the data of Table B-1, It can b e
seen from Appendix
B
that the autocorxelation is of the form,where A >, A 1 2 and -1 2
@
= T a n 3.
@
*When
g
is small ( < O.X),sin-
=5
and Eq, (23) becomes 2R(T) e
-"
p1
+
A
2 sinI
,,This dernonstxates that f c r small damping
the
autocorrelationfunction of the random vibrations of a single-degree-of-freedom system has an exponentially - d a m p e d amplitude directly proportional to the
damping.
The r e s u l t s obtained in this section indicate that it should be
possible to d e t e r m i n e damping values f r o m the autocor relation analysis of r a n d o m vibrations by m e a n s of the logarithmic decrement: approach s u m m a r i z e d in Figure 1 ,
RESULTS
The results obtained from the power spectrum analyses of the
fifth floor displacements f o r f i v e different values of damping are shown
in Figure 2. This g i v e s an indication of the relation that can be expected between t h e measured damping value and the actual damping value. The
measured damping values tend t o b e too high but the scatter most
probably is due to the statistical r e l i a b i l i t y of the results(there is very
little information a v a i l a b l e on this a s p e c t of the problem and i t requires further attention). The fourth and fifth modes did not appear on the
i
- 7 -
P o w e r spectrum analyses w e r e performed t o see how the
measured values of damping varied f r o m floor to floor when the
actual value s e t up on the simulation was the same for each floor,
T h i s comparison was achieved by a n a l y s i n g the displacements of
e a c h f l o o r with damping values of 1 and
5
per cent respectively,F o r 1 per c e n t damping the average measured value obtained from
the five floors was 2 . 0 3 per cent wiih a standard deviation of 0 . 1 9
per cent; in the case of 5 per cent damping, the average value w a s
2.68 per cent and the standard deviation w a s 0. I I per cent. This
indicates that when the same value of damping exists at each floor,
then the measured value is independent of the floor analysed.
In the field ineasurements made of the wind-induced
vibrations of buildings, the velocity of the vibrations i s measured.
It has been shown that far small values of damping the analysis of
displacement or velocity recordings should f u r n i s h the same -galue for the damping coefficient. A s the analysis of velocity r e c o r d i n g s
had been used for the f i e l d measurements i t was a l s o used f o r the analogue measurements as a check of the method, F i g u r e 3 shows the
results of the power spectrum analysis of the velocity recordings of the
fifth floor when the same damping coefficient was u s e d at each floor. It can be seen that there is reasonable agreement between the m e a s u r e d
value and the actual value up to 5 per cent critical damping. At 10 per cent critical damping the measured value is approximately one -half the
actual value; this could b e due t o scatter but s i n c e the higher values
of damping a r e a l s o underestimated b y the analysis of the displacements
the error may be due to a more basic cause.
T h e r e is every possibility that in the case of actual buildings
the damping coefficients may vary significantly from f l o o r t o floor,
In
order t o i n v e s t i g a t e this case, different damping values w e r e appliedto each floor of the analogue simulation; the values used were 5 , 4 , 3 ,
2 and
Z
per cent on the Ist, Znd, 3rd, 4th and 5th floors, respectively,The results of power spectrum analysis of velocity and displacement
and autocoxrelation analysis of velocity are shown in Figure 4. It
appears that the measured damping values tend to be an average value of the t r u e damping values. This couldbeexplained by considering the e f f e c t of the feedback terms f r o m t h e more lightly damped floors t o the
m o r e heavily damped floors.
In F i g u r e 4 the displacement results tend to increase with
increase in actual damping values; it s e e m s , however, that the analysis
The s o l i d s y m b o l s in F i g u r e 5 represent the results obtained
directly f r o m :he autocorrelation function b y means of the logarithmic decrement approach. The results w e r e generally obtained by taking the
average value of damping values obtained from different portions of the autocorrelation function. It can be seen that there is some correlation between measured and actual damping values but t h e tendency i s to undere s timate the t r u e damping value.
A c l o s e r investigation of the autocorxelation functians obtained
for the different damping values showed that there w a s a tendency for the amplitude of the function to remain at a constant level after a certain interval of time. T h i s feature of the autocorrelation functions is dernon-
strated in F i g u r e 6 where, after 8 s e c , the amplitude seems to remain constant. T h i s indicates the presence of an error s i g n a l , which may be caused by the r e s p o n s e of the fil:er to the signal that i s being analysed, A logarithmic pLot of the uncorrected amplitudes is shown b y solid
symbols in Figure 7. This plot has t w o distinct linear portions; the
slope of such a plot is proportional to the damping value and therefore
t h e r e should be only one linear section. A f u r t h e r indication that the autocorrelation function i s in error is provided by a consideration of
the t e r m Ae
-ed
i n the autocorrelation function. For the examplegiven in Figure
6 ,
the amplitude between 5 and b s e c should be of the order of 2 per cent of that at time,=
0; in actual fact it is 20 percent.
If it is a s s u m e d that the measured autocorrelation function
is of the f o r m ,
w h e r e K c o s ~ n t is the error signal, then the damping value can be
determined f r o m the ratio of successive peak values obtained by
s u b t r a c t i n g K f r o m the peak values of RET]. The c o s r e ~ t i o n level is
shown
in
Figure 6 and with this correction the value between 5 and 6sec is of the o r d e r of 5 per cent of that at 7 r 0. Furthermore, a
logarithmic plot of the amplitudes for this case is shown by open
symbols in F i g u r e 7 , and this g i v e s a single linear section. If the
constant level shown in F i g u r e 6 w e r e t o o great, the logarithmic plot would again have two s e p a r a t e s e c t i o n s , in which case the s e c o n d s e c t i o n
would possess a g r e a t e r slope than the first. This would be opposite to the c a s e for the uncorrected values when i t is assumed t h a t K = 0 .
The procedure of m e a s u r i n g a m p l i t u d e s w i t h respect to the
of the damping values investigated. The r e s u l t s for the measured
damping values obtained by this method a r e shown as the open symbols in F i g u r e 5. The a g r e e m e n t between the measured and
theoretical values is quite good f o r damping values up t o 10 per
cent of the critical value. Both the c o r r e c t e d and uncorrected r e s u l t s show that similar damping values are obtained from the
analysis of velocity or displacement; the damping values a l s o
appear independent of the f l o o r an which t h e analysis is made if
the same value of damping exists at each floor.
DISCUSSION O F RESULTS
The results obtained f rorn the power spectrum analysis of
the random vibrations of the theoretical five -storey building indicate
a reasonable correlation between the measured damping value and the true value, There is a certain amount of s c a t t e r in the results,
h o w e v e r , and this aspect of the problem I s worth further study. The
problem is probably b e s t approached by analysing a s e r i e s of 10-min r e c o r d i n g s f o r a fixed damping value. T h i s would then provide a
statistical distribution f o x the rneasur ed damping values. L e n g t h of
record also i n f l u e n c e s the reliability of the r e s u l t s , but in practice t h e r e is a limitation imposed on t h e record length by the analysis
equipment, because i t is not feasible to have an averaging time greater
than 5 s e c . Therefore, the b e s t approach f o r the p o w e r spectrum
method at this s t a g e appears to be a knowledge of the statistical
d i s t r i b u t i o n of damping values for original record lengths of the order of 1 0 rnin.
One of the disadvantages of t h e power spectrum method is
that for lightly damped systems the half-power bandwidth is small and
it must b e c a r e f u l l y measured from the r e c o r d . In the autocorrelation
method, however, the procedure of plotting the logarithmic values of successive amplitude peaks provides an accurate method of determining
the slope of a line that is d i r e c t l y proportional to the damping value.
E
enough time-delays a r e used i n t h e calculation of the autocorrelationfunction and a constant amplitude occurs, t h e n the correction which is
s u m m a r i z e d in F i g u r e s 6 and 7 provides v e r y good a g r e e m e n t between
the measured and actual damping values. Unlike the power spectrum
method, the autocorrelation method gives good correlation between
measured and a r t u a l damping up t o 1 0 p e r cent of the critical value. The
power spectrum method seems to u n d e r e s t i m a t e the damping between 5 t o 10 p e r cent of t h e c r i t i c a l value but t h i s a l s o needs further investi- g a t i o n because i t may bc due t o the statistical variations of the results,
T h e o r y indicates that t h e r e should be n a error involved
when velocity rather than displacement is analysed to provide damping values. The results that have been obtained support this statement,
and justify the methods that w e r e used to obtain the damping charac- teristics of buildings from measurements recorded from velocity
transducers. The results also indicate that if the damping value is the same on each f l o o r , then there is very little variation in the measured damping value obtained from the analysis of the different f l o o r vibrations,
If
there is a variation in damping through the structure, however, the analysis of the random vibrations of any floor provides an average damping value,In
order to determine the variation in damping, a controlled frequency response t e s t would have t o be performed. CONCLUSIONSActual damping values and measured values based on the
autocorrelation analysis a g r e e within about 1 5 per cent for values
up to at least 10 per cent of critical damping. The power spectrum
analysis of a similar record d o e s not give such good agreement and
the error is perhaps close to 20 p e r cent.
Both methods, however, will g i v e an estimate of the order
of damping from a single 10-min record. If vibrations are recorded for an hour or more, this record could be broken down into a series
of 10-min records to provide a more confident estimate of the damping, The results have shown that in the case of lightly damped systems ( l e s s than 10 per cent of the critical value), the damping may
be determined f r o m the velocity or displacement records.
R E F E R E N C E S
1 . C h e r r y , S . and A,G. Brady. Determination of structural
dynamic properties b y statistical analysis of random vibrations.
Proc, T h i r d W o r l d Conference on Earthquake E n g i n e e r i n g ; New
Zealand, 1965.
2. Crandall, S. H. and W, D. Mark. Random vibrations in mechantcal
A M P L I T U D E R A T I O
2
F I G U R E 1
C U R V E S TO P R O V I D E DAMPING V A L U E S FROM A U T O C O R R E L A T I O N
F I G U R E
2
D A M P
l N G R E S U L T S O B T A I N E D FROM
P O W E R
S P E C T R U M
A N A L Y S I S
OF
THE
F l
F T H
F L O O R
D I S P L A C E M E N T S
n
S p e c t r a l
D e n s i t y A n a l y s i s
of
V e l o c i t y
I
I
2
F
I
0
1
2
3
4
5
6:
A C T U A L
P E R C E N T A G E
O F
C R I T I
C A L
D A M P I N G
F I G U R E 4
DAMP
l
N G M E A S U R E M E N T S O B T A l
N E D
WHEN
D l FFERENT
V A L U E S
O FD A M P I N G
W E R EI N T R O D U C E D
A T
E A C H
A n a l y s i s o f 5 t h
F l o o r
Diisplacernent
D i r e c t V a l u e
f r o m
AuZocorrePation
A n a l y s i s
o f 1st
Floor
V e l o c i t y
D i r e c t
V a l u e f r o m A u t o c o r r e l a t i o n
A n a l y s i s
of
5 t h
F l o o r V e l o c i t y
C o r r e c t e d V a l u e , 5 t h F l o o r
0
1
2
3
4
5
6
7
8
9
1 0
A C T U A L
P E R C E N T A G E O F
C R l
T I
C A P DAMP
l
N G
F I G U R E
5
D A M P I N G R E S U L T S
O B T A I N E D F R O M A U T O C O R R E L A T ~ O N
A N A L Y S I
s
2
0 2.02 4.04
6.06
8.08 10.1T I ME, S E C Q N
OS
EFFECT OF A P P L V l N G C O R R E C T l O N LEVEL TO
T H E
A U T O C O R R E L d T l O N F U N C T I O N S H O W NIN
F I G . 6A P P E N D I X A
POWER
S P E C T R U M ANALYSISThe power spectrum SdCW ) , of a viscously damped single
-
degree-of-freedom system is d i r e c t l y proportional to the expression
given by
The expression in Eq. (13 represents the power spectrum of the displace-
ment as a function of the angular frequency, w . the natural angular
frequency, w , and the fraction of critical damping,
P.
0
The power spectrum of the velocity output, S,(w), of a single
-
degree-of -freedarn system is directly proportional to the expressiongiven in E;q. (21,
2
The value of SJW) is divided through by w, in Eq. (2) because
i t p e r m i t s t h e d i r e c t comparison of Sd(w) and
S,(,,)
a s a function ofThis comparison has been plotted in Figure A - l for the c a s e when U ] ~ '
8
= 0.0 5; i t demonstrates that there will be little difference in the half-power bandwidth determined f r o m the analysis of displacement or velocity, when a power spectrum anaLysis is used for lightly dampedF I G U R E A - l
C O M P A R l S O N
OF
R E S O N A N C E C U R V E S O B T A I N E D FROM1
4 4
e r e/ $ I z
= I
-
-&az.)
n I
( A ,
-8,)
=
C.
3 7 1 4 5+ J
. o , o r r )-
( - .
7 9 9 9 s + i u l ~ l g )=
I.
79947( p a
-
A,)
=
d.
9 7r r r
+;
.
a r ~ ~ s )
- [ -
.W
7 7 5 -i .o,D,g) *.
=
, O J 0 3 6 ; - .-(4,
-
a,)
( -
7 5+;
.
0 , 6 1 8 )-
(.
1 4 9 9 5-i
- o ) a ) ~ )F I G U R E
B - 1
A C T T O C O V A R l A N C E F U N C T I O N O F 1