Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

A Distributed-Element Model for Hysteresis

W. D. I W A N
Assistant Professor o f A p p l i e d M e c h a n i c s ,
and Its Steady-State Dynamic Response
C a l i f o r n i a Institute o f T e c h n o l o g y , A model for the hysteretic behavior of materials and structures is presented. This model
Pasadena, Calif. is based on an approach which views the system as consisting of a series of ideal elasto-
plastic elements. The steady-state dynamic response of the model when subjected to
trigonometric excitation is determined. The effect of the roundedness of the hysteresis
loop on the nature of the response is investigated, and steady-state results, predicted by the
theoretical model, are compared with experimental results from an actual structure.

TI ,HE phenomenon of hysteresis is present in many time. The most likely reason for this is the feeling that a dis-
physical systems. These range from systems composed of a tributed-element approach must of necessity lead to a complex
single element where material damping may be significant to formulation of the force-deflection relations. This, however, is
structures consisting of a number of separate elements. In the not the case. It is the intent of the present paper to show that
latter case, the system hysteresis can result from the hysteretic the distributed-element formulation can be used to generate a
behavior of one or more elements(as, for example, the plastic yield- relatively simple hysteretic model which exhibits the essential
ing of some portion of a structure) or from such mechanisms as features of the hysteresis phenomenon and can be readily applied
slip between different elements of the system. to investigations of the dynamic response of hysteretic systems.
In order to study the effect of hysteretic behavior on system
performance, it is desirable to have a mathematically tractable
model of the system in question. T o date, the most straight-
The Model
forward model proposed for the study of hysteretic systems is the As a starting point for the development of the present model, it
bilinear hysteretic model which includes the elastoplastic model is assumed that the system is composed of a series of so-called
as a special case. In the bilinear model, the force-deflection curve Jenkin's elements as indicated in Fig. 1. Each Jenkin's element
of the system is replaced by a piecewise linear approximation consists of a linear spring with stiffness k/N in series with a
consisting of line segments with two different slopes. The model coulomb or slip damper which has a maximum allowable force
follows naturally from the idea that the system can be thought f*/N, where N is the total number of elements. Each element
of as a series-parallel combination of two linear springs and a is therefore an ideal elastoplastic unit and will have a force-
Coulomb damper. As such, it is a physically motivated model. deflection diagram similar to that shown in Fig. 2 when cycled
Considerable work has been done using the bilinear model
[1-8] 1 and, in general, the results have been satisfactory. The \\\\\\\4<\\\\\\\\
biggest objection to this model is that very few real systems have
a hysteresis loop that is bilinear. Thus, fitting the model to real
systems can be difficult.
In an effort to overcome this difficulty, several hysteretic
models giving rounded hysteresis loops have been proposed [9,
10]. Usually, these models are based on the assumption that the
system restoring force will be an odd function of the deflection u
and tha.t the curves defining various branches of the hysteresis - N
diagram will be mathematically similar to the initial loading or N
backbone curve. Therefore, these models are more or less

1
mathematically motivated. The biggest difficulty with these y/1—| 7,
models lies in the area of cycling between other than fixed loads
• f
where additional mathematical assumptions must be made. Also,
it is sometimes difficult to associate the parameters of the mathe-
matical model with the physical parameters of a particular real
system.
The model considered in the present paper is a physically
motivated model. It is based on the assumption that a general
hysteretic system may be thought of as consisting of a large
number of ideal elastoplastic elements having different yield
levels. This approach was suggested by S. P. Timoshenko as
early as 1930 [11] but has received only little attention since that

1 N u m b e r s in brackets designate References at end of paper.

C o n t r i b u t e d b y the Applied M e c h a n i c s Division and presented at


the W i n t e r A n n u a l M e e t i n g , N e w Y o r k , N . Y . , N o v e m b e r 2 7 - D e c e m -
b e r 1 , 1 9 0 6 , o f T H E A M E R I C A N S O C I E T Y OP M E C H A N I C A L ENGINEERS.
Discussion of this paper should be addressed t o the Editoria, D e -
partment, A S M E , U n i t e d Engineering Center, 345 East 47th Street
N e w Y o r k , N . Y . 10017, and will be accepted until January 15, 1967.
Discussion received after the closing date will b e returned. Manu-
script received b y A S M E Applied M e c h a n i c s Division, June 29, 1965;
final draft, O c t o b e r 11, 1965. P a p e r N o . 6 6 — W A / A P M - 2 . Fig. 2 Typical elastoplastic element

Journal of Applied Mechanics DECEMBER 1 9 6 6 / 8 9 3


Copyright © 1966 by ASME

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


f

Fig. 3 Syslem hysteresis l o o p

between fixed deflection limits. Physically, these elements might If the total number of elements N becomes very large, equation
be thought of as representing a slip plane in a solid material or as (5) may be written in the equivalent form
representing a particular slip joint or yielding member in a
/»C0
complicated composite structure.
f*<p(f*)df* + kx <p(f*)df*; x > 0 (6)
Upon initial loading, the force-deflection relation for a par- 0 Jkx
ticular element of the system will have the form
where >p(f*)df* is the fraction of the total number of elements
/; = kx/N; x > 0, 0 < x < U*/k having/* < f* < / * + df*. In this case, it is no longer necessary
(1)
to know the value of the yield force for each particular member
ft = Ji*/N; x > 0, x > fi*/k
but only the relative occurrence or distribution of the individual
If the direction of loading is reversed after the element has yield forces. If the second term of equation (6) vanishes as
yielded, the force-deflection relation will become x—*• w, the system will have a total "yield force" or "ultimate
U = [kx - (kA - ft*)]/N; force" which will be given by
x < 0, A — 2f*/k < x < A (2)
=
ft = ~i*/N; x < 0, x < A - 2fs*/k f f * VU * )df * (7)
where A is the maximum deflection which occurred with x > 0 as
If the distribution function <p(f*) is continuous and fy is finite, the
shown in Fig. 2. An expression similar to (2) will be obtained
initial loading force-deflection curve or backbone curve of the
when the loading is again reversed so that x > 0. This process
system will have a general shape similar to that of curve ab in
may be carried on indefinitely.
Fig. 3.
If we now consider the force-deflection relation for the entire
If the direction of loading is reversed after initial loading, as
system shown in Fig. 1, it will be seen that the total restoring
along curve bed of Fig. 3, the total force will result from three
force upon initial loading is made up of contributions from two
different groups of elements: Those elements which were in a
sources. Let the elements be arranged in order of increasing
positive yield state after initial loading and have now changed to
yield force. Then, the contribution to the total force due to
a negative yield state; those elements which were in a positive
those elements which have yielded will be
yield state after the initial loading but have not yet changed to a
negative yield state; and those elements which were unyielded
E fi*/N (3) on initial loading and are still unyielded. Thus the restoring
«= i force along the path bed will be given by
where n is the number of elements in a positive yielded state; i.e.,
the number of elements with kx > f*. The contribution of those fc(A-z)
elements which have not as yet yielded will be
/ - J Q - f w r w
kx(N - n)/N (4)
fkA
Thus the total force on initial loading with x > 0 will be + [kx - (kA - f*)Mf*)df* + kx <p(f*)df*;
Jt(A-i) J kA
2
/ = E f*/N + kxiN ~ nVN
«= 1 x < 0, -A < x < A (8)

8 9 4 / D E C E M B E R 1 9 6 6 Transactions of the AS M E

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


The procedure just followed may be continued indefinitely It should also be noted that the elastoplastic model is a special
merely by keeping track of the number of elements in each of the subcase of the present system and is obtained when 0 = 0. This
yielded or unyielded states at any given instant. In this way, the gives a possible means of checking results from the present model
entire history of the hysteresis loop may be traced out without against those of a fairly well-studied model.
introducing any additional mathematical assumptions about the
manner in which the loop curves change after each reversal of the
direction of loading. Thus the present model has considerable Selection of M o d e l P a r a m e t e r s
advantage over the mathematically motivated models in the study One of the attractive features of the present hysteresis model is
of systems which are cycled between other than fixed deflection the relative ease with which the various parameters of the system
limits. can be determined when it is desired to study some particular real
system. The easiest parameter to determine is the spring con-
stant k. This may be obtained from a free-vibration test with a
Specification of <p(f*)
known load or from a static-deflection test. The value o f / „ could
Almost any function <p(f*) could be used in the foregoing be found by statically loading the structure or material to yield-
analysis. One possibility is to specify <p(f*) numerically so as to ing but, in general, this is unnecessary. From the equations for
agree with a given set of experimental data. This procedure the initial loading curve, it will be seen that
might be considered if it is desired to investigate the changes
which take place in the size and shape of the hysteresis loop with dj .
repeated cyclic loading and aging. The alternative to this ap- — = k; x < Zp.i.
dx
proach is the use of an analytic expression for <p(f*). Whiteman
[12] found that a logarithmic normal distribution with its origin
displaced was helpful in studying internal system changes, but T- = b l-W* + (fy + A//'2)A-]; *p.i. < x < xy (12)
dx Af
this type of function leads to rather complex expressions for
the force-deflection relations. In the present paper, we will re-
strict ourselves to an analytic representation of <pif*) which leads
to considerable simplification of the force-deflection relations and
which appeare to be attractive from the point of view of where xv = ( / „ 4- A//2)/fc. The selection of the model parame-
practical application. ters can therefore be accomplished if experimental data on the
Let the distribution function <pif*) be a band-limited function slope of the initial loading curve of the real structure or material
centered about / „ with width Af and an area equal to unity as can be approximated by the three straight-line segments indi-
shown in Fig. 4. T o simplify the notation, introduce a dimen- cated in equations (12). This probably woidd be done graphi-
sioniess load p, a dimensioniess deflection q, and a dimensioniess cally. Then the value of xy would be obtained from the inter-
parameter defined by section of the middle straight-line segment with the df/dx = 0
axis and could be extrapolated from data taken below the ultimate
P = fIfv yield level in any given case. In some cases, the initial loading
curve is not a very accurate representation of the hysteresis loop
Q = kx/f„, Q = kA/f„ (9)
of the real system, and it will be advantageous to consider experi-
0 = A//2/„ mental data obtained from the loop itself. The same three-seg-
ment approximation technique will still apply if \df/dx\ is plotted
Then, from equation (6), the initial loading or backbone curve for as a function of |( A' — z)/2| instead of as a function of x alone.
this system will be given by the relations
Whether or not the three-segment approximation is justified
p = q; ? < i - /3 will depend on the nature of the experimental data for the given
system. In some cases, this approximation appears to lie ac-
ceptable, as will be shown subsequently.
p = q - - a - &)]-; 1 - 0 < q < 1 4- (8 (10)

P = l; q> l + j8 S t e a d y - S t a t e Response
As an example of the application of the present hysteretic
The dependence of the initial loading curve on the parameter 0 is
model, consider a single-degree-of-freedom system which is acted
shown in Fig. 5. For values of 0 between 0 and 1, the system
upon by a periodic force F cos oil and has a hysteretic restoring
will have a definite proportional limit defined by
force f(x, x). The differential equation of motion for this system
qpA. = 1 - / 3 ; zp.i. = /„( 1 - 0)/k (11) may be written as

Journal of Applied Mechanics D E C E M B E R 1 9 6 6 / 8 9 5

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


mx + f{x, x) = F cos cot (13) The frequency-response equation may be obtained from equations
(26) by squaring and adding to eliminate \p. This gives
where m is the system mass, and x is the deflection of the system

{(0 - m i
from its equilibrium position. In this case, the restoring force CiA) I'A
(27)
Six, x) will be an explicit function of x only, but an implicit func-
tion of both x and the past history of the motion. Following the
method of equivalent linearization [13], equation (13) may be re- Although the method of equivalent linearization was chosen
written in the form for the formulation of the frequency-response equation in the
present case, the same result would have been obtained by the
mx + yx + KX + e(x, x) = F cos Cot (14)
Ritz-Galerkin method, the energy balance method, or the method
where of slowly varying parameters.

e(x, x) = f(x, x) — yx — KX (15)


Evaluation of C(A) and 5 ( 4 )
Then, e(s, x) gives a measure of the error which would be intro-
It will only be necessary to consider the motion from 8 = 0 to
duced into the nonlinear differential equation if fix, x) were re-
8 = IT in the evaluation of CiA) and S(.1) since the hysteresis
placed by the "equivalent linearized" restoring force yx + KX.
loop is symmetric for steady-state oscillation. Making use of the
In order to justify this substitution, y and K should be chosen
assumed nature of the individual elements of the present model
so as to minimize e(x, x). In the present case, we choose to
and using equations (24) and (25), it may be shown that
minimize the mean squared value of e(x, x) with respect to y and
K, which implies that kA nkA
CiA) = (2d* - sin 26*) <p(f*)df
2tr Jo
= 0 (16) CO
d? + t/1
kA I ,<pif*)df* (28)
J kA
d[e(x, x)] 2
= 0 (17) and
£>K

kA fkA
where the bar denotes a time average. If the minimization is (29)
SI A) = *<p(f*)df*
performed and e(x, x) is neglected in equation (14), the equation IT JO
of motion will be
where
mx + yx + KX = F cos cat (18)
= cos-'[l - 2f*/ikA)] (30)
T h e steady-state solution to this equation may be assumed to be
of the form Equations (27), (28), and (30) are valid for any arbitrary dis-
tribution function <p(f*). When <p if*) has the particular form
x = A cos l (19) indicated in Fig. 4, it is convenient to introduce the additional
where dimensioniess variables 77, r, c(Q), and siQ) defined by
d = o)< - \p (20)
T, = oo/ik/m)'^
and where A and ip are the constant amplitude and phase of the
motion. Substituting (19) and (20) into (18) and equating co- r = F/fv
efficients of terms in sin 6 and cos 6 yields the steady-state re- (31)
sponse equations dQ) = C(A)//„

siQ) = SiA)/fu
-vico2A + KA = F cos \p
(21) Then, the frequency response equation becomes
ooyA = F sin ip

[fjy
The equivalent linearized parameters y and K may now be
72 = ± (32)
evaluated using the assumed form of the solution (19) and con- ' Q
ditions (16) and (17). This gives
and
— If f(A, 8) cos Odd (22)
It -
A T JO

ciQ) = Q; 0 < Q < 1 0

y =
" f fiA, 6) sin 6d6 (23) c(Q)
tTTp L
4tt/3
7T + cos di - - sin 6i(2 +
3
cos 2 0i)

Autt J 0 Q
where the time average has been performed over one cycle of the + ^ ( 1 - Q + /3); 1 - ^ < Q < 1 +
motion and where fiA, 6) is used as a shorthand notation for

C(Q)=
5[ 0iCOS01
fiA cos 8, —coA sin 6). (33)
Let

i- r2 1
w f ( A , 6) cos Odd (24)
Jo
C(A) =
TV — sin 6,(2 + cos 2 >,) ~ d..'2 cos a2
O

i r2 fiA, d) sin ddd (25) + — sin 0 2 (2 + cos 2 62) ; Q > 1 +


"" Jo
S(A) =

Then, equations (21) for the steady-state response become where

-MW2A + C(A) = F cos \p 8, = c o s - 1 [ 1 - 2(1 - 13)/Q]


(26) (34)
- S I A ) = F sin ip 82 = cos 1 [1 - 2(1 + 0)/Q]

8 9 6 / D E C E M B E R 1 9 6 6 Transactions of the A S M E

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


I
0.4 0.8 1.2 1.6 2.0
FREQUENCY, v

Fig. 6 S t e a d y - s t a t e r e s p o n s e , j3 = 1

and expressions for both c(Q) and s(Q) are substantially simpler than
those obtained using the bilinear models. In this case,
s(Q) = 0; 0 < Q < 1 — /3
c(Q) = Q - OV4
Q < 2 (38)
s(Q) = -QV(3x);
(35)
1 - 0 < Q < 1 +/3 When Q is greater than 2, the expressions for c(Q) and s(Q) take
the same form as those of the elastoplastic model.
Steady-state results for 0 = 1 are shown in Fig. 6. It will be
s(Q) 1 + /3
noted that the response curves have the general soft character
which is typical of most hysteretic models. However, the
The maximum steady-state response of the system will occur present curves do not have the point of vertical tangency that is
when the two roots of (32) coalesce. This will be the case when present in the bilinear hysteretic model. Since the analysis used
in this paper gives the same results as the method of slowly vary-
s(Q„) = —r (36)
ing parameters, its accuracy may be compared to that of the
where the subscript p signifies that the variable is evaluated at the latter method. A comparison of approximate steady-state results
point of peak response. For large Qp, equation (36) reduces to obtained by the method of slowly varying parameters with results
obtained by an exact solution in the case of the strongly nonlinear
4(3 + elastoplastic system shows very close agreement (differences less
Q, = (37)
37r(4/7r - r) than 2 percent) for amplitudes up to Q = 2. For the more nearly
linear case of the bilinear hysteretic system, agreement is es-
Thus Qp approaches infinity as r approaches 4/tt, and the system sentially perfect at these levels [3]. Therefore, it seems reasona-
exhibits unbounded resonance with finite excitation just as in the ble to assume that the results of the present analysis are reliable.
case of the bilinear hysteretic models [8].

Effect of R o u n d e d n e s s
Results for p = 1 The elastoplastic model is often used in the analysis of hys-
In general, an attempt would be made to select the parameter teretic systems even though the hysteresis loop of the real system
/3 on the basis of static or dynamic tests of the real system as indi- has rounded corners instead of sharp ones. As the elastoplastic
cated earlier in this paper. However, such testing is often not model is a special case of the present model when 0 = 0, it is a
feasible, and arbitrary model parameters must be selected. This straightforward matter to compare the dynamic response of a
selection is usually based on a knowledge of the general shape of system with a rounded hysteresis loop to that predicted on the
the hysteresis loop of the real system in such a way as to make basis of an elastoplastic analysis. Figs. 7 and 8 show the effect of
succeeding analysis as simple as possible. For the present roundedness on the general character of the response and on the
statistical model, a particularly simple representation of the amplitude of peak response. From Fig. 8, it may be noted that
system is obtained when /3 = 1. In fact, if Q is less than 2, the the elastoplastic model gives conservative results for the peak

Journal of Applied Mechanics D E C E M B E R 1 9 6 6 / 8 9 7

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Fig. 7 Steady-state response, r = 0.4

EXCITATION LEVEL, r
Fig. 8 Peak a m p l i t u d e of response versus excitation level

898 / DECEMBER 1 9 6 6 Transactions of the AS M E

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Fig. 9 Experimental hysteresis d a t a for column-supported structure

response only if the level of excitation is low (i.e., r less than about just before the final dynamic test. It is evident from the data
0.16 for the present model). This fact should be kept in mind that a change of some sort took place in the structure between the
when the elastoplastic model is used for analysis or design. two static tests. According to the present formulation, this
change takes the form of a lowering of the proportional limit and
Comparison W i t h an Actual S y s t e m an increase in /3; i.e., the distribution function <p(f*) has been
flattened and moved to the left. In the physical model, this
Recently, R . Hanson [14] has made rather extensive experi-
mental studies of the postelastic static and dynamic behavior of corresponds to a deterioration of those elements which have
a single-story structure having structural steel columns. Raw undergone continual yielding with a resulting lowering of their
experimental data taken from two different static tests of this .yield levels. This changing of the structure during testing pre-
structure are shown in Fig. 9 in the form proposed earlier in this cludes the selection of any one set of system parameters which
paper. The first static test, a low-stress-level test, was per- accurately describe the system for all tests. However, it is
formed after the first dynamic test discussed subsequently. possible to select a set of parameters which do account for the
The second test, to somewhat higher stress levels, was performed "average" behavior of the structure during testing. This is done

Journal of Applied Mechanics D E C E M B E R 1 9 6 6 / 8 9 9

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


by making a three-segment approximation to the experimental of elastoplastic elements. As such, the model is physically
data where, in this case, the low-stress-level data are favored some- motivated.
what at low stress levels. The three-segment approximation used 2 The equations describing the steady-state response of the
here is shown as a series of solid lines in Fig. 9. The parameters model to trigonometric excitation have been obtained using the
which result from this approximation are /3 = 0.643,/„ = 1042 lb, method of equivalent linearization.
and k = 3100 lb/in. 3 Steady-state response curves have been presented for the
In the dynamic studies of the real structure, the excitation was case where the yield forces of the elements of the model are dis-
a harmonically varying force whose amplitude was proportional tributed uniformly over a finite band. In this case, the theory
to the square of the frequency of excitation. Thus the theoretical predicts the existence of an unbounded resonance with finite
frequency response equation (32) must be modified if a meaning- excitation just as in the case of the bilinear hysteretic system.
ful comparison with the real system results is to be made. The 4 The effect of the roundedness of the hysteresis loop on the
modified equation will be steady-state response has been examined, and it has been shown
that a sharp-cornered elastoplastic approximation to a curved
hysteresis loop will give conservative (higher) estimates for the
^ = (Q' -1 ^
/iJ)
± ft'MQ)]* - <Q2 - ftf)MQ)],},/'] maximum amplitude of response only if the amplitude of excita-
tion is small.
(39) 5 The present distributed-element model has been used to
obtain the steady-state frequency response of a real system. The
where results are sensitive to the choice of model parameters but, in
general, they appear to be in satisfactory agreement with experi-
magnitude of excitation force at natural frequency of system
mental data.
(40)
References
A comparison of the experimental and theoretical steady-state 1 L. S. Jacobsen, " D y n a m i c Behavior of Simplified Structures
response results for this structure is shown in Fig. 10. The tests up t o the Point of Collapse," Symposium on Earthquake and Blast
were performed in order of increasing level of excitation, and the Effects on Structures, 1952, pp. 112-113.
2 L. E. G o o d m a n and J. H . K l u m p p , "Analysis of Slip D a m p i n g
effective weight of the test structure varied between 4223 lb and W i t h Reference t o Turbine-Blade Vibration," JOURNAL OF APPLIED
4250 lb. It appears from the figure that the present theoretical MECHANICS, vol. 23, TRANS. A S M E , vol. 78, 1956, p p . 421-429.
model is capable of giving a fairly accurate estimate of the peak 3 W . T . T h o m s o n , " A n a l o g Computer for Nonlinear Systems
amplitude of response as well as the general shape of the response W i t h H y s t e r e s i s , " J O U R N A L OF A P P L I E D M E C H A N I C S , v o l . 2 4 , T R A N S .
A S M E , v o l . 79, 1957, p p . 2 4 5 - 2 4 7 .
curves for this system.
4 G . V. Berg and D . A . D a P e p p o , " D y n a m i c Analysis of Elasto-
An indication of the sensitivity of the statistical model to dif- plastic Structures," Proceedings of the American Society of Civil Engi-
ferences in the approximation of experimental data may be seen neers, vol. 86, N o . E M 2 , 1960, pp. 35-58.
by considering a second three-segment approximation as shown 5 T . K . Caughey, "Sinusoidal Excitation of a System W i t h B i -
l i n e a r H y s t e r e s i s , " JOURNAL OF A P P L I E D M E C H A N I C S , v o l . 2 7 , T R A N S .
by the dashed lines in Fig. 9. The model parameters which re- A S M E , vol. 82, Series E , 1960, pp. 640-643.
sult from this selection are /3 = 0.755, f„ = 1057 lb, and k = 3100 6 R . Tanabashi and K . Kaneta, " O n the Relation Between the
lb/in. If the steady-state response is calculated using these Restoring Force Characteristics of Structures and the Pattern of
parameters, the peak of the response curve corresponding to the Earthquake Ground M o t i o n , " Proceedings of Japan National Sym-
posium on Earthquake Engineering, T o k y o , Japan, 1962.
h = 0.019 curve in Fig. 10 is lowered by about 12 percent; the
7 W . D . Iwan, " T h e D y n a m i c Response of the One-Degree-of-
peak corresponding to that of the h = 0.064 curve is lowered by Freedom Bilinear Hysteretic S y s t e m , " Proceedings of the Third World
about 4 percent; the peak corresponding to h = 0.124 is un- Conference on Earthquake Engineering, 1965.
changed; and the peak corresponding to h = 0.211 is raised by 8 W . D . Iwan, " T h e Steady-State Response of a Two-Degree-of-
about 2 percent. There is no significant change in frequencies. Freedom Bilinear Hysteretic S y s t e m , " JOURNAL OF APPLIED M E -
CHANICS, vol. 32, TRANS. A S M E , vol. 87, Series E , 1965, pp. 151-156.
Thus the response is rather sensitive to the choice of model 9 P. C. Jennings, "Periodic Response of a General Yielding
parameters, especially for low stress levels. However, at the Structure," Proceedings of the American Society of Civil Engineers,
same time, it should be pointed out that variations of as much as v o l . 90, N o . E M 2 , 1964, pp. 131-166.
10 percent in the experimentally determined peak response be- 10 W . Ramberg and W . R . Osgood, "Description of Stress-Strain
Curves b y Three Parameters," N A C A T N 902, July, 1943.
tween the virgin and worked structure were not uncommon [14].
11 S. P . Timoshenko, Strength of Materials, Part 2, first edition,
D . V a n Nostrand C o m p a n y , New Y o r k , N . Y „ 1930, pp. 679-680.
12 I . R . Whiteman, " A Mathematical M o d e l Depicting the
Stress-Strain Diagram and the Hysteresis L o o p , " JOURNAL OF AP-
S u m m a r y and Conclusions PLIED M E C H A N I C S , v o l . 2 6 , T R A N S . A S M E , vol. 81, Series E , 1959,
The results of the preceding analysis may be summarized as pp. 95-100.

follows: 13 T . IC. Caughey, " R a n d o m Excitation of a System W i t h Bilinear


Hysteresis," JOURNAL OF APPLIED MECHANICS, vol. 27, TRANS.

1 A relatively simple model for the hysteretic behavior of A S M E , vol. 82, 1960, Series E, pp. 649-652.
14 R . D . Hanson, "Post-Elastic Response of Mild Steel Struc-
materials and structures has been developed. This model is based tures," P h D thesis, California Institute of T e c h n o l o g y , Pasadena,
on an approach which views the system as consisting of a series Calif., 1965.

9 0 0 / D E C E M B E R 1 9 6 6 T r a n s a c t i o n s of t h e A S M E

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

You might also like