:PSE REQUEST/ web delivery

Subject: PSE REQUEST/ web delivery

From: PSE Melvyl Request Account <[email protected]>
Date: Thu, 16 Apr 2009 16:56:43 -0700 (PDT)
To: Northern Regional Library Facility <[email protected]>

'
·t~

---------- Forwarded message----------

Date: Thu, 16 Apr 2009 16:33:09
From: "V'Boris Jeremic <jeremic"@ucd$vis.edu> 11 @pesto.lib.ucdavis.edu
To: [email protected] . , ...
Subject: Document Delivery Service Request [did Form]

A DDS Request was submitted on April 16, 2009, 4:33 pm by:

Name: Boris Jeremic ([email protected])

Dept!Schl: Civil and Env. Eng
Phone: 530.754.9248
Status: Faculty

Home Lib: Physical Sciences and Engineering

Timelimit: No Time Limit
Delivery: (not selected)
Web Delivery: yes

ARTICLE
Periodical Title: Journal of the Engineering Mechanics Division ..
Article Author: Newmark, N. M.
Article Title: A method of computation for structural dynamics ...
Volume: ASCE 85 , .\
Issue: EM3
Date: 1959
Pages: 67-94

Location: NRLF (Available at NRLF)

Gii5l
NRLF
Ariel

Date _ v l L...'1 1"J r,

'Pr ~r:. \-:r A

~~· 4117/2009 8:19AM
:{of 1 '"'
UNIVERSITY OF CALIFORNIA

Northern Regional Library Facility

ELECTRONIC DELIVERY COVER SHEET

WARNING CONCERNING COPYRIGHT RESTRICTIONS

The copyright law of the United States (Title 17, United States Code) governs the making of
photocopies or other reproductions of copyrighted materials.

Under certain conditions specified in the law, libraries and archives are authorized to furnish a
photocopy or other reproduction. One of these specified conditions is that the photocopy or
reproduction is not to be "used for any purpose other than private study, scholarship, or research."
If a user makes a request for, or later uses, a photocopy or reproduction for purposes ill excess of
"fair use," that user may be liable for copyright infringement.

This institution reserves the right to refuse to accept a copying order if, in its judgm~nt, fulfillment
of the order would involve violalion of copyright law.
 J oumal of the
ENGINEERING MECHANICS DIVISION
Proceedings of the American Society of Civil Engineers

ENGINEERING MECHANICS DIVJSION

EXECUTIVE COMMITTEE
Daniel C. Drucker, Chairman; Dan H. Pletta, Vice-Chairman; •
Egor P. Popov; ·Bruce G. Johnston; Edward Wenk, Jr., Secretary

COMMITTEE ON PUBLICATIONS
H. Pletta, Chairman; JohnS. Archer; W. Douglas Baines; Hans Bleich;
Albert G. H. Dietz; Robert J Hansen; Eivind Hognestad; Ernest F. Masur

CONTENTS

July, 1959

Papers

Page

27

51

67

95
!lm~inlme:,e.
Swelling and Creep in Cement
by A. Hrennikoff . . . . . . . . . . . . . . . . • . . . . . . . . . . . . . . . . 111
137

ji:;;;~~ht""w\Q by the American Society of Civil Engineers.

i
I i
2094 July, 1959 EMS

Journal of the
ENGINEERING MECHANICS DIVISION
Proceedings of the American Society of Civil Engineers

A METHOD OF COMPUTATION FOR STRUCTURAL DYNAMICS

Nathan M. Newmark,l F. ASCE

This paper is one of the group of papers which were the bases for oral
presentations at the Joint ASCE-IABSE Meeting, New York, October 1958.
The entire group, including those published in CIVIL ENGINEERING, will be
reprinted in one volume.

SUMMARY

This paper describes a general procedure for the solution of problems in

structural dynamics. The method is capable of application to structures of
any degree of complication, with any relationship between force and dis-
placement ranging from linear elastic behavior through various degrees of
inelastic behavior or plastic response, up to failure. Any type of dynamic
loading such as that due to shock or impact, vibration, earthquake motion, or
blast from a nuclear weapon, can be considered.
A method of numerical integration is described which for simpler cases
and· for a relatively small number of degrees of freedom is suitable for use
with desk calculators. However the method is developed particularly for use
with high-speed digital computers. Consideration is given to various types of
damping, and to nonlinear behavior. A description is given of a method of
treatment of elasto-plastic members in flexure, including the development of
yield hinges. By suitable means of application of the loading, and with the
introduction of enough damping to prevent indefinite oscillatory motion, the
procedure can be used to determine the '"static" behavior of a structure as it
progresses through various degrees of inelastic behavior up to collapse.
Although the general procedure described is suitable for use in the study
of the response of structures to earthquake motions, a modified procedure is
described for handling the problem of a structure having time-dependent .
boundary conditions, which permits the direct calculation of displacements in
Note: Discussion open until December 1, 1959. Separate discussions should be sub-
mitted for the individual papers in this symposium. To extend the closing date one
month, a written request must be filed with the Executive Secretary, ASCE. Paper
2094 is part of the copyrighted Journal of the Engineering Mechanics Division,
Proceedings of the American Society of Civil Engineers, Vol. 85, No. EM 3, July,
1959. .
1. Head, Dept. of Civ. Eng., Univ. of Ulinois, Urbana, Ill.
67
 July, 1959 STRUCTURAL DYNAMICS 69
:au EM3
the structure relative to the ground displacement, for convenience in inter- cture may be an integral one, either a solid body or an as-
preting earthquake phenomena. However• the sn:s and other elements, in which the nodes may only define
The paper discusses the pzoblem of structures with curved members and semblage of pla1 cture for convenience in the placement of loads and masses.
points on the stru th nodes be so chosen that the resistance of the structure
describes in detail a method of dealing with arched or curved structures in
which the members are not permitted to change in leagth. The treabnent of lt iS necessary t11f
theenodes can be determinable, and that, if desired, the
to displacement o c behavior or plastic behavior can be taken into account.
such inextensible structures presents some difficulties in the general
influence of melasti ide ed to be supported at its foundations at nodes also.
The structure is c~ ~ analysis can take into account a situation in which
treatment but methods are available for modifying the general procedure by
introducing a series of constraint relations which reduces the number of de-
grees of freedom permitted in the motion of the structure. . Although the m with ~e or with displacement, the procedure is presented
the masses clUUlge hi h main constant No major change in concept or in
herein for masses wdcto :al with theca~ of a variable mass-time relation-
The methods described herein have been used i>r the computation of the
dynamic response of structures of various degrees of complexity including procedure is require
arches, domes, stiffened rings, framed structures, and simple spring-mass ship.
systems, subjected to various types of loading including nuclear bomb blast,
earthquake foundation motions, random shock disturbances, wave action, and Method of Analysis
impact and dynamic effects from moving vehicles. However, examples of
these applications are not given. Step-by-step Integration Procedure
The work on which this paper is based was supported at the University of e structure shown in Fig. 1 which is made up of weight-
lllinois by the Mechanics Branch of the Office of Naval Research. Consider the plan t rting lumped masses. The deformable ele-
bu:t d fo mabie elemen s suppo 'th be s
less e r cular instance are shown as e1 er am .
ments of the structure in this parti However much more complex structur-
or bars which act under~:~ements s~wn are deformable and conse-
al types can be consider
INTRODUCTION

Only general principles and methods for dynamic analysis are considered
quently each mass such as M, i' E can move in both the vertical and hori-
D, rl such as at A and B, may also move
zontal direction. The poin~ su::;a rlicular these points may move in such
in this paper. The basic method of analysis is a general step-by-step method both vertically and horizond ,;{'esseP,t'are int~duced in the strucrure by rela-
of iAj:~ration of the equations of motion, and is applicable to any structure a way that deformations an r n the foundation. Any of the masses may
consi~ of a group or series of concentrated or "lumped" masses supported tive motion of the points of suppo: otion or component forces in·the-hori-
have acting on it a force in any ec • ,
on a deformable structure. For convenience, the structure may be considered
to be a fx:l!,mework with joints or •nodes" at which forces may be applied, or zontal and vertical directions. oved from the strucrure .1Jl Fig. 2. In
at which Illal!!ses may be placed. Any finite number of degrees of fre,.edom may The mass at Min Fig. 1 is sh~: re~s is indicated, and the positive di-
be considered, but it is essential in the procedure that the forces required to Fig. 2a the point of attachment o ~ the strucrure, R, and the dis~
produce a pattern of deflection of the framework must be determinable when rections of the resisting force exert di ;ed In Fig 2b the mass is shOwn as
the deflections of the nodes at which masses are placed are specified. n is placement of the structure, x, are in ca: f P R and. x indicated. Although ~s
not necessary that the framing behave elastically. a free body with the positive directio~ oa ~r~ontal force, in general it could
Because in an actual structure the mass is not really divided into separate mass is shown as being acted on by o Y n it also in which case it would
component parts connected together by flexible elements, the structure which have a vertical component of force acting o t f resistance in the vertical di-
is analyzed is only an approximation to the actual structure. However, a have a component of motion and a componen i o determined by the arbitrary
reasonable approximation can usually be made. For the actual structure, with rection. The sign convention that is c:sen Ps The positive directions of x
distributed masses, the number of degrees of freedom of motion of the choice of the positive direction of the th~c=t~cture are taken as the same as
structure is infinitely large. The replacement of the distributed mass by a and of the resisting force R acting 0~ isting i>rce R acting on the free
number of concentrated masses reduces the number of degrees of freedom , . for P • and the positive direction fotor u:.:~; the resisting force R acting on the
body mass M is opposite in sense
and affects the response of the structure in the higher modes.
The simplification of a structure for the purpose of making a dynamic
analysis involves engineering judgment. One must select the essential ele-
ments of resistance of the structure so as to arrive at an adequate and reason-
~*~
The positive directions of the accel~ w~n
placement x are all the same. In gene
ti
:m~~~~~the~
have for each possible com-
displacement velocity, acceler-
ably accurate expression of the resistance of the structure to motion. Also ponent of direction of motion of the mass ~ esisting forces R at any instant,
one must evaluate the masses of the structure and •lump" them together at ation, resisting force, and applied forceb t~t ~stem of forces which are re-
places where they can be considered to act. In general the number of masses of time are defined in such a way as to e t rure into a deflection configu-
which should be considered depends on the accuracy which one wishes to quired to pull the weightless deformable s r;~e displacement x at the same
achieve in the calculations. ration defined by the instantaneous values o
The structure may be made up of individual members connected together instant of time.
at joints which would then be considered as the nodes of the structure.
70 July, 1959
EM3 EM3 STRUCTURAL DYNAMICS 71
The sign convention and notation are chosen so as to make it apparent that
the masses M modify or filter the forces P and transmit them to the structure
lo in modified form as R. If the forces P are applied very slowly there is only
a small acceleration and R is approximately equal to P. If the forces are ap-
plied quickly the difference between R and P can be very large. All the
stresses in the structure are defined by the system of forces R. The
structure can be analyzed statically for these forces. In general the forces R
may continue to exist even after the forces P have dropped to zero. Similar-
ly, R may be defined by the foundation displacements as well as by the de-
flections of the masses, even when no force P acts on the masses.
In general all of the quantities so far defined are functions both of position
on the structure and of time. If we consider a time tn and designate the values
of all of the various parameters including displacements and forces at that
time with a subscript n, such as Rn, our problem becomes that of defining the
displacements Xn+1 and forces Rn+1 as well as the velocities vn+1 and ac-
celerations an+1 at a time tn+1, which differs from tn by the time interval h.
In the argument which follows we shall deal with a single mass in a single
direction, but we might just as well deal with the whole set of masses and
their possible displacements and designate the various directions with ad-
ditional subscripts m such as in Rn m· However in the discussion which
follows we shall drop the second subscript for convenience, remembering that
for each of the degrees of freedom for each mass we have a set of equations
similar to the general set presented here. The derivation which immediately
follows is described in terms of a situation where damping does not exist, for
convenience. Later the procedure will be revised to include damping forces.
In general at any time, (consequently omit the subscript n), the acceleration
is given by the relation:
a = (P - R) /M (1)
A B n is assumed that at time tn we know the values of the displacement and
Flo. I. General Type of Structure and the velocity as well as the acceleration, but we know nothing about the situ-
ation at time tn+1· Although there are methods of numerical integration which
Loading permit us to make estimates, at least for small time intervals, of the dis-
placements and velocities at the later time knowing only the situation at the
earlier time, these methods which do not take account of the change in resist-
ing force during the time interval are not as accurate as the method described
herein.
The method presented here was derived by the writer and first presented
in Ref. 1. The relations which follow are given in terms of the acceleration
at the end of the interval, an+1, although this is not in general known. A dis-

-x
R R ~ p
cussion of how the procedure can be handled in spite of this fact will be given
subsequently. Two parameters, 'Y and {3, are introduced to indicate how much

(a) Structure with

Mass Removed
(b::Vmoved
from Structure
of the acceleration at the end of the interval enters into the relations for ve-
locity and displacement at the end of the interval. The relations which are
adopted are given below:
(2)

Fig. 2. Detail of Mass as Free Body xn+l = xn + vnh + ct- ~) anh2 + ~ an+lh2 (3)
n will be shown subsequently that unless the quantity 'Y is taken as 1/2, there
is a spurious damping introduced, proportional to the quantity
 72 July, 1959
EM3 EM3 STRUCTURAL DYNAMICS 73
r- 112
Now for convergence in a sequence of calculations the quantity p must be
It can be seen that if I' is taken as zero a negative damping reSults, which will numerically less than 1. The critical value, for convergence, of the time
involve a self-excited vibration arising solely from the numerical procedure. interval he can then be computed from Eq. (9) by setting the right-hand side
Similarly, if I' is greater than 1/2 a positive damping is introduced which will numerically equal to 1, with the result:
reduce the magnitude of the response even without real damping in the
problem. Taking account of the fact that I' must equal 1/2, we can rewrite
Eq. (2) as follows:
~=ire ff (10)
Critical values of the convergence limit as a function of {3 are contained in
(4) Table 1.
For practical purposes the time interval would ordinarily be taken as
smaller than that which corresponds to pure oscillation, or p = -1, in order to
Application of the General Procedure inSure rapid enough convergence. If p = -0.32, the errors will be reduced to
In general unless fJ is 0 we may proceed with our calculation as follows: one per cent of their original value in four steps or four cycles of iteration.
(1) Assume values of the acceleration of each mass at the end of the inter-
n would appear that, since for small values of fJ the convergence is most
val. rapid, the lower values of fJ would be best to use. However other consider-
ations affect the choice. The most important of these is the matter of stability
(2) Compute the velocity and the displacement of each mass at the end of
which is discussed in the next section.
the interval from Eqs. (4) and (3), respectively. (Unless damping is
For a complex system, it can be shown that the rate of convergence is de-
present it is not necessary to compute the velocity at the end of the
interval until step (5) is completed.) pendent upon the frequency or the period of the highest mode of the system.
Consequently, the time interval used must be related to the shortest period of
(3) For the computed displacements at the end of the interval compute the
vibration, or the period in the highest mode of vibration, for the lumped mass
resisting forces R which are required to hold the structural framework system. Since stability also depends on a similar criterion, it appears that
in the deflected configuration.
the greater the number of masses into which a system is broken do~ the
(4) From Eq. (1) and the applied loads and resisting forces at the end of
the interval recompute the acceleration at the end of the interval. shorter will be the permissible time interval for numerical calculation of the
dynamic response of the system.
(5) Compare the derived acceleration with the assumed acceleration at the
end of the interval. If these are the same the calculation is completed.
Stability and Errors in Numerical Computation
If these are different, repeat the calculation with a different value of as-
sumed acceleration. It will usually be best to use the derived value as In order to study the stability of the numerical integration procedure, let
the new acceleration for the end of the interval. us consider the special case of a simple system, a mass with one-degree-of-
The rate of convergence of the process toward equality of the derived and freedom without external force acting on it. For such a condition, and for
assumed accelerations is a function of the time interval h. For a single- some initial displacement and velocity, the motion of the system should be a
degree-of-freedom system having a circular frequency of vibration w the pure oscillation, with a circular frequency of vibration as given by the re-
ratio of the error in derived acceleration to the error in assumed acdeler- lation
2 K (11)
ation, (where the error is the difference between a value and the "correct" w = M"
value), is given by the relation:
in which K is the spring constant and M the mass. The relationship between
error in derived acceleration 2 2 the acceleration and the displacement is given by:
error In assumed acceleration = p = - ~ w h (5)
(12)
It is convenient to define the quantity wh by the symbol (} according to
Eq. (6):
wh "'9 (13)
(6)
With this notation Eq. (5) can be rewritten as:

P=-~ 2 00
Since the circular frequency w is related to the period T by the relation X + -
n 1
(2 - ci) X
n
+ X
n-
l + (7 - -21) r:i (x n - X
n-
I) ;= 0
2rc
w =- (8) From the general relations between finite differences and derivatives, it
T
Eq. (7) can be rewritten as follows: can be seen that the last term on the left of this equation corre~nds to a
factor times the velocity of the system, and consequently can be 1nterpreted
2rch 2 as a viscous damping term even though the system was defined as having
P = - ~(r) (9)
74 July, 1959 EM 3 EM3 STRUCTURAL DYNAMICS 75
no damping. This spurious last term can be eliminated by the choice of TABLE I
1 = 1/2. In this ease the general difference equation can be rewritten as:
CONVERGENCE AND STABILITY LIMITS
xn+l - (2 - d-) x 0 + xn-l =0 (15)
The general solution of the finite difference equation, Eq. (15), can be
Values of fl
written in the ease where the quantity oc2 is less than 4. In this ease, define Item
a quantity q, by the relation: 0 1/12 1/8 1/6 1/4
a: = 2 sin t:JJ/2 (16)
Convergence Limit, h/T inf. 0.551 0.450 0.389 0.318
The solution of Eq. (15) can then be written in the form
Stabi I ity Limit, h/T 0.318 0.389 0.450 0.551 inf.
X = A cos t:!J t/h + B sin t:!J t/h (17)
This can also be stated in terms of a pseudo period T 80 and an initial dis-
placement Xo and a parameter B which is of the same form as a velocity. The The stability limit criterion, corresponding to a value of oc2 = 4, can be ex-
result is: pressed in terms of the time interval also. The relation between 01 and 8 1n
X= X cos 2~t/T + B sin 2~t/T (18) Eq. (13) can be expressed as:
0 5 5

This may be compared with the exact solution, i, which is given by Eq. (19): (24)
. v
- 0
x = x0 cos 2~t/T + ~ sin 2~t/T (19) from which the stability limit hs can be written as:
21ths 2 (25)
It can be seen that the approximate solution, Eq. (18), is similar to the
exact solution and gives precisely the same response for an initial dis-
-T- =.J1 - 41'l
placement, but gives a different period from that of the actual system. The which can be simplified to the form:
value of the pseudo period T s is given by the relation:
hs 1/1t (26)
T5 = 2~h/IJJ (20) r = ,J1 - 41'l
The relation between the pseudo period Ts and the true period Tis: Values of the stability limit are shown in Table 1 as a function of {3.
From Table 1, it can be seen that for values of {3 greater than 1/8, if the
~IT=~ ~u time interval is chosen for convergence the numerical procedure will always
The following approximate formula may be used for a simpler definition of the be stable. However, for values of fJ less than 1/8, convergence does not in-
relative value of the pseudo and the real period of vibration: sure stability. Lack of stability gives no warning of difficulty, but introduces
a spurious increasing oscillation into a system which may be in oseilla.Uoo
\IT = rv 1 - (I - 12fl) 9 2/24 - (17 - 120fl + 720fl2 ) e4 /5760 - • • (22) anyway. Therefore an inexperienced computer may not recognize the diffi-
culty. Moreover, an instability in the higher modes only may not even be ap-
The response of the system to an initial velocity is given by the second
parent to an experienced computer. Consequently it appears that unl!'ss other
term in Eq. (18). The relationship between this response and the true re-
sponse to an initial velocity, as shown by the second term in Eq. (19), is indi- steps are taken to insure stability, one should limit the time interval by the
cated in Eq. (23): stability criterion rather than by the convergence criterion.

B [ 1 2]-112 Interpretation of Parameter {3

vo/w = -I + (l'l - 4) e = """ 1 + 1/\- @ 92 + • • • (23)
A method very much similar to that described here for f3 = 0 has been dis-
If fJ is exactly equal to 1/4, the maximum velocity response is correct but if cussed in Ref. 2. A method corresponding in many respects to that for .
it is different from 1/4 there is an incorrect maximum velocity response.
Values of the errors in the period and of the errors in maximum response
.. {3 = 1/12 has been given in Ref. 3. However, the general treatment previously
presented is different from that given here, particularly in the treatment of
to an initial velocity are given in Tables 2a and b for several values of fJ the starting of the motion. A method similar to that for f3 = 1/4 was first
and for a range in values of h/T. There is also given in Table 2e the rate of presented by s. Timoshenko in Ref. 4. However, he did not carry the pro-
convergence for the corresponding tabular entries. For a system with a cedure to the pOint of generalizing it for other than simple Ol~e-degree-of­
number of degrees of freedom, the limits are expressed in terms of the freedom systems, nor did he develop the conditions on stability and
shortest period of the system. convergence.
When oc2 > 4, the solution of Eq. (15) oscillates without bounds, and the
calculation does not yield results even in remote agreement with the exact
solution. The solution is said to be "unstable".
 STRUCTURAL DYNAMICS 77
76 July, 1959 EM3 EM3

TABLE 2

EFFECTS OF LENGTH OF INTERVAL ON ERRORS DUE TO NUMERICAL PROCEDURE

.,._· ...
c
0
Values of f3 :;:
h/T
0 1/12. 1/8 1/6 1/4 .=.. 0
""Q)
(a) Relative Errors in Period Q;
()
0.05 -0.004 -0.0001 0.002. 0.004 0.008 ()
0.10 -0.017 -o.Q003 0.008 0.017 0.033 0
0.20 -0.076 -0.006 0.028 0.059 0.12.1
0.25 -0.130 -0.015 0.038 0.087 0.179
0.318 -0.363 -o .045 0.047 0.129 0.273
00.389 -0.220 0.035 0.170 0.382
0.450
* -0.100 0.195 0.480
* * Fig. :3. Consistent Variations of Acceleration
(b) Rei at i ve Errors ln Maximum Response to an Initial Velocity in a Time Interval
0.05 0.012 0.008 0.006 0.004 0
0.10 0.052 0.034 0.025 0.017 0
0.20 0.209 0.166 0.116 0.073 0
0.25 0.614 0.306 0.202 0.122 0
0.318 in f. 0-732 0.414 0.225 0
0.389 inf. 1 .ooo 0.414 0
0.450
* >~ i nf. 0.]32 0
*

0.05
0.10
0
0
(c) Rate of Convergence
0.008
0.033
0.012
0.049
0.016
0.066
0.025
0.099
-R2

0.2.0 0 0.132 0.197 0.263 0.395

0.25 0 0.206 0.308 0.411 0.617
0.318 0 0.333 0.500 0.667 1.000
0.389 i< 0.500 0.]50 1 .ooo 1.500
0 .1+50
* * 1.000 1 ·333 z.ooo

* Values indicated are beyond 1 lmit for stability.

It is interesting to note the correspondence between J3 and the variation in

acceleration during the time interval. Although a physical relationship is not
possible for all values, for at least four values of /3 it is possible to define
consistent variations of acceleration in the time interval. Three of these are
shown in Fig. 3. It appears that a choice of f3 = 1/6 corresponds to a linear - ' >

variation of acceleration in the time interval; a choice of /3 = 1/4 corresponds

to a uniform value of acceleration during the time interval equal to the mean
of the initial and final values of acceleration; and a value of /3 = 1/8 corre- Fig. 4. Structure with Absolute and Relative
sponds to a step function with a uniform value equal to the initial value for the
first half of the time interval and a uniform value equal to the final value for Damping
the second half of the time interval.
 78
July, 1959
I EM3 EM3 STRUCTURAL DYNAMICS 79
t can also be shown that f3 = 0 corre
at the beginning and end of the time inte~nds. to double pulses of acceleration (3) Now apply the net resistance, corresponding to the difference between

ring just before the end of the preceding int es

beginning of the next interval.
:e
of a part equal to 1/2 of the acceleration timWlth each double pulse consisting
time interval, one occur-
erv and the other just after the
the applied force and the damping force, to the structure and determine
the displacements of the structure at all points when the prescribed dis-
placements are put in at the points where masses exist.
(4) Compute the acceleration at the end of the interval that is required to
Treatment of Damping give the displacement determined in the preceding step.
(5) Compare the acceleration so computed with the one initially assumed
Damping of various kinds may exist in st
forces may be proportional to veloci . a ructural system. Damping and repeat if necessary.
they may be of a frictional type or /:' or to some power of the velocity, or This procedure is considerably more complex than that which is used when
portional to displacement or reiattve :~cases they may be even pro- masses exist at all nodes. Consequently, it may be better to put in an actual
lute", and depend on the particular ement. Damping may be "abso- small mass than to make the mass zero. However, if damping forces are not
spect to ground, or •relative" and dmotion or velocity of the mass With re- acting at points where there are zero masses, then it is relatively simple to
With respect to some other point :end on the motion or velocity of the mass .handle the problem directly.
points having damping forces is ~~wne~t~ture. A structure with several
different from zero then the damp • 4. If masses M1 and Mg are Recommended Time Interval and Choice of fJ
indicated on the right-hand side of~ fofigurces can be applied to the masses, as
the following way: e re, and the calculation is made in In any set of calculations in which an error in any one step makes all
After the acceleration at the end of th subsequent steps incorrect, it is desirable to have a self checking procedure.
the end of the interval and the displ e interval is assumed, the velocity at Consequently the procedure described herein works best where it is used with
(3). The complete motion of the ~ement are computed from Eqs. (4) and a converging system of approximations because under these conditions the
val, and regardless of the kind of~mingis now defined at the end of the inter- calculations in any one time interval are repeated several times with slightly
With the damping force placed on the p , the damping force is determinable. different values of the numbers involved. A close agreement between there-
celeration at the end of the int rval i mass in the proper direction, the ac- sults of successive calculations is in general a sufficient check on the accu-
damping force as Well as the a~plied ~ now computed taking into account the racy of the numerical work when the work is done on a desk calculator. (Such
different from the assumed accel at• ore~ and the resisting force. If this is a check may not be necessary on a high-speed computer.) It can be seen,
As in the case with zero dam er lOn e calculation is repeated. therefore, that a value of fJ =0, where there is no damping, is not necessarUy
structure are computed from th ping, the resisting forces acting on the a good choice because a separate check will be required in these cases to in-
the interval. The damping forcees ~~flacement of the structure at the end of sure accuracy.
the displacements or velocities of th ing on the masses are computed from Studies of the effect of damping and of negative spring constants such as
law used. e masses in accordance with the damping those corresponding to a decrease in load with increase in displacement, indi-
It can be seen that in the case of dam 1 cate that better results are obtained with values of fJ in the range from 1/6
advantages over any other choice of b p ng a Value of f3 = 0 presents no real to 1/4 than in the range below 1/8. In general, with a time interval of the
still has to be assumed at the end of~ ~cause of the fact that the acceleration order of 1/5 to 1/6 of the shortest natural period of vibration, the rate of con-
locity or other parameters which d t e Interval in order to compute the ve- vergence will be rapid enough for all practical purposes, and the errors will
the interval. e ermine the damping forces at the end of be small enough to be tolerable for every combination of damping or negative
For those cases where there are no ma spring constant that appears to be practicable.
or external forces are applied the sit t· sses at points where damping forces As a structure goes into the inelastic range, in general the periods of vi-
trates this by applying forces to th t ua lon is more complex. Fig. 4 illus- bration all become longer and the shortest period becomes longer as well.
sary in these cases that the resist~ ~o zero masses, M1 and Mg. It is neces- For purely plastic· resistance at the limit condition of an elasto-plastic
the external load and the dam i j orce be equal to the algebraic sum of structure, the period is infinitely long. Consequently the time interval can be-
plied to the structure and exc~p~o~~~=· dand this resisting force is then ap- come considerably longer as plastic action develops in the structure. It is
mains at this particular value r ependence on the damping force re- generally desirable to increase the time interval in accordance with the change
is now necessary to compute the egar~ess of the displacements. However it in the structural rigidity as the structure becomes inelastic. However, it is
the structure in order to find th ~cce erations and velocities of the point ~n not always convenient to compute the shortest period of vibration for a
following way for the points wh e amthping force. This can be done in the structure which goes partly into the plastic range. Consequently it is desir-
ere e mass is zero· able to establish the conditions which govern the choice of time interval on
(1) Assume a value of accelerati . some measurable behavior of the structure which is a natural function of the
is zero. on of the mass point even though the mass
method of calculation. Since in general it is not practicable to consider
(2} Compute the velocity and di modal behavior for structures which are not elastic, it is not desirable nor
damping force. splacement of the point and determine the
convenient to separate the individual modes of action of the structure.
With the use of a value of f3 greater than 1/8, the more convergence of the
sequence of calculations is sufficient to insure stabUity and the rate of
 81
STRUCTURAL DYNAMICS
80 July, 1959 EM3 EM 3 t o~~·te Either moment distri-
. . b pe mitting them o r ..... · t ·
convergence will be an adequate criterion for the time interval. Consequent- (2) Relax the Jomts Y r f alculation can be used to de ermme
bution or some other technique 0 c
ly, one can establish a rate of convergence based on the IRlmber of iterations the final moments in all the members. d
which it is desirable to make in a time interval and then examine the rate of w be computed in all the members, an
convergence of the actual calculations as the calculations progress. For ex- The shears and axial forces c~ no rti al components of the forces on the
ample, if it is desired to have three significant figure accuracy in dis- one can compute the horizont~ an ve . l~ced configUration. These forces,
placement and velocity, three significant figures in the acceleration are also pins which hold the structure m the disp fo . acting on the structure.
desirable. A rate of convergence such that the error is less than one part in reversed in direction, are the resisting r~: to make the calculation for any
a thousand at the end of three cycles would imply a rate of convergence of 10 If the structure is not elastic, it is posSl t d formation or curvature-
per cent or of 0.10. By reference to Table lc it can be seen that a ratio of axial-displacemen e d H -
known load-deformation or f the calculations is uncbange · ow
time interval to natural period of the order of 0.10 for f3 = 1/4 will insure this moment relationship. The technique o . al that one take account properly of
rate of convergence. Similarly, a time interval slightly longer will be ade- ever for nonlinear behavior it is essentll t so as to be sure that the ele-
' . mber or e emen
quate for f3 = 1/6. However, it appears that these time intervals will introduce the direction of loading in a me direction with the appropriate re-
errors of the order of 2 per cent to 3 per cent in the period. If this is not ad- ment is continuing to deform in the same is recovering or unloading with the
missible, then a faster rate of convergence can be established to keep the duced stiffness, and to determine whe~it f the direction of relative deformation
error in the period down as well. appropriate increased stiffness. A te 0
Let us say for example that a time interval consistent with a rate of con- can readily be made to determine thi_s. 5 where masses might be considered
vergence of 10 per cent is desirable and that three cycles of iteration will In a structure such as shown in Fig . .' usually neither convenient nor
normally be considered convenient so as to bring the results to an error of placed at all the joints of the structure, l~S t·on in which the masses can be
less than one part in a thousand at the end of the third cycle. Then a fourth desirable to consider the most general :U a \he reason for this is that the
displaced horizontally as well as vertic h y.izOntal direction is very short
cycle will verify that this is in fact the case. Under these conditions, then,
we can assume a time interval and run through the calculations several times. period of vibration of the masses in th~ ~r vertical direction. If one were
If the desired rate of convergence is not obtained, we have obviously assumed compared with the period of vibration m ed one would have to use an ex-
too long a time interval and we can shorten the length. Since the rate of con- to solve the problem with the general proce ~~e calculation tedious.
vergence is generally a function of the square of the time interval, the next tremely short time interval which w~~dt ~nsider a restricted type of defor-
estimate can usually be decided on from the course of the preceding calcu- In such cases it is usually desira e o be f degrees of freedom. This
lations. If the time interval is made smaller in accordance with the estimate mation of the structure to reduce the n~ther s~ructure by means of vertical
of convergence, and if convergence is now in fact obtained with about three to can be done by connecting the masses that nly vertical forces can act on the
four cycles of iteration, then this time interval can be used in subsequent links as shown in Fig. 5, and speci~~ v~ only vertically. The links are
steps. If however the time interval chosen leads to convergence in only two masses which in turn are considere o m~ long enough so that the hori-
or three cycles, we have probably taken too short a time interval and in the considered to remain vertic_al_ by r;;akin;.
zontal motions of the truss JOmts o no m r
te:.uce an angle into the link be-
next cycle in the calculations we may choose to use a longer time interval.
With this type of procedure, and with continual reexamination of the time tween the truss joint and the mas~. th ber of degrees of freedom with .a
interval in terms of the rate of convergence, one can take account of the In the structure shown in Fig. e nu~ the system shown only 5 degrees
change in the characteristics of the structure without loss in accuracy and ma ss at each node would be 20. However ~r th alculation and the number
th difficulty m e c in
without supplementary calculations. The procedure described above can of freedom are needed. Both e d tion of motion will be greatly -
readily be programmed for a high-speed digital computer. However, an upper of repetitions of calculations for a given ~~~e is used. However' some diffi-
limit on the time interval is usually desirable to avoid difficulty in terms of creased if the 20 degree-of-freedom struc i Fig 5 is encountered because
accounting appropriately for the variations in the applied loading. culty in computing the resistance functions ~ ~s in length of all the
it is not possible directly to determine the tc f :splacement of the lower chord
members from only the vertical componen o
Calculation of Resistance
of the truss. . . to roceed in a slightly different fashion
In cases of this sort, lt is posslble p . f the influence matrix for the
For the general case of a framed structure with either flexural or axial by determimng . elther
. di rectly, or by invers1ond othe given set of vertic al dis _
resistance, and with deformation in all of the members, it is a relatively structure, the set of forces required to pro uce arize the calculations as
simple matter to perform the calculation for the resisting forces if the dis- placements. In the general case, one can summ
placements are given. All that is required is to take the set of displacements
at a given time instant, pull the joints of the structure into the corresponding follows: f th . masses for individual
. . 1 ents of each o e
configuration, compute the axial forces in the members from the changes in (1) Compute the vertlcal dlSP acem h f the masses in turn. The set of
length between the nodes, and compute the moments in the members by a unit values of vertical force at e~ ~nfluence matriX for deflection of
process which involves two stages as follows: values obtained is designated as e
the structure.
(1) Consider the joints locked against rotation and determine the fixed end
moments in all of the members corresponding to the deformations.
 82
July, 1959
EM3 EM:3 STRUCTURAL DYNAMICS 83
(2) From these influence coefficients, compute the set of forces on the
structure for given values of the displacements. The calculation in-
volves solving a set of simultaneous equations for each set of values of
displacement. This can be systematized by '"inverting" the infiuence
matrix to obtain the stiffness matrix for the structure, which represents
the forces acting on each of the masses for a unit displacement of each
of the masses in turn. If this inversion is performed once, the matrix
Fig. 5. Structure with Reduced of values can be applied to any set of displacements to determine the re-
Number of quired resisting forces.
Degrees of Freedom (3) The coefficients in the influence matrix will change when any elements
in the structure become plastic. Consequently the calculation of the
influence matrix and the inversion of the matrix to obtain the stiffness
B matrix must be performed for each change in stiffness that occurs dur-
ing the history of the structure. It may be desirable to make the calcu-
lations by a •relaxation" procedure for more complex structures of this
type. However, in principle the structure can always be analyzed to find
the forces consistent with a given set of deformations.
F It might be pointed out that for a beam of the same span as in Fig. 5, with
a series of masses acting on it, it is not necessary to use the influence matrix
and. invert it to find the stiffness matrix. The structure can be analyzed di-
rectly by pulling the joints into the deflected configuration, locking them
against rotation, and then distributing the fixed end moments corresponding to
these deformations to find the shears in the members and the consequent
c A E
forces acting on the masses. Although in such a case there would be theoreti-
D cally 10 degrees of freedom, 5 of the degrees of freedom are associated with
horizontal axial deformations in the beam, and these are not •coupled" with
(a) Frame with Plastic the vertical deformations because of the nature of the system. This obser-
Hinges vation is true only for a straight beam. In a later discussion, the procedure
applicable to an arched beam will be described.
The structure in Fig. 5 is one of several special cases which require a
slightly modified form of treatment. Another case concerns any structure
with constraints on the deformation of some of the members. Such a structure
might be a polygonal arch which is considered to deform only in flexure and
which is not considered to have any axial deformation of the arched members.
In this case also the number of degrees of freedom are reduced, but the
treatment must be somewhat different from that for the truss because the
number of constraints may reduce the number of degrees of freedom even be-
low the number corresponding to a mass at each node prescribed to move in
a particular direction. Further discussion of this topic is contained in the
last chapter of this paper.
Both Figs. 4 and 5 show structures with zero masses at some of the nodes.
Kink" in
11
A general procedure for such structures can be discussed. In general, one
member at can handle the problem in the following way when damping forces are not
present. The modification for damping is readily made as has been discussed
joint previously.
(b) Detai I of Joint (1) Assume values of acceleration only at the joints where masses exist
Fig. 6. and in the direction in which the masses can move.
Elasto- Plastic Frame (2) Compute the displacements of the masses, but not the displacements of
the joints not having masses.
(3) Let the joints without masses be free to deflect but apply to these joints
the external forces which are applied at these points. If these external
 85
EM3 STRUCTURAL DYNAMICS
84 July, 1959 EM3
an interior load applied dynamically. Then we will in effect have a prescribed
forces are zero, no restraints and no loads are applied at the massless displacement of joint E which can be considered to be an external hinged sup-
nodes. port as shown. Consequently, our method of treatment is general and can deal
(4) We no:w have a ~ystem with prescribed deflections at some points only with internal loads as well as loads applied at the joints.
and Wlth prescr1bed loads at other points. It may be pointed out that at At some time. tn the structure has been analyzed with the result that we
some particular joint for example there may be a prescribed vertical have determined the plastic hinges which exist at that time and, as a conse-
displacement and a prescribed zero horizontal load, if we consider that quence, we have determined the velocities, displacements, and resisting forces
only vertical displacement of the mass is to be considered at that point. in the structure. We can also determine from the final configuration and
(5) We now require the calculation of the resisting forces at the masses in moments in the structure at that time the rotations of the joints (J j and the ro-
the direction of motion. In general this can be done by an influence tations at the plastic hinges ea. If we have analyzed the structure by moment
coe~ient procedure in which we apply unit forces at the masses which distribution, the fixed end moments can be computed and distributed and from
are displaced, one at a time, and compute the displacements at these a comparison of the fixed end moments and final moments the rotations at
~sses. We use these influence coefficients to write a set of equations each end of each member can be determined. The rotation of a joint is the
which in effect says that the product of each unknown reaction force at same as the rotation of the ends of all members meeting at that joint which do
each mass multiplied by the influence coefficient for deflection at each not have plastic hinges. The '"kink" or relative angle between a member hav-
of the masses, and summed for each of the masses individually leads ing a plastic hinge at the joint and the angle of the joint is measured by the
to a deflection at each mass equal to the prescribed deflection ~us difference between 8a and 8j_. The kink angle is of importance in determining
the. deflection in that direction at that mass due to the prescribed forces whether the member is loading or unloading at that joint.
acting at the nodes where no masses exist. n is considerably simpler to deal with the structure in each interval by
(6) The solution of this set of equations yields the desired results. We may making calculations of the change in the structure. In other words, one works
now have to complete the calculation by determining the displacements with the change in displacements during the interval as defining the fixed end
at the massless nodes of the structure if these are required for any moments at the end of the interval and the final moments so computed are the
other purpose. In most instances they will not be needed and need not changes in moment to be added to the original moments in the various
be dete~mined. Where damping forces exist, however, we will need members. In this process one can deal with the elasto-plastic hinges as actu-
these displacements so that we can determine the velocities and acceler- al hinges if the structure is loading during the interval and the hinges will be
ations at these points in order to check on the values of the damping removed if the structure is unloading during the interval.
forces dependent on these quantities. Three situations require consideration:
(1) If the total moment, including the original moment plus the increment
Non-Elastic Behavior in moment during the interval, at the end of any member exceeds the
plastic hinge moment, then a hinge must be placed in that me~er.
General Comments (2) If a hinge exists in a member and the increment in kink angle m the
~or trusses the problem of inelastic behavior of the members is a relative- member at .the joint at the end of the interval is in the same direction
~y s1mple ~d straightforward one. One need take into account only the change as the preceding total kink, then the hinge remains.
m deformation of the members in each time interval to determine whether the (3) If the kink angle in what was originallY a plastic hinge should decrease,
plastic action is continuing or the member is unloading elastically H :ving then the plastic hinge must be removed.
knowledge of the load-deformation characteristics of the member ·theare- The structure is modified by adding, keeping, or removing the plastic
quired ~rce in each member can be determined and the rest~ forces or binges originally assumed for the calculation at the end of the interval,-~ ac-
the res1sting forces in the structure are directly determinable. Although the cord with the above criteria, and the calculation repeated until the conditions
ge~eral concept for framed members is the same, there are some difficulties
which require consideration. obtained are consistent.
Only in unusual cases can there be plastic hinges in all members meeting
at a joint. Such a case is shown at Din Fig. 6.. In such an instance the joint
Elasto-Plastic Frames rotation can be assumed to remain at its value at the time the last members
The method for dealing with elasto-plastic behavior in frames is described at the joint become plastic. (It is clear that in such a joint at least two
here. .Consider a general framework consisting of members acting in flexure members must become plastic finally, and at least two members must simul-
as dsho;n in Fig. 6a. At some stage in the deformation, the moment at the taneously "unload" and have the hinges removed.)
~n s 0 the members shown as black dots in the figure has reached the plastic The only other special case that needs consideration is that shown by the
member F in which both ends develop plastic hinges. Such a member can be
edmtoit thmomplaent.. O~e ?r several of the members meeting at a joint may be load- considered to have end rotations the same as those which existed prior to the
e stic l1m1t moment.
dn·
i lS ~neral enough to consider members which are not loaded internally development of the last hinge in the member.
;:int: ·
wInch ~~ ~oments arise from a prescribed set of displacements of the
a Sl a 10n such as that shown at E in the figure we might have had
In some cases it may be desirable to study the formation of plastic hinges
in a structure as the loading on the structure is applied, more or less
 87
EMS STRUCTURAL DYNAMICS
86 July, 1959 EMS
'"statically", or applied and released in various ways. This can be done very
simply by specifying a relatively slow rate of application of the loading, corre-
sponding to a maximum being reached in not less than twice the natural period
of vibration in the fundamental mode. It is usually convenient to introduce
damping so as to avoid oscillatory motion during or following the loading
cycle. The masses can be taken arbitrarily small in such a calculation if we
are not concerned with the dynamic response, so as to make the fundamental
period relatively short.
F=My _ 1----+t-t.:.;.;M_ _1 X-
Time-Dependent Boundary Conditions
P= P-F
In general if the boundaries or supports of a structure move with time no
change is required in the method of calculation and the general procedures de-
u =x-y
scribed herein are directly applicable. However, there are instances where
it is desirable, or convenient, to deal with the motion of the elements of the
structure relative to the base or foundation of the structure, rather than in
absolute terms. Such situations arise commonly in earthquake motions. The y
method described herein .is useful primarily in those cases where the foun-
dation moves as a single unit and where motion of the foundation of the
Structure with Time- Dependent
structure with all masses equal to zero would not introduce stresses in the Fig. 7.
structural framework. Boundary Conditions
Consider the structure shown in Fig. 7 where the base can move as a unit
in the horizontal direction. Let the motion of the base be defined as a function
of time by the quantity y. The motions of any of the masses in the structure.
are defined by x. Let us assume that the axial deformations of the members
~~--~c~)---TG~)--~~
can be neglected and that we have only flexural deformations to consider. In
such cases, we can lump the masses at each floor level at one point as shown. (a) Structure with Weak or Zero
The more general case offers no difficulties, however. Between Sets of Modes
If now we apply to the structure with masses on it a force F, in accordance Coup I' no
with the relation
F =My (27)
then the structure and the foundation will move as a unit with no relative dis-
placement between the two.
If now we apply to the structure considered supported on a fixed foundation
a force P, defined by the relation
(28)
we have changed the problem from one in which there are external forces and
a time-dependent boundary motion into one in which we have external forces
and no boundary motion. The displacements for the modified forces P will be Modes
designated by u, where (b) Structure with All
U =X - y (29) Coupled
it can be seen, of course, that the total motions of the masses, x, are the sum
of
of the relative motions u plus the foundation motions y, and the total loads ap-
Fig. 8. Examples of Different Types
plied are the external forces P. Consequently, the procedure is valid.
For more complex systems, one must modify this procedure by taking into Coupling Between Modes
account the displacements at each of the several points of support, and de fin-
ing the quantity y for the motion of the masses as being the motion at that
point on a massless structure consistent with the foundation motions. In the
 STRUCTURAL DYNAMICS 89
88 July, 1959 EM3 EM3
case of a beam, for example, simply supported at one end and hinged at the
other where the hinged end is constrained to move vertically in accordance
with some time-dependent condition, the displacements of the masses along
the length of the beam are proportional to the end displacement and are equal
to that end displacement multiplied by the ratio of the distance from the simple
support to the mass divided by the length of the beam. The same method of
analysis applies, and the modified forces will give the relative displacements
of the beam, relative to boundaries fixed in position.
In this case, and in other cases where no stresses are introduced in the
weightless beam by the boundary motions, the entire solution is straight-
forward and all of the stresses can be obtained from the modified system.
However, in more complex cases where the boundary motions may .introduce (a) Arch Considered
stresses in the massless system, the general principles are applicable but the
total stresses must then be determined by adding the stresses in the massless
system with boundary motions to the stresses in the system with fixed bounda-
C
t~y
-Sx
ries and with modified forces. In such instances, the problem of plastic be-
havior introduces complexities which may be unwieldy and it is not desirable
to work with the replacement system. In all cases it is possible to work with
e~&v ~x
-w=-&h
&v
the original system directly with the general methods that have been de- 8h
scribed.
Constraint Relation for Zero
{b)
Curved Frames and Arches Extension of BC
General Procedure Fig. 9 . Extensible and lnextensible Arched
Consider the two structures shown in Fig. 8. In Fig. Sa, a beam is shown
with two masses. These masses may move vertically with consequent flexure Structures
of the beam or horizontally with consequent axial deformation. There is no
coupling be~een these sets of motions. There are two degrees of freedom in
each set or type of motion and four degrees of freedom altogether. The verti-
cal motions can be dealt with as if in a two-degree-of-freedom system and the
axial motions can be dealt with also as in a two-degree-of-freedom system
with neither being considered to· have influence on the other. However, in the
structure shown in Fig. Sb, neither pure vertical nor pure horizontal motion
of each mass is possible. In general, all four modes of motion or all four de-
grees of freedom are coupled, and the structure requires treatment as if it
were a four-degree-of-freedom system.
However, in the structure shown in Fig. Sb, if it is desired to avoid axial
deformation of the bars, only one mode of motion is possible, with the mass
to the left moving down to the right and the mass to the right moving up and to
the left. The system has only one degree of freedom, and can be analyzed as
such although the method of analysis requires some care in order to avoid
introducing motions that are inadmissible.
Loading Producing No
In.extensible Members-Constraint Relations Fig. 10. Symmetrical
Arch
Consider the structure shown in Fig. 9a. Here there are three interior Displacement in an lnextenslble
masses and if the members are considered inextensible there will be two de-
grees of freedom. We arrive at this number by considering the total number
of degrees of freedom of the system if the members were capable of defor-
mation, namely 6, and subtracting from this number the number of constraint
coDditions or in this case the number of inextensible members, namely 4.
 90 July, 1959 EM3
~or any one of these members, such as BC, the condition that the defor- EM3 STRUCTURAL DYNAMICS 91
mations at the joints correspond to no change in length of member BC can be symmetrical loading producing no displacement. Consequently, one can arrive
?xpressed in terms of the relative displacement of joint c compared with at a loading consisting of a downward load at C and equal horizontal loads with
Joint Band the slope of the member BC. For small deformations where 6x
c
and 6 Y represent the horizontal and the vertical motion of point with respect
to po~nt B, and where oV and OH represent respectively the vertical differ-
no upward load at joints B and D. There are two other independent loadings
which are anti-symmetrical. From combinations of the four loadings, various
types of loading patterns can be derived.
ence m elevation between C and B and the horizontal distance between c and The advantage of having the constraint conditions stated in this way is that
B, the relationship that the relative deformation involve no change in length now we can add any of these types of loadings to any set of resistances or of
can be expressed as follows: external loadings so as to preserve the number of degrees of freedom of the
6x 6V
structure at the proper number, in this particular case two. In other words,
6y = - 5H (30) after the two independent displacements are decided on such as the horizontal
and vertical displacement at joint Bin Fig. 9, and the corresponding dis-
This equation can be interpreted as meaning that the deformation is such that placements at C and at D are determined consistent with these, then both the
point C moves perpendicularly to member BC. external forces and the resistance can be modified by adding force patterns
A similar c_onstraint relation can be written for each of the bars. If the similar to those in Fig. 10, (plus the two anti-symmetrical loadings), so as to
total deformations are expressed in terms of the increments in the defor- make the accelerations in the various directions of all of the masses con-
~tion, one can find four relations among the three horizontal and three verti- sistent.
al displacements of point B, C, and D. These will be somewhat more com- Consistency in the vertical displacements during the interval requires con-
ple7tbut the relations can be written in any case. The number of independent sistency in the vertical accelerations at the end of the interval. Since the dis-
r a ions ~hich it is possible to write in this case will be equal to the number placements at the beginning of the interval have been made consistent, and
of constrru.nt_ conditions, namely four. With four relations among six dis- since the velocities and accelerations at the beginning of the interval are con-
placements, 1t appears that there are only two independent displacements re sistent, we need only insure that the accelerations at the end of the interval
maining. These can be chosen as any two components of displacement and - are properly consistent. In order for these to be properly consistent, they

a::
the calculation_s carried through in the usual way, but with only two rru:.sses
two directions of motion considered. By the use of the constraint relation-
~ P~, when the. two deflections which are independent are assumed the re-
must obey the same relationship that the displacements obey. In other words,
we have the same relationships among the accelerations in the various di-
rections at the joints or nodes as we have constraint relations corresponding
ainiting deflections can be readily computed, and the entire pattern of defor- to zero extension of the members. It is a simple matter to adjust the resist-
ma on of the structure established. ing forces by adding the proper components of loadings so as to make the net
th However, it is possible to develop a more direct procedure for determining acceleration at all of the masses consistent.
T~ relati~~ps for th: resisting force components and for the displacements. As a by-product of the results shown in Fig. 10, one can apply the principle
st s can one by noting that there are certain patterns of loading on the of virtual work to find relationships among the displacements at the various
. ructure which produce only axial stresses in the members. For these load- joints. For any pattern of loading which produces no displacement, the
Ing components _no deformations of the arch, can take place. It is obvious that product of the loadings in the pattern times the displacements in any set of
o;e type of loading would correspond to axial forces dir.ected along the length consistent displacements of the structure in the direction of the loadings in
~al~~done ~f the members. This can be expressed also in terms of the verti- the pattern must be zero. This follows from the principle of virtual work di-
res
ultanho~1Zontal components at the two ends of the member such that the
t hes along the length of the member
rectly. In the case of an arch the vertices of which lie in a parabola, as in
Fig. 9, the loadings at the three vertices, as shown in the upper part of
. By considering four component loadings c~rresponding to loadings produc- Fig. 10, will be equal. This implies that the sum of the vertical deflections at
Ing axial force in only one membe t t· · B, C, and D, taken as positive downward, must be zero. A similar relation-
load condition hi h d r. a a lllle, one arnves at four independent
can be combins w c pro uce no displacement in the arch in Fig. 9. These ship can be arrived at from the loadings in the lower part of Fig. 10 or from
tter ed in _various ways. One obvious combination consists in a each of the load conditions considered.
k re::.s~!~~y ve~lc:J forces at the interior nodes. These of course would
enter into th: ·t:ld
~et c horizontal reactions at the ends but these do not
in the u er plc ure '!- e ends do not move. This pattern of loading is shown
Reduction in Number of Effective Degrees of Freedom
The operations above suggest a modification for the approximate dynamic
F" 10 PP part of Fig. 10. A second pattern of loading in the lower part of
0
o:;r m~::!;!s ~f: e tw~ se~s~f loadings which produce tension only in the
analysis of an arch which is inextensible. Consider the arch shown in Fig. 11
which is the same as in Fig. 9 except that the masses are carried by vertical
B and D are made eq~~rcth th ~ structure is symmetrical, and if the loads links, which remain vertical, in the same manner as in the truss in Fig. 5.
loading pattern in the ' . en e oading pattern will be symmetrical. The Here now we constrain the masses to move only in the vertical dir.ection.
symmetrical structur:pp;~ part;f ~g. 10 will always be symmetrical for a There are then only three degrees of freedom for the structure. If the
types of s mmetric . . ese o oadings represent the two independent
arch of ~ 9 Of alloadings producing no displacement in the inextensible
structure were extensible, we have reduced the number of degrees of freedom
• · course any combination of these two loadings will also be a from 6 to 3, but we have lost the horizontal motion of the masses. If we know
that these are unimportant or feel that they might be, or if we wish to permit
92
July, 1959 EM3
EMS STRUCTURAL DYNAMICS 93
this approximation in order to find a quicker solution to the problem, as we
might in the case of a very flat arch, we can work with this problem in the
same fashion that we solved the problem in Fig. 5 earlier. We need now find

I~
only the influence matrix and the corresponding stiffness matrix in order to
find the solution.
If, however, the arch is ine:xt:ensible, there are really only two degrees of
freedom of the structure and we have a system which apparently has three.
However, by use of the fact that for the arch shown a set of loadings consisting
of equal vertical loads at each of the three joints will produce no deformation,
we arrive at the condition that the sum of the three vertical displacements
must be zero. Then we could use this as a constraint condition which will
permit us to work with only two independent vertical displacements. We can
define the third ve.rtical displacement in terms of the other two, and arrive at
a consistent set of forces by adding to any set that we compute at the two inde-
Fig. II. pendent joints a proper combination of three equal vertical loads so as to
Scheme For Approximate make the vertical accelerations at the three points consistent. Consistency
in the vertical accelerations means consistency in the increments in de-
Dynamic Analysis of an Arch flections, which means that the vertical accelerations must also obey the rule
that the sum of the vertical accelerations at points 1, 2, and 3 as shown in
Fig. 11 must be zero.
It can be seen that in general for an arch with N nodes, the number of de-
grees of freedom for an extensible arch will be 2N, and for an inextensible
arch, N-1. The number of degrees of freedom for vertical motion at the nodes
will also be N-1.

Extensible and Inextensible Rings

If the ends of the arch are free to move, additional degrees of freedom are
introduced. These represent no real difficulty and can be considered either
for the extensible or ine:xt:ensible case in a relatively simple manner. How-
ever, more complexities are introduced in the case of a complete ring.
Consider, for example, the hexagonal ring in Fig. 12. There are six mass-
es at the nodes of the ring. For other shapes or number of sides, the re-
lationships can readily be determined. In this particular case, it is clear that
the number of degrees of freedom for an extensible ring is 12. However three
of these are the rigid body motions for an object in a plane and produce no in-
ternal stresses. In the general extensible case of deformation, with two de-
grees of freedom for each mass, one would compute ordinarily 12 dis-
placements, one in each of the two directions at each of the 6 masses. The
statically required forces needed for consistent accelerations will automatical-
ly be achieved, and the system can be treated as if it had in fact 12 degrees of
freedom.
In the case of the inextensible ring, there are in the case of the hexagon
six constraint relations corresponding to the six bars. This means that there
are essentially only 3 degrees of freedom for the system rather than 9 for the
inextensible case. Here again the 3 rigid body motions correspond to no
stress in the system. However the system can be treated as if it had in fact
Fig. 12. six degrees of freedom since the rigid body motions will automatically be
Hexagonal Ring achieved. The six constraint relations must be used to insure consistent
values of acceleration and displacement.
 July, 1959 EM3
94 July, 1959 EM3 2095
Miscellaneous Problems

Space Frames and Space Structures Journal of the

Space frameworks or space structures of a more complex character than ENGINEERING MECHANICS DIVISION
those generally described herein can be handled by the same general princi-
ples. For a structure in space, there are in general 3 degrees of freedom for Proceedings of the American Society of Civil Engineers
each mass at a node, and consequently 3 components of external force andre-
sisting force which must be considered. This introduces complication only
in the order of magnitude of the calculations which are involved.
ULTIMATE STRENGTH CRITERIA FOR REINFORCED CONCRETE
Buckling Ladislav B. Kriz 1
For structures in which axial loading produces secondary effects because
of the defiections of the parts of the structure, thereby increasing the
moments and the defiections for a given loading, the calculations are more
complex. In general, the resisting forces for a given amount of presc:x:ibed
SYNOPSIS
deflection are reduced, in about the same ratio that the deflections for a given
loading are increased when buckling loads are present. This reduction in re- h f t t al members are derived by de-
sisting force means generally a greater acceleration and a consequent greater Criteria for ultimate strengt of ~....rue uercompression edge strain which re-
dynamic displacement in a structure in which buckling tendencies exist. These . . alytically the value o e..... rem 1
term1n1ng an 1 d The following rectangu ar
tendencies can be taken into account but the methods for doing so become rela- suits in a maximum value of moment or oa . s (b) reinforced concrete
tively unwieldy. One can compute the resisting forces on the assumption that members are consider~d: (a) hodmdoge~:~:c:~a:n~rete columns. In all three
the buckling tendencies are negligible, and then for these resisting forces de- beams and (c) eccentncally loa e rel t I'th tests and is a
' d · d · ·n agreemen w '
termine the relative defiections corresponding to the secondary effects of the )
cases, ultimate strength so enve lS 1 ression edge and properties of
forces produced by the defiection of the structure. Then by reducing the re- function of only the stress at the extreme camp
sisting forces in a more or less arbitrary manner, one can arrive at a set of the cross-section involved.
resistances which would account for the prescribed deflections. Although this
method is not entirely satisfactory, it does give a means of taking into account
approximately the buckling tendencies for complex structures. For simple INTRODUCTION
structures, the problems can sometimes be handled directly. The real diffi-
culty in the case of buckling problems is that buckling is a nonlinear effect b is commonly determined on the
and the methods of calculation used to handle the calculations in each interval Ultimate strength of structural mem. ers ain plane during bending and
in the process depend on linearity, within that interval at least. basis of the assumptions that pl~ne sectlO~S~e~e range of linear stress-
that stress is a function of stra:n o~y .. Wl:hr~u hout a member is determined
strain relationship the st~e.ss .dlstnbution s and ~f moments together with the.
REFERENCES from the equations of e~uill?n~m of fore~ If the stress-strain relationship
requirement of linear distnbution ?f stram~ the parabolic, trapezoidal or
1. N. M. Newmark, "Computation of Dynamic Structural Response in the is non-linear, simplifying assumptions, sue nlas. traduced or else the ultimate
Range Approaching Failure," Proceedings of the Symposium on Earthquake rectangular stress distribution, a:e comm? . y ; from t~st data.
and Blast Effects on Structures, Los Angeles, 1952, Published by Earth- strength design equations are derlved ~mf;.nc te ~trength may be studied ana-
quake Engineering Research Institute. By another approach, th~ problem o ~ ~m;ain load functions expressed in
2. S. Levy and W. D. Kroll, "Errors Introduced by Step-by-Step Integration
of Dynamic Response," National Bureau of Standards Report, February I lytic ally by finding the m~un.mum value o ce loaded member. This can be d~ne
terms of the internal reslstlng forces of th t . relationship of the inelastic
without defining mathematic~ly ~e str~:-s ~:1:trength may be derived which
l'
1951.
material. In this manner, cntena for Ima
3. L. Fox and E. T. Goodwin, "Some New Methods for the Numerical Inte- I To extend the closing date one mon~,
gration of Ordinary Differential Equations," Proc. Camb. Philos. Soc., Note: Discussion open until ~cember 1, 195:~utive Secretary, ASCE. Paper 209? lS
Vol. 45, pp." 373-388, 1949. a written request must be filed with the E . g Mechanics Division, Proceedmgs
part of the copyrighted Journ:U. of th~ Engine~~~ 85 No. EM 3, July, 1959.
4. S. Timoshenko, "Vibration Problems in Engineering," D. Van Nostrand of the American Society of Clvll Engmeers, 1. ' nt Section Research and
Co., New York, first edition, 1928, pp. 79-81. (and in later editions, see St t ral Deve opme '
1. Asst. Development Engr., rue u t Assn Chicago, lll.
for example 4th edition, 1954, pp. 143-145.) Development Div., Portland Cemen ''
95