You are on page 1of 9

A MCNEIL

~ONL~EAR
ANALYSIS OF CONCRETE
STRUCTURES AND C~MPA~SON
WITH FRAME TESTS?
BILLL. GUNNIN
Ellisor and Tanner, Inc., Consulting Engineers, 800 Expressway Tower, Dallas, TX 75206,U.S.A.

FRANZN.RAD
dent

of Each

and Applied Science, Portland State U~versi~, Portland, OR 97207,U.S.A.


and

Rrcm
Department of Civil Engineer&The

W. FURLONG

University of Texas at Austin, Austin, TX 78712,U.S.A.

Ahstraet-A computational technique is described for the general nonlinear analysis of large planar frames under

static loadiag.The analysis accommodatesboth material and geometricsour~~sof nonlinearity in a highly stable
numericalprocedure.For materialbehaviorthe moment-thrust-curvaturerelationshipsare reducedto polyaomials
made to representcommonstructuralmembers.The nonlinear influenceof axial shorter&gand P -A effects from
displaced joints are accommodated by the analysis. Beam-column effects that magnify moments by the inkluenceof
axial force acting through deformed members can be ap~o~~~
by introducing nodat points (or joints) at points of
maximum d~o~tion
between the ends of members.
Accuracy of the analytic procedure is demonstrated by comparing computed results with eight reinforced
concrete frames tested at the University of Texas at Austin.
NOMENCLMURE

y deformed shape of a beam


A disoiacement of a nodal paint
q endrotation
p radius of curvature
* delined as (III, /I,)/(EI&_.)

Crosssectional area
total area of steel iu a cross section
wsepxi
parameter
width (breadth) of a concrete cross section. Also a
parameter as used in Ramberg-Osgood polynomials
di.&nce from the compression face of a section to the
centroid of the tension steel
modulus of elasticity
modulus of elasticity of concrete
modulus of elasticity of stee1
eccentricity of column load
flexural sti@ness
flexural stigness of column
flexural stifkss of beam
standard cylinder strength of concrete
yield strength of steel
thickness of a concrete cross section
lateral force
moment of inertia
beam length, center to center of columns
colunm length, center to center of beams
clear height of column
bending moment
yield ~ment capacity
RambereOsgood parameter
axial load capacity of a short column in absence of
moment
applied axial load on column
applied axial load on leeward and windward columns
axial elongation
projection of L on the x-axis
length along a beam
projection of L on the y-axis

tPresented at the Second National Symposium on Computerk.d Structural Analysis and Design at the School of Engineering
and AppIied Science, George Was~on
University, Was~n~on
D.C., 29-31 March 1976.

Vectors
member displacements in local coordinates
initial strains
equivalent strains (in constant strain method)
member forces in local coordinates
reduced forces (in constant strain method)
total loads
applied loads
loads due to iuitial strains
restraint loads
x-coordinates of joints
y-coordinates of joints
displacements in global coordinates
incremental displacements
forces in global coordinates
Matrices

[K] composite stiflness matrix for the frame


[-r7_]composite stiffness matrix modified for boundary canditions
[S] member stiffness matrix
[Tl displacement transformation matrix
[TT] force transformation matrix (transpose of [Tj)
[T&] displacement transformation matrix for the member in a
displaced position
~ODU~ON

ultimate load analysis of strueM frames requires


the consideration of several sources of nonlinearity in the
load-displacement or load-force behavior of the struc-

The

ture. One source of nonlinearity


is the nonlinear
force-deformation relationship of the members. A second
source is the displacements of the joints during loading. A
third source of no~~ty
is the late& defo~ti~
of

the members, that is, the ~~~


251

e&t.

258

B. L. GUNNIN
et al.

Described in this study is a general method of nonlinear


analysis of planar frames under static loads. The word
general is used to describe several aspects of the
analysis. First, the analysis is not restricted to frames with
rectangular bays, i.e. frames consisting of vertical
columns and horizontal beams, but the method is
applicable to any configuration of a planar frame. All
displacements of the joints are considered. Not only are
joint displacements such as P-A effects accounted for,
but also the effect of axial shortening is accommodated.
Secondly, the word general suggests that the nonlinear
material force-deformation relationship does not restrict
the analysis to one type of material. This study
concentrates on the analysis of reinforced concrete
structures, but the method has been used for the analysis
of steel frames[l] and it could be extended to other
materials. Thirdly, the word general suggests that large
frames as well as small can be analyzed by the proposed
method. The capacity to handle large complex frames
effectively prohibits dividing up individual members into
many submembers, and thus precludes for member
behavior the use of finite element analysis, discrete
element analysis or the finite difference method. However, finite element analyses can be used as a source for
the general form of moment-curvature relationships used
in the basic analytic technique.
For the analytical method used in this investigation, the
moment-thrust-curvature relationships are reduced to
mathematical expressions which adequately represent
common structural members. The expressions are then
integrated mathematically to obtain end rotations, thereby
reflecting the spread of zones of yielding. The expressions
are differentiated mathematically in order to solve the
nonlinear moment-rotation equations for each member.
The moment capacity as well as the stiffness of a
reinforced concrete member is changed by the presence
of axial forces.
Deformations of the members are accommodated
approximately by dividing the members into a small
number of submembers. This subdivision is not to be
confused with the subdivision of members into many
elements required for finite element or discrete element
analysis. Subdivision into smaller elements is necessary
only for members where magnification of bending
moments by axial loads acting through deformed members causes zones of yielding between the ends of a
member.
The description of nonlinear moment-rotation characteristics by a mathematical expression requires that only a
few parameters be stored for each member and also
permits most members to be treated as single elements.
Thus the efficiency of the proposed method, with respect
to both computer storage requirements and computational
time, makes possible the analysis of large structures.
The computational scheme assumes that the material is
nonlinear but elastic, so that unloading of a member
follows the same path as loading. The method also
assumes that the response of the frame is independent of
time, but the method could be modiied to account for
member unloading and for creep.

displacements. The application of matrix methods to the


analysis of linear elastic frames will be described first in
order to define the notation, which is taken from
Wilson[2].
Matrix analysisof linear elasticframes
For any beam AB, as shown in Fig. 1, the end forces F
are related to the end displacements D by the linearly
elastic stiffness matrix [S]:
_
AE
P
-i_o
0
u
MA

zz

M,,

gI
L
gJ
L

El
L
gl
L.

IDa

i QH

or
F = [S]D.

(1)

Fig. 1. Forcesand deformationsof a beam element.


The displacement transformation matrix [ ZJ is constant
for small displacements. It relates member displacements
D in local coordinates to joint displacements A in a global
coordinate system (see Fig. 2):

X
LYO
i
Lx1 $0,
-iLx0 $1
--

-3

SO

Y
L
Y
-7L
-7

il

or
D = [ T]A.

METHOD OF ANALYSIS

The method of analysis described herein utilizes a


combination of techniques developed by other investigators. The linear elastic stiffness method of matrix
structural analysis is modified to accommodate the effects
of nonlinear material strains and the effect of finite joint

Fig.2. Geometryof a beamelement.

A generalnonlinearanalysisof concretestructures

The transpose of the displacement transformation matrix


[TT] similarly relates global forces cp to member end
forces F:
cp= [TT]F.

(3)

In the displacement (stiffness) method of matrix structural


analysis, the relationship between global displacements
and global end forces for each member is given by the
matrix product:
cp= [TTST]A.

(4)

The members are assembled into the complete structure


by joining all elements at their joints, applying in the
process the requirements of equilibrium and compatibility. This operation consists of adding all member stiffness
terms (TTST) which refer to the same joint, thus forming
the composite stifTness matrix [K]. IJen the composite
stilTness matrix [K] is modified to [K] by including the
e&&s of the boundary conditions. In the matrix analysis
of a linear, elastic structure, the system of simultaneous
linear algebraic equations
@]A = R.,

(5)

is solved for the joint displacements A as a function of the


applied loads R,,. Then the member forces F are obtained
by the matrix multiplication
F = [ST]A.

(6)

Nonlinearity in frame response due to nonlinear material


strains
In order to account for nonlinear material strains in the
analysis of a structure, the load-deformation relationship
for each member must be available, perhaps in the form of
a mathematical expression. A convenient and versatile
mathematical expression for the description of momentcurvature relationships is given by the Ramberg-Osgood
polynomial [3]:

where M, is the moment at some point x along the beam,


(d*y/dx*) is the curvature at that point, EI is the flexural
stiffness and M,., a and n are parameters. Functions of
eqn (7) are displayed in Fig. 3. By varying parameters a, n
and M,, a wide range of load-deformation relationships
may be descrii.
The shape of the moment-curvature
I5

jn=l

259

function approaches the idealized elasto-plastic state as n


and a are increased.
Wilson121 assumed that a structure behaves linearly
under small increments of load and used the RambergOsgood polynomial in an approximation of the chord
stiffness of each member for each small load increment.
For the beam segment shown in Fig. 1,
tMB);-MB.

M, (MA

Q-9

Combining eqns (7) and (8):

g=$ [1+@&I(M,
tMB,;-MBI-]
x

[I

(MAtMB);-MB

II
.

(9)

Integrating eqn (9) and evaluating the constants of


integration

q,=&

2hfzt-MB+

(n + l)(nY2)M:-

- IMBI(~+)MB
(n + l)(MAJ)Mat (n t2)IMAI(+)Ms
(MA+ MB)

(10)
Similar equations for @ are obtained by interchanging
the A and B subscripts.
Note that the use of eqn (10) implies that maximum
moment(s), and thus maximum curvature(s), must occur
at the end(s) of a member.
Wilson also used an approximation to the chord
stiffness derived from the moment-rotation relationship
(eqn 10) in an incremental method to predict frame
response.
Another approach to the treatment of structural
nonlinearity due to material properties is described by
Argyris[4]. He used the concept of initial strains to
describe strains in structural members due to temperature
changes or due to a prior loading. He also used the
concept to describe strains in structural members in
excess of the linear strains for a given load. For the
method described in this paper, the initial strains
actually are member end displacements, not strains. Since
the consideration of inelastic strains is a nonlinear
problem, the operations defined by the equation below
represent the kth step in an iterative scheme.
Argyris defines the initial loads R!!,!,, as a vector of
elastic loads arising at the nodal points of the m th
member if all initial strains D!$) are suppressed:
RI;!,,)= [ TTS]D!:?,.

(11)

The sum of the initial loads for all members at a node


XR@,j gives the joint loads due to initial strains. Then the
vector of total loads Rck at a node is given by
Rck= R(I+xRck
I(m)
V

OO

IO

20

f$-#

Defwmation

Fig. 3. Nondimensionalizedmomenkxrvature diagram.

(12)

where R, is the vector of applied loads. The vector of


total joint displacements A([is obtained by solving the
stiflness equations
(13)

260

B. L. GUNNYet al.

Then the vector of member forces F{$ is given by


F$) = [ST]A- [S]D$A).

(14)

That is, given an estimate of initial strains Dl&, the


member forces F[Z, can be computed. There remains the
problem of obtaiuing a new (and more correct) estimate of
the initial strains for the next cycle of iteration.
Lansing et a[.[51 describe two methods, the constant stress method and the constant strain method,
for computing a new estimate of initial strains if an
estimate of the member forces is known. Lansing
describes the application of these techniques to the
analysis of a pinconnected truss and to 6nite element
idealizations. The same techniques can be used for the
analysis of rigid frames. In the method proposed in this
study, constant stress and constant strain actually
refer to a member force and a member end displacement.
Lansing states that the use of the constant stress method
to estimate new values of initial strain will cause
divergence of the iterative scheme. The constant strain
method is convergent for moment curvature functions
such as the Ramsberg-Osgood polynomials. In the
constant strain method, a vector of equivalent strains
DT{zj is computed, where

such structures, an analysis which includes the effects of


geometry changes is needed.
One method for treating the problem of large displacements is described by Saafan and Brotton[6]. They
assumed that only the joints reflect changes in geometry,
while member deformations remain unchanged. That is,
they provide a splint for the beam-columns. This splint
sustains the axial load while the member itself carries the
shear and bending moment. Their method can be
described as follows:
Consider an unloaded frame with linearly elastic
materials (A = 0). The unloaded frame has the geometry
defined by the nodal coordinate vectors XOand Yo, and
the displacement transformation matrix [To] for each
member is a function of these coordinates. Let the frame
be loaded with the vector of applied loads R., and let the
frame be given an arbitrary vector of joint displacements
Acr).(The analysis of large displacements of structures is
nonlinear, so the superscript (k) represents the kth step in
an interative scheme.) The revised displacement transformation matrix T&(is formed from the joint coordinates
X&, and Yot A;), and the member forces are then
computed using the revised [ TA]:
F = [sTAk)]Ak.

(17)

For an arbitrary pattern of displacements, the forces F


generally will not be in equilibrium with the applied loads
R.. A vector of restraints can be computed from
and where 9P:= (L/6EI)[2M!& - M!&l + the element
of D& representing cphand where a similar expression
for cpl, is obtained by interchanging the A and B
subscripts. Then a vector of reduced member forces
Fe{ is computed from Df$& where
0
Wc,,

F${=
1

(16)

Rr(k)= _ R(I + OF

(18)

where PF represents the summation of forces from all


members at a joint. These restraints represent the forces
required to hold the frame in the displaced shape A) in
equilibrium with the applied loads R. Now let the
restraints be applied as loads on the displaced joints. The
resulting incremental displacements are obtained by
solving the set of equations

MWnl,

@Is = Rik.

(19)

The

evaluation of the reduced forces from the


equivalent strains involves the solution of two simultaneous nonlinear algebraic equations. Newtons Method is
used in the computational scheme described herein to
solve the two nonlinear simultaneous equations for each
member.
As the tinal step of the constant strain method, the new
estimate of the vector of initial strains D$$ is computed
by direct substitution of the elements of Fe{ into the
nonlinear portion of eqn (10).
Argyris initial strain concept and Lansings constant
strain method were used to evaluate the nonlinear
response of a frame for a single value of load. The
computational scheme was then used to predict the load
history of a frame by incrementing the loads to failure and
by using the results of the current and previous load
increments to compute extrapolated estimates of initial
strains for the next load increment.
Nonlinearity in frame response due to geometry changes
III many structures for which the material behavior is

linear, the displacements of the joints during loading are


small enough so that the displacement transformation
matrix [ZJ given by eqn (2) can be assumed constant.
However, if materials in structures exhibit nonlinear
behavior, large strains may cause large joint displacements during loading. For an ultimate strength study of

Then a new, and more correct, estimate of the frame


equilibrium position is given by
A(k+i)= A(_+
g(k).

(3-J)

The

iterative scheme described by eqns (17)-(20) has


converged when the vector tik (or alternatively 8Q is
essentially zero, or when the equilibrium of each joint is
essentially satisfied. For such a condition an equilibrium
position of the frame under the applied loads R,, has been
found.
The iterative scheme described by eqns (17)-(20) could
be used to determine the equilibrium position under a
vector of loads R. When employed for successively
changing load vectors it could be used to determine the
load history of a frame. A good estimate of the frame
geometry A for the next load increment could be obtained
by extrapolating from the values for the current and
previous load increments.
Combining
changes

nonlinear

material

strains

and geometry

For the computational scheme described herein, the


iterative approaches used by Argyris and Lansing are
combined with that of Saafan to form a generalized
nonlinear analysis. For each increment of load, two

261

A generalnonlinearanalysisof concretestructures
iterative schemes, one contained within the other, must be
sat&tied. In the inner iterative scheme, the geometry A is
held constant while the initial strains I)I are determined
for the initially constant geometry. In the outer iterative
loop, the geometry A is iterated until equilibrium of the
joints is achieved.
Note that the solution of the system of eqn (19) for
incremental joint displacements S makes use of stitTuess
equations which neglect material nonlinearity. Thus,
convergence of &fans iterative scheme to the correct
configuration of joint displacements A is slower than
would be the case if the stiffness equations did reflect
nonlinear material strains. However, good estimates of
joint displacements are obtained by parabolic extrapolation from the displacements during the three previous
load increments, so usually no convergence problems are
encountered until instability of the structure is impending.
In summary, the frame analyses described herein are
based on the following assumptions:
(1) The member elements are prismatic, and the joints
are perfectly rigid.
(2) Between nodal points, small distortion theory is
assumed, that is, shear distortions are neglected and the
curvature of a member is assumed to be proportional to
the second derivative of deformation with respect to
length

pared with data from eight experimental tests on braced


and u&raced reinforced concrete frames.
Thrust-moment-curvature relationship
A computer program, orighmlly developed by Breen[7l,
was utilized for the generation of thrust-momentcurvature data. The program assumes the Hognestad
stress-strain curve for concrete in compression[8] aud an
elastic-plastic stress-strain curve for steel in tension and
compression. The tensile strength of concrete is ignored.
Equal areas of steel are assumed to be placed, each at a
single level, symmetrically about the centroid of the
rectangular cross section. For given values of axial load
and strains on the edges of the cross section (a measure of
curvature), au iterative scheme is used to locate the
position of the neutral axis. Then the stresses are
integrated over the cross section to yield a value of
moment. By incrementing edge strains and axial load, a
complete family of thrust-moment-curvature curves can
be generated. Four such curves are shown as point data in
Fig. 4.

1 d2y
-=2
*
( P dx )
(3) All loads are applied at joints within the plane of
the frame, and all displacements of the frame are withiu
that plane.
(4) Failure of a member by local buckliug or lateral
buckling is neglected in the analysis.
(5) The moment-curvature relationship for the members can be described by the Ramberg-Osgood eqn (7).
(6) The yield moment MY used in the RambergOsgood equation can be changed by the presence of axial
load, as reflected by an interaction equation.
(7) The secondary effect of axial load acting through
member deformation caused by primary moments in a
member is neglected.
(8) The change in length of a member is caused only
by axial load; the difference between arc length and chord
length of a member is neglected.
(9) The response of the structure is independent of
time.
(10) There is no reversal of strain.
The increase in moment in a member due to thrust [see
(7) above] can be treated approximately by dividing the
beam-column into two, three or more segments. Nonlinear curvatures concentrate near a hinge or nodal point.
The concentration is greater with higher values of II in the
Ramberg-Osgood equation, that is, as the equation
approaches an elastioplastic relationship. The placing of
an extraneous joint at an anticipated location of a hinge
can reflect most of the secondary moment due to thrusts
acting on deformed members. Extra joints may
likewise be placed iu beams to permit the formation of
beam mechanisms, or to permit the application of
concentrated loads on the beams.
AFPLlCATlONS OF TEE NONLNAR ANALYSIS
To rlElNF0IlcF.D co1ycBETE SrRU~

In this section, the development of thrust-momentcurvature relationships is first discussed. Next, results
obtained from the general nonlinear analysis are com-

Cobmn section-test franc L3


beTcUeIfwse&mpopertles)

ooco5

00010

oca5

owal

I/in.

Cvwhre,

Fii.

P/P0

0
0

0.0
0.3

0.5

0.7

I
o.m5

4. Thrust-moment-curvaturerelationships.
Ramberg-Osgoodparameters
(XL)

2.80
1.27
1.37
1.71

(Xlb)

31.1
50.0
12.2
5.3
5.46
50.0
0.289 50.0

MY

Anax

(k-in)

(X 10-q

38.7
65.2
52.0
32.5

1plg
1.31
0.99

The solid lines in Fig. 4 show a form of the


Ramberg-Osgood polynomial

(21)
where b, c, My and n are functions of axial load. The
four parameters b, c, MY aud n required to define a
moment<urvature relationship for a given axial load are
computed iu the following manner. First, several points
with ordinates MY and abscissas P/PO on an interaction

B. L. GUNNIN
et al.

262

diagram for the section are determined by using the


thrust-moment-curvature
generator described above,
where the term MY denotes the maximum moment for a
specific P/PC, ratio. A third-order curve is fitted by the
least-squares method to these points on the compression
and tension portions of the interaction curve.
Next, the coel%cients b and c in the Ran&erg-Osgood
polynomiual are computed for a given value of MYand for
an assumed value of n by solving the normal equations
for a least-squares curve fit of eqn (21) through the
generated points [(d2y/dx2),Ml. Also computed is the
variance V,, (for the assumed value of n) which is a
measure of the goodness of the fit of the fitted curve.
The variance is thus computed for three assumed values
of n and a second-order curve is fitted by the
least-squares method to the points defined by (n, V.).
Then a better estimate to the value of n which causes the
variance to be minimized is computed by taking the
derivative of the fitted second-order curve. The process is
repeated with the new estimate of R and two of the old
estimates until n is essentially unchanged, and the
variance of the fitted curve (21) is a minimum. Thus,
values of the parameters b, c and n are computed for
several values of PIP,. These values are stored and, for a
given value of P/PO, values of b, c and n are obtained by
linear interpolation between the stored values.
The maximum curvature a section can sustain for a
given axial load is obtained from the thrust-momentcurvature generator. These maximum curvature values
are also stored, and the maximum curvature for a given
axial load is computed by linear interpolation between the
stored values. No limiting rotational capacity was placed
on beams or on columns with a thrust less than the
balanced axial load.
In order to investigate the behavior of long reinforcedconcrete coluIIIlIs, a series of experimental tests on
concrete frames has been conducted at The University of
Texas at Austin. Included in this series of investigations
are tests on rectangular sway frames, tests on rectangular
fcames with columns in symmetrical single curvature,
tests on three-bay frames with columns in symmetrical
single curvature and tests on sway frames under
combined loading. In each frame test, loads, deflections,
comer rotations and surface deformations at various
stations along each member were measured. The surface

deformation data were then reduced to average curvature,


and 6nally to bending moment values. In the following,
results of eight frame tests are compared with the
response predicted by the proposed analysis. For brevity,
the comparison is limited to load-deflection response for
each frame.
The average section properties of frame members are
given in Table 1. In all load-deflection plots, the
experimental data are shown as dotted lines and the
predicted response is indicated by solid lines. In the
analysis, the effective concrete strength was assumed as
0.85 f :.
Frame L3-sway

frame under column and lateral loads

Tests on eight rectanguhu frames which were subjected


to column and lateral loads are reported by Ferguson and
Breen[9]. The result of a test on specimen L3, which had
columns of height ratio I./h of 20, stiffness ratio Y of 2.0
and nominal eccentricity ratio e/h of 0.1, is compared
with the response predicted by the proposed analysis. The
column loads and the horizontal loads were incremented
proportionately to failure of the frame.
A plot of the column load vs lateral deflection of the
frame is shown in Fig. 5. The test frame failed at a column
load of 31.7 kips, and the load predicted by the proposed
analysis is 305kips. The flatness of the load-deflection
and load-moment curves indicates a stability failure.

2H

a4
0

in

IO
20
kkrimntd deflection at tap of cdwms,
m

Fii. 5. Load-deflectionstudy-frameL3.

Table 1. Sectionand materialpropertiesfor reinforcedconcretemembers


PraEQ
I,3

Rame

Ram

P2

prsme

PO

COklVkU*

Beams

cOl!JlRllSBeams

C0hSTIS

Ba*rns

CDlUmnS

seam

6.059

6.0*
4.0*
o.**
o.o1z33*
4300
54.9
28.5

6."*
4.')*
o.z*
0.0183*
3240
54.0
28.5

6.O*
4.01
0.2'
0.0667*
3240
54.0
28.5

6.01
4.0*
0.165*
0.0121*
5450
56.1
29.8

6.0*
5.030"
0.13.P
0.0168*
5450
70.5
29.8

width
b (in.)
Depth L (in.)
Cover/depth d'lt
Percent reinforcenentPt
Gzncrets strmgth f; (psi)
Steel yield stress f (ksi)
St..1 modulus EC103 1st)

0.046
0.1792
0.0192
3200
56.4
29.3

6.061
3.183
0.2618
o.oml
3200
58.5
29.3

Width b (in.)
Depth c (in.)
Cover/depth d'/t
PereCnt reinforcement ot
Concrete strewjth
f; (psi)
Steel yield stress f (kai)
Steel modulus E(103 i:
si)

6.000
3.975
0.2066
0.0184
2900
75.0
26.0

6.000
3.950
0.1978
0.0186
29w
75.0
28.0

6.006
5.003
0.1566
0.0146
44hO
75.0
28.0

6.0*
4.0*
0.2*
0.0667*
4300
54.9
28.5

6.000
3.985
0.1946
0.0185
4460
75.0
28.0

6.076
3.976
0.2073
0.0331
4460
60.0
31.2

6.062
3.973
0.2070
0.0183
6600
75.0
28.0

6.027
5.046
0.1633
0.0145
6000
75.0
28.0

6.111
4.090
0.2268
0.0320
6001
60.0
31.2

6.000
3.985
0.1902
0.0183
373s
77.8
28."

6.025
4.005
0.2077
0.0332
3580
69.6
31.2

30

A general nodi ear analysis of concrete structures


Frames F2 and F4-frames
with columns in single
curvature
Tests on seven rectangular frames loaded so that the
columns were bent in symmetrical single curvature are
reported by Furlong and Ferguson [lo]. The loading of the
frames produced conditions in the columns similar to
those produced by alternate-span loading of a multi-story
frame. Frames F2 and F4 of the test series were selected
for study using the proposed analysis. The beam loads aP
and the column loads P were increased proportionately to
failure of the frame.
Frame F2 had columns of height ratio b/h of 20,
nominal eccentricity ratio e/h of 0.106 and stiffness ratio
Y of 1.48. Frame F4 had columns of height ratio I./h of
20, nominal eccentricity ratio e/h of 0.222 and stiffness
ratio Y of 0.81. In the analysis, in order to account for
magnification of bending moment in the columns due to
axial load, an extra joint was placed at the midheight of
the columns.
The deflections at the midheight of the two columns of
Frame F2 and F4 are plotted against column load in Fig. 6.
Frame F2 failed at a column load of 61.6 kips, and the load
predicted by the proposed analysis was 64.6kips. For
Frame F4, the actual failure load was 52.5 kips, the load
predicted by the proposed analysis was only 44.9 kips.
Frame
70

I=$
1

L
124 in OOk6
68in 01500

60

E8

A4

fi
40 in
32 in

westuAlrm

East cdum

243

Frame B l--three-bay frame


Tests on three three-bay frames with the columns in
single curvature are reported by Blomeier [ 111.Frame B 1
was selected for study, which had columns of height ratio
I./h of 15, stiffness ratio Y of 1.0 and nominal
eccentricity ratio e/h of 0.1. The column loads P and the
beam loads UP were increased proportionately to failure.
In order to approximate the beam-column effect, an
additional joint was assumed at the column midheights.
During the test, beam hinging occurred at a column load
P(l + a) of 58.3 kips. The beam loading ram extensions
were exhausted at a column load of 66.2 kips. When the
sp&imen was retested later, collapse occurred at
91.Okips, but only the results of the first test are
considered in the present study.
Figure 7 shows a plot of column load vs the horizontal
detlection at midheight of one of the columns and the
vertical deflection at midspan of one of the beams. The
plot of measured deflection at midspan of the beam
indicates beam hinging at a column load of about 58 kips.
The predicted response indicates that the column
midheight section fails at a load of approximately 76 kips.
When the load-moment combination 6rst reaches the
interaction diagram, this condition is defined as material
failure in the proposed analysis. During the retest of
specimen Bl, the experimental data (results not presented
here) indicated that the column load-moment path
travelled on the interaction diagram, with increasing
axial load and decreasing moment. This explains the
difference between the predicted capacity and the
measured ultimate load when frame was retested.

50

bflection af calumn mid-height, in scale +O.l_l

Fig.6. Load-deflectionstudy-framesF2 and F4.

Frames AZA5-sway
frames under combined loads
Tests on five rectangular frames subjected to column,
beam and lateral loads are reported in Ref. [12]. Four
frames were selected for study, two with identical
columns, and two with unequal size columns.
In all frames, beam and column loads were applied in a
ratio held constant to produce a nominal eccentricity ratio
e/h of 0.1. Beam and column loads were increased until
75% of the predicted ultimate frame capacity was
reached. These loads were then held constant while lateral
load was applied until frame failure.
Experimental evidenceb, 121 shows that because of
increased strength of the beam-column joint block, hinges
do not occur at the center of the joint block. Instead, they

Haizantal deflection at

Ceflectii.

Fig. 7.

in

Load-deflectionstudy-frameB 1,

264

B. L. GUNNINet al.

occur in the weaker member near the face of the joint


block. The effect of the shift of hinge is to stEen the
smrcture,as wellas increasethe ultimatelateralload. The
sway frames in this study were analyzed twice, first
consideringthen neglectingthe effect of the hinge shift.
The shiftingof the hingewas accomplishedin the analysis
by assnmingadditionaljoints in the beam and column,
at the faces of the joint block. The additional joints
were connected to the beam column centerline intersection by members whose properties were about five times
stiffer than the frame member properties. The five fold
stiffness increase of the jolt-~rn~s
was the result
of an independentstudy using a finite element model of
the joint blockl.121.
Frames A2 and A5 had identicalcohrmnsof heightratio
L/h = 9.5, stiffness ratio Y of 2.0 and e/h = 0.1. Frame
A2 contained beams that were weaker than its columns;
while the beams in Frame A5 were stronger than its
columns.
Plots of the column load and lateral load vs lateral
deflection of Frames A2 and A5 are shown in Fig. 8.
Curves marked S designate the predicted responses
considering the hinge shift. The lateral deflection of
Frame A2 under symmetricalgravity loads was probably
due to a colitis
of accidental~cen~city of column
loads and slight initial crookedness of the frame.
The mode of failure of Frame A2 was by instability
after the formation of two hinges in the leeward comers
of the beams. The measured ultimate lateral load was
1lOOlb;and the predicted lateral loads were 13OOlband
IlOOlb for the frame with and without hinge shift
respectively.
Frame AS failed predominantIyas a result of leeward
column failure. The measured ultimate lateral load was
2300lb; the predicted lateral load capacitieswere 1910lb
and 163Olbfor the frame with and without hinge shift
respectively. For specimenscast in a horizontal position
use of an effective concrete strength larger than 0.85fE
has been suggested[7,121.In the analysis of Frame A5
when the effective concrete strength was assumed as
l.Of: (results not shown here), the predicted ultimate
lateral loads were 19Nllband 248Olbfor the frame with

and without hinge shift. These predicted capacities are


reasonably close to the measured value.
Frames A3 and A4 containedunequalcolumn sizes. In
Frame A3the windwardcolumnhad twice the stiffnessof
the leeward column. The cohunn hei~t/~c~ess I./h
ratios were 7.54 and 9.5 for the windwardand leeward
columnsrespectively; and the respecive stiffnessratios VI
were 4.0 and 2.0. In Frame A4 the stiffer column was
placed on the leeward side; thus interchangingthe ratios
givenabove. The nominale/h ratio was 0.1for the weaker
column in both frames. The beams in both frames were
weaker than the leewardcolumns,so hingeswere formed
in the beams.
Plots of column load and lateral load vs lateral
deflection of Frames A3 and A4 are shown in Fig. 9.
Curves marked S designate the predicted responses
consideringthe hinge shift. Frame A3 failed at a lateral
load of 25tlOlb;and the predicted loads were 2050and
165Olb for analysis with and without hinge shift
respectively. Moment data at the leeward comers
indicated that accidental eccentricity in the applied
leewardcolumnload existed duringthe test. The induced
moment was in the opposite direction of the lateral load
moment,thus increasingthe ultimatelateralload capacity.

-k
19.
15

5
-II 10
i
-1
05

0028

0.65

0.034

I x)

Lateml deflection,In scale t0l-l

Fii.

Fig. 8. ~d~efl~tion

study-frames A2 and A5.

9. Load-deflection study-frames A3 and A4.

Fig. 10. Test frame A4 after failure.

Ageneraluoobar analysisof concretestructures


This observation explains the rather large differences
between the measured and the predicted capacities.
Frame A4 failure load was 17OOlb,and the predicted
loads were 1700 and 1450lb for analysis considering and
neglecting hinge shift respectively. Fignre 10 shows the
frame after failme while the instrumentation is in place.
CONCLUSIONS

A generalized analysis for the complete nonlinear


behavior of planar frames has been described. Applications to reinforced concrete frames have been demonstrated. The method predicts nonlinear behavior of
concrete frames as determined from physical tests with
reasonable accuracy. The proposed method could be
extended to provide a method of analysis for other types
of structnres for which thrust-moment~urvature
data
can be generated.
Acknowledgements-This paper is based on portions of the
doctoral dissertations submittedby B. L. Gunnin and F. N. Rad at
The University of Texas at Austin. Portions of this investigation
have been supported by grants from the National Science
Fountation; Department of the Army, Civil Works Division; and
the Reinforced Concrete Research Council. Computing services
were made available by the Computation Center of The
University of Texas at Austin.
REFERENCES
1. B. L. Gmmin, Nonlinear analysis of planar frames. Ph.D.
Dissertation, The University of Texas at Austin (1970).
2. E. L. Wilson, Matrix analysis of nonlinear structures. Proc.

265

2nd Conf. EIectronic Cornput., American Society of Civil


Engineers (1%3).
3. W. Ramberg and W. R. Osgood, Description of stress-strain
curves by three parameters. National Advisory Committee for
Aeronautics. Tech. Note 902 (1943).
4. J. H. Argyris, Continua and discootinua. Proc. Conj. Matrix
Meth. Struct. Me&m., Wright-Patterson
Air Force Base
_
(1%5).
5. W. Lansing, W. R. Jensen and W. Falby, Matrix analysis
methods for inelastic structures. Proc. Conf. Matrix Met/z.
Struct. Mechan., Wright-Patterson Air Force Base (1965).
6. S. A. Saafan and D. hi. Brotton, Elastic tinite deflection
analysisof rigid frameworks by digital computer. Symp. on the
Use of Computers in Civil Engng (1962).
I. J. E. Breeo, The restrained long concrete column as a part of a
rectangular frame. Ph.D. Dissertation, The University of
Texas (1962).
8. E. Hognestad, A study of combined bending and axial load in
reinforced concrete members. Unioersityof Illinois Bulletin,
Engineering Experiment Station Bulletin Series No. 399
(1951).
9. P. M. Ferguson and J. E. Breeo, Investigation of the long
concrete column in a frame subject to lateral loads. Symp. on
Reinforced Concrete Columns, American Concerte Institute
SP-13 (1966).
10. R. W. Furlong and P. M. Ferguson, Tests of frames with
columns in single curvature. Symp. on Reinforced Concrete
Columns, American Concrete Institute SP-13 (1966).
11. G. A. Blomeier, Effect of yielding of restraints on slender
concrete columns in braced frames. Ph.D. Dissertation, The
University of Texas at Austin (1968).
12. F. N. Rad, Behavior of single story two-column reinforced
concrete frames under combined loading, Ph.D. Dissertation,
The University of Texas at Austin (1972).

You might also like