Computers & Structures Vol. 46, No. 1, pp. 83-97. 1993 CVM-7949193 s6.00 + 0.

00
Printed in Great Britain. 0 1992 Pergamon Press Ltd

INELASTIC ANALYSIS OF REINFORCED

CONCRETE SPACE FRAMES INFLUENCED BY
AXIAL, TORSIONAL AND BENDING INTERACTION
G. M. CQWHI and F. CAPPELLO
Universitri di Ancona, Facolta di Ingegneria, Istituto di Scienza e Tecnica delle Costruxioni,
Ancona 60131, Italy

(Received 7 October 1991)

Abstract-The inelastic analysis of spatial reinforced concrete frames under static loads is performed
according to a peculiar discrete truss model of the beam element behaviour having any mechanic,
geometric and reinforcing characteristics, based on the diagonal compression field theory, in the presence
of axial, bending and torsional interaction. The authors propose a theoretical iterative method of
calculating the nodal element actions at various inelastic nodal displacements and supply a suitable
computer procedure, based on the residual redistribution concept, in order to analyse the deformational
and stress state of any spatial frame at various load levels. Numerical analyses by a suitable computer
program show that the theoretical results are good in comparison with the experimental results and the
influence of the interaction effects on the cracking and ultimate behaviour of the structure is quite
significant in many cases.

1. INTRODUCTION moment is resisted by the circulatory shear flow in the

wall of the tube. Each element of wall works like a
Theoretical investigations in the field of inelastic
plane truss with concrete compression struts inclined
analysis of reinforced concrete frames, subject to
at an angle a with the longitudinal axis of the tube..
static actions involve significant problems and a
Longitudinal steel reinforcement and closed stirrups
considerable computational effort, since the stiffness
are the tension struts (Fig. 1).
of the structure depends on the stress and strain state.
By means of the theory of diagonal compression
In a previous study [l] the authors faced the prob-
fields [3], this model is extended to combined axial,
lem related to calculating the stiffness, above all in
biaxial bending and torsional components [4-61. The
space frames, using a suitable iterative procedure.
effect of the bending component about the y and z
The nonlinear force displacement relations were built
axes and of the axial component action along x, is
up by a polynomial interpolation algorithm.
that the values of a and t, become functions of the
This paper, as an improvement and extension of
coordinates x, y and z (Fig. 2b). The problem can be
the previous study, presents a suitable method for
faced, as indicated in [4], discretizing the hollow
analysis of reinforced concrete frames, which takes
cross-section in a sufficient number of rectangular
into account the nonlinear effects of axial, torsional
elements, having height bi. The elements must be
and bending interactions, according to the geometric
fairly small to ensure that ai and ti are constant in
characteristics of the cross-sections and the mechan-
each element. Figures 2(a and b) show the difference
ical characteristics of the materials. Only the viscous
between the pure torsion model (constant thickness t
effects are not taken into account. Useful compari-
and angle a) and the combined axial, bending and
sons with the results of experimental tests are also
torsional components model (variable thickness ti
provided.
and angle ai).
As indicated in [l], a general approach to the
The proposed model, as an improvement of that in
problem has been adopted by identifying three
[4], is also based on the following assumptions:
behaviour models of the cross-section, element
(a) no tension in concrete after cracking;
(beam) and structure.
(b) cross-sections remaining plane (the longitudinal
strains vary linearly over the cross-section and
2. ANALYSIS OF THE CROSS-SECTION
BEHAVIOUR MODEL
the shear strain is neglected);
(c) absence of warping restraints (the torsional
The behaviour of a reinforced concrete cross-sec- moment is resisted by the circulatory shear flow);
tion subject to torsion is provided, before cracking, (d) small deformations.
by the linear elastic Saint-Venant theory with reason- Given the stress-strain relationship of concrete and
able accuracy. After cracking, literature and codes steel, the method consists of determining the axial,
provide models that assume the cross-section acts like torsional and bending components N(x), M,(x),
a tube of constant thickness t [2]. The torsional M,(x) and M,(x) as functions of the axial defor-

a3
84 G. M. Coccm and F. CAPPELU)

Fig. 1. Space truss in pure torsion.

mation and curvatures about the x, y and z axes p(x). where &(x) is a suitable coefficient which accounts
L(X), a(x) and X,(X). for the application point of the compression stress
For each element as many equations of equilibrium resultant on element i.
and compatibility as the number of unknowns relat- The same expression may he obtained [2] calculat-
ing to the same must be written. ing the angular distortions due to ~,~(x), cSi(x) and
The following equation of compatibility can be ~Jx) in a rectangular wall element having unit height
obtained for each element i of the generic cross- and cotan a width (Figs 3a, band c) at the application
section of abscissa x, according to assumption (b): point of the compression resultant on element i. The
longitudinal component c,,(x) cotan a&), due to
Gilx) = VCx) - L(xl.Yt(x) - Xy(xk(x)s (1) E,~(x),and 6,(X)(1 -/Ii(x)) cotan a,(x), due to cd(x),
must be equated to the transverse component Ebb
relating the strain state to the cu~atures and axial tan CQ(X), due to LJx). and c&x)(1 -&(x))
deformation. Q(X) is the longitudinal strain in the tan ai( due to Ebb. This corresponds to a mini-
element i, the same being defined as positive for mum of the total angular distortion.
tension, q(x), X,(X) and x,(x) being positive accord- The following equation of compatibility relates the
ing to Fig. 2(b). J+(X) and .ai(x) are the coordinates torsional curvature L(X) to strains Q(X), gSi(x) and
of the average point of the external edge of the eci(x). It may be obtained by applying the principle
element. of virtual work to the equilibrated set of M,(x)
The compressed diagonal struts angle ai is component and the I”i(x) circulatory shear flow to the
assumed to he the same as the local principal strain congruent set of x,(x) curvature and r,(x) total
direction. It may be easily obtained [2] from the angular distortion, relative to the above strains in
stirrup strain EJx), the same being positive for every element i
tension, from the longitudinal strain c,,(x) and the
compressed concrete strut strain eci(x) the same being
positive for compression Mx(xkr(x) = i Ti(x)ri(x), (3)
i=l

tanz a.(x) = Mx) + ~AX)(l - B,(x))1

i (2)
I%tx> + %itx)tl - Bdx))l ’ where n is the total number of elements.
 Inelastic analysis of reinforced concrete space frames 85

(4

Y
f

Fig. 2. (a) Pure torsion model (constant thickness t and angle a). (b) Bending-torsion model (variable
thickness ti and angle a,).

Summing the above contributions due to Q(X), where the coefficient yi(x) accounts for the effect of
E&) and Eli (Fig. 3a-c) the r,(x) total angular stress variation in the thickness.
distortion is The following results from the cross-section
rotational equilibrium
T,(x) = c,,(x)cotan ai + &(X)tan ai(x)

+ [l - flj(X)]Cci(X)[COtan a,(x) + tan ai(X (4) Mx(x)= i: Ti(x)[4 - Bi(x)ti(xll* (6)

i-l
The circulatory shear flow Tt(x) is obtained from
equilibrium along y or z axis of the compression stress where d, is the distance of the element edge from
resultants on element i (Fig. 4) origin 0.
Substituting the values of r!(x) from eqn (4) Tj(x)
T,(X) = 2~,(x)y,(X)t,(X)U,,(X)b, COS a,(x)sin cLi(X),
from eqn (5) and M,(x) from eqn (6) in eqn (3) and
(5) by equating the left-hand terms to the right-hand
86 G. M. Coccyx and F. CHPELLO

(a)

@I ‘------wtg
r ----_---_-----------
a(
.#4

Fig. 3. (a) Longitudinal angular distortion due to Q. @) Transverse angular distortion due to 4.
(c) Longitudinal and transverse angular distortions due to cc,.

terms of the sum, after some algebra, the following from the equilibrium along y or z axes of the stress
results are obtained resultants in the truss (Fig. 4)

&)cotan a&) + &x)tan a,(x)

+ r1 - Mx)l&) x [cos a&)sin a,(x)/cotan a,(x)], (8)

sin a,(x)cos ai
where A, is the area of the stirrup.
=[ddi- Pi(~x)ti(x)lXx(xx)* (7) As shown in 133, the r,(x) thickness must
be calculated by mounting for the beading of
The equation relating the Q&S) stirrup stress to the compression struts caused by xX(x), X,(X) and
compressed concrete strut stress may be obtained x,(x) curvatures. Figures 5 and 6 show this effect on
 Inelastic analysis of reinforced concrete space frames 87

Fig. 4. Stress resultant equilibrium at the comer element.

element i due to curvature in direction a,(x) for the inelastic behaviour of diagonal compression
and stress and strain fields along the t,(x) thick- concrete struts. For small values of actions, the
ness. compression diagram is triangular and the coefficient
The pi(x) and y,(x) coefficients, previously indi- values are pi(x) = l/3 and n(x) = 3/4 (elastic
cated, may be calculated, according to the strain and behaviour).
stress compression diagrams along the t,(x) thick- The curvature about the y, axis, related to the
ness, by the well-known compatibility, rotational and above effect, may be obtained by the first and the
translational equilibrium relations [7), which account second partial derivatives, with respect to the variable

261(x)71
(x)t, b.)%I(~)c~~~I(x)
I

Fig. 5. Stress and strain fields due to curvature in direction ai(

88 G. M. Coccm and F. CAPPELLO

Fig. 6. Effect of flexural and torsional curvatures in direction ai(

xi, of the displacement function along z of the axes and to the signs of Fig. 2(b), the equilibrium
element yields

xY,(x) = xY(x)cos2 ai + 2XX(x)sin a,(x)cos ai(

N(x) = 2 Ajao - i T,(x)cotan ai(
(9) j=1 i=l

According to the strain diagram of Fig. 5 and from M,(x) = - 2 Ajarizj+ i T,(x)cotan ai
eqn (9) it follows thatt j= 1 i= I

x Izi - Bi(X)fi(X)CoS 4il9

cci(x)= ~dx)[&(x)co~*
a,(x) + x,(x)sin 2%(x)1.

(10) M,(x) = 2 Ajouyj - i T,(x)cotan a,(x)

j-I i=l

For assigned values of ?I, x,(x), x,(x) and X,(X), X [Yi - MxMx)sin 4il9 (11)
eqns (l), (2), (7), (8) and (10) together represents
a nonlinear system in the five unknowns ai( where Aj is the area, uri is the stress and zj, yj are the
r,(x), +(x), Q(X) and eci(x) for every element i, coordinates of the n, longitudinal bars in the cross-
because tensions a,,(x) and uSi may be obtained section $i is the angle between the outward normal to
from the uniaxial stress-strain curves for concrete the element edge and the positive direction of the z
and steel. axis.
The internal actions in the generic cross-section The stress state corresponding to assigned values of
of abscissa x may be obtained, from the solution of longitudinal deformation and curvatures may be
the above system, by means of eqn (6) and the calculated in a nonrectangular cross-section,
following flexural translational and rotational equi- described by any irregular set of vertices (Fig. 7a), by
librium equations. According to the xyz Cartesian means of the same equations. In fact, eqns (l), (2),
(7), (8) and (10) are related only to the ith element,
while eqns (11) are related to the element subdivision
number of the cross-section perimeter.
7 For a nonrectangular cross-section, as shown in Fig.
In particular, it is possible to apply this method to
7(a), the following value of the curvature about the parallel
axis to the element edge: x+,(x) = ~,(x)cos 4, + X,(x)sin Qi, hollow cross-sections provided that the ti(x) thick-
must substitute the x,(x) value in eqn (9). ness is checked for values greater than the real values.
 Inelastic analysis of reinforced concrete space frames 89

6)

Fig. 7. (a) Discretization of nonrectangular cross-section. (b) Discretization of composite cross-section.

More complex T or double T shaped cross-sections accounting for the tension stiffening or cracking of
can be accurately analysed, according to the hypoth- concrete [7], may be utilized for calculating the
eses of undeformed transversal cross-section profile internal actions.
and absence of warping restraints, subdividing these
into many rectangular elements as shown in Fig. 7(b).
3. SOLUTION ALGORITHM OF THE NONLINEAR
Each may be analysed according to the rectangular
SYSTEM
cross-section of Fig. 2(b) and its contribution to
internal actions may be calculated by means of eqns As is well known, the solution of a nonlinear
(11) and summed to the contribution of the other algebraic system of n equations in n unknowns may
elements. However a unique cross-section having be obtained by means of Newton iterative method, or
many vertices may be considered because, as it is well other similar current methods, according to the
known, the effect of shear flow is nearly negligible degree of nonlinearity of the system. The authors
along the common edges of various rectangular have noticed that calculating the solution in this way
elements. requires a considerable computational effort, which is
The reader should note that the nonlinear equation highly influenced by the accuracy of the initial trial
system in the five unknowns a,(x), ti(x), Q(X), Gus solution, related to the pure torsion hypothesis, as
and Gus, for each element i, cannot have a solution shown in [4]. Therefore an additional computational
for any q, xX(x), x,(x) and x,(x) values, as in the effort for calculating the initial solution is required.
presence of bending or biaxial bending with small At the same time prevailing bending or biaxial bend-
or null torsion, or in absence of cracking. If a ing effects cannot in any way be drawn from this
solution does not exist, the well-known expressions, initial trial solution.
90 G. M. Coocm and F. CAPPELLO

In order to avoid the above shortcomings the x’, is considered. The beam is subjected to the set of
authors have developed an iterative solution algor- end displacements and forces shown and is loaded by
ithm, eliminating a priori the unknown values, which a uniform set of pX,, p,,, and pzPcomponents.
were inconsistent with the real physical problem. According to the previous hypotheses (Sec. 2) and
Initially t&) is derived from eqn (7) of small displacements, the strain state of the generic
cross-section may be obtained from the nodal
~ ,txj = [4- Bi(x)ti(x)lXx(x) 6lii(x) displacement components [1]:
SI
tan ai -Gz$g

- 11- Bi(Xki(X)~ (12)

sin2 ai
x = _ c&d - 41x,)
Deriving ti(x) from eqn (10) and substituting the I L '

above expression in eqn (2), Q(X) is obtained:

~,,(x')=a(x')~,,,+6(x')~~~,+c(x')(W2- 41,

&i(X)
d,XJx)tan ai -26,(x) x,,(x')=a(x')~,,,+b(x')~,,- c(x')(u2-4), (15)
-
Bi(xMxNan ai
+ 211-ml where
~,(x)cos2 a,(x) + x,(x)sin 2a,(x)

(13) a(x’)= -g +s,

The value ai results from eqn (8)

b(x’)= -1 +f$,
ai = arcsin J usi(X)As

2/?i(x)yi(x)ti(x)uCi(x)~x .
(14)

(16)
The method consists of assigning small discrete incre-
ments to the unknown ai in the O-(x/2) range (for
As shown in [1], the smaller the length L of the beam
instance l/1000 of the interval) according to the
element, the more eqn (15) can be utilized successfully
desired accuracy. For each trial value of ai( Q(X)
for a nonlinear analysis, because they represent linear
is calculated from eqn (1) according to q, xX(x), x,(x)
compatibility relationships. The internal actions can
and x,(x). Q(X) is calculated from eqn (13), &(x)
therefore be drawn applying the proposed method to
from eqn (10) and +(x) from eqn (12), assuming the
any set of nodal displacement components. The nodal
pi(x) value found in the previous trial and checking
forces can be drawn from the following according to
the physical conditions Q(X) >, 0, t,(x) >, 0 and
the principle of virtual work 0’ = 1,2):
Q(X) 2 0. If at least one of these conditions is not
verified, the current a,(x) is discarded. In the first trial
(at(x) = 0), the elastic value of the coefficients pi(x) qx.=qj 0
N(x’) dx’ + fp,. L,
and y,(x) is assumed. The tension aCi(x) and the
coefficients /Ii(x) and yi(x) are calculated according L
to the uniaxial stress-strain relationship of concrete.
Then ai is drawn again from eqn (14), having
Map.= -(-lpi
I0
M,.(x’) dx’,

s
previously defined the tension a,,(x) according to the L

uniaxial stress-strain relationship of steel. If the value qy,= -(- 1)’ M,.(x’)c(x’) dx’+$p,.L
found differs, by a negligible amount, from the trial 0

value, this is assumed as the ith element solution and L

the contributions to the internal actions are calcu- Mi,,<= M,,,(x’)a(x’) dx’ - ( - l)j$pz,L2,
lated by means of eqn (11). If the above condition is s 0

not verified in any trial, according to the proposed L

behaviour model, the solution does not exist and the Q=(-ly’ M,.(x’)c(x’) dx’ + $p?L
internal actions are calculated as previously indi- s0

cated. The iterative algorithm is applied, in the same

way, to all elements of every cross-section. Mjzv= L M,.(x’)a(x’) dx’ + ( - l)j$p,,,L’. (17)
s0
4. DISPLACEMENT BEHAVIOUR ANALYSIS OF
BEAM ELEMENT
For every set of inelastic nodal displacements, the
nodal forces can be calculated by applying a suitable
The beam element of Fig. 8, having length L, numerical integration method to the first term of the
mechanical and geometric properties constant along right-hand side of eqn (17). In this way the overall
 Inelastic analysis of reinforced concrete space frames 91

Fig. 8. Loads, displacements and forces in the beam element.

equilibrium of the beam element is violated owing to means of the usual iterative Newton-Raphson tech-
the numerical approximations of the integration nique
method. In order to avoid this shortcoming, a heavy
iterative procedure may be used to correct the initial K,s(”+ ‘) = P + P*(“), n = 0, . . . , N, W-0
displacement values in order to satisfy the equi-
librium and compatibility relations at the same time. where s(”+ ‘) is the nodal displacement vector related
Alternatively, as shown in [l], a method of linearly to n + 1 iteration, K, is the elastic stiffness matrix of
interpolating the internal actions of the two end the structure and P is the nodal load vector. P*(“) is
cross-sections, may be utilized. The advantages of the fictitious nodal load vector, relevant to the pre-
this method consist in applying the procedure shown vious iteration n, expressed as
in Sec. 2 only to the two end cross-sections and in
globally accounting for linear noncracked behaviour
of intermediate cross-sections. p*Q) = K, i s;], (19)
i-0
5. DIBPLACEMENTBEHAVIOUR OF THE WHOLE
STRUCTURE
where $1 is the nonlinear nodal displacement vector
relevant to the ith iteration. As is well known, the
The displacement inelastic analysis of space frames latter can be determined in every iteration by apply-
can be performed discretizing the whole structure ing an unbalanced nodal load vector to the structure.
in a suitable number of beam elements and solving The components of this vector are calculated accord-
the matrix algebraic linear equation system [l] by ing to the difference between the elastic nodal forces
92 G. M. Coccm and F. CAPPELLO

and the inelastic nodal forces, relevant to the ~(“1 A suitable choice of the amplification coefficient
component vector of the previous iteration. The matrix is the product 6’01, where
procedure is applied through N iterative cycles up to
convergence. In this respect the authors have noticed gcn = IIF-F”- I)11
that the application of the above procedure to the IIF” _ Ii+ UII ’ (21)
present analysis may involve a considerable number
of iterative cycles owing to the great influence of axial
I is the identity matrix, F(‘) and F(‘- ‘) are, respectively,
internal action on the flexural and torsional stress
the vectors of nodal forces relevant to the ith and
state. This influence is emphasized in structures
(i - 1)th iterations. The above expression is a very
having a low degree of redundancy, owing to the low
good approximation and extension of eqn (20) to the
capacity of stress redistribution. To obviate this
multidimensional case, because the coefficient is pro-
shortcoming it is possible to apply a convergence
portional to the ratio between the unbalanced force
acceleration algorithm, based on the Aitken secant
method [8], which utilizes suitable amplification co- vector modulus, at the (i - 1)th iteration, and that
relevant to the variation of forces at the ith and
efficients of the s$ components. According to the
(i - 1)th iterations.
nonlinear uniaxial relationship between the load
component P and displacement component s, shown The above procedure may be applied to the analy-
sis of geometric nonlinear effects, by utilizing suitable
in Fig. 9, as is well known, the following amplification
corrective coefficients in the elastic stiffness matrix of
coefficient is obtained:
the beam element, as functions of the action values,
6(” = P Pci-l)
-
as shown in [9] and in [l].
~(0 _ p(i- U ’

which agrees with the ratio of the diagram secants 6. COMPUTER PROGRAMMING PROCEDURE
at the i - 1th and ith iterations respectively. In
The proposed method can be easily applied to
the present case of multidimensional nonlinear
structural analysis by a suitable computer program
relationship between P and s, it is necessary to
according to the following steps
determine a matrix of coefficients [8], calculated
according to the mutual differences among elastic 1. the structure is discretized by beam elements
and inelastic nodal forces at the (i - 1) and (i) having length L, which permits assuming the cross-
iterations. section, the longitudinal and transverse reinforce-

P’
ptt-1

Fig. 9. Iterative technique using amplification coefficient.

 Inelasticanalysisof reinforcedconcrete space frames 93

ment, the mechanical properties of concrete and steel stress-strain curve for steel. In this step, the
constant in each of these; existence of the system solution is verified. If the
2. for each heam element, or groups, the shape of solution does not exist the internal actions are
the section, by means of the vertex coordinates, the obtained by discretizing the whole cross-section
dimensions and the placement of lon~tudin~ bars, by means of a large number of elements in which
the area and the spacing of stirrups, the elastic the stresses and their resultant are calculated
moduli, the ultimate strength of materials and the according to eqn (1) and to the stress-strain
number of elements, which discretize the section, are curves;
given; 5. the unbalanced nodal loads and the s$ vector,
3. the elastic nodal displacement vector at 0 initial relevant to these, are calculated. In order to limit the
iteration {s(O) = s, = K;‘p) is calculated after having computational effort the 6” amplitication coefficient
assembled the stiffness matrix of the structure, ac- matrix is calculated according to eqn (21);
cording to the geometric elastic characteristic of beam 6. at the end of ith iteration, the global nodal
elements. If the analysis of geometric nonlinear effects displacement vector sci) is increased by the S%$
is required, the calculation of vector s(O)is performed contribution;
as shown in [9]; 7. steps 4,5 and 6 are repeated up to convergence,
4. according to the proposed method the nodal which is assumed as reached, according to the
forces are calculated in the ith iterative cycle, utili- following inequality:
zing the uniaxial stress-strain curve for concrete,
which accounts for softening and tension stiffening
effects [2], and the perfectly plastic uniaxial elastic

St - 0.31 cm21lS cm II II 43&n

As* 5.06+5.06 cm2

Aa = 1.98+1.98 cm

Eb =250,000 kg cm2
Rbk -246 kg/cm l
Rbk lono =4.400 kg/cm2
Rbc -=3.000 kg/cm*
Fig. 10. Analysed structural frame.
94 G. M. CoccHl and F. CAPPELU)
Porg)
30000

Q Tests (71
+ Theory

0 - VS (cm)

0 2 4 6 8 10 12
Fig. 11. Load deflection for specimen Sl.

with 6 = 0.0001. The above inequality agrees with the val of the unknowns ui which can be obtained quite
minimization of the ratio between the norm of non- accurately by means of a reduced number of trials.
linear iterative displacements and the norm of initial The reliability is also guaranteed as regards the
elastic displacements; solution of equation system (18), relevant to the
8. cci(x), cli(x) and csi(x) strains are compared with whole structure, because the above procedure con-
ultimate conventional strain of concrete and steel; sists in subsequent elastic analyses at fictitious load
9. the end nodal forces of the beam elements are conditions, caused by inelastic behaviour of beam
obtained according to the calculated global displace- elements.
ment vector s;
10. the above steps are repeated for every load
condition. 7. NUMERICAL RESULTS AND COMPARISON OF
THEORY WITH TESTS
The authors have noticed that the reliability of the
above procedure is perfectly guaranteed as regards The structure in Fig. 10 ([lo] specimen S l), consist-
the solution of nonlinear equation systems for every ing of a floor beam of length L, and a spandrel beam
cross-section, owing to the restricted existence inter- of length L,, is analysed. Both beams have rectangu-

&kg)
30000
1

Q Test (71
*Theory
+ Gross
+ Cracked

I--- I

1 2
Fig. 12. Load twist for specimen Sl.
 Inelastic analysis of reinforced concrete space frames 95

pw
30000 -

20000 -

Q Test 171
* m-Y
4 Gross
-9 Cracked

0 1000 2000
Fig. 13. Load torque for specimen Sl.

lar cross-sections. The elastic, geometric and re- P = 18,000 kg, utilized in [lo] to design the cross-sec-
inforcement properties are shown in Fig. 10. tions, is noticed. Beyond this value both diagrams
The structure is discretized by means of 17 nodes show identical behaviour of large displacements at
and 8 beam elements, of constant length, for the floor small increments of load, with a low percentage of
beam, and 8 beam elements for the spandrel. The end shiftings. It should be noticed that it is necessary to
restraints are described as those carried out in [lo]: remove the controls of the ultimate conventional
the allowable (1) and nonallowable (0) nodal dis- strains of concrete and steel, shown at step 8 in
placement components, for every restraint, are given Sec. 6, in order to describe the structural behaviour
in Fig. 10. at higher loads than the above, because the tests
The diagrams in Fig. 11 show the greatest vertical are performed up to the condition of concrete
displacement (- u of node 5) at various p load values, crushing.
given by theoretical analysis and tests. There is good The diagrams in Fig. 12 show the greatest torsional
agreement between theoretical and experimental rotation (4, of node 9) at various P load values. There
results. The typical slope variation at the cracking is good agreement between theoretical and exper-
load P = 4500 kg, up to the ultimate load imental results. A slope variation according to the

&kg)

1
30000

Q Test [7]
+ Theory

-VS(em)

1 2 3 4 5
Fig. 14. Load deflection for specimen S6.

CAS w-3
96 G. M. Coccm and F. CMPELL.O

ww
30000 1

20000 -

10000 -

-) Gross
+ Cracked

0 100 200 300 400

Fig. 15. Load torque for specimen S6.

previous behaviour is noticed. In the same figure, the P load values. Unlike the results of specimen S 1, the
straight lines relevant to the elastic linear behaviour theoretical and experimental diagrams agree well at
of gross and cracked cross-sections, described in [lo], every load value, owing to the lower influence of the
are shown. Inspection of the theoretical diagram torsional behaviour of the spandrel as a consequence
slopes shows that the torsional stiffness of the span- of the different length ratio. As regards the variation
drel initially agrees with the elastic gross stiffness up of the diagrams themselves, the observations made
to P = 4500 kg. It then approaches the cracked stiff- regarding Fig. 11, apply. The effect of variation of the
ness and finally is almost null for values of loads ratios between the flexural and torsional stiffness of
higher than the ultimate. The comparison between the beams is shown in the diagrams of Fig. 15, similar
the theoretical and experimental diagram shows a to those of Fig. 13. Contrary to the results obtained
lower stiffness of the truss model with respect to the for Sl, the theoretical behaviour model is closer to
real model with a small percentage of shiftings. the linear elastic behaviour than the experimental
Figure 13 shows the diagram of the ratio between the test.
torsional continuity moment X at node 9 and the
length L, of the floor beam, as a function of the P
value. This diagram is indicative of the flexural state 8. CONCLUSIONS
interaction with the torsional state of the spandrel
beam. It is shown that for a P value lower than the In this paper the authors studied the problem of the
ultimate, the theoretical X values are included among theoretical analysis of reinforced concrete space
those relevant to the cracked model and those rel- frames accounting for the interaction effects among
evant to the elastic model. Beyond this limit the the stress components, by means of a suitable exten-
diagram approaches the last model as confirmed by sion of the diagonal compression field theory, for
experimental results, owing to the variation of the pure torsion, applied to the general case of coupled
ratio between the flexural stiffness of the floor beam axial, torsional and bending actions. Major attention
and the torsional stiffness of the spandrel. As pointed was devoted to numerical problems in calculating the
out previously, the shiftings are caused by lower stresses and the strains in reinforced concrete cross-
torsional stiffness of the truss model. sections, subject to coupled actions, having any geo-
The proposed method is applied to the analysis of metric and reinforcing characteristics, and to the
the structure shown in Fig. 10, reversing the dimen- inelastic analysis of the whole structure, by means of
sional ratio between the lengths (L,= 290 cm, suitable iterative algorithms. The numerical results,
L, = 457 cm) under the same geometric cross- obtained by means of a computer program, have
sections and material strength characteristics. proved the reliability and the versatility of the pro-
For the floor beam, the longitudinal steel areas posed procedure and the good agreement between
are As = 0.71 + 0.71 cm*, As = 5.06 + 5.06 cm*, theoretical and experimental analysis of typical
stirrups 0.31/19cm2/cm, for the spandrel structures.
As = 0.71 + 0.71 cm*, As = 4 x 1.98 cm*, stirrups The comparison of the results has likewise proved
0.31/11 cm*/cm (specimen S6). The diagrams in the weight of a correct prevision of the above inter-
Fig. 14 show the theoretical and experimental values action effects, not only at the cracking state but also
of the vertical displacement (-v of node 5) at various at the ultimate state.
 Inelastic analysis of reinforced concrete space frames 97

REFERJSNCES 6. A. E. McMullen and J. Wanvaruk, Concrete beams in

bending, torsion and shear. J. Sttuct. Engng, ASCE
1. G. M. Cocchi and F. Cappello, Sul calcolo delle ossa- May (1970).
ture spaziali in regime non lineare. Giomale de1 Genie 7. R. Park and T. Paulay, Reinforced Concrete Structures.
Civile 7, 8, 9 (1987). John Wiley (1979).
2. T. T. C. Hsu and Y. L. MO, Softening of concrete in 8. K. J. Bathe and A. P. Cimento, Some practical pro-
torsional members-theory and tests. ACZ JI, Paper cedures for the solution of non-linear finite element
No. 82-25, May/June (1985). equations. Comput. Meth. Appf. Meek Engng 22,59-85
3. D. Mitchell and M. P. Collins, Diagonal compression (1980).
field theory-a rational model for structural concrete in 9. G. M. Cocchi and F. Cappello, Instabilita’ laterale di
pure torsion. ACZ Jl, 71, 396-408 (1974). strutture reticolari prefabbricate in regime transitorio e
4. M. B. Leung and W. C. Schnobrich, Reinforced con- di esercizio. Atti de1 Congresso CTE sull’In-
crete beams subiected to bending and torsion. J Srruct. dustriahzzazione Edilizia, Firenze, 9-l 1, November
Eegsg,ASCE i13,No. 2 (1987). (1984).
5.P. Lampert and B. ~mlim~n, Ultimate strength and 10. M. P. Collins and P. Lampert, R~st~bution of
design of reinforced concrete beams in torsion and moments at cracking-the key to simpler torsion design.
bending. International Association for Bridge and AC1 Publication SP35, March (1971).
Structural En8ineering, 31-1, Zurich (1971).