MICHAEL D. KOTSOVOSt
I would like to congratulate the authors for an interesting and comprehensive design proposal. In general terms, I share their view with regard to the significant contribution that the use of strut-and-tie models may have towards the development of a unified design concept for structural concrete. However, I believe that only models which provide a realistic description of structural behavior may lead to efficient design procedures. The authors, on the other hand, appear to hold the view that it is sufficient for design purposes to use the theory of elasticity (adapted, if necessary, to conform with practical considerations) as a guide for developing strut-and-tie models which will provide an internal structural system sutiable for modeling any type of concrete structure. They consider that a structure under increasing load will adapt itself to the assumed internal structural system and that such a system can easily lead to an
* PCI JOURNAL, V. 32, No. 3, May-June 1987, pp. 74-150. ? Department of Civil Engineering, Imperial College of Science and Technology, London, England.
understanding of structural behavior. The basis of their argument appears to lie on the widely accepted view that the truss analogy provides a rational tool for the design ofreinforced concrete beams in flexure, shear and torsion. Their paper, therefore, describes an attempt to generalize the truss analogy in order to apply it in the form of strut-and-tie models to other structural members. However, the truss analogy does not appear to provide, in general, a realistic description of the causes of the observed behavior of structural concrete. There has been recent published experimental evidence which indicates that a reinforced concrete beam subjected to combined flexure and shear does not have to behave as a truss in order to attain its flexural capacity. 28 This is clearly indicated in Figs. A and B which show that, in contrast with predictions based on the truss analogy, the unconventionally reinforced concrete beams D, in Fig. C, and C and D, in Fig. D, exhibit strength and deformational characteristics similar to those of Beams B (also shown in Figs. C and D) which are reinforced in compliance with the truss model.
171
a 1.0
0 L
b 0.8
I-
0 0
0.6
a o 0.4 U 0
U U 0 0
0.2
10
15
20
25
0
0 0 x 0.8 1.0
0
0
v
I-
0.6
0 0.6 0 u
U
0.2
10
12
deflectionmm Fig. B. Load deflection curves of beams with aid = 3.3. 172
135
1
A
203 03/40
Li_H'
203 03/40
203 403/40
-I
60 I.-
500
1-
'^
x:300
400
, 11.6/30 B
206
306
01.6/30 t
01.6/30
160
-^ 500 ,
173
It could be argued, however, that although the unconventionally reinforced concrete beams could not have behaved as trusses, the above results cannot disprove the view that Beams B (which were designed in compliance with the truss anal-
ogy) behaved as such. The truss analogy forms the basis of current code provisions for shear design and it is generally expected that compliance with these provisions always leads to safe design solutions. And yet, there is published ex-
Z Y
o
goo
-j 600
400
/ 200 /
10
0 ,J 600
400
/
/ ---- ACt 318-83 Codes o experiment FE analysis
200
24
10
Horizontal Reinforcement (mm/m) Fig. E. Relation between shear capacity and amount of transverse reinforcement. 174
perimental evidence which indicates that this is not always the case.29'3 Fig. E indicates that code provisions for earthquake resistant design 31 predict a linear increase in shear capacity with increasing percentage of transverse reinforcement, from a lower level representing the contribution of concrete to shear resistance, to a "plateau" corresponding to "crushing" of the concrete struts of the truss model. The figure also includes experimental values obtained from tests on structural walls 29,30 and clearly indicates that, for certain percentages of tensile reinforcement, the code provisions overestimate considerably the wall capacity for the upper range of values of the percentage of transverse reinforcement. It may also be interesting to note in the figure that
the experimental results correlate very closely with results obtained by finite element analysis.32 The above results shed doubt on the validity of current concepts with regard to not only the ability of the truss analogy to realistically represent structural behavior, but also the view that a structure under increasing load will respond in the way prescribed by the physical model used to design it. Designing in compliance with the truss analogy appears to lead to overreinforced (in the transverse direction) structural members in which the presence of excess steel is likely to give rise to high secondary stresses within concrete for compatibility of deformation purposes. The development of such stresses in critical regions may be the cause of "prema-
- -. -
20
25
Specimen
Flange p fd
pv
web
ph 0.007 0.0025
A B
0.033 0.033
0.011 0.011
Fig. F. Effect of web reinforcement on the strength and deformational characteristics of structural walls. Note that pf,, p v , and Ph denote percentages of vertical flange, and vertical and horizontal web reinforcement, respectively.
PCI JOURNAL/November-December 1988 175
ture" failure similar to that exhibited by the walls with a high percentage of transverse reinforcement discussed above. The amount of transverse reinforcement, assessed on the basis of current code provisions, appears to be excessive even for the case where the use of the truss analogy yields safe design solutions. This is clearly indicated in Fig. F which shows that a reduction of the amount of web reinforcement by more than 60 percent has virtually no effect on the deformational and strength characteristics of the structural wall with geometry and tension reinforcement as indicated also in Fig. F. 33 Similar results have also been obtained from tests on both reinforced and prestressed concrete beams subjected to combined flexure and shear during an ongoing research program concerned with the development of design methods.34 It may be concluded from the above that structural concrete should not be expected, in general, to behave in the way it is modeled to behave. It would appear that the main reason for this lies in the fact that modeling is usually based on concepts which do not give full consideration
to important aspects of concrete behavior. For example, the theory of elasticity, proposed by the authors as a basis for the development of strut-and-tie models, cannot provide any detailed information with regard to the strength and deformation characteristics of concrete at the material level. Despite the vast amount of such information produced to date, only recently attempts have started to use it in design.3437 I believe, therefore, that design methods have reached the stage where no improvement is any more possible without attempting to implement in design information on the properties of concrete as a material. Such an attempt has already begun and led to the introduction of concepts which appear to provide a realistic description of the causes of the observed behavior of structural concrete. 33.37 These concepts are currently discussed in CEB Task Group 24 and research work 34 is in progress which explores possibilities of using them as a basis for the development of physical models which may be employed in design in a way similar to that stipulated by the authors.
REFERENCES
28. Kotsovos, M. D., "Shear Failure of RC Beams: A Reappraisal of Current Concepts," Contributions to a Joint Meeting on Fundamental Developments in Design Models, Organized by CEB Commissions II and IV, Karlsruhe, November 1986; CEB Bulletin d'Information No. 178/179, March 1987, pp. 103-111. 29. Cardenas, A. E.; Russell, H. G.; and Corley, W. G.; "Strength of Low Rise Structural Walls," Reinforced Concrete Subjected to Wind and Earthquake Forces, ACI SP-63, American Concrete Institute, Detroit, Michigan, 1980, pp. 221-241. 30. Maier, J., and Thttrliman, B., "Bruchversuche an Stahlbetonscheiben," Institut for Baustatic and Konstruction, Eifgenossishe Technische Hochschule Zurich, January 1985, 130 pp. 31. ACI Committee 318, "Building Code Re176
32.
33. 34.
35.
quirements for Reinforced Concrete (ACI 318-83)," American Concrete Institute, Detroit, Michigan, 1983, 111 pp. Lefas, 1. D., and Kotsovos, M. D., "Behaviour of RC Structural Walls A New Interpretation," Colloquium on Computational Mechanics of Reinforced Concrete, Delft, August 1987. Lefas, I. D., "Behaviour of Reinforced Concrete Structural Walls," PhD Thesis, University of London (in preparation). Kotsovos, M. D., "The Use of Fundamental Properties of Concrete for the Design of Prestressed Concrete Beams," SERC Research Grant, GR/E/07333, Civil Engineering Department, Imperial College, London. Kotsovos, M. D., "Consideration of Triaxial Stress Conditions in Design: A Necessity," ACI Structural Journal, V. 84, No. 3, May-June 1987, pp. 266-273.
36. Kotsovos, M. D. "Compressive Force Path Concept: Basis for Reinforced Concrete Ultimate Limit State Design," AC! Structural Journal, V. 85, No. 1, JanuaryFebruary 1988, pp. 68-75.
37. Kotsovos, M. D.; Bobrowski, J.; and Eibl, J.; "Behaviour of Reinforced Concrete T-Beams in Shear," The Structural Engineer, V. 65B, No. 1, March, 1987, pp. 1-10.
quate applications of the truss model were in fact the starting point for expanding it to a more general method. Some comments on individual paragraphs of the discussion may support what is said above: Beams D in Figs. C and D are not conventionally reinforced with stirrups in the shear span and therefore the regular truss model or corresponding code rules do not apply. A code which gives safe results for given requirements such as distributed stirrup reinforcement should not be blamed for showing wrong results if these requirements are not fulfilled. The more flexible strut-and-tie method, as contrasted with the rigid truss model, is capable of dealing with such irregularities as well. If the truss model is adjusted to the chosen reinforcement, the tested beams' capacity can be explained. A model shows in contrast to a thoughtless application of a code formula that shear forces can, for example, be transferred by an inclined strut over a considerable length of a beam without any stirrups. Furthermore, a second load path with concrete ties contributes to the shear capacity as shown for B-regions in Fig. 30 of our paper. However, the load bearing capacity depends also on the nodes joining the struts and the chords: compatibility stresses may cause a shear crack to penetrate into a node and separate there the chords from the web (see Fig. 34). The importance of a minimum stirrup reinforcement to prevent such separation is outlined on page 115 of the paper. The stirrups in the node region explain the difference between failure loads of Beams A and D.
177
Fig. G. Crack pattern of Specimen S2. Professor Kotsovos derives from comparison of test results and code provisions that the truss analogy gives unsafe results. First of all it must be stated, that the test panels 29.30 33 mentioned in the discussion
are D-regions for which the standard truss model normally does not apply. So it is wrong to argue on this basis. We investigated the three square panels30 using the method described in our paper. The predicted failure loads are not beyond 84 percent of the measured ultimate load for Panel Si and 90 percent for Panels S2 and S3. Hence, all predictions according to the strut-and-tie method are safe, contrary to Professor Kotsovos' implication. A numerical analysis for Panel S2 will be given here as an example. Fig. G shows the crack pattern of the test specimen and Fig. H gives an optimized, yet simple strut-and-tie-model. (Other models, especially the standard truss model, give more conservative predictions.) The nodes are not critical (smeared nodes in the panel). From the geometry of the model and the capacity of the ties T1 and T2, the ultimate load Fh and all the strut and tie forces can be analyzed consecutively from simple
826kN
826kNj C8
LII __ cg/
4% 12/T3 C5 C /I 32'
/ / T4
C7
T1
100J J'
1000
100
f 9 As1 = 57,4 kN/cm2. 5,5 cm2 T1 = T1 = 316kN T2 = T2 = 57,4.6,0= 344 kN C8 = T2 tan 41 = 299 kN T3 = C8 = 299 kN < T3i T3 = 57,4 . 6,0 = 344 kN T4 = Tl T3 cos45= 17kN <T4 Fh = C8 + (T4 + 826 kN) tan 32,5 0 = 836 kN C5 = T2 + 826 kN = 1170 kN -== C5, C5= 0,8f A, s = 0,8.36400.0,1.0,4= 1165 kN Cs = 1000 kN =C6= 0,8.36400.0,1 0,35 = 1027 kN C7 = 424 kN < C7 = 0,8. 36400.0,1.0,2 = 582 kN
The predicted load Fh = 836 kN corresponds to 90 percent of the measured ultimate load F = 928 kN. The two structures
compared in Fig. F of the discussion carry the shear force by two load paths similar to those drawn in Fig. H: A direct strut Cs and the strut-tie combination C 9 , T3 , C7. No horizontal reinforcement beyond the minimum reinforcement is necessary for the direct strut. If the struts are strong enough, the behavior of the beam is controlled by the chord reinforcement and vertical reinforcement and not much difference can be expected in their behavior if these reinforcements are the same as given in Fig. F. The authors believe that this discussion supports their view that strut-and-tie models (correctly applied) provide a rational and practical tool for the design of reinforced and prestressed concrete beams and that experimental evidence supports their view. It also shows that code rules may be misinterpreted if their basis is not clearly defined.
179