ere Title no. 81-38 ee eee Size Effect in Shear Failure of Longitudinally Reinforced Beams by Zdengk P, Bazant and Jin-Keun Kim Consequences of recent fracture mechonks studies af concrete for atyngcagonel shar fll of lonitacinally reinforced beats or ‘one slabs without sheer enforcement were studied. The rack {ng produced by shear was essumed fo propagate with diperse tone of microcracks atthe facie font. Dimensional anal of the ‘neray release rate then shows thatthe mona shear ste a fall- tre should not be a constant bu shoud very at (1 + W/8, AJ in nhich d= beam depts, 4, = maximum ogeregte size, and dy constant, For releively smal beans, representing the gre majority =f those tested in the leborateris, the nominal stress a failure ts seary constant; homever, or much deeper beams i considerably de ‘lines with inresing ie. Ths wend i confirmed by previous exe lnnema resus fn ation tothe sz ef, 2 ation Jorma for {he effet of sect ratio and shear span is derives. Compartons Sith exiting test data involving nary 300 tes indicate tht, compared 10 ‘he formulas inthe current bung codes, the coofcent of vale. ‘ion of deviations from the formule is rece to les thon one-half. eswordc: beams (upper) balding ends: ekg racing dine sional als tate rfrcedcoersey shee roman ‘ssc anal STATEMENT OF PROBLEM Predicting brittle failures of concrete structures due to tensile cracking of concrete is much more difficult than predicting ductile failures. The diagonal shear failure of longitudinally reinforced beams and one-way slabs without shear reinforcement is a good example. Although great progress has been achieved in various theoretical studies'* and extensive experimentation,!*3* the scatter of the deviations of test results from the formulas in current building codes is enormous (see Fig. 7.13 in Reference 29) and is much larger than the scatter of tensile strength or fracture energy. ‘The current formulas are based on the concept of ten- sile strength. However, this concept is theoretically jus- tified only in the case of ductile failures governed by the theory of plasticity. For failures in which the stress decreases after reaching the strength limit, as is the ease for tensile cracking, the sirength concept is inconsistent when applied in a continuum analysis. For example, fi nite element analysis of cracking based on the strength criterion can exhibit a strong spurious dependence on the choice of mesh size.*™” As is well known from frac- ture mechanics, a theoretically consistent approach ‘must be based on an energy criterion of failure, 456 Having realized this fact, Reinhardt” recently ana- lyzed certain test data for diagonal shear failure on the basis of the classical (linear elastic) fracture mechanics and found a relatively good agreement with these test data. However, when some other available data are ‘considered, the agreement with the linear elastic frac- ture mechanics is not very good and does not seem ‘much better than for the strength criterion. This is not surprising, since the linear elastic fracture mechanics has been found to be inapplicable to concrete, a8 dem- “onstrated, e.g., by Naus aad Kesler. ‘The last few years, however, witnessed an increased intérest in basic studies of concrete fracture, and a new form of fracture mechanics which appears applicable to concrete has emerged. This new approach does not treat fracture as a point phenomenon, but recognizes that in brittle heterogeneous materials such as concrete the fracture propagates with a relatively large fracture process zone in which progressive microcracking grad- ually reduces the tensile stress to zero.” The aim of this paper is to explore the consequences of this new, nonlinear fracture mechanics for diagonal shear fai ure. The main purpose of fracture mechanics is the prediction of the effect of structure size, and it will be seen that a considerable improvement can be brought about in this regard. STRUCTURAL SIZE EFFECT IN FRACTURE The structural size effect may be illuminated by con- sidering structures of different sizes but geometrically similar shapes, e.g., beams of the same steel ratio and the same ratio of depth to shear span. The strength cri- terion may be stated as oy sile stength of concrete and o, ure. For reasons of dimensionality, oy = ¢P/bd where P = given load, d = characteristic dimension of the structure, e.g, the depth-of beam, b = thickness, and y= constant. Thus, if one considers the plot of log ey “Rey. er mon Race crac ae sc, ‘Born erinesageuean il be publ eS he Falgrauqun 1963 ACL ACI JOURNAL | September-October 1984:‘Zea P. Batam FACI a pofeser and deco, Cen Jar Conere fend Geomatrit, Nortbweom Unveny. Dr, Bats ese sr Ire eng, seis os conan Yo Argonne Nana Lateran et = ra othe fr ed ton eit Bnd 0 ir Jara Fe sre at Inav of BILEM Commitee TCD on cree, af ASCEEMD Commie or ‘reper of Matera ond IA SHIRT Dion H. Ha wor on cote sd ‘sonatas inate hair, fc and aby hve bee eed by {TRILEN madal, ASCE Heber Prize and TY. Ln Anat 1100 Asad ‘Gaggntuin Fallows, Ford Famdation Flow, ond election a Palos of Americon Academy of Mechanic “ACI meer Dan Kn ea gaduase roach tnt and PRD cndite in cat ens at Nerthertom Unis. He aiid hi BS ond MS “dezresin the Deprmet a Arete fem See! National Univer. He fas worked a techie aston Stn! Neon! Unies on ae ter et Uison tee of Teno, Keres anda actual deer Biz ‘search iret cadets Dhar an trea ene ed ‘orc corre 5 well a ares retool versus log d (Fig. 1), the locus of all failure states is a horizontal line, epardles of whether one uss elastic, plastic, or some other strength-based analysis. The only Aiference between these types of analysis is the level at which the horizontal line i 0 be drawn. Fig. 1 shows examples for some elementary situations, such as bend- ing, shear, and torsion of unrenforced beams. For linear elastic fracture mechanics, the size effect completely different, As is well known,"* oy varies wersely asd, 0 the plot of log a, versus log dis 8 straight line of slope - 14; see Fig. 1 However, except for extremely large structure sizes, this size effect is senerally too strong for concrete structures, a8 the sub sequent analysis will confirm, Due to the dspersed nature and progresive devel- opment of cracking in concreic, the structural sie ef- feet may be described as” @ 400. 60)= FE = 7) O in which d, is the maximum ageregate size, 2 is a rela- tive structure size, and 2, is an empirical constant. The foregoing equation may be derived for various simple situations and may be also deduced in general by di ‘mensional analysis based on a hypothesis charact ing the dispersed nature of cracking™” (See Appendix). For structures of a small size relative to the size of ‘aggregate, i.c., for small A, the value of ein Ea. (1) may be neglected in comparison to unity, and then we. have 4) = Land oy = f/ = constant. This indicates ‘that the strength criterion (horizontal fine in correct for small size structures, which happens to be the case for most structures tested in laboratories. For structures of a very large size compared to the size of aggregate, 1 may be neglected in comparison to 2/2 in Eq. (1), and then ay = FQN)", which is the size ef- fect of linear elastic fracture mechanics described by the inclined straight line in Fig. 1. Obviously, Ea. (1) rep- resents a gradual transition from the strength criterion for small structures to the linear elastic fracture me- chanics for very large structures. For \ < hy the strength criterion dominates, and for > y the frac- ture mechanics aspect of failure dominates. AGI JOURNAL 1 September-October 1964 strength. or yield ertarton ot ocating tts rnonilogor fracture mechanics Nominal Streee at Pasture, tog fn) log (eine a) Fig, 1 = Mlustration of sie effect according (0 various theories EFFECT OF STEEL RATIO AND SHEAR SPAN Eq. (1) could now be combined with some existing formula for the diagonal shear failure and compared to test data. However, the huge scatter of the data is ‘ot ofily to the size effect, but elso to the manne’ in Which other factors, such as the steel ratio p and the shear span a (Fig. 2) are taken into account. Thus, to derive full benefit from a better formulation for the size effect, we should also try to improve the analysis for other influences. For that purpose we will try to use some rational, albeit crude and simplified, arguments, Consider the end segment of a beam shown in Fig. 2{a), in which a constant shear force V acts throughout the shear span a, In general, the shear span may be de- fined asa = A/V where ¥ = shear force, and M bending moment in the same cross section. The bend- ing moment at any distance x from the support may be expressed as M = Tid where d = depth of the beam, T = Ty) = tensile force resultant acting at the cen troid of longitudinal reinforcement, and J = f(x) variable coefficient. The shear force may then be ex- pressed as V = dM/dx, and the derivative of the prod- luct may be written as a sum of two terms. aq, a Vane My ei = Tt @ ‘As known from various preceding studies,” component V; is due to a ccinposite beam action and arises from the transmission of a tensile force into the steel bars by ‘means of bond stresses, and component V; represents what is known as arch action, sincs it arises from an arch-like variation in the height of the location of the compressive resultant C = T. 3 457.G4 ce al ia. te Fig. 2 — Notation for the analysis of diagonal shear We need a simple description of the function /(x), ‘and choose for this purpose @ in which jy is a constant, defining the location of the ‘compression resultant C at the end of the shear span, x = a. According to the classical bending theory of rein- forced concrete beams with only tensile reinforcement and with a negligible tensile capacity of concrete, we would have h ‘le/d), cfd = (arp? + 2np)% — np (4) in which p = steel ratio, n = E/E, = ratio of elastic ‘moduli of steel and concrete, and c = depth to neutral axis at x = @. Eq. @) is, however, unnecessarily com- plicated and may be replaced by the following simpler expression d= hem o in which & and m are certain constants. These con- stants can always be chosen so that the values given by Eq, (4) and (5) are almost undistinguishable; see Fig. 3. 458 en 2 Fig. 3 — Comparisons of two formulas for the effect of steel ratio on the arm of internal forces For the composite beam action contribution, we may xpress the rate of change ofthe tensile force Tin terms of the bond stress transmitted from concrete t0 the steel bars, i.e., dT/dx = ¢, ( E Dyut,), in which c= some constants, and ED, is the sum of the diameters (ofall bars in the eross section. This relation expresses the equilibrium condition of a unit segment of the see ‘bar in the longitudinal direction. If the concrete cross section is kept the same and the amount of reinforce. ment is varied, E D, is proportional to ve. Further- ‘more, the ultimate bond stress is roughly proportional to fi where q = 05, 30 we have dT/dx = cub J". in Which c, is some constant. Substituting into Eq. 2), wwe thus get - Y= kermpeod o To express the arch action contribution to shear, we may set T = 0, pbd, and substituting into Eq. (2) we obtain he "ane oo Eq. (7) then reduces to @ t= (2) ota ® Considering the steel stress 9, as constant, and substi- tuting Eq. (5), we further obtain o wa wag bd o which ¢, is some constant, Finally, summing the contributions from composite beam action anid arch action, V =, + Vs, and caleu- ACI JOURNAL / September-October 1984lating the nominal shear stress at failure as v = V/bd, we acquire the formula Kor (# +h Or 9) (ray, in which ky = oy/k;. This formula is similar to that used in the ACI Code,* but itis to a greater extent based on mechanics analysis and contains more empirical pa- rameters to be found from test results; these parame. ters are ki, ky, P, 9, andr. ‘The foregoing analysis did not take into account the size effect appropriate for brittle failures due to con- ‘rete cracking. According to Eq. (1), we should there- fore multiply the nominal shear stress at failure by function 40) ensuing from dimensional analysis of the ‘energy release by fracture. Thus, we finally obtain the ae reno (see) (oe ZY" on in which X, represents an additional empirical parame- ter. STATISTICAL ANALYSIS OF EXISTING TEST DATA Shear failure of beams is one problem for which ex- tensive statistical information has been accumulated over the years. This statistical basis was exploited by Zsutty"” for the development of a very simple predic~ tion formula, the best one proposed up to now. His statistical analysis, however, did not particularly cover the size effect and was made before some of the impor- tant test results on the size effect became available. Eq. (11) proposed here has been compared to essentially all important experimental evidence, both that with regard to the effect of steel ratio, shear span, and concrete strength, and the more limited one with regerd to the effect of sine (beam depth), The text data used included those of Moody et al.,” Diaz de Cossio et al Mathey,” Van den Berg,” Taylor,™ Rajagopalan,” Kani, Leonhardt and Walther," Walraven,” Tay- Jor, Risch et al.,* Bresler and Scordelis,® Krefeld and Thurston," Bhal,’ Mattock," Placas and Regan,” and ‘Swamy and Qureshi. First consider the size effect, ic., the effect of d/d, at constant @/d, p, and f The necessary data, requir- ing tests of geometrically similar beams of different depths, have been generated by Kani, Leonhardt and ‘Walther, Walraven, Taylor, Risch, Bhal, and Swamy. For geometrically similar beams, Eq. (11) may be writ- ten as (in Fig. $, C, appears as C) (12) in which C, is a constant. This may also be rearranged to the following linearized form @ td ili ig ent a stated © ‘The Intter of these equations yields C,-?as the vertical axis intercept, and C,-#/2, as the slope of the regres- sion line. Data that exist for one particular concrete can be easily and very closely fitted with Eq. (13), as ex ‘emplified in Fig. 4 for Walraven’s and Kani’s test re- sults. This plot clearly indicates that the sizeindepen- dent strength criterion, currently implied in code for- smulations, contradicts experimental evidence. So docs the linear fracture mechanics, which corresponds in Fig. % to the straight line of slope - 4. ‘None of the available data for one particular con- ‘rete and fixed p and a/d are, however, sufficiently ex- tensive to allow statistical analysis. For that purpose, all the aforementioned available data for the size effect must be analyzed collectively. This cannot be done by linear regression alone, since coefficients he Ds q, andr of Eq. (11) are involved nonlinearly. Trial and error approach coupled with nonlinear optimization (Mar- quardt-Levenberg algorithm)" has been used to de- termine the optimal values of these coefficients. The optimum fit of the data, achieved with Eq. (It), is shown in Fig. 5(@) as the linearized regression plot [Ea. (wormen ae Ase Fetes on (©) Ken top (@/a) Fig. 4 — Comparison of Eq. (11) to Walraven’s and Kani’s test results ACI JOURNAL / Soptember-October 1984Tete (ne one 210 VF 63000677) tag (4/8) Fig. 5 — Comparison with existing test data for beams of different sizes (13)), and in Fig. 5(b) as the plot of the logarithm of the nominal shear strength versus the logarithm of rel- ative size d/d,, Due to combining test data for differ- ent concretes, different beams, and from different I ‘oratories, the scatter is now much larger; however, the sie effect is clearly confirmed, The strength criterion, implied in the current design approach, would corre: spond to a horizontal line in both Fig. 5(a) and 5(), which would obviously contradict the test data, On the ‘other hand, a straight line of slope ~ 4 in Fig. 5(b) would give too strong a size effec, in clear disagree iment with test results. ‘Note also that if the range of beam sizes were re- duced to about one-half, no clear size effect would be ‘apparent and the strength criterion would represent the data for the smaller beams as well as the present the- ory. This accentuates the need of testing beams of widely different sizes if any evidence on the size effect, should be obtained. Fig. 5(b) also gradual transition from a strength criterion that applies for small beams to a linear fracture mechanics criterion ‘that would no doubt be applicable to very large beams. To obtain more extensive evidence, and also to ver- ify the dependence on p, a/d, and f?, numerous further data were inchided, although each of them, taken alone, provides no information on the size effect as such, In their majority, these further data correspond to the smallest beams admissible for a given agsregate, and thus they cannot be expected to strengthen the evi- ence on the size effect. The values of all six coeffi- cients in Eq. (11) have been optimized with regard to this complete statistical evidence, and the results have ‘been plotted as the measured value of nominal shear strength v, versus the value given by the optimized for- mula (Eq. (DJ. If the formula were perfect, the plot would have to be a straight line of Slope 1, passing through the origin, Thus, the deviations from & straight 460 line plot represent the errors. A plot of this type is shown, for the proposed Eq.11), in Fig. 6(@. The standard deviation of the vertical erors with regard £0 the regression line, andthe corelation soctient, ae listed in the ig 1 should be menloned thatthe vale ues of f were considered in psi (1 psi = 6895 Pa), and that the cubic strength of concrete f..(/,.is in psi), when ‘Sadiated, was converted to the linia strenathac- cording tthe formula f= [0.76 40.2 198 Us! 28403) f® ‘Asa result of all thse statistical comparisons, the following formuta is proposed for the mean ultimate nominal shear strenath IFT + 3000 ve/(a/d)") (14) For the sake of comparison, the formulas used in ACI Standard 318-77° and in the CEB-FIP Model Code 1978 have also been used to fit these data. How ever, the coefficients of these formulas had to be dis- regarded because they are not intended to give the mean values of the ultimate nominal shear strength but the values for the initiation of cracking, These formulas may be written as v= Min(ks VP + bso XH, 3.5.72) cach ca reel + Kye) (CEBFIP) (16) in which f! and v for ACK are in psi, v and ry for CEB-FIP are in MPa and fora < 2d AGI JOURNAL 1 September-October 1984oan as / Vf s05E H! scsee arorooe's tome noise VERE Nominal Ultimate Shear Steve Measured «tog vy. (0s Sa'PHt) Roménal Uiimate Shear Stress Calcul cog mes (ein pet) Fig. 6 — Comparison of various formulas with the bulk of the existing test data on ‘the ultimate shear strength 0.01 fy + 0.06 for fy < 20 0.008 fig + 0.1 for fu: > 20 = Max(1.6~4, 1) p= Min(A,/bd, 0.02) for CEB-FIP an In Eq, (17), d is in meters, but fin the present form must be given in MPa. Coefficients ky and &, in these Formulas have been optimized to obtain the best fit of all the 377 data points. The resulting optimum fits are shown for the ACI and the CEB-FIP formulas in Fig. 6(a) and (b), and the values of the optimum coeffi- cients are listed for these formulas as well as the pres- cent model in Table 1. The scatter apparent from these figures and quantified by the values of standard devis- tion and the correlation coefficient in these figures is obviously much larger than the seatter for the pro- posed formula, especially in the ease of CEB-FIP for- mula, It must be kept in mind, however, that these for- ‘mulas are not intended to give the ultimate strength but the initiation of diagonal shear cracking (as loosely as it may be defined). The best previous formula is doubtless that of Zsutty,™ which reads a a) wa” ACI JOURNAL 1 September-October 1984 Table 1 — Coefficients obtained by nor regression for ¥, Nome of] Model lable | @ be neue | a7 | 1.6efra ‘Gare | Wh | esi} zouny {e1dp2$| 26 |s84 loss | 03s. lass erdzas| “at [aio s4| ~0:0050|1'50 Proposed model | 296 _| 723]322¢/2s[029 | 02 [ps1 in which ky, r, p, and q are four empirical constants. ‘The values of these constants have been optimized again to obtain the best possible fit of the 377 data points used in Fig. 6, The resulting plot is shown in Fig. 6(c). It may be noted that this formula agrees with the data nearly as well as the proposed formula. However, the size effect evidenced in Fig. 5 is not modeled by this formula, although it could be introduced by multiply- {ng the formula with the function (4). Note also that, in contrast to the present formula, Zsutty’s formula purely empirical, not based on some mechanics analy- is. ‘As another useful statistic, one may consider the population of the values of ¥ = (Vix/¥i) ~ 1, in which ‘ais the calculated value and »,,, is the measured value of nominal shear strength. These values are plotted as 4614 function of the logarithm of relative size in Fig. 7. ‘The proposed formula (Fig. 7(d)] appears best. Zsut- ty's is nearly a5 good, although it does not reflect the declining trend of the data as a function of d/d,, ap- parent from Fig. 7(c) as well as Fig. 7(d). The scatter for the ACI formula (Fig. 7(a)] is much larger, and even more so for the CEB-FIP formula [Fig. 7(b)]. De- spite this larger scatter, a declining trend with regard to the size is noticeable in Fig. 7(2) and (6). DESIGN PHILOSOPHY: CRACK INITIATION OR FAILURE? ‘The philosophy of the present design codes is to achieve a certain safety not against the ultimate load in diagonal shear failure but against the load for which the initiation of diagonal shear cracks is observed, ‘Therefore, comparisons have also been made with the (2wuch less numerous) available test data on the nomi- nal cracking shear stress v.. Fig. 8 shows such a com- parison in terms of the measured v, against the calcu- lated ¥,, a5 obtained by optimizing the coefficient the formula in Eq. (11), (15), (16), (17), and (18). (Note that all the existing test data for the initiation of diag- onal shear cracking are confined to the values of d/d, between 10 and 24.) In Fig. 8 the proposed type of for- mula again gives the best agreement with test data, al- though only slightly better than Zsutty’s formula. The improvement compared to the ACI formula is, in terms of ¥, not very significant, but it is more significant ‘compared to the CEB-FIP formula. The values of the ‘optimized coefficients used in plotting Fig. 8 are listed in Table 2. ‘The statistics for the nominal cracking shear stress ¥, may also be worked out for the variable Y = (%/Ya) = 1. These values are plotted for all the four formulas against the logarithm of the relative size d/, in ig. 9. Itis apparent that, for crack initiation, the size effect is ‘much lee pronounced, nearly undetectable. In fact, for the «rue crack initation, the size effect should be non- ‘existent, since the beams do not contain any initial stress concentrator (a notch). The fact that any size ef- fect seems (0 be apparent indicates that the observed Values of ¥, did notin fact correspond to the true crack initiation, which, of course, is very difficult to define as the cracking begins by a gradual formation of invisible microcracks ‘The fact that no significant size effect on the nomi- nal shear stress at cracking is observed while at the same time the size effect is clearly confirmed by tess of the ultimate nominal shear strength raises @ question with regard to the present design philosophy of design. {ng against crack initiation rather than ultimate fallure Since the ultimate value of y decreases with size, there ‘obviously exists a certain sufficiently large size for which the ultimate » ceases to be larger than the value of » for crack initiation. This is, of course, natural to expect. It is known from nonlinear fracture mechanies that the strength reserve due to stable erack growth be- comes smaller as the size increases and vanishes when a certain size is exceeded, ‘Thus, designing against the crack initiation rather than ultimate failure does not assure a uniform safety margin, The safety margin decreases with increasing /da, and for a sufficiently large d/d, the safety margin or cose veranpl soem tog (00/0) wae cot e456) “Verso MEEpaa for a/az28 * coiso Vip @aF tor 0/425 co10 Ye OR +9000vp707a") .* og (4/44) Fig. 7 — Same data as in Fig. 6 compared with formulas in a different manner ACI JOURNAL / September-October 1984ce tes (ost wean woxwny Nominal Crashing a as as ip sass Nominal Cracking Shear Streee Colrtated log vy, (van pet) Fig. 8 — Comparison of various optimized formulas to existtig data on the initiation of diagonal shear cracking rust become completely wiped out. Therefore, in the writers’ opinion, the design approach in the codes should be changed to one based on safety against ulti- mate failure, or even better, a combination of both eti- teria, ‘The aforementioned insufficient safety margin for very large beam sizes cannot be, however, detected from the available experimental evidence. For the di- agonal shear failure, and even more $0 for other types of brittle failure suck as punching shear, torsion, etc., the available experimental evidence is confined, in its majority, to structures of about the smallest size that is admissible with the given aggregate. Actual structures are, however, for the most past, of a much larger size. ‘The extrapolation from tests on small size structures to much larger actual structures is, of course, the main purpose of fracture mechanics, since testing of very large structures would hardly be economically feasible. PROPOSED DESIGN FORMULA In view of the shortcomings of a design based on the shear stress at crack initiation, a design formula should bbe obtained by appropriately scaling down the formula for the mean ultimate strength (Eq. (14)]. The scaling down should be such that for smaller beam sizes the safety margin should $s about the same as for the pres- ‘ent ACI formula. To get an idea of the safety margin, ‘we plot in Fig, 10(a) all 296 data points in comparison, ACI JOURNAL J September-October 1984 Table 2 — Coefficients obtained by nonlinear sgression for v, : ot Eouty 0.29 Proposed mods! 12k to the ACI formula; the ordinates represent the mea- sured values of v, at the ultimate state, and the abscis- sas the values predicted by the ACI formula {Eq. (15). Perfect agreement would correspond in Fig. 10 to a straight line of Slope 1. The huge scatter in Fig. 10(a) ig due to the fact that a comparison to ultimate load data is now being made for a formula originally devel- ‘oped for crack initiation rather than failure. An even larger scatter is seen in a similar plot [Fig. 10(b)] based ‘on Eq. (16) for the CEB-FIP formula. From Fig. 10(a) and 10(b) we see that the present design formulas pass near the lower limit of the existing data points, al- though some data points lic, disturbingly, well below the formula {Fig. 10(a),(b)), especially for the ACI for- mula [Fig. 10(a)]. ‘Various scaling factor values have been tried with Eq. (14) to determine the value for which only a few data points would lie below the formula, and only slightly below it. This led to the following design for- 463Tira “ yy caf “Hf t it (0) ae (o) cere ~ i fal | i ee nif see were yen i :” PRoeOSCO FORMA, i Tey zsu : Py i z | . i | J SEER stg se Nomisat (imate Shear Strst Colulated by Mele Fig. 10 — Comparison of the design formulas with the bulk of existing data ACI JOURNAL / September-October 1984ssh is pope 2 ena fre ans aces ~ BES (ur + 0 J) ai a)" 1 25d, va witha = as) Here a = a/d for the case of concentrated load (Fig. 2), and a = ¢/4d for that of uniform load (here fmust be in psi. This formula is shown by the straight line in Fig. 10(¢). Unlike Fig. 10(2), no data points fall signif- ‘cantly below this formula. Note that the band of data points based on this formula becomes much narrower than for the preseat code formulations TFig. 10(a),(b)) and is also somewhat narrower than for Zsutty’s for- ‘mula shown in Fig. 10(¢), This formula was also scaled (replacing &, in Eq. (18) by ky = 0.75k,) so that only @ few data points would lie below the formula, ‘The fact that no data points in Fig. 10(4) for the present formulation lie high above the straight inclined line means that the proposed formula is overall eco- nomic. The economy may be quantitatively character- ined by factor $, = E,A/ny where ¥ ~ ordinate of data centroid, A, = vertical deviation of data points from the straight line (/ = 1, 2,... m), anda = number of all points. Only the points lying above the straight line are counted. The smaller is ¢., the better the economy. Calculations yield 4, = 0.595, 0.844, 0.448, and 0.290 for Fig. 10(a), (b), (€), and (), respectively. The num- bers of points that lie above the inclined straight line are n = 249, 289, 286, and 286, respectively. Fig. 11 shows the same comparisons as Fig. 10, but in different variables. The ordinates are, similarly to Fig. 9, the values of ¥ = (Yu4/va) = 1. The compari- sons lead to similar conchusions as Fig. 10. (Note that only 25 data points of the 77 data points of Kani” are plotted in Fig. 7 and 11, although all data points agree with the formula well this is because they all refer to ‘the same beam size and are all crowded in such a small spot that they could not be graphically distinguished.) From Fig. 11(a) we see that the points laying signifi- cantly below the ACI formula correspond indeed to large beam sizes, which again confirms our previous argument about the size effect and how it affects the safety margin. REMARK ON THE EFFECT OF SHEAR REINFORCEMENT is certainly a reasonable design approach to as- sume, a8 is done in current codes, thatthe ukimate load {in presence of shear reinforcement (stirrup) is a sum of the ultimate load in absence of the shear reinforcement plus the addtional capacity due to the shear reinfrce- ment alone, obtained by plastic analysis. This ap- proach, however, is not as easy as it might seem, since the presence of shear reinforcement mitigates the size effect, as shown in Reference 34 by a similar dimen- sional. analysis. No meaningful experimental evidence Seems to be available for the size effect in presence of shear reinforcement. Iti, nevertheless, theoretically evident thatthe reduction inthe loss of safety maggin or cnn svaseno'i east weer 7 109 (/0) .# ia ’ “coer ESTAR tr afacas enefe0 VEEN ar aves tos (4/4) Fig. 11 — Same comparison as in Fig. ACI JOURNAL / September-October 1984 - 10, plotted in a different variableWith the increasing size, which has been demonstrated here for the presently used strength-based formulas for the cracking nominal shear strength, may be consider- ably milder or even insignificant when shear reinforce- ‘ment is present. This question is of great interest for very large structures and cals for further investigation, IMPLICATIONS FOR PLASTICITY ANALYSIS. Recently, in an effort to replace purely empirical de- sign formulas by some rational mechanics analysis, it hhas become popular to apply plastic limit analysis not only to ductile failures of concrete structures due to yielding of reinforcement but also to brittle failures ue to failure of concrete. The results of the present study shed further doubts on this approach. Plasticity yields an incorrect size effect for the britle shear fail- ure of beams and is likely to do the same for other brite failures such as torsion of beams or punching shear of slabs. This is, of course, not surprising since the stress-strain relation of concrete has no yield pla- teau and exhibits strain softening, which causes that the limit stress state cannot exist simultaneously along some postulated failure surface, as required by plastic limit analysis, but is reached successively at various points of the failure surface, In some recent investigations it was concluded chat plastic limit analysis “‘works” for punching shear fail- tures of slabs. However, to make it “work,” the tensile strength had to be considered to be 1/200 of the com pressive strength. This is about 20-imes less than the actual tensile strength. How can it be so low? A likely ‘explanation is the fracture mechanics size effect, and if this is so it means that plastic limit analysis does not work, Various apparent successes of plastic limit anal- ysis of brittle failures of concrete structures, recently presented in the literature, are likely due to the fact that really large structures have never been tested and the size effect has never been checked, With the exception of conditions of very high hydro- static pressure and of structures that fail primarily due to yielding of reinforcement, plasticity is not the cor- rect theory for concrete. Fracture mechanics, of the proper type, is. IMPLICATIONS FOR FINITE ELEMENT ANALYSIS OF CRACKING Finite element analysis based on the strength crite rion does not yield any size effect, ie., it corresponds oa horizontal line in Fig. { or 5(b) when failure loads for geometrically similar structures, obtained with sim= ilar meshes, are compared. The fact that the experi ‘mental results in Fie. 5(b) disagree with the horizontal fine means that this type of finite element analysis can- not be used for what is its main purpose, that is, to predict the failure of a real size structure after caibrat- ing the finite element code by means of laboratory-size tests. Therefore, fracture mechanics-type cracking eri- teria must be used for this purpose in finite element codes. Not, of course, linear elastic fracture mechan- 466 ies, because such finite element anlaysis would give the size effect according to the straight line of downward slope —¥4 in Fig. 1 or 5(b), but nonlinear fracture me- chanies. A finite element model of this type, which is suitable for large-scale computation and leads to the curved transition diagram shown in Fig. 1 or 5(b), has been presented in References 35 and 33. ‘CONCLUSIONS 1. For diagonal shear failure of reinforced beams and one-way slabs without shear reinforcement, i is appro- priate to consider the size effect which theoretically re- sults from a dimensional analysis of the energy release rate in the propagation of fractures that have a dis- persed cracking zone at their front. 2. A size effect of this type agrees with available test data far better than either the lack of size effect corre- sponding (0 the strength criterion or the size effect of Jinear elastic fracture mechanics. 3. AS a function of the ratio of beam depth d to maximum aggregate size d,, the nominal shear strength exhibits a gradual transition from the strength criterion (which prevails for d/d, < 25) to an energy criterion for fracture (which prevails for d/d, > 25). For ex- tremely large beam depths, the size effect of linear elastic fracture mechanics is approached asympto cally. Most of the existing test data are confined to the range for which the strength criterion dominates. 4. The present practice of designing against the inti ation of diagonal shear cracks rather than ultimate failure does not yield a uniform safety margin when different beam sizes are considered, and the safety margin becomes completely wiped out for a sufficiently large size. From the viewpoint of the size effect, only a design Tormuta based on the ultimate failure load [EQ, (19)] can provide a uniform safety margin against cat- astrophic failure for all structure sizes. 5. A rational, mechanies-based formula for the ef fect of steel ratio and relative shear span can be ob- tained by superimposing the shear forces transmitted by composite beam action and by arch action. This yields a formula which is similar to that presently used in ACI code but isin better agreement with test data, ACKNOWLEDGMENT Some ofthe theoretical facture dies relevant to the present work hve been supported by the National Science Foundation under Gratt No, CEE-E305148 10 Novthwestern University. Thaaks ate due «2 Mary Hil for her invaluable secretariat! asstace, REFERENCES |. ACLASCE Committe 326, "Shear and Diagonal Tension, Pars, 1 and 2°" ACL Journat, Proceedings V. 9, NO, 1 J30. 1882, pp. and No.2, Feb 1952, pp. 277-334, “Shear, Torsion and Punching,” Bullen "Information No, 146, Comite Buro-tnterational dy Biton, Pai, Jan. 1982, 25 pp. 3 Fenwick, RC. and Paulsy, Thomas, “Mechanics of Shear Re sistance of Concrete Beams,” Proceedings, ASCE, V. 54, STIO, Ox 1968, py, 2325-2350, 4. Henan, 0. nd Losberg, A. ‘Structural Desien of Concrete Consrucion Based on Shear Faire,” Nord Retong (Stockholm), No. 195, pp. 19.29, Gin Swedish) ACI JOURNAL / September-October 19845. Kan, G.N. J, "The Ride of Shear Fai ‘AGI Journat, Proceedings V. 61, No.4, Ape. 1964, pp. 441-67. 6. MasGrefor, J. G., and Gergely, P, "Suggested Revisions to ‘ACI Building Code Clauses Dealing with Sheat in Beams," ACL ‘oumnat, ProcedingsV.74, No 10, Ot. 197, po. 453-00 7. Relahardt, H. W., *Masstabscinfos be Sehabverschen im Lh der Bruchmechan," Beton und SteMbetonbau (Ben), No 1, 1981, pp. 19-21, . Teylor, Howard P, J. “The Shear Serength of Large Beams,” Proceedings, ASCE, V. 98, STH, Nov, 1972, pp. 2473-2490, 9, Bhal, N.S, "Uber dea Eifuss der Balicnboke auf Sehubtra- ‘ihigheit von cnfeldrgen Stabetonbalken mit ued ofne Sehub- Dbevehrung,” disertation, Univers Stuart, 1968, 124. 10. ester, Hors, and Seordalis, AC., "Shear Steag of Rein forcea Concrete Beams,” ACI Jouns, Proceedings V. 6, Jan, 1963, pp. 51-54 U1. Forel, C., “Tests of Shear Strength and Shear Relaforce- rent of Concrete Beams (Schubesigkelt und Schubbewebrune der Beionbalken),” Handlingar No. 78, Kungliga Tekniska Hoasholans, Stockholm, 1984, pp. 31.32, 12. Kani, G. N. J, “ase acts Concerning Shea File,” ACL ounnat, Proceedings V. 63, No. 6 June 1956, pp. 675-82. 13. Kanl, G.1N. J. “How Safe Are Our Latge Relaforced Con- ‘rte Beams?" ACI Jounnat, Procadings V6, NO. 3, Mar. 1967, pp. 128-41 14 Koch, R., and Rostasy, FS, “Sshubtraefihigket von Platten as Stahileichtbeton obme Schubbewehrung,” Beton: wad Sihlbe- ‘onbau (Bedi), ¥. 73,1978, pp. 2-46. 15, Krefeld, Wil J., and Thursion, Charles W., “Studies ofthe ‘Shear ad Diagonal Teasion Surength of Simply Supported Rein- forced Concrete Beams," ACI Jousat, Proceedings. 63, NO. 4, Apr 1966, p. 41-476. 16. Leonhardt, F. and Waliter, R., “Bekrige sur Behandlung der Schubprobiome im StabTbetonbau Beton- und Stehbetonba (Bet- Ui), V. $6, No, 12, 1961; V. 51, No.2, 3, 6 7 8, 19525 and V. $8, No. Sand, 1963. 17 Mathey, Robert ., and Watsein, Davi, “Shear Strength of Beams Without Web Reaforcment,” ACL JOURSAL, Proceed V. (0, No.2, Feb, 1963, pp 15-208, Mattock, Alan H., “Diagonal Tesion Cracking in Concrete earns with Axial Fores." Proceedings, ASCE, V. 95, ST, Sept. 1968, pp. 1887-1900. 19. Moody, K. G.: Vist, LMU: Ener, B.C. an Hopnested,E. “shear Stengih of Heaferced Conrete Beams, Parts 1 and 2," ACL JOURNAL, Proceedings V. SI, No.4, Dee. 1984 5, Jan. 195, pp. 417-46, 20, Ojia, 8. K., "The Sexe Suength of Uniformly Loaded Beams Without Web Reinforcement.” Magazine of Concrete Research London), ¥. 25, No. 75, une 1971, pp. {118 21, Plaas, Alexander, and Regan, Paul E., “Seat Paice of Reinfored Concrete Beams,” ACI ouRNAL, Proceedings V. 68, No. 10, Oct. 1971, pp. 78-73. 22. Rajagopalan, K. 8., and Fesguson, Phil M. “Exploratory Shear Tests Emphasizing Percentage of Longtuial Steel." ACI Journal, Proceedings V. 68, No.8, AUg. 1968, pp. 634-63. 23 Disa de Cosi, Roger, and Sies, Cheer P., “Wehavoe and. ‘Strength in Shear of Beams and Frames Without Web Reinforce: sent," ACL JOURNAL, Proceedings V. 56, No.8, Feb. 196, 9p. 695- Bs. 24, Rusch, Hubert; Haugll, Flan Robert: and Mayer, Horst, ‘Schobversuche sn Stahibeton-Rechteckbalken mit gleich vere teiter Beast,” Bulletin No, 145, Deutscher Ausichuss fr Sahl beton, Belin, 196, pp. 430. 25. Swamy, R. Nerayan, and Qureshi, Shamsuda A., "Suen, ‘Cracking, and Deformation inte in Reinforced -Beams Undet ‘ending and Shear," ACI JounwaL, Proceedings V. 68, No. 3, Mat. 197, pp. 187-195, 26, Taylor, R., "Some Shear Test on Reinforced Concrete Beams ‘Without Shear Relaforcement”" Magasine of Concrete Research Condon), V-12, No 36, Nov. 1960, pp. 145-154. 27. Van den Berg, P. 3 "Shear Sueapthof Reafored Concrete Beans Without Web Reinforcement, Part Factors Affecting Load ‘AGI JOURNAL / September-October 1984 332, and No. 1 Diagonal Cracking." ACI Journas, Proceedings V. 59, No: 11, Nov. 1962, pp. 1557-1600. : 28. Walraven, 1. C., “The Influence of Depth on the Shear ‘Strengih of Lighevelgh’ Concrete Beams Withoat Sent Reiforee- sent." Sievin Laboreory Report No, 578-4, Dalft Unverty of ‘Technology, 1978, 36 pp. 29, Park, Robert, and Pauley, Thomas, Reinforced Concrete ‘Structures, Soha Wiley & Sens, New York, 1975. 769 pp. 30. Babant, ZdentkP., and Cedali, Lag “Fractre Mechanles of Reinforced Concee,* Proceedings, ASCE, V. 106, EM, Dec. 1980, pp. 1287-1306, 31. Ratan, 2dentkP., and Cedoin, Lug), “Finite Bement Mod- sling of Crack Band Propagation," Journal of Siuctural Engineer fing, ASCE, V. 108, No, 1, Jan. 1983, pp. 6252. 32, Reinhard, , W., "Sint of Brite Fracture of Strse tural Concrete” Advanced Mechanics of Reinforced Conerete, TARSE Cotloquium, Det, 1981, pp. 201-210. 33, Batant, 2. "Crack Baia Niodel for Fractere of Geomater- Inis." Proceedings, 4th International Conference on Numerical “Methods in Geomechanics, University of Alberla, Edmonton, 1982, V3, pp. 137-112 “, Butan, 2. P., “Mechanics of Fracture ad Progressive Crack: in Concrete Structures," Report No. 83-2/428m, Center for Concrete and Geomaterials, Northwestern Univery, Evanson, Feb. 1983, 130 pp. Also, to appeat im Frocure Mechanies Applied t0 Concroe Srctures, Martius Nijhoff Publishers, The Hager, 38. Haant, Zdenek P. and Oh, B. H, "Crack Band Theory for Fracture of Concrete," Materias and Structures, Research and Test- ‘ng (RILEM, Pai), V. 16, No. 93, May-June 1983, pp. 15-177. Metal,” Report No, 83-2/658, Center for Conerete and Geomater- als, Nosthwesern University, Evanston, Feb. 1983, 4 pp. Also, to ‘ppear'in Journal of Structural Engineering, ASCE, V. 110, NO. TEMS, Apr. 198, pp. 518-35, 38: Zsutty, Theodore C., “Beam Shear Strength Prediction by ‘Apalyl of Existing Dats,” ACL Jounnat, Proceedings V. 65, No, 1, Nov. 1968, pp 99-951. 438. Zeutty, Theodore, “Shear Strength Prediction for Separate CCaegerie of Simple Beam Tess," ACI outa, Proceedings V. 62, No.2, Feb 1971, pp. 138-143. “40. Marquardt, D. W., “An Algorithm for Lees Squares Estima. ton of Nonlinear Parameters,” Journal, SIAM, V. I No. 2, 1963, p. 1-4 ‘1, Brown, K. Mand Denes, J.B, “Derivative Free Analogues ofthe Levenberg-Marguardt and Gauss Algoritims for Nonlinear Least Squates Approximations," Numersche Mathematit No. 18, 1972, pp. 288297. 42, Nell, Adam M., “A General Relation for Strensts of Con- crete Specimens of Different Shapes and Sie,” ACL JOURNAL, Pro fvedingsV. 63, No. 10, Oct. 1966, pp 1095-119. “3. ACI Commie 318, “Building Code Requirements for Rei forced Concrete (ACI 318-79," American Concrete Tasttute, De- tlt, 1977, 102 pp. ‘44. CEB-FIP Model Code for Concrete Stractures, 34 Edition, (Comité Euro-nternationl du Bton/Fédération Interationale dela Précontrainte, Pars English Béton, Cemest end Concete Associ ation, Werham Spies), 1978, 348 pp. APPENDIX For seades'convesince we summarize the dimensional analyse from Reference 37 that lads to Bq. (1). To take the dispersed and progrestve nature of cracking at the Fracture frot ino secant the Tollowing hypothesis may be introduced: The teal potenalenerey release W caused by facture na given structure sa funtion of bot () the length ofthe acute a, and @) the area of the eracked zone and, Here mt = material constant characterizing the width ofthe erack- ‘ng zone atthe fracture front.” Under Pat (1) of the bypethess we dertand the pat of energy thats ekeased from the uneracked re sons ofthe src to he race fon 467Variables «and amd are not nondimensional, They ar, however, allowed to appear ony ina nodimensonal frm, Ts fore ven by the following variables eo representing the nondimensonal facture length snd the nondimer- Sonal aree of the cracked sons. Furthermore, Wimust be propor. tional to volume d° of the structure (where & = thickness) and to the characterise energy dens 03/26, hich o, = Pb ~ nom nal stress at fire, P= given applied oad, and’ = characteris ‘dimension of the structure, Consegenty, we must have (PV sat see m E(@)wreen oe in which f isa certain continuous an coninvonsly differentiable positvefuneion, and parameters reptesen alo ofthe race simensons characterizing the geomtrical shape of the sretere. For similar stractres,& re constant. The condition for the facie to ropagatels Lop 2 Paes fn which Gis the fracture energy, a mateial property chassterizing the energy consumed per unit extension of the factre, pet ual thickness, Consider now geomeucaly similar structures, for which parame: ter are constant and only the characteristic cimension d vates, Ac cording othe easel of eiferentation, 9/30 =f (Bag) ¢ @u/2a) in whch we inode the notation f. = Banfi = Dog Thus, substitution of Eq 2) ito Bq. 2) sks (i. , fmt) (4a) 2 oe on Furthermore, the fracture ensay may be express asthe area under the complete tense sets-staia cure, lnhading the strein softening down to zero tes, des the width ofthe caching ont md a\L mE ow ‘in whic £, i the iia late modules of consrete,£,sthe mean srain-soflening meduls, which ip negative, and fs he dec en- sil strength of cone, Substituting Bg. (2) and P Eat acaran we m(s 2) os In which B = KI = B/EQ/fI" and X, « miffy B and, ate con- ants when geometrically simile stictures of different ssc are considered. Thus, Bq @5) proves our starting equation, Ex. (I). ACI JOURNAL / September-October 1984