Wason Medal for Most Meritorious Paper: Myths and Fallacies in Earthquake Engineering — Conflicts between Design and Reality thas long been recognized that our codified approach cs for seismic design bear a comparatively tenuous re- lationship to expected performance. Design isbased on ‘static “snapshot” simulation of the dynamic event, using methods extrapolated from approaches felt to be ade. quate and conservative for gravity Eoad design. A major dif- ference between gravity load effects and seismic response is that ultimate strength should never be developed under grav- ity load, while itis almost certain to be developed under seis- ‘mic response, typically at a level of excitation that may be a fraction of the design level for seismic attack. Further, ale though ductile seismic response implies greater dependency ‘on displacements than forces, we still, a5 a matter of conve- nicnee and adition, design for specified force levels, and ‘reat displacements in a comparatively cursory way This paper wets originally submitted at the Thomas Pavlay cent Developments in Lateral Foree Trans- erin Buildings,” hetdin La Jolla, Calif, in September 1993 sponsored hy ACI the University of California-San Diego. lund the Earthquake Bngiaeering Research Institute. The pa- per was published in the ACT symposium volume wit the vaume itl, as Special Publication 157, 1995, pp. 281-254, with Priestley, Michiel P. Collins, and Frieder Seible as ed- ors of he voleme. At the recent convention in New Orleans, La, the Tasti- tute's Bosrd of Direction approved the awarding of the Wasi Medal for Most Mertorious Paper to Priestley for ‘authorship of this paper. The citatian on the inedad will re ‘ognice Priestley for is paper “examining aspects of the cur- Trent philosophy and procedures of seismic analysis and design, | Alva in New Orleans, the Board authorised the Arthur B Anderson Award to Priestley for his “contributions 10 the evelopment of a new philosophy of seixnic design based on ‘mathematical analysis and targe-scale experimental re- search” anil she application of the same 10 new concrete onstruction, precast stractures, and retrofitting of existing ‘eonerete spans ‘These 10 honors are stated to be presented atthe Awards ‘Breakfast during the Apel convention in Seail, Wash. The Wason Medal for Moxt Meritorious Paper was founded in E917 ana is given annually for an ouastandling paper pub- Tished iy the Insitute. The Anderson svar was extablish in 1972 to recorgnice an ACI past president and is fo standing contributfons tothe eudvancement of knowledge of ‘concrete as a constriction macerial.” When the design approach is bused on carefully consid- ered philosophy, as in the capacity design approach pio- neered in New Zealand snd gradually becoming accepted in many other seismic regions, excellent results are to be ex- pected because the design structure is comparatively insen- sitive to the various assumptions made, IL uppears, however, that the enormous approximations in- volved in seismic design are perhaps becoming less appreci- ated, rather than more, as sophisticated analytical techniques become specified by cocles and accepted into common dk sign practice, as a matter of routine. In the United States. and T suspect elsewhere, this has resulted in a tendency for the functions of analysis and desien to be separated and per- formed by different specialisis, The analyst is responsible for ‘modeling the structure and running the lateral force analysis ly a3-D modal analysis process. Results from the sis are presented lo the designer who determines mem ber sizes, reinforcement quantities (if reinforeed cone: construction), and detailing aspects. The analyst is typically more involved in the analytical process than in the correct simulation of member characteristics, with potential dan- gers, The rest of the separation of design and analysis tends to be that analysis drives the design process, rather than the reverse, which might seem to be more appropriate “There isalso room to examine current design and detailing practice, much of which is also extrspolated trom gravity Joad considerations. Occasionally, this process can lead us in directions that are inappropriate for seismic behavior. Even when tenets of structural performance are based on purely dynumic characteristics, such as energy dissipation under cyclic response. the directions we are accustomed to taking are not necessarily the best for survival and damage control. and In this paper. some of the accepted seismic desi analysis procedures are identified as “myth” or “fallacy an overstatement perhaps, @ make a rather dry lapic seem more interesting. Nevertheless, a critical examination of the bases of ous design processes is ulWays appropriate, since the origins of these are often obscure, und lost in the history of design practice, or worse, in code committee minutes. Some of the points to be made are well known, athers perhaps le <0. The elastic spectral analysis fallacy ‘The fundamental basis of seismic design is still the assump: tion that an elastic (or modified elastic) acceleration re- SS Concrete international‘Spectral acceleration Perlod Fig. 1 — Design acceleration response spectrum. sponse spectrum provides the best means for establishing required performance of a structure. The limitations of the approach are well known, and accepted because of the de- sign convenience and the lack of a Viable design alternative ‘A case can be made that this is a fallacy, and that viable de- alternatives exist, or could be developed with compara- tive ease ‘To summarize the limitations: 1. Response is based on a “snapshot” of structural 1re- sponse; that is, response ut the moment of peak base shear for ‘an equivalent elastically responding structure. Duration ef- fects. which tend to be petied-dependent. with short-period structures suffering a greater number of response cycles than long-period structures, are not considered. The merits of us- ing: modal combination rules to provide some insight into higher mode effects seems hardly worthwhile when these will have to be considered by largely empirical rules later in the capacity design process, 2. The relationship between peak displacement response of clastic und inelastic systems is complex, and more variable than commonly accepied. Various rules, such as the “equal energy’* and “equal displacement” rules, are commonly em- ployed, but without much consistency of logic. If we consid er u typical elastic acceleration spectrum, as in Fig. |, four distinet zones can be identified. At zero period, displacements af elastic and ductile sys- F ~_> (a) Structure Fig. 2— Bridge tent example February 1997 tems cannot, by definition, be related, The structure will be subjected to peak ground ductility capacity, and wil f corresponding to PGA is provided. In the rising portion of the acceleration spectrum, displacements of inelastic sys- tems are greater than those of elastic systems with equivalent initial stiffness, and the “equal-energy” relationship. has some application. Inthe initial stages of the falling portion of the acceleration spectrum, clastic und inelastic displacement responses. are often similar, leading to the “equal-displacement” rule. As the structural flexibility increases stil further, the “equal- displacement” rule tends to beeome increasingly conserva- tive. At very long periods, there is essentially no structural response to the ground motion, aud the concept of an abso- lute displacement (indepenclent of period or ductility) could beadvanced, where the relative displacement ofthe ceater of mass of the Structure is equal to the absolute peak ground displaceme Different codes rely on startlingly different relationships between elas inelastic displacements, In the United States, the relationships between force reduction factors (A), seismic load factors (1.4), and design displucements (A; x 2R,) included in the Uniform Building Code! ap- proach cau be interpreted as implying that inelastic displace- hients are expected 10 be about 50 percent of equivalent clastic displacements. At the other end of the spectrum, many Central and South American codes rely on the equal energy approach, wit a base ductility of about 4, This im- plies inelastic displacement about 150 percent of equivalent elastic displacements. Although we accept displacement ea- pacity to be more fundamental to seismic response than strength, it appears that different groups of experts cannot agree within a factor of 3 as to what these should be, from a given clastic acceleration spectrum. Although these points have been recognized and partially considered in codes that define inelastic spectra with vari- able ratios between clastic and ductile coordinates, such as NZS 4203; confusion is still almost universal, 3. The elastic acceleration approach plices excessive em- ‘phasis om elastic stiffness characteristics of the structure and its elements, As discussed subsequently, we are less careful than we should be in determining these characteristics. The question remains as (0 whether belter altematives might be OP he (&) Plastic displacement 55considered. [believe that they can, and thata more consistent approach may be achieved by complete inversion of the desiga proce: I should be noted that although many researchers have acement-based design, the processes. de- scribed are in fact still stength-based. Mochle for example, usses the relative merits of ductility-based and displace: inent-based design, but in his comparison, the starting point is still a given strength and stiffness (and hence period) with the difference being whether displacements or ductilities are checked. As acknowledged by Moehle. when properly car- ried out, the two approaches are directly equivalent. In the approach described here, strength and stiffness are the end product of the design process, rather than the starting point, The procedure is initially illustrated by reference to the simple multicolumn bridge picr shown in Fig. 2, for which displacement-based design is comparatively straight for- ward, A set of elastic displacement response spectra for dif- ferent levels of equivalent viscous damping are required Fig. 3a). These can be generated in much the same way as elastic acceleration response spectra, The shape, with a res- ‘onant region as shown in Fig. 3a, with reducing displace- ment response at large petiods is churacteristic of displacement spectra (Fig. 2b). The sequence of operations involves the following steps: 1. An initial estimate for the structural yield displacement 4, is made, Since final results are not particularly sensitive to the valve assumed, A, could be based on typical drift gle of about 8,= 0.005, For 2 building design, a lower value would generully he appropriate. 2. The Timit to acceptable plastic rotation of critical hin: is determined. This will be a function of the importance of the structure, the section geomeiry, and an acceptable level of transverse reinforcement. For the bridge bent of Fig. 2, hinges are assumed to develop at top and bottom of the columns. 3. The maximum acceptable structural plastic displace- ment A, al the center of seismic force, corresponding to the Plastic fotation limit of the most critical hinge, is Found from considerations of meckunisms ceformation 4. A first estimate of total acceptable structural displace- tisthus A= A+ A. ‘An estimate of effective structural damping is made, based on the implied displacement ductility level L,—A,/A, (lig. 4), where curves are given based on typical liysteresis, characteristics for structures with beam hinges or column hinges, respectively (@) Design displacement spectra Fig. 3 a and b — Displacomont response spectra for design. 6. With reference to elastic response spectra for the site (eg., Fig. 3a), the effective response period can now be esti- mated. ‘The effective stiffness of the substitute structure at ‘maximum response can thus be found from: Eq. 1 and the required structure yield strength, or base shear cups inyi An Eq 7. With a knowledge of the required shear capacity, the member sizes cin now be proportioned, and aa initial esti- ‘mate of reinforcement made. The elastic stiffness can thus be calculated, and a refined estimate of the yield displacement obtained 8. The total displacement, structure ductility. and hence ef fective sructurul damping are thus revised, and steps 4 -7 ne- peated until a slable and satisfactory solution is obtained. Individual flexural strength requirements for potential plas- tic hinges are finalized, based om statics, The approach outlined has considerable flexibility, since plastic hinge rotational capacity can be related to transverse Aolailing (OF vice versa), and the design is not dictated by somewhat arbitrary decisions abour force-reduction factors. It would appeur that this method of displacement-based dlesign could also be applied to multi-story frame or shear wall buildings, provided some additional assumptions are made. The two critical pieces of information required are (1) the relationship between maximum interstory drift and strue- tural displacement at the height of the center of seismic force: and (2) the shape of the lateral force vector 1o be ap~ plied. These aspects are illustrated in Fig. 5 for an idealized frame of m stories each of equal Reight f. The center of sei mic force is approximately at2/3of the building height, and the maximom displacement at this height can thus be ex- pressed as: A=A42 nhO,K Eq.3 Damping Retio Displacement (em) 20 (b) Actual eleplacoment spectra (Brawley 1979, 315°) 56 Conerete Internationaltan Ven ampieg (5) ati « 3 ‘ @ @ fs spent Duet Fig. 4 — Equivalent viscous damping vs. displacement duciity evel. where KS | defines the nonuniformity of drift up the build ing height and 8, is the maximum acceptable rotation of the plastic hinges ard hence the maximum story deift angle, On the basis of inelastic amalyses of frames, Paulay and Priestley# cecommend thatthe distribution of drift should be assumed to be that shown in Fig, Sc, where the drift in the lower half of the stories is equal to twice the average drift at the roof level. Assuming further that this distribution cam also be applied to the plastic component of drift, Eq. 3 can be simplified to: A=, +05 nh0, Eq. 4 Thutis, K=0.75 It is suggested that improved estimates of plastic drift would be obiained by elastic analysis of a substitute steuc- lure,’ where stiffness of members containing hinges is re duced in proportion 10 their expected ductility. Hence, if eam hinges are expected lo have rotational ductilities of Uy = 7 (which might correspond to « structure displact ductility of jt, =4) then the appropriate stiffness for the team in the clastic analysis would be: K, = K,fll,= 0.14K,. The adequacy of the design can thus be checked hy a lateral analysis of the substitute structure, ie displaced shape can be approximated by Fig. Sc, it follows that the vector of lateral inertial forces to be upplied to the structure:should also take the same shape The displacement-bused design approach appears attra tive in principle, but will need to be checked by specific ex- amples covering a wide range of structural types and periods, The refined analysis myth Iwas noted at the start of this paper that structural analyses for design purposes huve become more sophisticated in re~ cent years, with the consequences that the analysis and de- Tunctions are frequently separated and carried out by different people. The reason for the increased sophistication in the analysis is principally related to the availability of powerfull computers rather than a perceived inadequacy of earlier and simpler analysis techniques. Although 3-D modal analysis is undoubtedly useful in siractures with unusual or irregular geometry, it is doubtful ifit produces better resulls than those obtained from simpler methods — say lateral analysis based on an assumed lateral force distribution. The myth here, then, is that refinement of the analysis produces more “accurate” results, 1is appropri ate to consider the refinement of the analytical process in light of the approximations still remaining. Elastic modal analysis essentially relies on the equal: placement approximation, since it is not feasible to use dil- ferent foree-reduction faclors associated with different modes of elastic response. As noted previously in reference to Fig. 1. this is appropriate for a comparatively narrow band ‘of periods, Deflection profiles from elastic modal analyses tend to un- derestimate drift levels in the lower stories of a building. As Foree vector (@) Structure Fig. 5 — Maximum response of a frame building (o) Inelastic response Displaced shape (©) Approximate plastic displacement profile, February 1997 57noted earlier, its felt that agreement could be improved by use of a substitute structure approach. Elastic analyses are generally hased on approximations of member stiffness that should be considered gross, even in the clastic range. As an example of this, let us consider thet the columns of the lowest story of the frame shown in Fig. Sa are reinforced concrete 600 x 600 mm (24 x 24 in.) columns, re inforced with 8 —D28 (28 mm {1-1 in.] diameter) hars with, i¢ld strength of 455 MPa (70 ksi). Concrete strength is /, = 31 MPa (4.5 ksi), It will be normal in the analysis to a5- sume that all three columns at the Lowest level have the same stiffness, though itis possible the central column might be allocated a slightly higher stiffness because of increased ax- ial load. It would seem to be impossible to allocate different stiffnesses to the two outer columns when multi-modal re- sponse is considered. or the sake of argument, itis assumed that the outer col- umns carry gravity loads Of Pip. .)=0.2 fA, and that scis- mic axial forees of P, = 40.2 fA, can be expected. Fig. 6 shows moment-curvature relationships for the compression and tension columns, including the effects of varying axial force as the lateral base shear increases, und the relationship assuming a constant axial force of P = 0.2 f./A,. Onset of first yield of tension reinforcement and attainment of a coni- pression strain of €, = 0.003 are also noted. ‘Taking the yield condition to correspond to a curvature of about 0,0084/m (0.03/11) — which is the yield point for a bilinear approximation to the P=0.2 fA, curve — we find that the effective stiffness of the compfession column is more than twice that of the tension column, As a conse- quence, the distribution of elastic Forces in the lower stories is likely to be substantially different from that predicted by the “refined” elastic analysis. Note, however, that it woul be comparatively straight forward to consider these effects in an equivalent lateral force approac! In some design codes (e.g. ref. 2) it is permissible to re- distribute up to 30 percent of moment from a tension column a. 20 A, A A ‘A lomaet) O sotridderenccerstar sao Mamta) & ‘oro Fig. § — Influence of axial foree on moment curvature relationship for rectangular column (1 kNm = 8.85 kip-in) toacompression column. tis perhaps of interest to note that using the 30 percent redistribution rule, force levels pred ed by constant-stiffness analysis for the tension column could be reduced by 30 percent of the average of the tension and compression column capacities. This results in a mini- mum permissible strength of 600 kNm (442 kipseft) for the tension column, or about 15 percent higher than the capacity. Design based on variable stiffness would, however, not re- quire any redistribution beyond that naturally resulting from the stiliness difference. It should also be noted that the redi tribution limit of 30 percentis set to avoid excessive ductility, demand, Although it will be seen that the tension column does in fact reach yield at a curvature 16 percent lower than the P=0.2 (2A, case, the onset of crushing (very consery tively estimated at e, = 0.003) occurs at a curvature more than twice that for the P= 0.2 ¢,A_ column, and more than three times that for the compression ‘column, Ultimate eurva~ tures are similarly affected. Thus, at least so far as the ten- sion column is concemed, the 30 percent limit to redistribution would seem quite unreasonable and the eritical condition is likely tobe the compression column, whose duc- tility demand we imagine to be reduced by the redistribution ‘process (herein appears another fallacy). The strength/ductility trade-off fallacy ‘The current design emphasis on force-based design, together -with the general adoption of the equal displacement approx- mation Teuds us to the natural conclusion that required strength, S, and displacement ductility demand, j., are relat- ed by the expression: Seq = constant Eq.5 fora given structure or critical element, with the usual caveat that for short-period structures, the approximation may be inappropriate. ‘The fallacy of this observation becomes obvious when w invert the logic, Consider that we are designing a structural element, say a bridge column, and we decide that, as de- signed, the ductility capacity (and hence in a more basic sense the displacement capacity) is inadequate, A quence, we decide to increase the strength to reduce the duc- tility demand, We do this by increasing the longitudinal steel ratio, and keeping the section size constant. Have we really ows nor OOS mS ny Fig. 7 — Dimensionless change in structural response of Cantilever bridge pier as a function of longitudinal reinforcement ratio. 58 Conerete International(@) Elastic-pertectly plastic (b) Degrading suitfnass (Takeda) {r-e. column hinge} (c) Nonlinear elastic [unbonded prostressing) Fig. 8 — Response for different hysteretic modols including P-A eftect. improved anything? Probably not — the equal-cisplacement approximation still says we require the same ultimate dis- placement capacity, even thoush the ductility demand has gpparenily been reduced, and its certainly not clear that in- creasing the longitudinal reinforcement ratio has increased the ultimate displacement Infact, quite the opposite is more likely. Fig. 7 plots the re- sults of varying longitudinal reinforcement ratio for circular columns with asial load ratio 0.1 f/A,, Results for moment capacity, stiffness, ductility eapacity and ultimate displace ment are expressed in dimensionless form by reference to the value pertaining to a “standard” longitudinal ratio of p, 0.0015, Iwill he seen that as p, is increased, the ultimate ‘moment capacity (Mf) increases almost proportionately, but the ultimate displacement duces even more. Of course, the argument also contains a fallacy, since the stiffness (K_) has increased almost as much as the strength has, and thus the period will have changed. Nevertheless, it is unlikely that this stiffness variation w hhaye been included in the original calculations, and the r quired inerease in ultimate di achieved. It can even be reasonably argued that if we were worried about the ductility or displacement capacity, we ‘would have been better off reducing the reinforcement ratio, and hence the stren; The energy myth One of the more pervasive myths in earthquake e is that energy dissipation should be maximized 0 obtain op timum seismic response. In this myth, itis supposed that we should strive towards obtaining hysteretic characteristics as closely approximating elastic/perfectly-plastic (EPP) re- sponse as possible. Although there are situations, particular- ly those involving very short period structures where this is, in fact, desirable, there are many cases where better response can be oblained with apparently less desirable Loop shapes. Fig. 8 compares three idealized hysteresis loop shape clastiefperfectly-plastic, degrading stiffness model typical of areinforeed concrete column hinge,’ and the bilinear elastic characteristic theoretically sppropriate for a plastic hinge with unbonded prestressing tendons.” Response is shown both with and without PA, Consider first the response without P-A effects, I'response isin the “equal-displacement” domain of Fig. 1, the peuk re- sponse displacements of the three systems are likely to be very similar if each hus the same initial stiffness, In actual fact, the displacements of the FPP system will on average be ititile smaller than the other two, but the difference will not be great. Let us assume that a maximum drift of 2 percent, corresponding to a displacement ductility factor of 11, = $ is dobiained. After the earthquake, the restdual drift of the EPP system could he as high as 1.6 percent, that of the column hinge about 0.9 percent, and the bilinear sysicm of Fig. 8¢ ‘will return to its initial position. Which system has exhibited beuer response? Its at least arguable that residual displace- tenis are ultimately more important than maximum dis- Placements, given the difficulty of straightening «bent building after an eurthquake. Consider now the influence of hysteresis loop shape on re sponse as affected by P-A effects. MacRuc has rather con Vineingly shown? that the tendeney for instability under P-A, is strongly related to the shape of the loop. In Fig. 8, the in- fluence of P-A moments on the inelastic response shape is shown by dashed lines. With the EPP loop of Fig. 8, re sponse ala given instant of the seismic response has resulted in 2 residual deformation corresponding to point B. ‘The sinicture will oscillate with the elastic stiffness about this point until response acceleration sufficient to develop the Yield strength occurs. Ax will he seen in Fig. 8a, the acceler- ition required to make the system reach the upper yield line igmuch less than that for the lower yield line. ILis thus prob- able that plasticity will develop in the direction of increasing, rather than reducing residual displacement. With a long du- ration and an EPP loop shape, the system is inherently unstae ble under P-A effects. With the degrading stiffness model of Fig. 8b, and a resid- ual deformation corresponding to point B, the lower yield line is closer to the zero acceleration line, and is thus more likely to be attained than the upper yield line. The system is, thus inherently stable, since the probabilities of inelastic de- formation favor decreased residual set. MacRae® has demon- raed the validity of this argument witha very large number of dynamic inelastic time-history analys Since the system in Fig. 8c is elastic nonlinear, there are no residual displacements to be considered, and the system is stable for P-A effects. It should also be noted that for longer period structures, where P-A effects are likely to be significant, the equal dis- February 1997 59placement rule would indicate that A effects are unlikely to significantly increase the macimum displacement of the stable systems significantly. This is also supported by time- history analyses, For the FPP system, which js inherently un table auuctedl peubecitie cqial chaslaeeseat nor erates ergy rules can be applied, since the inerease in maximum and residual displacement is strongly influenced by the duration of the earthquake record. One should not, of course, dismiss the value of hysteretic energy absorption. However, it is clear that current emphasis on the loop shape is overstated. Steel structures, with defor- mation characteristics approximuting EPP loops, have a greater tendency for undesirable deformation response than the equivalent reinforeed concrete structure The distribution of flexural reinforcement fallacy the discussions to this point have largely related to analysis issues, However, itis cleur that many aspects of design and detailing could also bear eritical review. The remainder of this paper will examine a few issues specifically related tw reinforced concrete design, though more could be identified, both with reinforced concrete and with other materials. One of the most pervasive fallacies relates to the way we distribute reinforcement in beams of ductile moment-resist- ing frames, By use of moment redistribution, we frequently end up with positive and negative mosnent demands that are equal, or nearly so. We then plaice reinforcement in two bands, as elose lo top and bottom respectively of the bean, as shown in Fig, 9a, in the mistaken view that this provides, the most efficient distribution. Wong et al.” have shown that essentially the sume moment capacity can be achieved by distributing the total amount of reinforcement down the sides of the beam (Fig. 9b). Fig. 10 compares the flexural strength of the alternative distributions of Fig. 9, asa function of the mechanical reinforcement ratio p_f./ /°;. where p = A /bhis the total reinforcement ratio. The’ sirength differences. be- tween the distributions are i ‘There are, however, good reasons for adopting the distri bution of ion at beam-column joints of wwo- ‘way frame construction is considerably eased, greater pro portion of the joint shear foree can be associated with the di- agonal concrete strut, thus reducing the demand for joint shar reinforcement, and flexural overstrength result from strain-hardening of reinforcement is reduced. This lat ter point could be taken advantage of by reducing the over- strength ratio used 10 develop member forces in the capacity design process, thus resulting in design efficiencies. [ee wae + eo 1 ee (@) Conventional roinforcoment (6) Vertically distributed reinforcement Fig. 8 — Arrangomonts of longitudinal reinforcement in beams. Design efficiency could also he improved by more appro- priate use of strength reduetion factors in the basic capacity design equstions, which can be generally stated as 05,20, 0,5, Ey. 6 where 5, is the nominal strength of a particular action (flex ure, shear, etc.) S_is the strength required for thal zetion ro- sulting from the basic analysis assumptions, 6, is strength reduction factor applied to S, to provide @ dependable strength, and «9, and 0, are dynamic amplification and over- strength factors, which again relate to the analysis assump lions, design efficiency, and action considered. Currently in seismic design? we associate a strength Fe- duction factor to the basie Hextral strength of plastic hinges, but not to members or actions protected by capacity design principles, on the basis of perceived conservatism in the val ues oF @, and @,, currently specified It would appear that the logic may have thus been inverted. tis clear that we do not need a flexural strength reduction factor for plastic hinges. sinee srmall variations in strength from the specified value will only result in small variations in ductility demand, As noted in relation to Fig. 7, nereasing, the reinforcement content, wich is the end result of applica- tion of a flexural strength reduction factor, may not improve overall safety. However, if we wish to totally proscribe nonductie inctas- fic deformation (e,g.. shear) we need a high degree of assuri- ty that the dependable strength of that mode cannot be exceeded. If we believe strength reduction factors need to be associated with that action (e.g., shear) as a result of possible noneonservatism of design equations. or possible materi under-strength, then these factors should be utilized in the capacity design process. It the product «,, is felt to be such {hat no strength reduction factor is needed this implies that 04, is t00 high, and shoul be reduced This may seem to be a matter of sernanties, since the end result would probably be little different in current design practice. However, if design is to be permitted in accordune ‘with more advanced analytical processes (e.g, time-hisiory analysis to determ strength of plastic hinges has been determined) then values for «@, and @, might become determined by the results of the analySis process. Ih this ease. it would be better o have the ne expected dynamic influences once 018 a a Cr) pit att Fig, 10 — Dimensionless flexural strength for beams with conventional and vertically distributed reinforcement. Concrete Internationalvariability of desis correct plirameters Of course, the process could be greatly simplified if we did away with Strength reduction factors completely, as is the cease in Japanese design, The value of 6 would this be inher- ent in the equations for stength. It is hard 10 see that we would Lose much in the process. n strength properly associated with the The shear myth(s) Shear design of reinforced concrete is So full of myths, flla- cies, and contradictions that it ishard to know where co begin in an examination of current design, Periaps the basic myth central to our inconsistencies in shear desig is that of shear itself. It has been argued that we tie ourselves into intellectual knots by sepurating Mexure and shear, and considering them essentially independent entities. Compression field theory as, developed by Collins et al"is an attempt to integrate the 3 tions. Similar attempts have been made elsewhere." The fat remains, however, that iL is very convenient to separate the Flexural and shear actions, The more fundamental approsch- es are not only inconvenient from a design viewpoint. but also do not produce notably better agreement will experi- mental resulls — particularly when shear strength of ductile linear members characteristic of framed structures is considered, Our understand plastic hinge regions s of the mechanisms of shear transfer in icitlarly {b) The 45 degree truss mechanism Fig. 11 — Shear transfer in beam plastic hinges. ‘adjacent toa column. as depicted in Fig. 1, the sumption is that concrete shear mechanisms such as egate interlock. dowel action, and transfer ure undependlable, because of the potential presence of a full depth Flexural crack (see Fig, Ha), Hence shear must be entirely transferred by 2.45 degree truss mechanism, involving wansverse re nent (Vig. 11), pon reflection, the 1wo halves of this assumption (ide full depth crack: truss mechanism) are mutually incompati: bie. The truss mechanism of Fig, 11, whether based on 45 degree or some other angle, relies on the development of di- sugonal compression struts, stabilized by vertical tension in the stirrups or ties, and changes in the longitudinal heam ten- sion and compression resultants at the “nodes” (formed by intersection of the tie and di forces), vertical section is taken at any position. the vertical component of the diagonal compression lores continous ‘across the section must equal the shear force transferred by the truss mechanism. At the critical section at the column face, there is a full depth flexural erack formed by inelastic action in the wo opposite directions oF response. As a con sequence, there can be no diagonal compression forces crossing this section, and the shear carried by the truss mech- anism must also be zero. This apparent, and obvious, dilemma has been "rational- ized” hy the assertion that the intersection of diagonal eracks, a J 1a) allows the cracks to dilate and close in the middle section of the beam, thus permitting diagonal compression struls to develop. There are two concerns to adopting thi solution: First, if diagonal compression can develop across this crack, then perhaps the assumption that V, =0 should not be made Siice atleast part of the rationale for discarding Y, has the section, Consider equilibrium of the stress resultants in Fig. 11b. The resullant of the diagonal compression forces may be as- sumed to act al midheight. [Fa 45 degree truss is assumed. its vertical and horizontal components must bath be equal to V s shown, ‘The flexural compression force acts at oF near the center of compression reinlorcemeal Without the diagonal compres the snoment capacity M,-Td—d) Fq.7 If the diagonal compression force is inelucled, the moment capacity reduces to: M,=1d-a)- ve (a-a") Eq. 8 ecting the dead load shear, assuming equal moment capacity at opposite ends of the beam, and a beam length of xd - ds a eevee 4.9 Va Tae) A Mu hence M, = Nd =a) -=* or, M, Ey.10 Ina deep beam with (s 4, this implies « 20 percent reduction in moment capacity. which is nol supported by ex- perimental results. Even with more slender beams (sa 10) the moment reduction shomld he evident. Note tat when February 1997 61dead-load shear is added, negative moment capacity should be further eroded. if the truss mechanism were comrect. Since this behavier is aot apparent in experimental results, it would appear that relitnce on alternative mechanisms is, needed, Ifa Tull depth crack can develop, itis difficult to es. ‘cape the conclusion that all slicar must be carried by dowel action. IF this is the case, the primary shear transfer Function of transverse reinforcement in the plastic hi jon must be Loredluce the sinsupported length over which dowel action of the Longitudinal reinforcement occurs, hence incressing. the shear that ean be uansferred Rational models for the amount of shear reinforcement re- quited can be developed, which are very different from those resulting from the intss model. Critical aspects include diam- eter of longitudinal reinforcement (the bigger the better), and locution of the first stimup from the column face — which should be as stuull as possible. ‘As mentioned, there are many other inconsistencies in sheur design, A particularly troublesome ane is the Way in which we treat the en strength by axial compression. Our codes indicate thut the depree of enhance ment depencls on whether the structural element is a heam, a column, ora wall, and whether the axial compression comes from applied gravity forces, or from prestress, Differences in influence of more than 100 percent are possible. Since the uilferences appear to result from semantic definitions, a de sree of skepticism is appropriate Development of reinforcement fallacy ‘The final example eonyidered is the way in which we consi er development af reinforcement. Development is «catch-all term used to describe embedmentor anchorage, splicing, and (a) Column base lap splice (©) Knee joint (e) Column top (@) Column mid-height ‘ap splice Fig. 12— Anchorage and splicing of reinforcemant. Neural bond, The ssime equations, with modifiers in some cases, are used to describe the three situations, Although this, is convenient for design, itis unlikely to be realistic in prac- lice, considering the wide number of possible situations o curring. as illusirated for a bridge column ia Fig. 12. A ;hunber of rather different conditions exist depending on lo- cation, whether or not confinement is present, and whether ‘or nor the splitting eracks assumed in the basic development Jength equation can actually develop. he fundamental basis of the basic development lens equation can also be questioned, Until recently in the ACI code, this length was given by f= O0L9A, LAF. Lg. Where A, is in mm’ and f,and_f are in MPa. Modifiers are included to represent the influence of cover, bar spacing, lo cation in the conerete pour, confinement, exe. Hg. 13 shows the ratio of basic development length to bur diameter ¢, for USS. har sizes, for f,= 41 MPa (60 ksi), and ("= 22.5 MPa G.3 ksi). In Lig. 13, the equation definitions are ia American Standard units, Tt will be seen that the basic development th increases in terms of number of bar diameters as the neter increases, In fact, considering the range of bar sizes used in the United States (#3 - #18, 375 - 2.26 in, [10 - 57, mnt] di), Eq. EL implies the dimensionless development lenyth increases by a factor of six, oF the actual! development length increases by a factor of 36. Tests on scale models do ‘ot support a scale dependency of this nature. To some extent, of course, the modification lactors applied to Les teduce the apparent scale dependency, but it would ap- pear from Fig. 13 that the fundamental basis of our develop- ment calculations may he flawed, Returning to more fundmental approaches results in equations that are more intellectually satistying,!? and fit the data better, Conclusions A somewhat irreverent examination oF aspects of seismic analysis and design has been presented. In onler to establish inconsistencies and inaeeuracies commonly accepted by the dlesign profession, u number of the examples have been de liberatcly overstated. A critical review of this. puper will show that some of the arguments are at best simplistic, if not flawed. It should be emphasized that i is not contended that cue rent seismic design practice, particularly the version adopted in New Zealand, is unsafe, If this paper has a message or Act 918-89 ata = “CALTRANS. Aer 318-83 6 0s 10 1s 20 25 Dar Diameter, dy Fig. 13 — Basic development length, for deformed bars in tension (F7, 60ksi). f. 62 Conerete internationalconclusion, itis simply the following cautionary note related to the tendency for increased complexity in analysis: Given the wide range, and occasional gross nature of the ‘sumptions ancl approximations inherent in seismic design, ‘we might be better keeping the design and analy'sis processes simple enough so that We still understand what we are doing. A second, and more serious (perhaps!) point brought out at, the start of this paper related to displacement-based design: If we accept that displacements are more important than forces, it is time we started basing our designs on displace. ‘ment, rather than acccleration spectra. References, 1 Unie sulang Corte, tteretional Coatere 1988, 926 pp. New Zealand Standard NZS 423, Code of Practice for Genera tural Design and Design Loses for Buildings, Standards Asoc New Zealand, 1984 3, Moehle. J.P, “Displacement-Based Design of RC Structure tw Eantiguakes,” darthquake Specta V. 8, No, 3, August 1992, pp. 403. 4, Paulay,, and Priestley, MIN., Seismic Desige af Concrete aut Max sony Structures, Joba Wiley arid Sons, New York, NV, 1992, 7 pp, 5, Shibata, A, and Sozen, M.A., “Substitute-Suucture Metiod for Seis- tic Design in Reinforeedl Conerste" Journal af the Siractural Division, ASCE, V, 102, No, STI. 1976, pp. L8. 6, Tike, 3 Soren, MLA.sand Niclsen, NN, “Reintoreed Caner sgonse to Simulated Esethouakes” Journal of the Siructural Dis ASCE, V,96, No, S112, Desombsr 1970, pp. 2557-2573 7, Priestley. MLN, aad Tian Ren Two, “Seismic Response of Precast Pre ssromied Concrete Frames with Partially Dehonded Tendans" PCT Journal of Bulking Oficial, Re. V. 38, No. 1, January/icbrvsry 1993 pp, Si-69 JN nd Jian Ta0,“P-A Design in Seismic sctural Systems Research Project, Report No, SSRP 9305, University of Califor, San Diego, May 1993, 115 pp, Wong, PKC. Priestley, MILN2 and Park, R., “Seismic Resistance of Frames with Verscally Distt Langlidinal Reinforcement ia Bean, ACE Structural Jouarsal, V.87. No.4, SulyFAngust 1990, pp. 488-498, 10. Collins, M.P, and Necchio, F ing the Response of Reit- forced Concrete Heams Sbjsted tS ine Mositied Diagonal Compression Field Theory,” ACT Siri i V.85, May/Tune 1988, 1, Hsu, TTC ACT Sin 12. Prestley. MJLN.. “Seismic Assessment of Existing Conorete Bridye es" in Seienic Arseasment and Revi of Brides, Stustoral Systens Re scarch Project, Report No, SSRP 9103. University of California, San Diego, Jnly 1991, pp. 841 Reveived und nevieweu under Intute puieation policies MJ, Nigel Priestley is a professor of strictural engineering at the University of California, San Diego. He was edu- ‘cated in New Zealand, and spent 10 years on the faculty of the Department ivil Engineering, University of Can- terbury, before moving to the United States in 1986. His research injerests Include seismic response of bridges and buildings It’s traditional with CONCRETE INTERNATIONAL that our expert authors come from the ranks of our readers. So, why not become a C/ author too? You may well qualify. It makes sense. You probably have many years of experience in concrete design and constru In, and have something to write about that others will be interested in’ + A recent job your firm completed, * A success you had with a technical problem you and your colleagues faced, + Or, you just might have a valid opinion how important matters confronting the industry can be addressed ‘What better, more accurate and knowledgeable reporter can there be than you — someone who is expert in and has been right on top of what is being published. So, let the world know about what you are involved with ‘through the pages of CI! During any single year, the types of articles we publish in Cf will include: design and construction of concrete fiber reinforced concrete: formwork; computer use in concrete design and construction; mixing and placing of concrete; admixtures: repair and rehabilitation . .. you name it! GAIN WORLDWIDE RECOGNITION FOR YOURSELF AND YOUR ORGANIZATION Give Cl a try — with an article on a topic of your choice. Call attention to the work you and your people are doing: help others facing the same or similar problems and challenges! For more information on how to publish in CI, contact: Bill Semioli, Editor & Associate Publisher,CONCRETE INTERNATIONAL, P.O. Box 9094, Farmington Hills, MI 48333, Phone (810) 848-3797, fax (810) 848-3701, ‘TDD# (810) 848-3823, February 1997 63