Vous êtes sur la page 1sur 98

European School for Advanced Studies In Reduction of Seismic Risk

Myths and Fallacies in Earthquake Engineering, Revisited The Ninth Mallet Milne Lecture, 2003

M.J.N.Priestley
Co-Director

Rose School, Collegio Alessandro Volta, Via Ferrata, 27100, Pavia Italy May 2003

THE NINTH MALLET MILNE LECTURE


Peter Merriman Chairman of SECED, The Institution of Civil Engineers One Great George Street, Westminster, London, SW1P 3AA, UK The biennial Mallet Milne Lectures were begun by the Society for Earthquake and Civil Engineering Dynamics (SECED) in 1987 in honour of Robert Mallet (1810-1881) and John Milne (1850-1913), the nineteenth century pioneers of seismology in Britain who made fundamental contributions to the observation and understanding of earthquakes. The prestige lectures are given by internationally recognised experts in the field of earthquake engineering on topics close to their own professional interests but also of relevance to the broader earthquake engineering community. The Mallet-Milne Lectures have covered many different aspects of earthquake engineering primarily from a technical perspective including the broader issues of policy for disaster prevention. Three of the Lectures in the series have been dedicated to issues relating to engineering seismology and seismic hazard assessment. In the First Mallet Milne Lecture Engineering Seismology, Professor Nicholas Ambraseys described a new approach to the assessment of liquefaction potential and a re-evaluation of twentieth century seismicity in Turkey. The Fifth Mallet Milne Lecture From Earthquake Acceleration to Seismic Displacement by Professor Bruce Bolt of the University of California at Berkeley focussed on the destructive nature of near field ground motions containing high energy pulses that are usually referred to as fling a term that the lecturer himself was responsible for coining. In the Eighth Mallet Milne Lecture Living with Earthquakes: Know your Faults Dr James Jackson of Cambridge University addressed the identification and characterisation of active geological faults. Dr Jackson illustrated the major advances made in the ability of seismologists to determine source parameters for earthquakes, in the understanding of the relationship between crustal deformations and geomorphology and in the developments of remote technology that allow rapid and accurate measurements of deformation of the Earths surface. The other five Mallet Milne Lectures have considered how the seismic community may use technical knowledge with pragmatism to ensure effective design. In the Second Mallet Milne Lecture 1989 Coping with Natural Disasters, Professor George Housner of the California Institute of Technology addressed the challenges and perspectives of the International Decade for Natural Disaster Reduction on the eve of which the lecture was presented. For The Third Mallet Milne Lecture Reduction of Vibrations Professor Geoffrey Warburton of Nottingham University illustrated the application of isolation and active control to limit the forces on structures subjected to earthquake loading. The Fourth Mallet Milne Lecture entitled Simplicity and Confidence in Seismic Design was delivered by Professor Tom Paulay of the University of Canterbury. He used his extensive design experience to address the concepts that can be employed to ensure that reinforced concrete buildings can be expected to perform, in Professor Paulays words,

as they were told to. The issue of expected and actual behaviour of buildings and bridges was revisited in the Sixth Mallet Milne Lecture in 1997. Professor Roy Severn of the University of Bristol presented Structural Response Prediction Using Experimental Data drawing on his life-time experience in the dynamic testing of large structures around the world and the application of the results to earthquake engineering. In the Seventh Mallet Milne Lecture The Road to Total Earthquake Safety Professor Cinna Lomnitz of the National Autonomous University of Mexico addressed the dynamics of seismic wave propagation, the response of soft soils and the coupling of ground response with structural response. In an eclectic presentation Professor Lomnitz stressed the need to understand the interfaces between the many technical disciplines to determine what kind of structure would survive the next earthquake regardless of the precise location of the fault break. In the Ninth Mallet Milne Lecture Nigel Priestley of the Rose School, Italy, presents Revisiting Myths and Fallacies in Earthquake Engineering. Nigel has been a leading critic of current seismic design methods. Over ten years ago he thoroughly examined a number of the fundamental principles for the seismic design of structures and concluded that in many cases current practices, particularly as entrenched in design codes, are based on unrealistic concepts and approximations. A number of these concepts are revisited and re-examined in the light of ten years research and development. The practice of allocating strength of reinforced concrete elements in proportion to stiffness is shown to be inappropriate. Use of 3D multi-modal elastic analyses is shown to be inadequate to describe higher mode or torsional effects of inelastic systems, despite these being the rationale for the use of this design approach. The superiority of performance of inelastic structural systems with high inherent damping is shown to be a fallacy. Finally, Nigel outlines the progress to develop a simple and rational seismic design procedure based on displacement rather than strength considerations. Overall the paper is a challenge to the many researchers and engineers working on structural analysis and design ~ a timely reminder that there is still room for a considerable improvement in our understanding of the seismic performance of structures. Nigel Priestley is well known throughout the seismic community ~ having worked in New Zealand, the USA and lately in Italy. Nigel Priestley is Emeritus Professor of Structural Engineering of the University of California at San Diego, and is co-director of the Rose School, and is a practicing structural consultant. His research is focused on the seismic design of concrete and masonry structures, and on seismic design philosophy. He has published more than 650 books, technical papers and reports, mainly related to seismic design, and has received more than 30 international awards for his research. He is a fellow of the ACI, the IPENZ, NZ Society for Earthquake Engineering, NZ Concrete Society, and is an Honorary Fellow of the Royal Society of New Zealand. SECED are delighted that the IUSS Press have agreed to publish the Ninth Mallet Lecture in its entirety in this special issue. This will ensure that the many interested individuals who could not attend the formal presentation will also be able to benefit. Together with the previous Mallet Milne Lectures we are confident that this will be a valuable resource to earthquake engineers for years to come.

TABLE OF CONTENTS
CHAPTER 1. INTRODUCTION: A 1993 VIEWPOINT .1 1.1 The Fallacy of Design to Elastic Acceleration Spectra ...1 1.2 The Refined Analysis Myth ...4 1.3 The Strength = Safety Fallacy 5 1.4 The Myth of Maximized Energy Absorption .5 1.5 Myths and Fallacies Related to Reinforced Concrete Design and Detailing ...6 1.6 Scope of the Report ..8 CHAPTER 2. THE ELASTIC STIFFNESS OF CONCRETE MEMBERS ..9 2.1 Introduction .9 2.2 Elastic Stiffness of Circular Columns ..12 2.3 Elastic Stiffness of Rectangular Columns 16 2.4 Elastic Stiffness of Rectangular Walls ..19 2.5 Elastic Stiffness of Flanged Beams ..21 2.6 Yield Drift of Frames ..24 2.7 Conclusions 30 CHAPTER 3. MORE PROBLEMS WITH MULTI-MODAL ANALYSIS ..32 3.1 Introduction 32 3.1.1 Higher Mode Effects in Equivalent Lateral Force Design ...32 3.1.2 Torsional Effects in Equivalent Lateral Force design ..34 3.2 Design Higher Mode Effects by Multi-Modal Analysis ...34 3.2.1 Higher Mode Effects in Cantilever Walls from Time-History Analyses ..36 3.2.2 Modified Modal Superposition for Design Forces in Cantilever Walls 44 3.2.3 Consequences for Capacity Design .48 3.2.4 Higher Mode Effects in Frame Buildings from Time-History Analyses ..48 3.3 Torsional Response of Building Structures .53 3.3.1 Torsionally Unrestrained Systems .. 54 3.3.2 Torsionally Restrained Buildings 56 3.4 Conclusions 58 CHAPTER 4. THE ENERGY MYTH SEISMIC RESPONSE OF PRECAST PRESTRESSED STRUCTURES .60 4.1 Introduction ...60 4.2 Seismic Design Philosophy for Precast Concrete 61 4.2.1 Force-Based Design Recommendations for Strong-Connection Frames .63 4.2.2 Force-Based Design Recommendations for Ductile-Connection Frames 66 4.3 Unbonded Prestressing in Precast Design ...68 4.3.1 Theoretical Considerations ..68 4.3.2 Cyclic Tests of Precast Beam-to-Column Units with Unbonded Tendons .. 73

4.3.3 Precast Structural Walls with Unbonded Post-Tensioning and Additional Damping 76 4.4 Seismic Performance of a Large-Scale Precast Prestressed Test Building .78 4.4.1 Structural Systems in the Test Building ....78 4.4.2 Test Procedure 83 4.4.3 Structural Performance of the Test Building ....84 4.4.4 Modelling Structural Performance ..89 4.5 Steel Frames with Unbonded Post-Tensioning ....90 4.6 Conclusions 92 CHAPTER 5. DIRECT DISPLACEMENT-BASED DESIGN. A RATIONAL SEISMIC DESIGN APPROACH ....93 5.1 Introduction ....93 5.2 Basic Formulation of the Method ....94 5.3 Single-Degree-of-Freedom (SDOF) Structures ....96 5.3.1 Design Displacement for a SDOF Structure ....96 5.3.2 Yield Displacement .98 5.3.3 Equivalent Viscous Damping ..98 5.3.4 Design Displacement Spectra 100 5.3.5 Design Base Shear Equation ..101 5.4 Multi-Degree-of-Freedom Structures 102 5.4.1 Design Displacement 102 5.4.2 Displacement Shapes 103 5.4.3 Effective Mass ..107 5.4.4 Effective Damping 107 5.4.5 Distribution of Design Base Shear Force ...110 5.4.6 Analysis of Structure under Design Forces 110 5.5 Capacity Design for Direct Displacement-Based Design ...112 5.6 Some Implications of Direct Displacement-Based Design 113 5.6.1 Influence of Seismic Intensity on Design Base Shear Strength ..113 5.6.2 Influence of Building Height on Required Frame Base Shear Strength ..114 CHAPTER 6. CONCLUDING REMARKS ...116 7. REFERENCES ...119

CHAPTER 1. INTRODUCTION: A 1993 VIEWPOINT


Ten years ago, we critically examined[1] a number of the fundamental principles on which the seismic design of structures is based. It was suggested that the current emphasis on strength-based seismic design, based on elastic structural characteristics, modified by ductility capacity or behaviour factors leads designers in directions that are not always rational. The holy grail of maximized energy absorption was shown to be at least a mixed blessing. It was concluded that in many cases current practices, particularly as entrenched in design codes, were based on unrealistic concepts and approximations which we termed, with some hyperbole, myths and fallacies. The more relevant points made in the original paper are briefly summarized in this chapter. In subsequent chapters the arguments are revisited and re-examined in the light of ten years research and design development to determine what, if any, progress has occurred in developing more rational design approaches. 1.1 THE FALLACY OF DESIGN TO ELASTIC ACCELERATION SPECTRA As noted in reference [1], ten years ago the fundamental basis for seismic design was the assumption that an elastic acceleration spectrum provided the best means for assessing the seismic response of a structure. This has not changed in the last 10 years, though as will be shown in subsequent chapters, alternative approaches are emerging. The assertion that this assumption is in fact a fallacy rested on the following points: Design ignores duration effects and condenses response into snapshots of transient behaviour at maximum modal response. These modal maxima are combined by modal combination rules of dubious relevance for inelastic structural response. It is assumed that maximum transient response is more relevant than the final at rest condition of the structure after the earthquake, which is not considered in the design process. It is generally accepted that damage can be related to material strains, and that material strains can be related to maximum response displacements, but not to response accelerations. It would thus seem important for design procedures to emphasise the importance of estimating peak displacement response. In fact, the situation pertaining ten years ago was characterized by uncertainty in the estimation of displacement of inelastic systems. The normal approach involved modification of the displacements of the corresponding elastic system of equal initial stiffness and unlimited strength. The most common assumption was the

Priestley: Myths and Fallacies in Earthquake Engineering, Revisited

equal displacement approximation, which states that the displacement of the inelastic system is the same as that of the equivalent system with the same elastic stiffness, and unlimited strength. Thus, the design displacement is estimated as

max,duct = max,elastic =

T2 a(T ) g 4 2

(1.1)

The equal displacement approximation is known to be non-conservative for short-period structures. As a consequence, some design codes, notably in Central and South American, and some Asian countries, apply the equal energy approximation when determining peak displacements. The equal energy approach equates the energy absorbed by the inelastic system, on a monotonic displacement to peak response, to the energy absorbed by the equivalent elastic system with same initial stiffness. Thus the peak displacement of the inelastic system is

max,duct = max,elastic

R2 + 1 T2 R2 + 1 = a g T ( ) 2R 4 2 2R

(1.2)

where R is the design force-reduction factor. In the United States, where until recently the dominant building code for seismic regions has been the UBC[2], design displacements have been estimated as

max,duct = y

3R 8

(1.3)

where y is the yield displacement corresponding to the reduced design forces, and can be interpreted as

y = max,elastic / R
Hence the design displacement is given by

max,duct = max,elastic

3 8

(1.4)

It would appear that it should be a comparatively simple matter to determine, on the basis of inelastic time-history analyses, which rule is correct. Unfort-

Chapter 1: Introduction: A 1993 Viewpoint

Fig. 1.1 Displacement Spectra for Different Levels of Damping unately, we find that all are correct at some part of the period range of structural response, and all are wrong at other periods. This was explained[1] with reference to typical displacement spectra, reproduced here as Fig.1.1, where the influence of inelastic response was represented by a lengthening of the effective period of response, with hysteretic damping being represented as equivalent viscous damping. For short period structures, the increase in displacement response from period elongation is less than the decrease resulting from increased damping. For medium period structures, the two effects almost balance each other. For long period structures the period elongation does not result in significant displacement increase, and the influence of increased damping is to reduce the overall displacement response. For very long period structures, the displacement is equal to the ground displacement, independent of period and damping. It was also noted that the elastic acceleration approach placed excessive emphasis on the elastic stiffness characteristics of the structure and its elements. For reinforced concrete and masonry structures the estimation of these stiffness values varies greatly between different design codes. Further, these elastic characteristics only pertain to low level seismic response, and are permanently modified s soon as the structure exceeds yield.

Priestley: Myths and Fallacies in Earthquake Engineering, Revisited

1.2 THE REFINED ANALYSIS MYTH Structural analysis for design purposes had, by 1993, become more sophisticated than in previous decades, as design codes tended towards specifying 3-D modal analysis as a routine design tool. It was noted[1] that a consequence was that seismic design and analysis functions were frequently separated in the design office, and carried out by different people. It was argued that the increased sophistication in analysis was principally related to the availability of powerful computers rather than to a perceived inadequacy of earlier, and simpler analysis techniques. An undesirable side effect was that instead of analysis being used to verify design, it was driving the design. Sophisticated analysis were based on unrealistic estimates of member stiffness, with the designers role being reduced to that of detailer. The myth, then, was that refinement of the analysis process improved the end result. The following points were made: The use of modal combination of elastic response components together with a force-reduction or behaviour factor was of doubtful validity, since it was not clear that higher modes of response would be affected by ductility in the same way as was the fundamental mode. Deflection profiles from elastic modal analyses tended to underestimate drifts in the lower stories of frame buildings responding inelastically to seismic input. For reinforced concrete buildings, the approximations of member stiffness were typically so grossly inaccurate that any advantage from using advanced analysis techniques would be negated. The example was given of a reinforced concrete frame building (Fig.1.2(a)), where the stiffness of the outer columns at the ground floor level differed by a factor of two at first yield of flexural rein-

C Moment B A A B (a) Structure C Curvature (b) Column Stiffness

Fig. 1.2 Influence of Seismic Axial Force on Column Stiffness

Chapter 1: Introduction: A 1993 Viewpoint

forcement, as a result of the axial force variation caused by seismic axial load under first mode lateral response (see Fig. 1.2(b)). Because this influence does not extend to higher modes, where the compression column from first mode response now has equal probability of tension from any of the higher modes, it would seem to be impossible to allocate different stiffnesses to the columns when multi-modal analysis is considered. It must thus be expected that forces predicted in the columns by elastic modal analysis will be grossly in error. 1.3 THE STRENGTH = SAFETY FALLACY

It was shown that for many structures, including most reinforced concrete and masonry buildings and bridges, increasing the strength does not automatically improve safety at the design level of seismic response. If the strength increase is obtained merely by increasing the flexural reinforcement ratios while keeping member dimensions constant, then the displacement capacity of the structure will actually decrease as the strength increases, as a result of reduced ultimate curvature capacity. Since displacement capacity is more fundamental to damage control than is strength, it was argued that safety may well be improved by reducing, rather than increasing strength. 1.4 THE MYTH OF MAXIMIZED ENERGY ABSORPTION

As mentioned earlier, it has long been felt by designers and researchers that superior seismic performance will be obtained when the hysteretic force-displacement response of a structure absorbs the maximum possible energy. Thus emulating elastic/perfectlyplastic response as closely as possible has been seen as the ideal structural solution. In the original Myths and Fallacies paper[1], this attitude was disputed, noting that there are many cases where better response can be obtained with apparently less desirable loop shapes. Figure 1.3 compares three idealized hysteresis loop shapes elastic/perfectly-plastic, a degrading stiffness model typical of reinforced concrete, and a non-linear elastic response appropriate for a plastic hinge with unbonded prestressing tendons. Response is shown both with, and without P- effects. It was shown that for different loop shapes, but equal initial stiffness, the maximum response displacements were likely to quite similar, though the displacements of the elastic/perfectly-plastic would generally be smallest, and those of the non-linear elastic would generally be largest. However, when residual displacements were considered, the non-linear elastic was clearly the best solution, since it had zero residual displacement, and the elastic/perfectly plastic was the worst, since residual displacements for this hysteretic rule could be as much as 80% of the peak displacement. Residual displacements with the Takeda degrading stiffness model were less than those with the elastic/perfectly-plastic rule, as a consequence of the reduced unloading stiffness.

Priestley: Myths and Fallacies in Earthquake Engineering, Revisited

Fig. 1.3 Response Characteristics for Different Hysteretic Models Including P- Effect When P- effects (indicated in Fig. 1.3 by the dashed lines) were included in the analyses, it was found that the elastic/perfectly-plastic loop was inherently unstable under long-duration earthquake records, as a preferential direction of inelastic response was established once the first inelastic excursion had occurred. This can be seen in Fig. 1.3(a), where after an initial inelastic excursion and unloading the residual displacement is at point B. In subsequent elastic cycles, the yield envelope is asymmetrical because of the P- effect, and as a consequence, there is a higher probability that further inelastic response will occur in the same direction as the existing residual displacement. The degrading stiffness model (Fig. 1.3(b)) was found[3] to be much less susceptible to this creeping form of residual displacement, and of course the non-linear elastic hysteretic rule was completely insensitive to P- effects insofar as residual displacements were concerned. It was argued that provided damage levels are not unacceptably high, residual displacement are ultimately more important than maximum displacements, given the difficulty of straightening a bent building after an earthquake. Although the non-linear elastic concept of Fig 1.3(c) was only a theoretical concept in 1993, we will show later in this report that recent experimental results support these findings. 1.5 MYTHS AND FALLACIES RELATED TO REINFORCED CONCRETE DESIGN AND DETAILING Various aspects of reinforced concrete design and detailing were considered in the original paper to draw attention to various illogical practices. It was shown that the then (and still current) practice of concentrating flexural reinforcement in the top and bottom of beams in frame buildings (see Fig. 1.4(a)) made little sense when design moments were

Chapter 1: Introduction: A 1993 Viewpoint

(a) Conventional Reinforcement

(b) Vertically Distributed Reinforcement

Fig.1.4 Arrangements of Flexural Reinforcement in Beams dominated by seismic effects. The same flexural strength could be achieved by distributing the same amount of flexural reinforcement down the two sides of the beams, with the benefit of reducing congestion, improving joint shear performance, and also providing better control of shear deformations in beam potential plastic hinge regions. The use of artificially low estimates of dependable flexural strength of potential plastic hinge regions, by specifying flexural strength reduction factors, or partial material factors was shown to result in unnecessary cost increases without improving safety. Given that the moment capacity of plastic hinges is expected to develop at seismic intensities much lower than the design intensity, (as a consequence of design using force-reduction or behaviour factors), protection against yielding by use of conservative flexural strength estimates will be completely ineffective. A whole raft of problems with design for shear strength was identified. In particular, it was shown that incompatible assumptions were made for the seismic shear capacity of plastic hinge regions. In ductile beams of frame buildings, it was shown that the purpose of transverse reinforcement was not to aid in the development of a truss mechanism for shear transfer, but was to stabilize the flexural reinforcement against buckling, thus increasing its dowel action, which was the only available mechanism for shear transfer at high ductility levels. This was not, and is still not recognized in design equations for shear strength. Design equations for development of reinforcement were also shown to be of doubtful applicability in seismic regions. Partly this was due to the common attempt by codes to describe anchorage, splicing, and flexural bond by the same catch-all term of development. It was suggested that earlier, and simpler design equations based on simple multiples of bar diameter provided better fits with experimental data for many seismic applications[4].

Priestley: Myths and Fallacies in Earthquake Engineering, Revisited

1.6 SCOPE OF THE REPORT In Chapter 2,the elastic stiffness of reinforced concrete walls, columns, beams and frames is examined in some detail, basing the stiffness on conditions at effective first yield of the longitudinal reinforcement, or a limiting concrete compression strain above which nonlinear response of the concrete becomes excessive. This is effected by examining the yield curvature of typical reinforced concrete (or masonry) elements. Simple design recommendations are presented. Chapter 3 examines dynamic amplification of shear forces and moments in reinforced concrete structures, with primary emphasis on cantilever wall buildings. Current multimodal, and capacity design approaches are critically reviewed on the basis of non-linear time-history analyses. The potential for improved seismic performance resulting from the use of structural systems consisting of precast concrete elements connected by unbonded prestressing is examined on the basis of theoretical considerations and experimental evidence in Chapter 4. Of particular interest is the level of damage, at specified drift limits, compared with that for reinforced concrete structures, and the self-centering characteristic imparted by the use of unbonded prestressing. The concept of Direct Displacement-based Design, which was advanced in the original paper as a potential means for resolving many of the philosophical problems inherent in current force-based design is developed in Chapter 5, in much greater detail than in the original paper. It is emphasized that the method, which emphasizes design for a specified displacement, and considers strength as a secondary parameter, is extremely simple to apply in practice, when compared with current seismic design requirements. Some of the implications of the design approach are contrasted with corresponding, and contrasting conclusions from force-based design.

CHAPTER 2: THE ELASTIC STIFFNESS OF CONCRETE MEMBERS


2.1 INTRODUCTION In section 1.2, problems with the recent tendency towards increased complexity in seismic design were briefly discussed. Amongst these was the incompatibility of refined analysis techniques, such as multimode analysis, with crude approximations to member stiffness, particularly for reinforced concrete and masonry members, where the effects of cracking significantly reduce the stiffness below the initial uncracked value. It was pointed out that the elastic cracked-section stiffness of concrete columns was a function of the axial load ratio, and for columns of a frame structure can vary by as much as +/50% or more during seismic response. As a consequence, modal analysis, based on specified, and constant stiffness is unable to provide accurate estimates of seismic forces, even within the elastic range of response. Calculated elastic periods are likely to be in error, and more importantly, the distribution of forces through the structure, which depends on relative stiffness of members, can be grossly in error. These comments refer to elastic response. It would appear that even greater errors will be induced when the response is inelastic, since some members will remain elastic, while others respond in ductile fashion. Consequently, relative stiffness variations will diverge from the elastic values as inelastic response develops. It would thus seem that what has been termed the Principle of Consistent Crudeness has been violated. There is little point in sophisticated analysis if it is based on very coarse and inaccurate data. The coarseness in stiffness estimates used in seismic design is apparent when different design codes are considered. In many cases, the stiffness is still taken as the gross section (uncracked) stiffness. This would appear to be inappropriate, since cracking of critical elements such as beams will normally have occurred under gravity loading. Even if no previous cracking has occurred prior to the design seismic level of excitation (unlikely, since the probability is that the design level of seismic excitation will be preceded by a number of events of lesser intensity), cracking will occur early, and the stiffness will rapidly reduce. The uncracked stiffness will never be fully regained during, or after the seismic response, and thus is not a useful estimate of effective stiffness. In other codes, some recognition of the influence of cracking is made, by specifying reduction factors to be applied to the gross-section stiffness. Typically a blanket value of 0.5 is used. In a few codes, notable the New Zealand concrete design code[5], different adjustment factors are specified for different member types. Thus a reduction factor of

10

Priestley. Myths and Fallacies in Earthquake Engineering, Revisited

0.35 is applied to the gross-section stiffness of beams, while a value between 0.4 and 0.7 is specified for columns. This is clearly an improvement on the use of gross-section values, but is still inadequate to represent stiffness to a degree of precision adequate to justify modal analysis, since the influence of flexural reinforcement ratio is not included. Regardless of what assumption is made, the member stiffnesses are assumed to be independent of flexural strength. The section stiffness can be assessed from the moment-curvature relationship in accordance with the beam equation:

EI =

MN y

(2.1)

where MN is the nominal moment capacity of the section, and y is the yield curvature of the equivalent bilinear representation of the moment-curvature curve. This can be explained with reference to Fig. 2.1, which shows a typical moment-curvature relationship
50000
Mu

40000 MN
Moment (kNm)

30000 My 20000

10000

'y y
0 0.02
Curvature (1/m)

u
0.04 0.06

Fig. 2.1 Moment-Curvature Relationship and Bi-Linear Approximation

Chapter 2: Elastic Stiffness of Concrete Members

11

together with a bilinear approximation. The curve in Fig. 2.1 is the moment-curvature response of a square 1.6m x 1.6m column, with moderate axial load and reinforcement content, and might be appropriate for a bridge column, or the base of a tall building. It has become accepted by the research community that the most appropriate linearization of moment-curvature relationships is by an initial elastic segment passing through first yield, and extrapolated to the nominal flexural strength, MN, and a postyield segment connected to the ultimate strength and curvature. First yield of the section is defined as the moment, My and curvature y when the section first attains the reinforcement tensile yield strain of y = fy/Es , or the concrete extreme compression fibre attains a strain of 0.002, whichever occurs first. The nominal flexural strength MN develops when the extreme compression fibre strain reaches 0.004, or the reinforcement tension strain reaches 0.015, whichever occurs first. Thus the yield curvature is given by

y = ' y (M N / M y )

(2.2)

Member stiffness depends on the distribution of curvature along the entire member length not just at the critical section where yield will occur. However, experimental evidence indicates that basing the member stiffness on the assumption that curvature varies linearly with moment along the member is adequately accurate. This is partly because slight reductions in curvature, below the linear assumption, for sections with reduced moment are balanced by increased deformation due to shear effects and strain penetration into supporting members. Also, after one or two cycles at or close to yield, the moment-curvature response of all cracked sections up to yield approaches linearity. Examination of Eq (2.1) reveals that the common design assumption that member stiffness is independent of strength implies that the yield curvature is directly proportional to flexural strength:

y = M N / EI

(2.3)

This assumption is illustrated in Fig. 2.2(a). Experimental evidence, and detailed analytical results, indicate that the assumption of stiffness being independent of strength is not valid. In fact, yield curvature is effectively independent of strength, and hence the stiffness is directly proportional to the flexural strength, as is seen from Eq (2.1) with y a constant. The correct relationship is thus illustrated in Fig. 2.2(b). As a consequence of these findings, it is not possible to perform an accurate analysis of either the elastic structural periods, nor of the elastic distribution of required strength through the structure, until the final member strengths have been determined. At the very least, this implies that conventional seismic design, based on elastic member stiffnesses and force-based considerations should be an iterative process, where the member stiffnesses are upgraded during each iteration. Data justifying the simplistic representation of Fig. 2.2(b) are presented for columns walls, beams and frames in the following sections.

12

Priestley. Myths and Fallacies in Earthquake Engineering, Revisited

M1 M M2 M

M1

M2

M3

M3

y2 y1 y3 (a) Design Assumption (constant stiffness)

y (b) Realistic Conditions (constant yield curvature)

Fig.2.2 Influence of Flexural Strength on Moment-Curvature Relationship 2.2 ELASTIC STIFFNESS OF CIRCULAR COLUMNS Circular reinforced concrete columns are the most common lateral force-resisting elements for bridges in seismic regions[4]. In order to investigate the effective stiffness of circular columns, a parameter analysis was carried out varying the axial load ratio and flexural reinforcement ratio for a typical bridge column. The following basic data were assumed: Column diameter D = 2m Cover to flexural reinforcement = 50 mm Concrete compression strength fc = 35 MPa Flexural Reinforcement diameter = 40 mm Transverse reinforcement: spirals = 20mm at 100mm spacing Reinforcement yield strength fy = 450 MPa Axial Load Ratio Nu/fcAg = 0 to 0.4 (9 levels) Flexural reinforcement Ratio l/Ag = 0.005 to 0.04 (5 levels) Analyses were carried out with a program Cirman4 which represents the section by a minimum of 30 parallel layers. Concrete strains in each layer were calculated from stressstrain relationships that differentiated between cover and core concrete, and considered

Chapter 2: Elastic Stiffness of Concrete Members

13

the effects of lateral confinement on the concrete compression stress-strain relationship. Reinforcement stress strain characteristics included a yield plateau and strain-hardening effects. Note that concrete confinement and reinforcement strain hardening have comparatively minor effects on the yield curvature or nominal moment capacity, and mainly influence ultimate curvature and ultimate moment (see Fig. 2.1). A selection of the computed moment-curvature curves is shown in Figs. 2.3 for two levels of flexural reinforcement ratio, and a range of axial load ratios. Only the initial part of the moment-curvature curves has been included, to enable the region up to, and immediately after yield to be clearly differentiated. Also shown in Fig. 2.3 are the calculated bi-linear approximations for each of the curves. Note that the apparent overestimation by the bi-linear representations of the actual curves is a function of the restricted range of curvature plotted. If the full moment-curvature curve is represented, the agreement is similar to that shown in Fig. 2.1. It will be seen that the moment capacity is strongly influence by the axial load ratio, and also by the amount of reinforcement. However, the yield curvature of the equivalent bi-linear representation of the moment-curvature curves does not appear to vary much

50000
Nu/f'cAg = 0.4 Nu/f'cAg = 0.3 Nu/f'cAg = 0.2

40000
Nu/f'cAg = 0.4

40000

Nu/f'cAg = 0.1 Nu/f'cAg = 0

30000
Moment (kNm)

Moment (kNm)

Nu/f'cAg = 0.3 Nu/f'cAg = 0.2

30000

20000

Nu/f'cAg = 0.1

20000

Nu/f'cAg = 0

10000

10000

0 0 0.002 0.004 0.006

0 0 0.002 0.004 0.006

Curvature (1/m) (a) Reinforcement Ratio = 1%

Curvature (1/m) (b) Reinforcement Ratio = 3%

Fig. 2.3: Selected Moment-Curvature Curves for Circular Columns (D=2m, fc= 35MPa, fy = 450MPa)

14

Priestley. Myths and Fallacies in Earthquake Engineering, Revisited

between the curves. Data from the full set of analyses for nominal moment capacity, and equivalent bilinear yield curvature are plotted in dimensionless form in Fig. 2.4. The dimensionless nominal moment capacity is defined as

M DN =

MN f 'c D 3

(2.4)

and the dimensionless yield curvature as

Dy = y D / y

(2.5)

where y is the flexural reinforcement yield strain. The influence of both axial load ratio and reinforcement ratio on the nominal moment capacity is, as expected, substantial in Fig. 2.4(a), with an eight-fold range between maximum and minimum values. On the other hand, it is seen that the dimensionless yield curvature is comparatively insensitive to variations in axial load or reinforcement ratio. Thus the yield curvature is insensitive to the moment capacity. The average value of dimensionless curvature of Dy = 2.25 is plotted on Fig 2.4(b), together with lines at 10% above and 10% below the average. It is seen that all data except those for low reinAve. +10%
l = 0.04

Dimensionless Moment (MN/f'cD3)

l= 0.04 l = 0.03 l = 0.02 l = 0.01

0.16 0.12 0.08 0.04 0 0 0.1

Dimensionless Curvature (yD/y)

0.2

2.5 2 1.5 1 0.5 0 0 0.1

Ave. -10%
l = 0.005

l = 0.005

Average yD/y= 2.25

0.2

0.3

0.4

0.2

0.3

0.4

Axial Load Ratio (Nu/f'cAg) (a) Nominal Moment

Axial Load Ratio (Nu/f'cAg) (b) Yield Curvature

Fig. 2.4: Dimensionless Nominal Moment and Yield Curvature for Circular Bridge Columns

Chapter 2: Elastic Stiffness of Concrete Members

15

forcement ratio coupled with very high axial load ratio fall within the +/- 10% limits. It should be noted that though the data were generated from a specific column size and material strengths, the dimensionless results can be expected to apply, with only insignificant errors, to other column sizes and material strengths within the normal range expected for standard design. The results would not, however apply to very high material strengths (say fc>50MPa, or fy>600MPa) due to variations in stress-strain characteristics. The data in Figs. 2.3 and 2.4 can be used to determine the effective stiffness of the columns as a function of axial load ratio and reinforcement ratio, using Eq. (2.1). The ratio of effective stiffness to initial un-cracked section stiffness is thus given by

EI / EI gross =

MN y EI gross

(2.6)

Results are shown in Fig. 2.5 for the ranges of axial load and reinforcement ratio considered. It will be seen that the effective stiffness ratio varies between 0.13 and 0.91.
1

0.8
Stiffness Ratio (EI/EIgross)

l = 0.04 l = 0.03

0.6
l = 0.02 l = 0.01 l = 0.005

0.4

0.2

0 0 0.1 0.2
Axial Load Ratio (Nu/f'cAg)

0.3

0.4

Fig. 2.5: Effective Stiffness of Circular Bridge Columns

16

Priestley. Myths and Fallacies in Earthquake Engineering, Revisited

It should be noted that for convenience in computing the stiffness ratios of Fig 2.5, the gross stiffness of the un-cracked stiffness has been calculated without including the stiffening effect of the flexural reinforcing steel. That is

I gross =

D 4 64

(2.7)

Since the reinforcement increases the un-cracked section moment of inertia by as much as 60% for the maximum steel ratio of 4%, the stiffness ratios related to true uncracked stiffness would be lower, particularly for the higher reinforcement ratios. The value of the concrete modulus of elasticity used in computing Fig. 2.5 was

E = 5000 f ' c

(MPa)

(2.8)

2.3 ELASTIC STIFFNESS OF RECTANGULAR COLUMNS A parallel study to that described in the previous section was carried out for rectangular concrete columns. Ductile rectangular columns can occur in bridge design, and at the base of columns of multi-storey frame buildings. For the purposes of this study the special case of a square column with flexural reinforcement evenly distributed around the perimeter was investigated. The following basic data were assumed: Column dimensions B = h = 1.6 m Cover to flexural reinforcement = 50mm Concrete compression strength fc = 35 MPa Flexural reinforcement diameter = 32 mm Transverse reinforcement: hoops = 20mm dia./5 legs Reinforcement yield strength fy = 450 MPa Axial load ratio Nu/fcAg = 0 to 0.4 (9 levels) Flexural reinforcement ratio l/Ag = 0.005 to 0.04 (5 levels) The analyses for rectangular sections were carried out with a program Recman2 which was based on similar principles to the program Cirman4 used for the circular analyses of Section 2.2. A selection of the computed moment-curvature curves for the rectangular sections is presented in Fig. 2.6 for two levels of reinforcement ratio, and a range of axial load ratios. As with the curves for circular columns presented in Fig. 2.3, only the initial part of the moment-curvature curve, up to about 10% of ultimate curvature, but significantly beyond initial yield, has been included. Also shown in Fig. 2.6 are the bilinear representations of the moment-curvature curves based on the principles outlined in Section 2.1 and Fig. 2.1. The strong influence of axial load ratio and reinforcement ratio on moment capacity apparent for circular sections is also apparent with rectangular sections, as expected.

Chapter 2: Elastic Stiffness of Concrete Members


40000
Nu/f 'cAg = 0.4 0.3 0.2

17

30000
0.1

Moment (kNm)

25000 20000
Moment (kNm)

Nu/f'cAg = 0.4 Nu/f 'cAg = 0.3 Nu/f 'cAg = 0.2 Nu/f 'cAg = 0.1

Nu/f 'cAg = 0

20000

15000 10000 5000 0 0

Nu/f 'cAg = 0

10000

0 0.001 0.002 0.003 0.004 0.005 0 0.002 0.004 0.006

Curvature (1/m) (a) Reinforcement Ratio = 1`%

Curvature (1/m) (b) Reinforcement Ratio = 3%

Fig. 2.6 Selected Moment-Curvature Curves for Rectangular Columns (b=1.6m, hc=1.6m, fc= 35MPa, fy = 450MPa) Also, as noted for circular sections, the effective yield curvature of the equivalent bilinear approximation to the moment-curvature relationship does not appear to be significantly dependent on axial load ratio or reinforcement ratio. Data from the full set of analyses for nominal moment capacity, and equivalent bilinear yield curvature are plotted in dimensionless form in Fig. 2.7. The dimensionless nominal moment capacity is defined as

M DN =

MN f ' c bh 2

(2.9) The dimensionless

where b and h are the column width and depth respectively. curvature is

Dy = y h / y

(2.10)

Trends for the rectangular columns, apparent in Fig. 2.7, are similar to those displayed in

18
Dimensionless Moment (MN/f'cbh2)

Priestley. Myths and Fallacies in Earthquake Engineering, Revisited


0.3 2.5 2 1.5 1 0.5 0 0 0.1 0.2 0.3 0.4
Average yh/y = 2.10

l = 0.04 l = 0.03

Dimensionless Curvature (yh/y)

Ave. + 10% Average Ave. - 10%

l = 0.04

0.2

l = 0.02 l = 0.01

l = 0.005

0.1

l = 0.005

0 0 0.1 0.2 0.3 0.4

Axial Load Ratio (Nu/f'cAg) (a) Nominal Moment

Axial Load Ratio (Nu/f'cAg) (b) Yield Curvature

Fig.2.7 Dimensionless Nominal Moment and Yield Curvature for Large Rectangular Columns Fig.2.4 for circular columns. Nominal moment capacity is strongly dependent on both axial load ratio and reinforcement ratio, with approximately an eight-fold increase in moment capacity from minimum axial load and reinforcement ratio to maximum axial load and reinforcement ratio. Dimensionless yield curvature is only weakly dependent on axial load ratio and reinforcement ratio, thus implying that the yield curvature is insensitive to the nominal moment capacity. The average value of dimensionless curvature of Dy =2.10 is plotted on Fig.2.7(b), together with lines at 10% above and 10% below the average. It is seen that all data except those for l =0.005 at both low and high axial load ratio fall within the +/- 10% limits of the average value. As with the circular columns, the dimensionless results of Fig.2.7 can be expected to apply to other column sizes and material strengths within the normal range of material strengths. Small errors can be expected for small column dimensions, where the ratio of cover to core dimensions will be significantly larger than for the data presented here. As with the circular column data, results should not be applied to rectangular columns with very high strength concrete or reinforcement. The data of Figs.2.6 and 2.7 have been used to develop curves for the effective stiffness ratio, based on Eq.(2.6). Results are presented in Fig.2.8. For ease of application of the results, the gross-section un-cracked stiffness was computed ignoring the stiffening effect of flexural reinforcement as

I gross =

bh 3 12

(2.11)

Chapter 2: Elastic Stiffness of Concrete Members

19

0.8
Stiffness Ratio (EI/EIgross)
l = 0.04 l = 0.03 l = 0.02

0.6

0.4

l = 0.01 l = 0.005

0.2

0 0 0.1 0.2
Axial Load Ratio (Nu/f'cAg)

0.3

0.4

Fig. 2.8 Effective Stiffness of Large Rectangular Columns Concrete modulus of elasticity was computed using Eq.(2.8). The range of effective stiffness is from 0.12 to 0.86 times the gross section stiffness, calculated in accordance with Eq.(2.11), indicating the strong dependence of effective stiffness on axial load ratio and reinforcement ratio. Clearly the common assumption of a constant section stiffness independent of flexural strength is entirely inappropriate. Results from Fig.2.8 can be applied to other column sizes than those used to generate the graph by appropriate substitution of section dimensions into Eq.(2.11). 2.4 ELASTIC STIFFNESS OF RECTANGULAR WALLS Similar calculations to those reported in the previous two sections can also be carried out for rectangular structural walls, which are common lateral-force resisting elements in seismic design of multi-storey buildings. These have been fully presented elsewhere[6]. Analyses considered two separate cases one where the flexural reinforcement was distributed uniformly along the wall length, and the second, more common case where most of the flexural reinforcement was concentrated at the wall ends, with a comparatively light reinforcement distributed along the wall length. It is emphasized that

20

Priestley. Myths and Fallacies in Earthquake Engineering, Revisited

though the latter is the more common case, this is largely because of the misconception that concentrating the flexural reinforcement at the wall ends increases the flexural capacity when compared with the same total amount of reinforcement distributed uniformly along the wall length. In fact, the flexural strength of the two distributions will be almost identical, as simple trial calculations will show. As with the distribution of flexural reinforcement in beams of frame buildings, discussed in Section 1.5, shear performance will be enhanced by uniform distribution of reinforcement, which should thus be preferred over concentration at the wall ends. Results are presented in Fig.2.9 for the dimensionless yield curvature for walls with different axial load ratios and reinforcement ratios, in similar fashion to the above analyses for columns. However, lower ranges of axial load ratio (with a maximum of Nu/fcAg= 0.12), and reinforcement ratio ( with a maximum of l = 0.02) were adopted since these were considered to be practical upper limits for structural walls. A reinforcement ratio of 0.005 was considered to be uniformly distributed along the wall length, with the remainder concentrated near the wall ends. In Fig.2.9 the yield curvature has been made dimensionless by multiplying by the wall length lw, and dividing by the yield strain y of the flexural reinforcement, in similar fashion to columns in Eqs.(2.5) and (2.10).

Fig. 2.9: Dimensionless Yield Curvature for Rectangular Walls

Chapter 2: Elastic Stiffness of Concrete Members

21

An average value of

Dy = y l w / y = 2.0

(2.12)

is plotted on Fig.2.9, together with lines at average +5% and average 5%. It is seen that except for the case of l = 0.005 (which in this case means uniformly distributed reinforcement without additional end reinforcement), coupled with low axial load, all data conform to the average +/- 5%. Analyses where all the flexural reinforcement was distributed uniformly along the wall length resulted in an average dimensionless curvature approximately 10% higher than given by Eq.(2.12), with about twice the scatter of Fig. 2.9[6]. The effective stiffness for rectangular walls can thus be calculated, as a fraction of gross wall stiffness as

EI / EI gross =

MN 1 3 ( Dy y / lw ) E (tlw / 12)

(2.13)

where t is the wall width, and MN is the calculated nominal flexural strength. In Eq. (2.13), Dy is given by Eq. (2.12) for walls with most of the flexural reinforcement concentrated at the wall ends, or by 1.1 times the value from Eq. (2.12) when all the reinforcement is uniformly distributed along the wall length. 2.5 ELASTIC STIFFNESS OF FLANGED BEAMS In order to investigate the influence of flexural reinforcement content on the yield curvature and effective stiffness of beams incorporated in ductile building frames, the typical beam section of Fig. 2.10 was considered in an earlier study[7]. Top and bottom

Fig. 2.10: Beam Section Considered for Analysis

24

Priestley. Myths and Fallacies in Earthquake Engineering, Revisited

effective stiffness ratio is highly dependent on the reinforcement ratio, though axial load is not considered relevant for beams in frame design. 2.6 YIELD DRIFTS OF FRAMES The information provided in the previous sections is sufficient to enable realistic estimates of frame member elastic stiffnesses to be made, provided reinforcement content and axial load ratios are known. It will be noted that since dimensionless yield curvatures can be considered to be constant for a given section type, the effective stiffness can be directly estimated, with considerable accuracy, once the nominal flexural strength is determined, using Eq.(2.1). This would enable refinement of the analysis process for seismic design, when force-based procedures are used. The comparative invariance of dimensionless yield curvature indicates that yield drift of frames might similarly be essentially independent of reinforcement ratio and strength. Fig.2.11(a) shows a typical beam/column subassemblage extending half a bay width either side of the joint, and half a storey height above and below the joint. This can be considered a characteristic element of a frame building. Since bay width will normally exceed storey height, and column curvatures will typically be less than beam curvatures as a consequence of capacity design procedures[8], beam flexibility is likely to be the major contributor to the deformation. The deflected shape is shown in Fig.2.11(b). The yield drift y can be expressed as

y = by + jy + 2 c / lc + 2 s / lc

(2.15)

where by and jy are the rotations of the joint center due to beam flexure and joint shear deformation respectively, c is the flexural deformation of the column top relative to the tangent rotation at the joint center, and s is the additional deformation of the column top due to shear deformation of beams and columns. To allow for strain penetration of longitudinal reinforcement into the joint region, it is assumed that the yield curvature in the beam develops at the joint centroid, and reduces linearly to zero at the beam midspan, as shown in Fig.2.11(c). The yield drift due to beam flexure is thus:

by =

y (0.5lb ) y lb = 3 6

(2.16)

Ignoring strain-hardening, and thus substituting from Eq.(2.14(a)):

by = 0.283 y

lb hb

(2.17)

Chapter 2: Elastic Stiffness of Concrete Members

25

Fig. 2.11: Elastic Deformation Contributions to Drift of a Beam/Column Joint Subassemblage

26

Priestley. Myths and Fallacies in Earthquake Engineering, Revisited

Typical calculations based on a storey height/bay length ratio of 0.533 (storey height = 3.2m, bay length =6m) and a maximum column curvature of 0.75y indicate column displacement c will add about 40% to the yield drift in Eq. (2.15). It is further assumed, based on experience, that the joint deformation and member shear deformation add 25% and 10% respectively to the yield drift. As a consequence, the yield drift is predicted to be

y = (1.0 + 0.4 + 0.25 + 0.1) * 0.283 y


= 0.5 y
lb hb

lb hb
(2.18)

Equation (2.18) is compared in Fig.2.12 with the results from 46 beam/column test assemblages which included a wide range of possibly relevant parameters[7], including Column height/beam length aspect ratio (lc/lb) Concrete compression strength (fc) Beam Reinforcement yield strength (fy) Maximum beam reinforcement ratio (As/bwd) Column axial load ratio (Nu/fcAg) Beam aspect ratio (lb/hb) : 0.4 0.86 : 22.5MPa 88MPa : 276MPa 611MPa : 0.53% 3.9% : 0 0.483 : 5.4 12.6

Fig. 2.12 Experimental Drifts of Beam/Column Test Units Compared With Eq.(2.18)

Chapter 2: Elastic Stiffness of Concrete Members

27

Test units with equal, and with unequal top and bottom reinforcement ratios were considered, as were units with and without slabs and/or transverse beams framing into the joint. Note that Eq.(2.18) only includes two of these parameters (beam reinforcement yield strength (which dictates the yield strain y = fy/Es), and beam aspect ratio), on the assumption that the other parameters are not significant variables. Clearly this is an approximation, and some error is expected. It should also be noted that a few test results were discarded because of lack of joint reinforcement and premature joint failure, resulting in excessive yield drifts. Also, a few units with unacceptably high ratios of beam bar diameter to joint depth (db/hc), resulting in slip of beam bars through the joint prior to yield were also discarded. Full details of the test units are available in [7]. As is apparent from Fig.2.12, the agreement between experimental drifts and predictions of Eq.2.18 is reasonable over the full, and rather wide range of yield drifts. The average ratio of experiment to theory is 1.03 with a standard deviation of 0.16. Considering the wide range of parameters considered, the comparatively narrow scatter is rather satisfactory. In order to examine the sensitivity of the ratio of experimental drift to theoretical drift (/T) to different potentially significant parameters, it is plotted against reinforcement yield strength, maximum beam reinforcement ratio, beam aspect ratio, and test unit aspect ratio in Figs. 2.13(a), 2.13(b), 2.13(c), and 2.13(d) respectively. In no figure is a strong trend apparent, though there appears to be a trend for E/T to increase with increasing test unit aspect ratio, and to decrease with increasing beam aspect ratio and reinforcement yield strength. Most of these trends are expected: increasing column height relative to beam length will increase the contribution of column deformation to yield drift, and shear deformation will increase for low beam aspect ratio. The extent of the influence is, however, not sufficient to justify increasing the complexity of the very simple Eq.(2.18). Sensitivity to concrete compression strength and column axial load was similarly low[7]. It is of interest to investigate the probable levels of yield drift to be expected in modern design. Examination of Fig.2.12 indicates a range from 0.44% to 1.5%. These are generally much larger than assumed for seismic design. However, many of the test units investigated in Figs. 2.12 and 2.13 have proportions that would be considered unusual for current design practice. If we consider a minimum probable bay length of 6.0m, and a typical beam depth of 600mm, then Eq.(2.18) will predict yield drifts of 1.03% and 1.25% for frames with beam reinforcement of 414MPa (North American practice) and 500MPa (European practice) respectively. Since it is common to specify maximum allowable drifts of between 2.0% and 2.5% under the design level earthquake, the realistic maximum storey drift ductility demand must be in the range of = 1.6 to 2.5. Since structural displacement ductility demand will always be less than the drift ductility demand of the critical storey, design ductility levels should be less than these values. As a consequence it will almost never be possible to make full use of the behaviour factors (displacement ductility limits) specified in most seismic design codes, if the drift limits are to be maintained, and realistic estimates of elastic stiffness used.

28

Priestley. Myths and Fallacies in Earthquake Engineering, Revisited

(a) Beam Reinforcement Yield Strength (fy)

(b) Beam Top Reinforcement Ratio (As/bwd) Fig. 2.13 Sensitivity of Experiment/Theory Drift Ratio for Beam/Column Units to Various Parameters, Based on Eq.(2.18)

Chapter 2: Elastic Stiffness of Concrete Members

29

(c) Beam Length/Depth Ratio (Lb/hb)

(d) Test Unit Column Height/Beam Length Aspect Ratio Fig. 2.13 (cont.) Sensitivity of Experimental/Theoretical Drift Ratio for Beam/Column Units to Various Parameters, based on Eq. (2.18)

Chapter 2: Elastic Stiffness of Concrete Members

31 (2.19a) (2.19b) (2.19c) (2.19d)

Circular column: Rectangular column: Rectangular cantilever walls: T-Section Beams:

y = 2.25 y / D y = 2.10 y / hc y = 2.00 y / l w y = 1.70 y / hb

For reinforced concrete frames, the yield drift can be expressed, with adequate accuracy as

y = 0.5 y

lb hb

(2.20)

It is noted that the analyses reported in this chapter identify a fundamental problem with current force-based design: Design: Strength is allocated in proportion to stiffness Analysis: Stiffness is proportional to strength

CHAPTER 3: MORE PROBLEMS WITH MULTI-MODAL ANALYSIS


3.1 INTRODUCTION In Chapters 1 and 2 we identified major problems with multi-modal analysis as a basis for seismic design. Principal amongst these was inability to define stiffness prior to completion of design, and also inability to model stiffness changes associated with seismic-induced variations in axial force. It was also mentioned that the concept of using a relatively sophisticated elastic analysis procedure to describe the inelastic response of a structural system could be of doubtful validity. In light of these concerns, it is of interest to examine the principal advantages cited in support of use of multi-modal analysis. These are: It provides improved representation of higher mode effects, which are inadequately considered in simple equivalent lateral force procedures. When used in 3-D structural modeling, it enables torsional response to be included when determining design forces. Before examining these claims, it is relevant to briefly outline how these aspects (higher mode effects; torsional response) are considered in equivalent lateral force procedures. 3.1.1 Higher Mode Effects in Equivalent Lateral Force Design In equivalent lateral force design, the required flexural strengths at potential plastic hinges are determined from a simplified representation of the first-mode inelastic force distribution. Typically, and conservatively, 100% of the structure mass is considered to act in the first mode. Required strengths at other locations, or due to actions other than flexure, are found from capacity design considerations[8]. Basic strengths SE for these locations and actions corresponding to the first-mode force distribution are amplified by an overstrength factor o to account for potential flexural overcapacity at the plastic hinge locations resulting from higher than anticipated material strengths, and by a dynamic amplification factor to represent the potential increase of required strength due to higher mode effects. The required strength is thus

S R = oS E

(3.1)

Chapter 3: More Problems with Multi-Modal Analysis

33

For the required moment capacity of cantilever walls, the base moment is amplified to account for material overstrength, and a linear distribution of moments is generally adopted up the wall height to account for higher mode effects. As is apparent from Fig. 3.1(a), this implies higher amplification of moments at midheight than at the base or top of the wall. Reinforcement cut-off is determined by consideration of tension shift effects. This is achieved by vertical offset of the moment profile. The design moments thus do not exactly follow the form of Eq. (3.1). Shear forces corresponding to the Equivalent Lateral Force distribution are amplified by the flexural overstrength factor, and the dynamic amplification factor v directly in accordance with Eq. (3.1), as shown in Fig. 3.1(b). This factor has been obtained from previous research, and is presented elsewhere[8] in the following form:
V = 0.9 + n / 10 V = 1.3 + n / 30
for n 6 for 6 < n < 15

(3.2)

where n is the number of stories in the wall, and need not be taken greater than 15. For frame structures, beam shear forces, and moments at locations other than potential plastic hinges are amplified by the flexural overstrength factor. Since higher modes are not normally considered for beam design, the dynamic amplification factor is not normally included. However, it should be noted that vertical response is essentially a
1 0.8 0.6 0.4 0.2 0 0 0.4 0.8 Dimensionless Moment 1.2 1 0.8
Dimensionless height

Dimensionless Height

Tension Shift

Linear + Overstrength Design Forces

0.6 0.4 0.2 0 0

+ Dynamic Amplification

Design Forces

Overstrength

0.5 1 1.5 2 Dimensionless Shear Force

2.5

(a) Moment Profiles

(b) Shear Force Profiles

Fig. 3.1 Dynamic Amplification of Design Forces for Equivalent Lateral Force Design Of Cantilever Walls

34

Priestley. Myths and Fallacies in Earthquake Engineering, Revisited

higher mode, and may amplify the gravity moments considerable. A strict formulation of capacity design would take this into account. Column end moments and shear forces are amplified for both beam plastic hinge overstrength and dynamic amplification. For one-way frames, upper limits for the dynamic amplification of column moments of 1.80 have been recommended, with 1.3 for column shear forces. Further amplification for beam flexural overstrength is required[8]. 3.1.2 Torsional Effects in Equivalent Lateral Force Design The consideration of torsion in the design of building structures is well established[8], and is only briefly summarized here for design to the equivalent lateral force method. The steps in the design process are as follows: 1. Determine the direct storey force to be distributed at each floor level, between the different lateral-force resisting elements. 2. Determine relative elastic stiffnesses of all lateral-force resisting elements, and allocate the storey force to the elements in proportion to their stiffnesses. 3. Determine the plan location of the center of elastic stiffness. 4. Determine the torsional polar moment of inertia, considering all parallel and orthogonal lateral-force resisting elements. 5. Determine the calculated eccentricity e between the centers of mass and stiffness for the storey under consideration. Consider variations in e to account for accidental torsion. 6. From 1 and 5, determine the maximum and minimum torsional moments. 7. Determine the lateral forces in the lateral-force resisting elements to resist the maximum and minimum torsional moments. 8. For each lateral-force resisting element determine the worst possible combination of direct and torsional lateral force, and design the element for this. 3.2 DESIGN HIGHER MODE EFFECTS BY MULTI-MODE ANALYSIS When multi-modal analysis forms the basis of seismic design, the higher mode effects are directly considered in the analysis, since all significant elastic modes, rather than just the fundamental mode, are included when determining the design forces. The individual response of each significant mode, in terms of lateral forces, element bending moments or shear forces, and displacements are found from direct modal combination of the relevant action or displacement. The procedure may be outlined as follows: The mode shapes im and periods Tm of each significant mode m are calculated, where i are the individual floors or mass locations. The portion of the total structural mass m participating in each significant mode is calculated as

Chapter 3: More Problems with Multi-Modal Analysis

35

im mi m =
i ,m =1 N i ,m =1

1
N

(3.3)

mi
i =1

2 im

mi

where mi are the masses at locations i. obtained as


N

The modal base shear for each mode is then

VBm = Sa m g m
i =1

mi

(3.4)

where Sam is the acceleration coefficient corresponding to mode m, and g is the acceleration of gravity. Then, the modal base shear is distributed up the structure following
Fim = V Bm

im mi
N

(3.5)

im mi
i , m =1

The modal displacements may be found using Equation (3.6), where Tm is the period of vibration of mode m:
m = Sa m g
2 Tm 4 2

(3.6)

Finally, the modal displacement at each floor i can be computed as


N

im mi m
im = im
i , m =1 N 2 im mi

(3.7)

i , m =1

Elastic-response modal moments and shear forces are calculated from the modal forces given by Eq. (3.5), and are then combined in accordance with established modal Adopting the SRSS combination for combination rules such as SRSS or CQC[9]. simplicity, the elastic forces and displacements are thus given by
N

Fie =

2 Fim , M ie =

2 M im , Vie =

2 Vim , and ie =

2im

(3.8)

i ,m =1

i ,m =1

i ,m =1

i ,m =1

36

Priestley. Myths and Fallacies in Earthquake Engineering, Revisited

In Eqs. (3.4), (3.7), and (3.8), N is the number of significant modes considered. The design forces are then determined by dividing the elastic forces by the behaviour (force-reduction) factor, R, and the design displacements are taken as equal to the elastic displacements. Thus:

FiD =

Fie M V , M iD = ie , ViD = ie , and iD = ie R R R

(3.9)

The assumption inherent in the above formulation is that the behaviour factor R applies equally to all modes. A corollary of this is that design forces of a designed structure are unaffected by the seismic intensity, provided that the intensity is sufficient to induce inelastic response. This is because (say) doubling the seismic intensity will be equivalent to doubling the behaviour factor. Although the elastic response forces, moments and shears will all double, the strengths of the plastic hinges will remain as designed, and hence the effective behaviour factor for the increased intensity must increase proportionately. Since the increased behaviour factor applies to all modes, the design forces, given by Eq. (3.9) are unchanged. Since the displacements are equal to the elastic displacements, however, the displacement response (and hence the displacement ductility demand) will be increased in direct proportion to the increase in seismic intensity. 3.2.1 Higher Mode Effects in Cantilever Walls from Time-History Analyses An analytical research project was recently carried out[10] to investigate the influence of seismic intensity on the higher mode response of cantilever structural walls, and hence to investigate, indirectly, the validity of design forces determined by multi-mode analysis. Six walls, from 2 storeys to 20 stories were designed to the Eurocode 8 elastic acceleration response spectrum of Fig.3.2, which is compatible with a peak ground acceleration of 0.4g, and a medium soil condition (subsoil class B, characterized by deep deposits of medium dense sand, gravel, or medium stiff clays with thickness from several tens to many hundreds of metres). The walls (see Fig.3.3) all had the same tributary floor mass of 60 tonnes, and gravity load of 200 kN at each level, and were designed in accordance with direct-displacement based design principles[11] to achieve maximum drifts (interstorey displacement divided by interstorey height) of approximately 0.02 at the critical top storey, under the design levels of seismic intensity. Wall lengths (lw), widths (b), reinforcement contents (l) and bar sizes (db) varied from wall to wall in order to satisfy the design displacement criteria. Details are listed in Table 3.1, which also includes the calculated plastic hinge length (lp), the expected displacement and curvature ductility demands (, ), the effective period at maximum displacement (Teff, approximately equal to shear force and bending moment (Vb, and Mb).

Tel . ), and design base

Chapter 3: More Problems with Multi-Modal Analysis

37

1.1 1.0 0.9 0.8

spectral Acceleration (g)

0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0.0 0.3 0.5 0.8 1.0 1.3 1.5 1.8 2.0 Period (sec) 2.3 2.5 2.8 3.0 3.3 3.5 3.8 4.0

Fig. 3.2 Elastic 5% Acceleration Response Spectrum for Cantilever Wall Study.
P P P

WALL F
m m m m m m m m m m m m m m m m m m m m

WALL E
P P P m m m m
20@3m

P P P P P P P P P P P P P P P P P

WALL D
P P P P P P
12@3m

P P P P P
16@3m

m m m m m m m m m m m m

m m m m m m m m m m m m

WALL C
P P P P
8@3m

m m m m m m m m

P P P P P P P P

WALL B WALL A
4@3m

P P P P P P

P P P P

P P P P

m m m m

P
2@3m

m m

Fig. 3.3 Idealization of Cantilever Walls

Chapter 3: More Problems with Multi-Modal Analysis

39

12 20 8

Height (m)

Height (m)

15

SRSS/

10

DBD

SRSS

ELF

SRSS

DBD

ELF

0 0 2000 4000 Moment (kNm) (a) 4-Storey Wall 6000

0 0 4000 8000 12000 16000 20000 Moment (kNm)

(b) 8-Storey Wall

30

40

20

Height (m)

Height (m)

30

20

10
SRSS SRSS/ DBD ELF

10

SRSS SRSS/

DBD

ELF

0 0 10000 20000 Moment (kNm) (c) 12-Storey Wall 30000

0 0 10000 20000 30000 40000 50000 Moment (kNm) (d) 16-Storey Wall

Fig. 3.4 Comparison of Wall Design Moments from Different Design Procedures

40
12

Priestley. Myths and Fallacies in Earthquake Engineering, Revisited

20 8

Height (m)

Height (m)

15

10
SRSS DBD SRSS/ ELF

DBD SRSS

ELF

5
SRSS/

0 0 200 400 600

0 0 200 400 600 800 1000 Shear Force (kN)

Shear Force (kN) (a) 4-Storey Wall

(b) 8-Storey Wall

30

40

Height (m)

20

Height (m)

30

20
SRSS/ SRSS DBD ELF

10
SRSS DBD SRSS/ ELF

10

0 0 400 800 Shear Force (kN) (c) 12-Storey Wall 1200

0 0 400 800 1200 Shear Force (kN) (d) 16-Storey Wall 1600

Fig. 3.5 Comparison of Wall Design Shear Forces for Different Design Procedures

Chapter 3: More Problems with Multi-Modal Analysis

41

indicated by DBD. Results from the multi-mode analysis are indicated as SRSS, and the equivalent lateral force procedure by ELF. In both SRSS and ELF designs, a behaviour (force-reduction) factor of R=4 was adopted to reduce the elastic response. The ELF procedure always produces higher design moments and shear forces than the other two methods, and except for short-period structures, the multi-mode procedure results in lower forces than the DBD procedure. In this study, the aspect of greatest interest was the shape of the moment and shear force envelopes, as compared with inelastic time-history results. Since the walls analyzed were the displacement-based designs, the multi-mode designs were factored to produce the same design base moment as the displacement-based designs, thus ensuring approximately equal displacement ductility demand. This implies a reduction in the design ductility for the multi-modal analysis from the design value of R=4. The resulting modified design forces for the multi-mode analyses are indicated in Figs. 3.4 and 3.5 as SRSS/. Note that, as required, the base moments for this design set in Fig 3.4 are the same as the displacement-based designs, but the SRSS/ base shear forces are significantly higher than the displacement-based shear forces. This is because the multimode analyses provide at least some direct consideration of higher mode effects. For displacement-based design, a dynamic amplification factor would be applied to determine the design shears, in accordance with the approach outlined in connection with Fig. 3.1. The influence of higher mode effects are also apparent in the shapes of the SRSS/ moment profiles above the base, particularly for the taller walls, and similarly in the upper levels of the wall shear force distributions. The DBD wall designs (and hence also the SSRS/ designs) were analyzed using a suite of five spectrum compatible earthquake records, scaled to different multiples of the design intensity, from 0.5 to 2.0. The computer code Ruaumoko[12], was used for the analyses, with hysteretic behaviour of potential plastic hinge regions represented by a realistic modified Takeda hysteresis rule. Note that higher mode effects are rather sensitive to the hysteretic model assumed, and earlier work carried out on the less realistic simplified elasto-plastic hysteresis provide somewhat different results. Moment and shear force envelopes from the time-history analyses at different seismic intensities are compared with the SRSS/ design values in Figs. 3.6 and 3.7 respectively. In these figures, IR indicates intensity ratio, or the multiple of the design seismic intensity. Thus IR = 1 indicates the design intensity, IR = 0.5 indicates 50% of design intensity, and so on. Referring first to Fig.3.6, we see that the time-history analysis results indicate only small increases in base bending moment with increasing intensity, as expected, since the increase, once the nominal moment capacity has been reached is only the result of the post-yield stiffness of the moment-curvature characteristic at the wall base. However, at levels above the base, and particularly at wall mid-height, moments increase very significantly with increasing intensity, especially for the eight and sixteen storey walls. Note that as multi-modal analysis is currently formulated, in accordance with Eq.(3.9), the only increase in moments with increasing intensity would be in proportion to

42
12

Priestley. Myths and Fallacies in Earthquake Engineering, Revisited

20 8
IR= 0.5 1.0 1.5 2.0

Height (m)

Height (m)

15

IR=

0.5

1.0

1.5

2.0

10
SRSS/

SRSS/

0 0 1000 2000 3000 Moment (kNm) 4000

0 0 2000 4000 6000 8000 10000 Moment (kNm)


(b) 8-Storey Wall

(a) 4-Storey Wall

30

40

20

Height (m)

Height (m)

30
IR=0.5 1.0 1.5 2.0

IR= 0.5

1.0

1.5

2.0

20

10
SRSS/

10

SRSS/

0 0 4000 8000 12000 16000 20000 Moment (kNm) (c) 12-Storey Wall

0 0 10000 20000 Moment (kNm) (d) 16-Storey Wall 30000

Fig. 3.6 Comparison of Multi-modal Moment Envelopes with Results of Inelastic Time-history Analyses for Different Seismic Intensity Ratios

Chapter 3: More Problems with Multi-Modal Analysis

43

12 20 8

Height (m)

Height (m)

15

10

4
SRSS/ IR= 0.5 1.0 1.5 2.0

SRSS/ IR= 0.5 1.0 1.5 2.0

0 0 200 400 600 800 1000 Shear Force (kN)

0 0 400 800 1200 1600 2000 Shear Force (kN)

(a) 4-Storey Wall

(b) 8-Storey Wall

30

40

Height (m)

20

Height (m)
IR=0.5 1.0 1.5 2.0

30

20

10 10
IR=0.5 1.0 1.5 2.0 SRSS/

0 0

SRSS/

0 0 1000 2000 Shear Force (kN) 3000

500 1000 1500 2000 2500 3000 Shear Force (kN)

(c) 12-Storey Wall

(d) 16-Storey Wall

Fig. 3.7 Comparison of Multi-modal Shear Force Envelopes with Results of Inelastic Time-history Analyses for Different Seismic Intensity Ratios

44

Priestley. Myths and Fallacies in Earthquake Engineering, Revisited

the strain-hardening increase at the wall base. It is also apparent that the multi-modal analysis (SRSS/), is non-conservative at the design intensity, and increasingly so at higher intensities. For the four and eight storey walls, where the design displacement ductilities exceed 2 (see Table 3.1), the multi-modal moment envelope is nonconservative even at 50% of the design intensity. In Fig. 3.7, the multi-modal and IR=1 time-history shear force envelopes are indicated by solid lines, with time-history results for other intensities indicated by dashed lines. Again it is seen that the shear force envelopes are strongly influenced by seismic intensity, and that the multi-modal design envelope is significantly non-conservative. For the four and eight storey walls, the time-history base shear force at IR=1 is almost twice the multimodal value, and for these two walls, the shear profiles at IR=0.5 exceed the design profile at all heights. At intensity ratios of IR=2, base shear force is between 2.5 and 3.7 times the multi-modal design envelope. The discrepancies between the multi-modal and time-history shear forces are more problematic than the corresponding moment discrepancies of Fig. 3.6. Although unintentional plastic hinging (which would be the consequence of designing to the multimodal moment envelopes) at levels above the base is undesirable, some limited ductility demand should be sustainable without failure. However, the consequences of the imposed shear demand exceeding the shear capacity could be catastrophic shear failure. 3.2.2 Modified Modal Superposition for Design Forces in Cantilever Walls Examination of Fig. 3.7 indicates that at an intensity ratio of IR=0.5, where ductility demand is low, or non-existent for all walls, the shape of the shear force envelope is well predicted by the modal analysis procedure. This suggests that it might be possible to predict the shear force and moment envelopes by simple modification of the modal superposition (multi-mode) approach. A basic and simple modification to the modal superposition method is available by recognizing that ductility primarily acts to limit first mode response, but has comparatively little effect in modifying the response in higher modes. If this were to in fact be the case, then first mode response would be independent of intensity, provided that the intensity was sufficient to develop the base moment capacity, while higher modes would be directly proportional to intensity. This approach is very similar to that proposed by Eibl and Keintzel[13] as a means for predicting shear demand at the base of the wall. This modified modal superposition approach is clearly an approximation to response. Although the first mode inelastic shape is very similar to the elastic shape, and hence the approximation should be reasonably valid for the first mode, it is clear that the higher modes will be modified to some extent by the first mode ductility, since a basic feature of the modified higher modes will be that, when acting together with ductility in the first mode, they cannot increase the base moment demand, which will be anchored by the moment capacity of the base plastic hinge. The approach suggested below attempts to extend the basic method of Eibl and Keintzel for shear forces to the full height of the

Chapter 3: More Problems with Multi-Modal Analysis

45

wall, and also to provide a method for determining the appropriate capacity design moment envelope. Shear Force Profiles: To investigate the appropriateness of a simple approach based on the above arguments, shear force profiles were calculated based on the following assumptions. First mode shear force was equal to the shear profile corresponding to development of the base moment capacity, using the displacement-based design force vector. However, for low seismic intensity, where plastic hinging was not anticipated in the wall, simple elastic first mode response, in accordance with the elastic response spectrum was assumed. Higher mode response was based on elastic response to the acceleration spectrum appropriate to the level of seismic intensity assumed. In other words, the forcereduction factor normally applied to higher mode response was not adopted. The basic equation to determine the shear profile was thus:
2 2 0. 5 Vi = (V12 i + V2 Ei + V3 Ei + ...)

(3.10)

where Vi is the shear at level i, V1i is the lesser of elastic first mode, or ductile first mode response at level i, and V2Ei , and V3Ei etc are the elastic modal shears at level i for modes 2, 3 etc. Moment Profiles: It is clear that the simple modal combination used for shear forces cannot be directly used for moments. This is because such an approach would predict increasing base moment with increasing intensity ratio, whereas the base moment is anchored to the capacity of the wall. As a consequence, the higher mode moment patterns are likely to deviate significantly from the elastic mode shapes, particularly in the lower regions of the wall. A number of trial combinations were considered. The most successful adopts a simple modal combination, similar to that of Eq. (3.10), but multiplied by a factor of 1.1, over the top half of the wall, with a linear profile from mid-height to the moment capacity at the base of the wall. The combination equation over the top half of the wall is thus:
2 2 0.5 M i = 1.1 ( M 12i + M 2 Ei + M 3 Ei + ...)

(3.11)

where Mi is the moment at level i, M1i is the lesser of the elastic first mode moment and the ductile design moment, and M2Ei and M3Ei etc are the elastic modal shears at level i for modes 2, 3, etc. Comparison with Time-history Results: The predictions of this Modified Modal Superposition (MMS) approach are compared in Figs. 3.8 and 3.9 with time history results for moment profiles and shear force respectively for the four, eight, twelve and sixteen storey walls. For clarity, comparisons are only provided for IR=1 and IR=2. Full comparisons for all six walls considered, and all four levels of seismic intensity are provided in reference [10]. Also included in Figs. 3.8 and 3.9 are the shear force or

46
12

Priestley. Myths and Fallacies in Earthquake Engineering, Revisited

20 8
SRSS/ THA MMS 1.0 2.0 = IR MMS

Height (m)

Height (m)

15
SRSS/ 1.0 2.0=IR

10
THA

4 5

0 0 1000 2000 3000 Moment (kNm) 4000

0 0 2000 4000 6000 8000 10000 Moment (kNm)

(a) 4-Storey Wall

(b) 8-Storey Wall

30
MMS

40

Height (m)

Height (m)

20

30

MMS IR=1.0 2.0

IR=1.0 THA

2.0

20
SRSS/

THA

10

SRSS/

10

0 0 4000 8000 12000 16000 20000 Moment (kNm) (c) 12-Storey Wall

0 0 10000 20000 Moment (kNm) (d) 16-Storey Wall 30000

Fig. 3.8 Comparison of Modified Modal Superposition Moment Envelopes with Results of Inelastic Time-history Analysis

Chapter 3: More Problems with Multi-Modal Analysis

47

12 20 8

Height (m)

Height (m)

15

SRSS/

10
SRSS/

MMS THA 1.0

5
2.0

MMS THA IR=1.0

0 0

IR=

0 0

2.0

200 400 600 800 1000 1200 Shear Force (kN)

400 800 1200 1600 2000 2400 Shear Force (kN)

(a) 4-Storey Wall

(b) 8-Storey Wall

30

1.0

2.0=IR

40

1.0

2.0=IR

Height (m)

20

Height (m)
SRSS/

30

20
SRSS/
MMS

10

10
MMS THA

0 0

0 4000 0

THA

1000 2000 3000 Shear Force (kN) (c) 12-Storey Wall

1000 2000 3000 Shear Force (kN)

4000

(d) 16-Storey Wall

Fig. 3.9 Comparison of Modified Modal Superposition Shear Force Envelopes with Inelastic Time-history Analyses

48

Priestley. Myths and Fallacies in Earthquake Engineering, Revisited

moment envelopes corresponding to conventional modal superposition (multi-modal analysis), for the actual base moment capacity, indicated as SRSS/. In Figs. 3.8 and 3.9, the MMS results are indicated by dashed lines, and the time-history results by solid lines. Note that with the MMS approach, higher mode effects will become increasingly important as the seismic intensity increases, since the response coefficients for the higher modes increase in proportion to the intensity ratio, while the fundamental mode coefficient is locked at the value corresponding to the base plastic hinge capacity. It is seen that the MMS approach provides a good representation of the time-history moment profiles at the design intensity, IR=1 (see Fig. 3.8). Except for the four-storey wall, the MMS profiles tend to be slightly conservative, but in all cases, the shape agreement is good. At the maximum intensity considered of IR=2 the agreement is also good, though it becomes increasingly conservative as the number of storeys increases. Similar behaviour is apparent for the shear force envelope comparisons of Fig. 3.9. At the design intensity, the agreement between the MMS and time-history profiles is extremely close. At an intensity ratio of IR=2, the agreement is still adequate, though the MMS approach becomes increasingly conservative as the number of stories increases. The full comparisons of Reference [10] indicate that excellent agreement was obtained for all walls at values of IR 1.5. 3.2.3 Consequences for Capacity Design: The results presented in the previous section indicate that there are significant problems when predicting the moment and shear force envelopes of cantilever wall buildings using conventional multi-mode analyses. It will also be recognized that the time-history results indicate problems with capacity design, as currently implemented in design. One of the claims of proponents of capacity design (including the author) has been that it desensitizes the structure to the characteristics of the earthquake. Although this is generally true, it clearly does not apply to uncertainties in the seismic intensity. Comparisons of dynamic amplification factors implied by the time-history results presented here with current design values indicate that the dynamic amplification factors need to be increased, and made intensity-dependent. Critical facilities such as hospitals should be designed for reduced annual probability of occurrence of the design level earthquake, implying increased higher mode effects, and higher dynamic amplification of shear force and moment. 3.2.4 Higher Mode Effects in Frame Buildings from Time-History Analyses A limited study has been carried out to investigate whether the conclusions derived for cantilever wall buildings could also be applied to reinforced concrete frame buildings. In frame design, dynamic amplification is normally applied to the column shear forces and bending moments to ensure that shear failures or unintended column plastic hinges are proscribed[8]. Three reinforced concrete frame buildings, described in Fig. 3.10, were designed by displacement-based principles for maximum design drift levels of 2.5%, in

Chapter 3: More Problems with Multi-Modal Analysis

49

Building 1: 4-storey frame: 2000kN/floor/frame Beams 750x400, Columns 600x600 3@3.5m Ec = 25GPa, fy = 420 MPa

4.0m

7.5m

7.5m Building 2: 8-storey frame: 2000kN/floor/frame Beams 750x400, Columns 750x750 Ec =25GPa, fy = 420MPa

7@3.5m

4.0m 7.5m 7.5m Building3: 3-storey tube frame Beams: Level2: 1000x400 Level 3: 900x400 3@ Level 4: 800x400 3.5m Columns: 600x600 Weights:Levels 2,3: 3500kN/floor/frame Level 4: 2500kN/frame Ec = 25GPa, fy = 420MPa 3@6m

Fig.3.10 Building Frame Details

50

Priestley. Myths and Fallacies in Earthquake Engineering, Revisited

the lowest storey. The first two frames (4-storey and 8-storey) were comparatively flexible, with yield drifts of 0.0105, calculated using Eq. (2.18), meaning that the design ductility was only 2.38 significantly less than force-reduction factors used in most frame designs. To investigate a stiffer structure, a 3-storey building with shorter beam spans and deeper beams was adopted for the third example, to ensure a design ductility level close to 4 was achieved. Note that the third building had different masses at the different levels, and also different beam depths at different levels, to introduce additional complexity to the response. The structures were each designed for a peak ground acceleration of 0.533g, using the EC8 spectral shape for medium ground (essentially Fig. 3.2 scaled up by 33%). The designed structures were then subjected to a suite of five spectrum-compatible earthquake records. The 4-storey frame was analyzed at three different levels of seismic intensity, corresponding to 50%, 100% and 150% of the design intensity. Results for the envelopes of storey column shear force (i.e. the sum of the shears in the columns at a particular level) are shown in Fig. 3.11. Note that despite the low design ductility level of 2.38 for this frame, it still developed a plastic hinge mechanism at 50% of design intensity (IR = 0.5). It will be seen that the storey shears, as with the cantilever walls, are dependent on

12

Height (m)

8
IR=0.5 1.0 1.5

0 400 800 1200 1600 2000

Shear Force (kN)

Fig.3.11 Storey Shear Forces for 4-storey Frame at Different Intensities

Chapter 3: More Problems with Multi-Modal Analysis

51

the seismic intensity, though the dependency does not seem to be as pronounced as was the case for the cantilever walls. Results for the time-history analysis (THA) storey shear forces at the design intensity levels are shown in Fig.3.12, and compared with conventional multi-modal analysis (designated in Fig.3.12 as SRSS/), and the modified modal analysis (MMS) approach advanced in the previous section for predicting shear force envelopes in cantilever walls. It will be recalled that this involved combining the shears from the design inelastic first mode with the elastic shears of the higher modes, using a SRSS combination. It is apparent from the results presented in Fig. 3.12 that the MMS approach, while working well for wall structures, provides inconsistent, and generally excessively conservative predictions of storey shear forces for these frames. Agreement is reasonably good for the three storey frame, but poor for the four and eight storey frames. On the other hand, the traditional elastic modal superposition approach is also inconsistent, and generally non-conservative. It is initially puzzling why the MMS approach should work so well for cantilevers but not for frames. It would appear that the answer to this may lie in the effect of ductility on the higher mode characteristics particularly the periods. A series of analyses on the wall and frame structures considered in this study was carried out where the stiffnesses for members with plastic hinges were reduced to 0.1 times their initial elastic values, implying member ductility factors of about 10. For the wall structures the second mode period was increased by less than 35%, while for the frame structures the second mode period was increased by at least 125%. It will also be recalled that the ratio of first to second mode periods is approximately 1:5 for cantilever walls, and approximately 1:3 for frames. As a consequence of these two effects, the response of the critical second mode tends to stay on, or rise to, the peak plateau of the response spectrum for cantilever walls as the wall responds inelastically, and hence the initial elastic response provides a good estimate of the higher mode effects. With frames, the higher mode periods lengthen to the extent that their response slides down the constant velocity slope of the response spectrum, reducing the higher mode significance as the structure develops ductility. Note that the behaviour of both walls and frames analysed in these studies support this explanation. For stiff short-period walls, it was found that the MMS approach was slightly non-conservative. This would be expected as the higher mode periods lengthen slightly and rise to the constant acceleration plateau. For the very long period walls, the MMS approach was slightly conservative, as higher modes exceeded the constant acceleration plateau. Response of the three-storey tube frame, which was shorter and stiffer than the other two frames was quite well predicted by the MMS approach, which gave worst representation of the most flexible eight-storey frame. Additional work is required to formulate useable design methods based on these findings. Initial indications are that the MMS approach is acceptable for wall structures, but that a more simple approach, based on the traditional dynamic amplification factor proposed in [8] modified for the level of ductility would be more appropriate.

52

Priestley. Myths and Fallacies in Earthquake Engineering, Revisited

SRSS/

SRSS/ THA THA MMS

12

20
Height (m)

MMS

Height (m)

10

0 0 400 800 1200 1600 2000 Shear Force (kN)

0 0 1000 2000 Shear Force (kN) 3000

(a) 4-Storey Frame

(b) 8-Storey Frame

SRSS/ THA

8
Height (m)

MMS

0 0 1000 2000 Shear Force (kN) 3000

(c) 3-Storey Tube Frame

Fig. 3.12: Comparison of Storey Shear Envelopes at Design Intensity Level (THA=time-history analysis, MMS = modified modal superposition, SSRS/= elastic modal superposition reduced by ductility demand)

Chapter 3: More Problems with Multi-Modal Analysis

53

Current capacity design recommendations[8] for frame column design also include dynamic amplification to avoid unintended column hinging. Results from the frame dynamic analyses, where the columns were modeled as elastic elements except at the potential base hinge, did indicate significant amplification, by up to 70%, of column moments above the design level, due to higher mode effects. To check the significance of this, the eight-storey frame was reanalyzed with ductile columns for the bottom four storeys. The strength was put equal to the column base design strength. The concern with ductile columns is that a soft storey mechanism might develop with simultaneous hinging at the base and at some higher level. Such behaviour results in excessive ductility demand in the columns. In fact the analyses with the ductile columns indicated that Lateral displacements were increased by only a few percent. Column ductilities, except at the column base, were extremely low. Beam ductilities were slightly increased in the lower storeys, but were reduced in the upper stories. Column shear forces were almost unaffected. It thus seems that a simple design rule would be appropriate, such as requiring the sum of the column moment capacities at a beam/column intersections to exceed the sum of the beam moment capacities by a fixed percentage, say 20%, as is commonly adopted in the United States. 3.3 TORIONAL RESPONSE OF BUILDING STRUCTURES In Section 3.1, the ability of multi-modal analyses to incorporate torsional effects by full 3-D elastic modelling of building structures was mentioned as a perceived reason for adopting the method. While it is undoubtedly true that this a valid advantage when elastic response is considered, it is less obvious that elastic modal analysis provides accurate simulation of inelastic torsional response. Recent work by Paulay[14,15], is used as a basis for the following arguments. There are two basic criticisms that can be directed at elastic multi-modal analysis for torsional response of inelastic systems. The first is that it relies on stiffness estimates of the various lateral force resisting elements. As was established in Chapter 2, these will not be known until design strengths, as well as member dimensions have been allocated. Also, the member stiffnesses are generally substantially in error in conventional design. As a consequence, any increase in accuracy resulting from increased sophistication of analysis technique is likely to be illusory. More important is the second criticism. Paulay[14,15] has shown that once inelastic action develops in the major lateral force resisting elements, the eccentricity due to elastic stiffness distribution is largely irrelevant, and strength eccentricity is of fundamental importance. Paulay differentiated between torsionally restrained and torsionally unrestrained systems, and we consider each type of system separately in the following discussions.

54

Priestley. Myths and Fallacies in Earthquake Engineering, Revisited

3.3.1 Torsionally Unrestrained Systems Consider the plan view of a regular building shown in Fig. 3.13(a). In the direction con-

V1, k1

C.M C.R

V1, k1

(a) Plan View of Building V1

(b) Wall Force-Displacement 2y inelastic elastic (s 1)ys y yield ys (c) Lateral Displacement Profiles Fig. 3.13 Response of a Torsionally Unrestrained Building

Chapter 3: More Problems with Multi-Modal Analysis

55

sidered, seismic resistance is provided by two nominally identical structural walls one at each end of the building. In the long direction of the building two identical walls on the building centerline provide seismic resistance. As a consequence of the symmetrical location of nominally identical walls, the center of mass (C.M.) and center of stiffness (C.R.) are both nominally at the center of the building. Accidental eccentricity could occur due to unintended variations in material properties between the two end walls, in terms of stiffness, k1, or strength, V1, (see Fig. 3.13(b)) or due to mass eccentricity. For the sake of example, we consider only mass eccentricity, and assume that the center of mass is displaced to the right of the central position indicated in Fig. 3.13(a). It is further assumed, to simplify the argument, that the lateral force-displacement response of the end two walls can be characterized by the elastoplastic response of Fig. 3.13(b). The structural system for this building may be represented, in plan, by the four-spring model shown in Fig. 3.14(a). It will be recognized that the two longitudinal springs perpendicular to the applied shear force V have no influence on response, and the shears induced in the two end walls may be found from simple statics, as shown in Fig. 3.14(a). The stiffnesses of the end walls are irrelevant in calculating the shear forces, which must V(B-x)/B Vx/B

(B-x) V (a) Torsionally Unrestrained k3 k1 C.R k4 (b) Torsionally Restrained Fig. 3.14: Spring Plan models of Building Stiffness C.V. e V C.M. k2

56

Priestley. Myths and Fallacies in Earthquake Engineering, Revisited

always be in the proportion x:(B-x). Since the maximum force that can be developed in the right-hand wall is V1, it follows that the maximum force that can be developed in the left-hand wall is V1(B-x)/x. Since x>(B-x), it follows that the left-hand wall will never yield. Elastic analysis, based on multi-modal or hand methods will predict an elastic lateral displacement profile indicated by the dashed line in Fig. 3.13(c), when the accidental mass eccentricity is included in the analysis. When reduced down by the behaviour, or forcereduction factor, the structural yield displacement profile is as indicated by the solid line marked yield. The structural yield and design displacements at the center-of-mass are ys, and sys, respectively. As indicated by the argument presented above, and as illustrated in Fig. 3.13(c), the left wall cannot yield, and hence the maximum displacement for this wall is y.(B-x)/x. The inelastic displacement profile at design intensity is thus as shown by the solid line marked inelastic. Since the center-of-mass displacements at maximum response will still be essentially the same as predicted by the elastic analysis, the displacement at the right-hand wall, at 2y, will be much greater than predicted by the elastic analysis. Thus the comparatively sophisticated multi-modal analysis produces displacements grossly in error, and has the potential for producing an unsafe design. However, although the argument presented is generally correct, it is a little simplistic, as it does not consider the influence of mass rotational inertia, which has been found in inelastic timehistory analyses to slightly modify the results from the simple predictions presented here. 3.3.2 Torsionally Restrained Buildings The second example considered is shown in Fig.3.15. In this example, wall 1, in the Y direction, at the left side of the plan view shown in Fig.3.15(a) is stronger and stiffer than wall 2, at the right-hand side. In the x direction strong walls of stiffness kx1 and kx2 are located on the plan perimeter. In addition to the centers of mass and stiffness (C.M. and C.R. respectively in Figs 3.15(a) and 3.14(b)), a center of strength (C.V.) is defined based on the position of the resultant of the wall yield strengths Vy1 and Vy2. The center of strength could be, but will generally not be, at the same location as the center of stiffness. In Fig. 3.15(c) the elastic displacement profile, based on multi-modal analysis at the design seismic intensity, and the yield profile, factored down from the elastic profile by the force-reduction factor, are indicated by dashed and solid lines respectively. Regardless of how carefully the seismic design is carried out, inevitable variations in material properties, or mass distribution will mean that one of the two end walls will yield first, as was posited for the unrestrained case. In this case, however, the stiffness and location of the x-direction longitudinal walls (see Fig. 3.14(b)) will ensure that the lateral force can continue to rise until both end walls yield. The total building forcedisplacement response is then as suggested in Fig. 3.15(b). Note that once both walls have yielded, the effective stiffness in the direction considered is zero.

Chapter 3: More Problems with Multi-Modal Analysis

57

kx1

C.R. Vy1, ky1

C.M. Vy2, ky2 C.V.

kx2 (a) Plan View of Building Vy1+Vy2

(b) Building Force-Displacement

elastic inelastic 1y1 sy1 yield y1 ys (c) Lateral Displacement Profiles Fig. 3.15 Response of a Torsionally Restrained Building When both walls have yielded, the center of the resisting force has moved from the center of elastic stiffness (C.R.) to the center of shear strength (C.V.) (though as men(s-1)ys

58

Priestley. Myths and Fallacies in Earthquake Engineering, Revisited

tioned above it is conceivable that these might be coincident). Referring to Fig.3.14(b), the torsional moment at this yield stage of response is thus

M T = Ve = (V y1 + V y 2 )e

(3.12)

where e is the eccentricity between the centre of strength and the mass centre. Since neither of the end walls have incremental stiffness as this stage, this torsional moment is resisted entirely by the x-direction walls, which are assumed to remain elastic, and which therefore dictate the torsional rotation (twist) shown in the yield profile of Fig.3.15. As the structure continues to displace laterally under increasing seismic excitation, the force in the two end walls remains constant. The torsional moment therefore remains at the level given by Eq.(3.12), and the twist remains unchanged. That is, after full yield is achieved, the structure translates without additional rotation, and thus the final inelastic displacement profile, shown in Fig. 3.15(c) by the solid line, is parallel to the yield profile. In this case, it is seen that the elastic estimate of the final displacement profile overestimates the twist. Since the displacements of the center of mass should be essentially independent of the twist, the elastic multi-modal analysis will underestimate the ductility demand on the critical stiffer wall 1. That is, 1 >s. Again it is seen that multi-modal elastic analysis is unable to capture the inelastic response of the structure, and is unsuitable for representing torsional response. 3.4 CONCLUSIONS Multi-modal (elastic dynamic) analysis depends on accurate assessment of the elastic stiffness of members providing the lateral-force resisting mechanisms. In Chapter 2 it was established that member stiffness depends on strength, and hence accurate elastic analyses could not be carried out until final member dimensions and strengths had been decided. In this chapter, we investigated some of the deficiencies inherent in the multi-modal approach when used to estimate the inelastic response of structures. It was found that, as currently implemented in design practice, multi-modal analysis seriously underestimates higher mode effects in structures that respond inelastically to seismic excitation. For cantilever walls, a comparatively simple modification can be adopted, based on applying the design behaviour factor only to the first mode moments and shears. However, for frame structures (where, fortunately, higher mode influences are less severe than with cantilever walls), correct simulation of higher mode effects is less straightforward. The time-history analyses presented in this chapter also indicated that structures are more susceptible to higher mode effects from earthquakes of higher intensity than the design level. It had previously been felt that capacity design procedures desensitised structures to intensity effects, except insofar as displacement ductility level is concerned. It was also shown that inelastic torsional response of structures is poorly represented by elastic analysis methods, including multi-modal analysis. Torsional response of

Chapter 3: More Problems with Multi-Modal Analysis

59

torsionally unrestrained structures is underestimated by elastic modal analysis, while inelastic torsional response of torsionally restrained structures is overestimated by elastic modal analysis. It can be concluded that elastic modal analysis is an inadequate tool for inelastic seismic design because of Stiffness inadequacies (distribution between members, and strength-dependency). Higher mode representation inadequacies. Torsional response inadequacies.

CHAPTER 4: THE ENERGY MYTH SEISMIC RESPONSE OF PRECAST PRESTRESSED STRUCTURES


4.1 INTRODUCTION In Chapter 1 we noted that conventional seismic design wisdom seeks to maximize the hysteretic energy inside hysteretic force-displacement response characteristics of structures, with elastic-perfectly plastic response seen to be the ideal. Based on arguments presented in the original Myths and Fallacies paper[1], however, it was pointed out that elastic/perfectly-plastic response has some undesirable side effects. First, residual deformation the final deformation state of a structure after the earthquake was likely to be larger than with alternative, and apparently less desirable hysteretic shapes, and second, time history analyses show that the elastic/perfectly-plastic characteristic is more sensitive to P- effects than other loop shapes. Figure 4.1
200 EP 171.8 mm FS 175.7 mm 100 TK 162.6 mm EP 44.4 mm TK 17.6 mm

Displacement (mm)

0 FS 0.0 mm 100

200

10

15

20

25 Time (s)

30

35

40

45

ELASTOPLASTIC 1000 500 Force (kN) 0 44.4 mm 1000 500 0

TAKEDA 1000 500 Force (kN) 0

FLAG SHAPED

Force (kN)

-500 -1000 -200

-500 -1000

17.6 mm

-500

0.0 mm

-1000 0 Displacement (mm) 200 -200 0 Displacement (mm) 200

0 Displacement (mm)

200

-200

Fig. 4.1 Response of Three Hysteretic Systems to 1.5*Loma Prieta,1989[16]

Chapter 4: The Energy Myth - Seismic Response of Precast Prestressed Structures

61

provides a typical comparison of three SDOF hysteretic systems to 1.5 times the Capitola record of the 1989 Loma Prieta earthquake[16]. The Takeda (TK) hysteresis rule is a good approximation to reinforced concrete, while the flag-shaped (FS) rule is characteristic of un-bonded prestressed sections with additional damping, which will be discussed in more detail later. In all cases, a mild P- influence was included. It will be seen that the maximum displacement response of the three different systems is similar, at about 170mm, but the residual deformation remaining at the end of the record shows marked differences, with the elasto-plastic rule, at 44.4mm, having the largest residual deformation. Pampanin et al [16], from which Fig. 4.1 is taken, investigated the practicality of including residual deformation as an alternative type of damage index to the more commonly used indices based on maximum displacement or energy absorbed. Precast concrete, and particularly prestressed precast concrete structures have a poor reputation for seismic response. In part this results from poor performance in a number of earthquakes (e.g. [17,18]), but also because it is perceived that the low energy absorption capacity of prestressed concrete structures results in higher response displacements, even if severe damage or collapse does not occur. In this chapter we discuss research results relevant to precast concrete seismic response that have become available since the original Myths and Fallacies paper. 4.2 SEISMIC DESIGN PHILOSOPHY FOR PRECAST CONCRETE It is comparatively straightforward to design precast concrete structures to satisfy a reinforced concrete emulation requirement. A precast frame or wall structure whose connections have been designed to have sufficient strength to remain elastic under seismic response, with plastic hinges forming elsewhere in the structure, should perform in an identical fashion to an equivalent reinforced concrete structure. Various design details involving the placement of in-situ concrete in precast frame connections have also been shown by test to provide reinforced concrete emulation, even when the plastic hinges form in the connection region[19]. This approach is widely accepted in New Zealand, where precast reinforced concrete is the structural material of choice in most multi-story buildings, and has recently gained favour in Japan[20]. Recent research in the U.S.A. under the auspices of the Precast Seismic Structural Systems (PRESSS) Program[21] has been directed towards the development of design recommendations for precast structures that do not satisfy the reinforced concrete emulation requirement of current building codes, and which would permit the construction in seismic zones of precast structures which utilize connection details that take advantage of the potential efficiencies of precast construction. In these designs, ductility is deliberately intended to occur in the connection. Many of these connection details utilize un-bonded prestressing. A fundamental difference in philosophy exists between the two approaches: reinforced-concrete emulation (which is sometimes termed strong joint precast) and

62

Priestley: Myths and Fallacies in Earthquake Engineering, Revisited

ductile joint design. In conventional strong-joint design, capacity design measures are used to protect the joint from inelastic action, ensuring that the plastic hinges occur in the reinforced concrete members, away from the joints. In ductile-joint design, the capacity protection is reversed: the members are protected by capacity design measures to ensure that they remain essentially undamaged during the design level of seismic response, ensuring that the inelastic action occurs only in the connections. It will be seen that this approach has the potential for reducing damage under design seismic attack. When ductile-connection precast concrete is used, a revision of the design approach is necessary. Ideally this should involve the adoption of drift-based seismic design rather than the more traditional force-based design approach, and the use of expected-value and structural reliability concepts for determination of appropriate estimates of member strength and deformation capacity. This approach, which is developed in some detail in Chapter 5, appears to be particularly attractive for ductile-connection precast concrete construction for the following reasons: Hysteretic characteristics for ductile-connection systems will often be very different from the elasto-plastic characteristics which have typically been assumed in the development of current force-reduction or behaviour factors. The principles of drift-based design, outlined in Chapter 5, permit the energy absorption characteristics of different hysteretic responses to be directly and simply incorporated into the design process. In many cases, it is found that reduced energy dissipation, if it exists, is not a significant factor in seismic response. Figure 4.2 illustrates four basic categories of ductile-connection hysteretic response: linear elastic (LE), nonlinear-elastic (NLE), tension/compression yield (TCY) and coulomb friction (CF) which have been investigated in the PRESSS program [21]. Two further categories (NLE+TCY, LE+CF) are formed by combination of two basic classes. Note that the NLE+TCY category has the zero residual displacement of the NLE system, with significant damping provided by the TCY elements. This will be discussed in more detail shortly. The emphasis on displacements, or drift, rather than on force, recognizes that displacement is more fundamental to damage limit states than is strength. This is obvious, since most damage models for steel or concrete relate to strain limits which can be directly related to displacement by geometric relationships. Thus, (e.g.), if a damage-control limit state is to be defined, it is much more meaningful to relate this to strain, and hence displacement, than to behaviour factors. The use of expected-value estimates of member capacity should be beneficial to competitiveness of precast construction, because of tighter control of material properties and dimensional tolerances, particularly for factory constructed precast. Thus strength reduction and overstrength factors, both of which are needed for capacity design, should show less variation from the expected value, than for, say, equivalent monolithic reinforced concrete structures. This means that capacity protected elements of a precast structure could be designed for lower strengths, with economic benefits.

Chapter 4: The Energy Myth - Seismic Response of Precast Prestressed Structures

63

Fig. 4.2 Hysteresis Characteristics for Precast Structures Although drift-based design is clearly the preferred option for precast design, it will be some years before it is fully codified. In the interim, it is important that designers have the opportunity for designing precast structures within the framework of existing codes. Some aspects of implementation of precast systems within existing code frameworks are briefly outlined in the following, based on more complete recommendations, presented elsewhere[22]. 4.2.1 Force-Based Design Recommendations for Strong-Connection Frames Connections for precast frames with strong connections (PF-SC) can be classified according to the following list: Beam-to-column Column-to-beam Beam-to-beam

64

Priestley: Myths and Fallacies in Earthquake Engineering, Revisited

Column-to-footing Column-to-column Figure 4.3 illustrates the different possible connection categories. Plastic hinge locations: Locations of plastic hinges in PF-SC frames should be chosen to ensure the development of a strong-column/weak-beam deformation mechanism when inelastic response is expected under seismic attack. For the different connection categories defined in Fig. 4.3, hinge locations should be further limited in accordance with the following criteria:

Fig. 4.3 Strong-Connection Categories for Precast Frames

Chapter 4: The Energy Myth - Seismic Response of Precast Prestressed Structures

65

Beam-to-column connection: The critical section of the plastic hinge should be located not less than a distance hb from the connection, where hb is the beam depth. Column-to-beam connection: The critical section may be located anywhere within the beam length, between the column faces. Beam-to-beam connection: The critical section of a plastic hinge may be located anywhere within the beam length between the column faces, but not less than a distance hb from the connection. Column-to-column connection: The critical sections may be located anywhere within the beam length, between the column faces. Column-to-footing connection: Where moment-resisting connections between columns and footings are provided, and a column-base plastic hinge is required to complete the inelastic deformation mechanism, the critical section in the lower column should be located a distance not less than hc from the columnto-footing connection, where hc is the column width in the direction considered. The design assumption is that plastic hinges will form in beams and at the column base only, and that the critical (relatively weakest) section of the plastic hinge must be separated by some distance from the connection. Considerable freedom is given in deciding the locations of the plastic hinges along the beam. However, it must be recognized that the further the hinges are separated from the column faces, the larger are the plastic rotation demands. To an extent this will be compensated by the larger effective length of the plastic hinge, when plasticity can spread to both sides of the critical section. Note that a wet-joint connection at a plastic hinge location is not covered in this classification, as it would be a ductile connection, dealt with subsequently. Required Connection Strength: Connection moment and shear strength should not be less than that given by Eq.4.1:

d S DN o D

(4.1)

where D is the design action at the connection (moment or shear) corresponding to nominal strengths of the plastic hinge mechanism, d is the dependable strength reduction factor, SDN is the nominal strength of the connection under action D, o is the plastic hinge overstrength factor, and is a dynamic amplification factor. Equation (4.1) may be considered a generalized capacity design equation. Appropriate values for d, o, and should be based on analytical and experimental results for the form of structural system being considered. For precast frames with confined reinforced concrete plastic hinges the following values would appear to be adequately conservative: d =0.9 o =1.4 =1.0 Beam-to-column:

66

Priestley: Myths and Fallacies in Earthquake Engineering, Revisited

d =0.9 o =1.4 =1.4 Column-to-beam: d =0.9 o =1.4 =1.0 Beam-to-beam: d =0.9 o =1.4 =1.4 Column-to-column: o Column-to-footing: d =0.9 =1.4 =1.0 As was discussed in Chapter 3, the overstrength factor o is dependent on the ductility factor, and hence the values above are approximate only. Also, different values should strictly be provided for moment and shear. However, for frame structures the values listed are considered to be reasonably conservative. Note that higher mode effects do not significantly affect beam actions (unless vertical acceleration response is considered), and hence =1.0 for these. For columns, higher mode effects can considerably influence the distribution of beam moment input between the column moments immediately above and below the joint. With European reinforcing steels (e.g. temcore), which have rather low strain hardening, and a tight control over yield strength, lower overstrength factors than listed above may be appropriate. For column-to-column connections made in the central 50% of clear column height, the connection strength should also satisfy Eq. (4.2):

d S 0.4 o De

(4.2)

where De is the larger of the appropriate design actions (flexure or shear) at the top or bottom of the column. Equation (4.2) is intended to cope with higher mode effects at a midheight column connection, where Eq.(4.1) might indicate near-to-zero moment, based on structural analysis. Note that a deliberately simple approach has been adopted. More precise procedures are possible, as, for example, suggested for wall buildings in Chapter 3. For frames, moments should strictly be related to equilibrium at the joint centroid, with reductions occurring at the critical column section according to distance from the joint centroid, as suggested in [8]. Considering the approximations in and o suggested in this simplified approach, this is not felt to be necessary for regular frames. Behaviour (Strength-reduction) Factors for Strong-connection design: The connection requirements provided above are intended to ensure monolithic reinforced concrete emulation. Consequently, behaviour factors for reinforced concrete design in existing design codes would be directly applicable. It should be noted, however, that very significant differences in behaviour factors (up to a factor of 2 or more) exist between design codes of different countries for the same structural and occupancy type. This is a consequence of the basic inadequacy of force-based design, for which clear relationships between behaviour factor and damage potential cannot be defined. 4.2.2 Force-Based Design Recommendations for Ductile-Connection Frames With a ductile-connection system, the intention is to provide inelastic deformation

Chapter 4: The Energy Myth - Seismic Response of Precast Prestressed Structures

67

within the connection, and to inhibit the development of inelastic deformation elsewhere by the use of capacity design procedures. The most direct and frequently utilized expression of this approach is where in-situ wet joints are provided at the ends of beams, often extending into the beam-column joint region. This procedure has been widely implemented in New Zealand, based on experimental work by Park and others[19]. Since the construction procedure essentially ensures that the structural system will perform as a monolithic reinforced concrete system, no special design rules are needed, and in New Zealand, such structures are designed without special constraints. Connection (Plastic hinge) Locations: Except for structures of three stories or less, ductile connectors for frames should only be placed in the following locations: In beams: at any location At column bases: to complete an inelastic mechanism Ductile connectors should not be located in columns of frames of more than three stories height, except at the column base, because the inelastic deformation mode will involve soft-story response, with unacceptably high local ductility demand. No restriction is placed on the location of ductile connections in beams. However, when connections providing inelastic response by rotation are used, these should be located as close to the columns as practicable to minimize the connector plastic rotation demand. Connectors placed at midspan would normally be designed to provide inelastic frame deformation by inelastic shear deformation of the connector. Ductile connections in columns may provide inelastic rotation or shear displacement. Note that the special case of shear-ductile connectors at the base of columns, with an entirely elastic structure (strong connection) at higher levels is a description of a baseisolated structure. Required Member Strength: Member strength in flexure and shear should comply with the basic capacity design approach described by Eq.(4.1), where in this case D is the design action in the member corresponding to nominal strength of the plastic hinge mechanisms involving the ductile connectors. Values for ductile connector overstrength factors and dynamic amplification factors will depend on individual system design. Behaviour Factors for Ductile Connection Design: One of the problems in adopting force-based philosophy for ductile-connection design is that lack of reinforced concrete emulation makes it difficult to set appropriate behaviour factors. Many systems do not have definable yield displacements, and hence displacement ductility factors, on which behaviour, or force-reduction factors are based, are not readily defined. Because of this difficulty, drift-based design is more appropriate for ductile-connection design. Where force-based design is adopted, it is advisable to base behaviour factors directly on experimental results, and to verify performance by inelastic time-history analyses, using hysteretic characteristics appropriate for the connection system utilized.

68

Priestley: Myths and Fallacies in Earthquake Engineering, Revisited

4.3 UNBONDED PRESTRESSING IN PRECAST DESIGN 4.3.1 Theoretical Considerations Although prestressed concrete is generally not favoured for seismic resistance, tests on both in-situ[23] and precast[24] beam-column subassemblages with fully grouted tendons have developed ductility comparable to monolithic reinforced concrete. However, after moderate ductility levels had been achieved, these subassemblages suffered excessive stiffness degradation at low displacements. This reduction was caused by a reduction in effective prestress clamping force through the column, resulting from inelastic strain of the prestressing tendon at the critical section. This behaviour is shown in idealized form in Fig. 4.3(a) which refers to a typical prestressing steel stress-strain curve.

(a) Prestress Loss due to Inelastic Response

(b) Force-displacement Hysteretic Response[24]

Fig. 4.4 Inelastic Behaviour of a Precast Prestressed Beam-to-Column Unit In Fig.4.4(a), fsi is the initial steel stress after prestress losses. During low-level seismic response, fluctuations of the steel stress will be within the elastic range, and no loss of prestress will result. At a ductility level of (say) =2, represented by point 2 in Fig. 4.4(a), the maximum prestressing steel response is expected to be on the inelastic portion of the stress-strain curve. On unloading, the steel follows a linear descending branch essentially parallel to the initial elastic portion. Hence, when the structure returns to zero deformation, the effective steel stress is reduced to fs2. On unloading from higher ductility levels, involving larger inelastic strains as indicated by point 3 in Fig.4.4(a), the entire prestress may be lost. This is clearly undesirable, particularly for precast connections where the surfaces between the beam and column are planar, rather than a rough naturally forming crack as may be the case for monolithic

Chapter 4: The Energy Myth - Seismic Response of Precast Prestressed Structures

69

reinforced concrete construction. Prestress shear-friction, which would be relied on to transmit gravity shear forces from the beam to the column would then be ineffective. The result of this behaviour is excessive pinching of the force-deflection hysteresis loops, as indicated in the typical response of Fig.4.4(b)[24]. Note that had the gravity loads been modeled in this test, the performance would have been even less desirable. The use of unbonded, or partially unbonded tendons, where the tendon is unbonded through the joint and for some distance on either side, as indicated in Fig.4.5(a) has been

Fig. 4.5 Beam-to-Column Connection with Partially debonded Tendons

70

Priestley: Myths and Fallacies in Earthquake Engineering, Revisited

suggested as a means for improving performance[25]. The following advantages were proposed: If the length of debonding was sufficient, the required ultimate displacement could be achieved without exceeding the limit of proportionality of the prestressing steel. Consequently there would be no loss of prestress on unloading from the design level of displacement. Shear friction on the beam-tocolumn interfaces would be maintained at all response levels, and support of gravity loads would not be jeopardized. Design of the beam-to-column joint region would be simplified, since the joint shear forces would be transferred by a diagonal strut, as shown in Fig.4.5(b). This is a consequence of the prestress forces on either side of the joint being equal for each tendon, because the tendons are debonded through the joint. Thus, reduced levels of special joint reinforcement would be needed. The response would be essentially elastic, though nonlinear, as indicated in Fig. 4.5(c). Although it was recognized that this would have possibly undesirable consequences for energy absorption, it has the merit that, following response to the design level earthquake, the structure would return to its original position without residual displacement, and the initial stiffness would be restored. Lateral Force-Displacement Characteristics: Force-displacement characteristics of a typical debonded precast beam-to-column prestressed concrete subassemblage are indicated in Fig.4.6(b), based on the forces and displacements measured at the top of the beam-to-column sub-assemblage, as shown in Fig.4.6(a). Assuming no tension capacity across the beam-to-column interface, nonlinearity of response will initiate at point 1 (Fig.4.6(b)) when precompression at the extreme fibre is lost, and a crack starts to propagate. Assuming further that the prestress centroid is at midheight, the corresponding moment is

Fig. 4.6 Force-Displacement Response for a Partially Debonded Prestressed Beam-to-Column Subassemblage[25]

Chapter 4: The Energy Myth - Seismic Response of Precast Prestressed Structures

71 (4.3)

M cr = Ti hb / 6

for a rectangular section, where Ti is the initial total prestress force, and hb is the beam depth. For the dimensions shown in Fig. 4.6(b), the corresponding lateral force F will be:

F=

2 M cr lb (lb hc )lc

(4.4)

where hc is the column depth and lc is the column height between contraflexure points. The corresponding displacement can be found from simple linear elastic analysis based on uncracked section properties. When the crack has propagated to the centroidal axis, corresponding to point 2 in Fig.4.6(b), the moment will be:

M 2 Ti hb / 3

(4.5)

Deviation from the initial elastic stiffness between points 1 and 2 will be minimal, unless the average prestress level is fairly high (fpc > 0.25fc). Above point 2 the precise force-displacement relationship is less simple to determine, since steel and concrete strains are not linearly related. However, since the aim will be to develop a bi-linear approximation to the force-displacement response (see Fig. 4.5(c)), it is sufficient to compute the conditions at maximum displacement. In terms of design requirements, this may be taken as corresponding to the limit of proportionality of the prestressing steel (point 3, Fig.4.6(b)). At this point, the tendon force, and also the tendon elastic extension above the initial condition are known. Assuming confined conditions for the concrete in the plastic hinge region the corresponding moment capacity M3 can thus be found from standard section analysis. Further, assuming that the opening of the crack at the beam-column interface is directly related to the tendon extension by simple geometry, the displacement at point 3 can be shown[25] to be:

3 = 2

(0.5hc + x ) M 3 lc + ( f slp f si ) M 2 Es (0.5hb c )

(4.6)

where fslp is the steel stress at the limit of proportionality, x is the unbonded length of tendon on one side of the joint (see Fig. 4.5(a)), and c is the depth from the extreme compression fibre to the neutral axis in the beam plastic hinge region (found from the known confined concrete stress-strain parameters). More complete details are provided in [25]. The increased displacement demand resulting from the use of a bilinear elastic hysteresis rule was also investigated in [25]. Five different natural accelerograms were scaled to the same peak ground acceleration, and used to investigate response of SDOF

72

Priestley: Myths and Fallacies in Earthquake Engineering, Revisited

Fig. 4.7 Force-deflection characteristics for Dynamic Analyses of Precast Designs systems designed in accordance with UBC[2] requirements. As shown in Fig. 4.7, three different hysteresis rules were adopted, all with the same envelope curves, and were compared with linear elastic response with the same initial stiffness. The bilinear elastoplastic rule represents an idealized response, the bilinear elastic is representative of unbonded prestressed systems, and the bilinear degrading is representative of bonded prestressed systems. Note that conventional reinforced concrete response would be intermediate between the bilinear elastoplastic and the bilinear degrading. Results of the analyses are shown in Fig.4.8. It is seen that there is very little difference in response between the bilinear elastic and the bilinear degrading. Both result in amplification of the peak displacement response compared to linear elastic and bilinear elastoplastic. The amplification is greatest at very short periods. For periods of 0.8 sec. and larger, the bilinear elastic is about 30% higher than the linear elastic response, on average. The displacement amplifications presented in Fig. 4.8 are artificially high, for two reasons. First, the amplification should be compared with a hysteresis rule more representative of reinforced concrete, such as the Takeda degrading stiffness rule. As mentioned above, this should produce results intermediate between the bilinear elasto2.5
Bilinear elastic

Ratio to Linear Elastic

2 1.5 1
Bilinear elasto-plastic Bilinear degrading

0.5 0 0 0.4 0.8 1.2 1.6 2

Period T (seconds)

Fig. 4.8 Amplification of Displacement Response for Different Hysteresis Rules

Chapter 4: The Energy Myth - Seismic Response of Precast Prestressed Structures

73

plastic and bilinear degrading results, implying displacement amplifications of less than 20% for medium and long period structures. Second, precast prestressed structures of the same physical dimensions as reinforced concrete structures of identical strength will have much higher initial stiffness than the reinforced concrete equivalent, and hence reduced elastic and inelastic displacement response. It is thus inappropriate to compare reinforced and prestressed concrete systems with the same initial stiffness and strength. The combined influence of these two effects is that very little, if any, displacement amplification can be expected from precast prestressed structures with unbonded tendons, compared with equivalent reinforced concrete structures. See (e.g.) Fig.4.1 4.3.2. Cyclic Tests of Precast Beam-to-Column Units with Unbonded Tendons In order to test the apparent advantages of precast frames connected with unbonded prestressing, a preliminary test program consisting of two large-scale beam-column subassemblages was carried out. The units represented interior and exterior beam-to-column joints of a perimeter seismic frame. Figure 4.9 shows reinforcement details for the interior unit. Full details are available elsewhere[26]. The test units modeled the region from midcolumn height below the joint to midcolumn height above the joint, and from midspan to midspan of beams on either side of the joint. The unit dimensions, shown in Fig.4.9 are intended to represent prototype

Fig. 4.9 Details of a Precast Prestressed Interior Beam-to-Column Subassemblage

74

Priestley: Myths and Fallacies in Earthquake Engineering, Revisited

dimensions at a two-thirds scale. Nominal concrete unconfined compression strength was 45 MPa, (closely matched in the experiments), and transverse reinforcement had a nominal yield strength of 420 MPa. The exterior joint was similar to the interior joint except that one beam was replaced by a 500 mm-long beam stub, containing the posttensioning anchorages. In these tests the prestressing tendons were unbonded for the full length of the beams. Transverse reinforcement in the beams and columns was substantially less than permitted by existing design codes. For the beams, this was because the prestress provided a dependable compression zone in the beams at the column face, and hence the concrete compression contribution to shear strength, which is normally ignored in plastic hinges of beams, could be relied on. In fact, only nominal transverse reinforcement was provided over the beam length. An exception was in the potential plastic hinge region, where interlocking spirals were placed in the beam compression zones to confine the concrete against the anticipated high localized compression strains (see Fig.4.9). Column transverse reinforcement was designed according to advanced shear strength procedures[27], and was also less than required by existing design codes. No special reinforcement was provided in the joint region (i.e. the column transverse reinforcement continued at the same spacing for the full column height, including through the joint), since theoretical arguments indicated that the joint shear force could be taken by the diagonal strut formed at the beam/column compression zones (see Fig.4.5(b)). The test units had design maximum displacements corresponding to 2% drift. They were subjected to three-cycle bursts of displacements to drift limits of 0.25%, 0.35%, 0.5%, 0.75%, 1.0%, 1.5%, 2.0% and 2.5% . The interior joint was then subjected to three additional cycles at 2.8% (40% above the design drift), and the exterior unit was subjected to three additional cycles at each of 3% and 4% (100% above design drift). Figure 4.10 show condition of the interior test unit at the design drift of 2%, and the exterior test unit at 4% drift twice the design drift. For the interior unit (see Fig.4.10(a)), beam flexural cracking was limited to a single wide crack at the column face. Minor horizontal splitting cracks associated with incipient spalling of the cover concrete first developed at a drift of 0.75%. At the design drift limit of 2%, the spalling of cover concrete extended a maximum distance of 200 mm from the column face. At the maximum test drift of 2.8%, cover spalling extended back about 300 mm from the column face. Very little other damage occurred in the beams. Only minor cracking developed in the column and joint region. As can be seen in Fig.4.10(a), the joint cracks were well distributed, and oriented approximately parallel to the corner-to-corner diagonal. On unloading from maximum displacement, these cracks closed up, and were virtually undetectable. The exterior joint subassemblage displayed a similar lack of damage to the interior unit (see Fig.4.10(b). At the design drift limit of 2% very minor spalling occurred, and at 4% drift, the cover spalling extended about 300mm from the column face. It was evident that the double interlocking spirals were very effective in confining the beam compression zones and maintaining the integrity of the plastic hinge regions. It was also evident that the spirals extended an unnecessary length into the beam, and a length of

Chapter 4: The Energy Myth - Seismic Response of Precast Prestressed Structures

75

(a) Interior Unit at 2% drift

(b) Exterior Unit at 4% drift

Fig.4.10 Damage to Precast Prestressed Test Units at High Drifts half the beam depth (i.e. 400 mm) would have been adequate. Lateral force-displacement hysteresis loops for the interior test unit are shown in Fig.4.11, together with the computed simplified force-displacement envelope calculated in accordance with the procedure outlined in the previous section. The theoretical envelope only extends to the drift of 2.6% corresponding to the limit of proportionality strain of the prestressing steel. Experimental response for both units indicated stable hysteresis, with comparatively little strength degradation between successive cycles to a given drift level. The predicted force-displacement response is a reasonable envelope to the experimental response. The initial stiffness was less than the theoretical prediction due to some softness in the reaction frame[26], but the post-yield response stiffness agrees well with the theoretical value. The maximum experimental strength for the interior unit was 3% higher than the predicted value, while the exterior unit attained maximum force levels about 8.5% above the predicted strength at the limit of proportionality. Theoretical predictions, based on material presented in the previous section imply unloading down the loading force-displacement curve, implying zero energy absorption. This behaviour was not obtained in the tests, due to inelastic compression strains developing in the concrete compression zones of the beam. This resulted in a softening

76

Priestley: Myths and Fallacies in Earthquake Engineering, Revisited

Fig.4.11 Lateral Force-Displacement Response of an Interior Precast Prestressed Beam-to-Column Subassemblage of the unloading and reloading stiffness. Residual drifts were, however, very low. The combination of low damage, low residual drift, high retention of stiffness at low displacements and stable hysteresis loops makes this form of construction attractive for seismic zones. 4.3.3 Precast Structural Walls with Unbonded Post-tensioning and Additional Damping.

It is obvious that the concept of unbonded prestressing can also be applied to precast wall buildings. Fig. 4.12 shows two possible applications. In the first (Fig.4.12(a)), wall elements are stacked vertically and post-tensioned to a foundation. This essentially creates a structural system that behaves in the same way as a prestressed beam-to-column connection, discussed in the previous section. A single crack can be expected to form at the critical section at the wall base, and will have non-linear elastic force-deformation characteristics. Note, however, that the footing must be of sufficient size and weight to ensure that rocking does not develop at the soil/footing interface. Although this can be an acceptable seismic mechanism very similar to unbonded prestressing, special measures are needed to ensure that plastic deformation to the soil at the corners of the footing is not excessive. Note also that the gravity weight contributes to the flexural resistance of the wall base in exactly the same way as the prestressing. Figure 4.12(a) also shows additional mild steel bars running through the wall/footing interface. The normal way this would be achieved would be with bars cast into, and protruding below the lowest precast wall element, grouted into preformed holes in the

Chapter 4: The Energy Myth - Seismic Response of Precast Prestressed Structures

77

Post-tensioning

Ductile Shear Links

Mild Steel bars

(a) Single Wall

(b) Linked Walls Fig.4.12 Precast Post-tensioned Walls

foundation. These mild steel bars are expected to yield in tension and compression as the base crack forms in much the same fashion as conventional reinforcement. Since the gap opening at the wall base can be large, it is normal to de-bond the mild steel bars for some length into the precast panel, to ensure strains at maximum displacement response are kept to acceptable limits normally less than 3%. Provided that the combined axial force at the wall base provided by gravity weight and prestressing exceeds the compression yield force of all the mild steel bars crossing the base interface, the residual displacements at zero lateral force will always be zero. In this case the mild steel bars provide additional damping to the bilinear elastic response, resulting in the flag-shaped hysteresis loops of Figs.4.1 and 4.2(d). An alternative configuration, also providing additional damping, is illustrated in Fig.4.12(b). Two parallel in-line precast walls are separately prestressed to the same foundation structure, and are linked by ductile shear links. Under lateral force, each wall rocks about its compression toe, resulting in relative vertical deformation across the shear links. If these are designed to have low yield displacements, considerable energy can be dissipated, and at the same time the overturning capacity of the complete system of two walls is enhanced above the sum of the capacities of the individual walls. This system was adopted for the PRESSS five-story test described in the next section. It should be noted that the concept of additional damping, in the form of yielding elements crossing the critical interface, can also be applied to beam-to-column connections in precast frame design. The yielding forces contribute to both yield moment and damping, and are thus beneficial to structural performance. It is a matter of economics whether it is better to limit structural response by increased levels of

78

Priestley: Myths and Fallacies in Earthquake Engineering, Revisited

strength, provided by higher prestress levels, or by a lower level of strength provided by the so-called hybrid system of prestressing plus tension/compression yield. 4.4 4.4.1 SEISMIC PERFORMANCE OF A LARGE-SCALE PRECAST PRESTRESSED TEST BUILDING Structural Systems in the Test Building

As the key element of the final phase of the PRESSS research program[21] a 60% scale five-story precast concrete building was constructed and tested under simulated seismic loading at the University of California, San Diego. The design of the building was based on concepts developed and tested in the earlier two phases of the program[21], and is described in detail by Nakaki et al[28]. A general view of the building under test, and the layout of the floor plan in the upper two floors are shown in Fig. 4.13. The purpose of the experiment was to test as many as possible of the ductile connection systems that had been developed in the early phases of the PRESSS program, in an environment which represented the full complexities of real structural interaction between lateral-force resisting elements and floor systems, and between orthogonal lateral-force resisting systems. As shown in Fig.4.13(b), the lateral-force system in one direction was provided by frames, while in the orthogonal direction, a combination of structural walls and gravity frames was used.

15 - 0

15 - 0

PreTensioned Frame

Topped Hollow Core

Jointed Wall System

Gravity Frame

Actuator Connection Panel

Gravity Frame

Topped Hollow Core

Actuator Connection Panel

TCY Frame

(a) Building Under Test (b) Floor Plan (top two floors) Fig.4.13 Five-Story Precast Prestressed Test Building (1 =304.8 mm)

15 - 0

15 - 0

Chapter 4: The Energy Myth - Seismic Response of Precast Prestressed Structures

79

The floor system for the lower three floors consisted of precast prestressed pretopped double-tees spanning between the two seismic frames, and connected to the centre of each of the two in-plane wall panels comprising the structural wall. In the upper two floors, topped hollow-core slabs spanning between the gravity frames and the central wall provided the floor system. The hollow-core slabs were connected to each other, and to the seismic frames with a cast-in-place topping. Although the plan dimensions of the test building were comparatively modest, it should be realized that the tributary floor area contributing inertia mass to the lateral force resisting systems in the prototype building on which the test building was based, was substantially larger than shown in Fig.4.13(b). Seismic Frames: The two seismic frames were different in concept, one being based principally on prestressed connections, and the other based principally on tension/ compression yield systems (see section 4.2). As shown in Fig.4.14(a), the lower three floors of the prestressed frame consisted of hybrid connections (see Fig.4.15(a)), with single-bay beams connected between continuous multi-floor columns by unbonded post-tensioning. Additional flexural strength and damping was provided by mild steel reinforcing bars slid through corrugated ducts in the beams and columns and grouted solid, with short unbonded lengths in the beams to reduce peak strains under seismic response. Behaviour of these connection was thus a combination of Nonlinear elastic and tension/compression yield, leading to flag-shaped hysteretic response as explained in Section 4.3.3, and shown in Fig.4.2(d).

15 ft (4.57 m)

15 ft (4.57 m)

15 ft (4.57 m)

15 ft (4.57 m)

PreTensioned Connection

TCY Connection

(a) Prestressed Seismic Frame

(b) Tension/Compression Yield Frame

Fig.4.14 Frame Connection Layout for Five-Story Precast Test Building

5 ELEVETED FLOORS @ 7 - 6 (2.28 m) O/C = 37 - 6 (11.43 m)

TCY Gap Connection

Hybrid Connection

80

Priestley: Myths and Fallacies in Earthquake Engineering, Revisited


1 - 6

(a) Hybrid Prestressed Connection

(c) TCY/gap Connection

(b) Pretensioned Beam Connection

(d) TCY Connection

Fig.4.15 Seismic Frame Connection Types[29]. In the upper two floors of the prestressed concrete frame, the precast beams were continuous for the full length of the frame, and were prestressed with pretensioned strands bonded in external stub beams, but debonded over the interior beam length between the outer column faces (see Fig.4.15(b)). No additional mild steel reinforcement was provided across the beam-column interfaces. The connections were thus column-tobeam connections (see Fig.4.3), with non-linear elastic hysteretic characteristics. The full-length pretensioned beams were threaded over column reinforcing bars extending from the top of the columns, with reinforcing bar splices in the upper column providing column continuity. Moment-resistance over the lower three floors of the non-prestressed frame was provided by the TCY-gap connections illustrated in Fig.4.15(c). At the bottom of the

Chapter 4: The Energy Myth - Seismic Response of Precast Prestressed Structures

81

connections the beams were clamped to the columns with unbonded post-tensioning threaded bars reacting through fibre grout pads over the lower partr of the connection only. Over the top two-thirds of the connection a 25 mm gap was provided. Mild steel reinforcement slid through corrugated ducts and was grouted solid in the upper part of the beams and in the columns. This reinforcement was designed to yield alternately in tension and compression under negative and positive moments respectively, without unloading the precompression across the bottom grout pad. As a consequence, all inelastic deformation in these connections occurred by opening and closing of the gap at the top of the connection. Over the top two floors of this frame, moment capacity was provided by the TCY connection using mild steel reinforcing in grout ducts at both top and bottom of the connection (see Fig.4.15(d)). Structural Wall: In the orthogonal direction, the structural wall consisted of four precast panels, each two and a half stories tall (see Fig.4.16). The panels were vertically connected to each other and to the foundation by unbonded vertical post-tensioning, using threaded bars. A horizontal connection across the vertical joint between the parallel and in-plane panels was provided by stainless-steel energy dissipating U-shaped flexure plates (UFP connectors), welded to embedded plates in both adjacent wall panels, thus acting as shear connectors, as discussed in relation to Fig.4.12(b). The two gravity frames on the periphery of the test building parallel to the structural wall were designed with hinged connections between the beams and columns, and thus did not contribute significantly to seismic resistance in the wall direction. Precast Column Bases: The precast column bases for the seismic and gravity frames were connected to the foundations by post-tensioning. The columns were thus capable of acting as vertical cantilevers in the out-of-plane direction, with a small, but significant contribution to seismic resistance. This was discounted in the design phase. Structural Design: Structural design was carried out in accordance with Direct Displacement-Based Design (DDBD) principles, which are discussed in more detail in the next chapter. The chosen design criterion was that the structure should achieve a maximum drift of 2% in either principal direction (frames or walls), under seismic excitation corresponding to a peak ground acceleration of 0.4g, on an intermediate soil. Erection of the test building was carried out inside the Charles Lee Powell Structural Laboratory at UCSD. Despite the restricted space and special constraints resulting from instrumentation requirements, erection of the building was carried out within 12 working days. A photograph of the test building during construction is provided in Fig.4.17. The four wall panels have been placed, and behind them, the double-tee floor beams for the lower three floors are visible. The double-tees for the front bay of the building have not yet been placed, nor have the hollow core slabs for the upper two floors. The frame to the left of the photo is the non-prestressed frame (TCY frame), and that on the right is the prestressed frame. At the back of the photo the upper two beams of the back gravity frame are visible, and behind that, we see the strong wall used for reacting seismic forces in the frame direction of loading. Note that in the prototype building the seismic frames do not extend to the corners of the building, and hence the beam stubs of the prestressed

82

Priestley: Myths and Fallacies in Earthquake Engineering, Revisited


TCY Frame UFP Connectors Prestressed Frame

18 - 9

Open

Open

Wall

Panel 3

Wall

Panel 4

Open

Open

18 - 9

Wall

Panel 1

Wall

Panel 2

Open

Open

15 - 0

15 - 0

Fig.4.16 Elevation of Structural Wall, showing Connections.

Fig.4.17 Five Storey Precast Test Building Under Construction

Chapter 4: The Energy Myth - Seismic Response of Precast Prestressed Structures

83

frame, which appear clumsy in the photograph, would be incorporated in the gravity beams extending from the end of the seismic frame to the corner of the building. 4.4.2 Test Procedure The objective of the experiment was to test the structure under a series of different earthquake levels. As shown in Fig.4.18, four different earthquake levels were defined, corresponding to 33, 50, 100, and 150 percent of the design level earthquake. The principal method of testing the building was psuedodynamic testing, using spectrum-compatible earthquake segments, scaled in frequency and amplitude from recorded accelerograms to match the appropriate excitation spectrum at 5% damping. The concept of psuedodynamic testing is described in Fig.4.19. A computer controlled the displacement input to each of two servo-hydraulically controlled actuators at each floor level. The computer contained a simple five-degrees-of-freedom model of the building, whose stiffness coefficients were initially calibrated by low-level flexibility tests. This model was subjected to inelastic time-history analysis under the specified earthquake record, to determine the initial displacements to be applied at the different floor levels. The forces required to apply the required displacements were recorded, and used to update the stiffness model describing the test building. Thus, as the stiffness of the test building was modified by inelastic action or strength degradation, the analytical model recognized the stiffness changes, and modified the required displacement input to the actuators accordingly. The mass used in the computer analysis represented the correct scaled tributary mass for the prototype building, rather than the reduced actual mass of the test building.
1.6 1.4 1.2

EQ-IV Zone 4, Soil Sc, Damping 5%

Acceleration (g)

EQ-III

1 0.8 0.6 0.4 0.2 0 0 1 2

EQ-II

EQ-I
3 4 5

Period T (s)

Fig. 4.18 Earthquake Excitation Levels for Five-Storey Precast Building Test

84

Priestley: Myths and Fallacies in Earthquake Engineering, Revisited

Fig.4.19 Concept of Psuedodynamic Testing, Applied to Test Building 4.4.3 Structural Performance of Test Building

Time-histories of floor lateral displacements of the test building when subjected to EQ IV excitation (i.e. 150% of design earthquake) in the wall direction of loading are shown in Fig.4.20(a). The response, which covers the period of strong motion from the scaled accelerograms, is referenced to model time scale. In prototype scale, the test duration is 10 seconds of strong motion. On the basis of displacement response, it would appear that the behaviour is dominated by first-mode response, with little contributions from the higher modes. However, when storey shear forces are examined (e.g. Fig.4.20(b)), it is seen that higher mode effects are very apparent, particularly in the first half of the response record. The behaviour exhibited in Fig.20 is characteristic of response in both the wall and frame directions of response. It was found that the higher mode effects were more pronounced at higher levels of excitation, particularly for the wall direction of testing, conforming to the theoretical findings of higher mode effects discussed in Chapter 3.

Chapter 4: The Energy Myth - Seismic Response of Precast Prestressed Structures

85

(a) floor displacements

(b) Storey Shear Forces Fig.4.20 Wall Direction Response at 1.5xDesign Excitation It is of interest to compare the design levels of moment, shear and floor force with peak values recorded in the experiment. Such a comparison is shown in Fig.4.21 for the wall direction of loading under design excitation (EQ III). The amplification of structure base moment capacity is primarily due to the contribution of the seismic frame columns acting as vertical cantilevers in the out-of-plane direction. At higher levels, the effects of higher mode effects are apparent in the shape of the moment profile. Storey shear forces (Fig.4.21(b)) show much more pronounced higher mode effects, particularly in the upper levels of the building. This conforms to behaviour predicted by the dynamic analyses reported in Section 3.2.1.

86

Priestley: Myths and Fallacies in Earthquake Engineering, Revisited

Fig.4.21 Comparison of Design and Peak Experimental Response Profiles Experimental floor force levels show the greatest amplification over design values particularly in the lower floors. This is of concern, in that precast design normally has discrete connections between floors and lateral-force resisting elements, which must be designed for the actual, rather than design, floor force levels. In fact, the high floor force

Chapter 4: The Energy Myth - Seismic Response of Precast Prestressed Structures

87

levels experienced in the experiment (and later confirmed theoretically by time-history analyses) were substantially higher than those for which the floor/frame and floor/wall connections were designed. As a consequence, at EQ III (design level) and EQ IV (150% of design level), it was necessary to artificially introduce increased elastic damping for the higher modes, into the computer model of the structure controlling displacement input. As a consequence, the higher mode effects shown in Figs4.20 and 4.21 are actually underestimates of the values that would have occurred if the same level of damping (5%) had been associated with the higher, as well as the fundamental mode. This raises an interesting question What is the appropriate level of damping to be associated with higher mode response? There is reasonable evidence to support the commonly adopted value of 5% in the first mode, but as far as we are aware, there is little to support the same, or any other value, in the higher modes. Since the dynamic amplification effects are found to be strongly dependent on this damping value, the issue is of some importance. A further issue of importance related to damping is the way in which elastic damping for the first mode is represented in inelastic time-history analyses. If the damping coefficients are related to initial estimates of stiffness, then there are grounds for believing that the elastic damping forces during inelastic response will be grossly overestimated. Systematic theoretical and experimental research is needed to resolve both of these important issues. One of the most important issues from the test program was the level of physical damage sustained by the test building. Fig.4.22(a) shows the only significant damage

(a) spalling at wall toe (b) Vertical movement at UFPs Fig.4.22 Wall Condition at 150% of Design Excitation

88

Priestley: Myths and Fallacies in Earthquake Engineering, Revisited

sustained by the structure during the wall direction loading. This consisted of minor spalling of cover concrete in the vicinity of the wall toes, even at 150% of the design seismic intensity. The condition of the structure at this level of response would have satisfied a serviceability limit state performance criterion, in that all damage was cosmetic, and could be repaired without disruption to the normal usage of the building. The condition of the structure when tested in the direction parallel to the seismic frames was also excellent. Figure 4.23 shows the condition of connections representative of the four connection categories comprising the two seismic frames, after testing to more than twice the design drift limit of 2%. The hybrid connections behaved in similar fashion to the unbonded beam/column subassemblages described in Section 4.3.2, with controlled joint cracking and only minor spalling of beam cover concrete adjacent to the

Fig.4.23 Condition of Different Connection Types after Testing to 4.5% Drift

Chapter 4: The Energy Myth - Seismic Response of Precast Prestressed Structures

89

columns. Thus, inclusion of the mild steel reinforcing bar energy dissipators did not result in increased damage. The pretensioned beams in the upper two floors of the prestressed frame also performed extremely well, with minimal damage, as illustrated by Fig.4.23(b). The critical cracks formed naturally in preformed rebates in the beams in line with the column faces. More damage was experienced by connections of the non-prestressed frame, and particularly those with TCY-gap connection details, as may be seen in Figs.4.23 (a) and (b). The level of damage, at twice the design intensity, however, was less than would be expected from a conventional reinforced concrete frame at 100% of design intensity. 4.4.4 Modelling Structural Performance It is clearly of considerable importance to be able to accurately model the response of precast structures. As has already been mentioned, we see this importance as being related to design verification, rather than as a tool to be used for driving design. We have also pointed out, in Chapters 1 and 2, that conventional analysis tools have in the recent past been more complex than accurate in modeling reinforced concrete behaviour. It was thus a key part of the PRESSS program in general, and the five-story test in particular to develop simple, yet adequately accurate analytical models for the five-story building response. In order to give a fair test to these models, they were developed prior to the experiment, and all predictions of response were carried out prior to the relevant experiment being run.

90.0

Igross = 839808 in4

78.1

91.0

UFP Spring
Lumped Nodal Mass

90.0

Rigid Link

80.0
Base Rigid Link

Gravity Weight & PT Forces Compression Only Spring

PT Spring

49.6

53.0

49.6

53.0

Fig.4.24 Analytical Model for Wall Direction Response Prediction

90

Priestley: Myths and Fallacies in Earthquake Engineering, Revisited

The analytical model used for wall direction response is represented in Fig. 4.24. The two walls were separately represented by elements at their vertical centerlines, and given structural properties based on the uncracked section. The walls were supported at the base on rigid horizontal beams supported at each end by stiff, compression-only springs located at the calculated position of the centre of compression at each wall end under rocking response. As shown in Fig. 4.24, these springs were preloaded by forces representing the gravity weight and unbonded prestressing force. As the wall rocks, the unbonded prestressing threaded bars extend, resulting in an increased clamping force at the wall base. This was modeled by an additional spring under the wall element, as shown in Fig.4.24. The two wall elements were connected by vertical springs acting between rigid horizontal links extending at each floor level. These inelastic springs represented the characteristics of the U-shaped flexural yielding energy dissipators (UFPs). The properties for these springs were determined by analyses, and confirmed by tests on spare dissipators, prior to the building test. It should be mentioned that additional lateral resistance of the building in the wall direction resulting from the strength and stiffness of the frame columns acting as vertical cantilevers in the out-of-plane direction was ignored in the initial design-confirmation analyses. An example of the comparison between predicted and experimental response, using the model of Fig.4.24, is shown in Fig.4.25 for the maximum excitation level EQ IV, corresponding to 150% of design excitation. The results show that the agreement between the predicted and experimental values is gratifyingly close for the displacement time-history shown in Fig.4.25(a). This is especially true when correction is made for the initial experiment offset displacement, indicating that the model accurately captured the fundamentals of the first mode, including inelastic action, and was capable of excellent prediction of peak displacement response. The agreement for base shear force (see Fig.4.25(b)) is also very good, with the level of base shear contributed by the first mode being predicted with an accuracy equivalent to that for the displacements. The component of base shear due to higher modes was also well predicted. A comparison between experimental and predicted values for the base moment timehistory (see Fig.4.25(c)) is also satisfactory. Note that the base moment is less affected by higher mode effects than is the base shear force. It is clear, however, that the strength was underestimated by about 20%, due to reasons previously discussed. Similar methods were used to predict the response in the frame direction, with a similar level of success as obtained for the wall direction. More complete information is available in [29,30,31]. 4.5 STEEL FRAMES WITH UNBONDED POST-TENSIONING

The Northridge earthquake of 1994[17] revealed steel frames with welded beam-tocolumn connections to be suspect under inelastic seismic response. More than 100 steel frame buildings were found to have weld fractures, or parent metal fractures adjacent to

Chapter 4: The Energy Myth - Seismic Response of Precast Prestressed Structures

91

Fig.4.25 Comparison of Predicted and Observed Wall Response At 1.5 times design excitation.

92

Priestley: Myths and Fallacies in Earthquake Engineering, Revisited

welds. Christopoulos and Filiatraut[32] have recently investigated applying the concept of unbonded post-tensioning which was developed for concrete frames, to steel frames. Extensive analytical and experimental work, including shake table testing, has shown that the concept is economical, results in zero residual displacement, and avoids problems of flange buckling which is almost inevitable when plastic hinges form in the beams. Their work clearly indicates that much better performance can be guaranteed when the fat hysteresis loops of steel frames, which most closely resemble the ideal elasto-plastic response, are replaced by nonlinear elastic hysteretic loops with minimal energy absorption. 4.6 CONCLUSIONS

The work described in this chapter was initiated because of a suspicion, expressed in the earlier Myths and Fallacies paper, that elasto-plastic hysteretic response was not the holy grail that earthquake engineers should seek. This work has indeed confirmed that suspicion. It has been found, largely based on analyses and experiments involving precast prestressed concrete structures that Structures with high hysteretic energy absorption tend to have larger residual deformations than structures with less desirable hysteretic characteristics. Structures with nonlinear elastic hysteretic characteristics, such as provided by unbonded prestressing, and/or gravity loads (in walls) have zero residual deformation, and generally peak displacement response is similar to that of systems with higher energy dissipation characteristics. This is particular the case when the increase in elastic stiffness of the prestressed structures is considered in the analyses. Precast (or conventionally constructed) structures with unbonded reinforcement need less transverse reinforcement for shear strength, and less special joint reinforcement in beam/column joint regions of frames than is required for conventionally reinforced concrete structures. Experiments on subassemblages, and on a near full-size test building have established that the level of damage to be expected with structures utilizing the unbonded prestressing concept is much less than with conventionally reinforced structures. This conclusion applies to steel frames as well as to precast concrete frames and wall structures. The performance of precast structures connected with unbonded prestressing can be enhanced by including special energy dissipating elements in the design. When properly designed these contribute to the lateral strength of the seismic system as well as increasing the effective damping (thus reducing, generally, the peak displacement response), without sacrificing the highly desirable characteristic of residual displacement, inherent in the unbonded prestressing concept.

Vous aimerez peut-être aussi