General report on local ductility

General Report on Local Ductility André Plumier Department of Civil Engineering, University of Liege, Belgium

1. INTRODUCTION The Northridge (1994) and Kobe (1995) earthquakes have been destructive for many steel and composite steel concrete joints, essentially in moment frame structures. The connections were believed to be ductile and capable of withstanding the repeated cycles of large inelastic deformation explicitly relied upon in the building code provisions for the design of these structures. A wide spectrum of unexpected brittle connection fractures did occur, ranging from minor cracking, observable only by detailed non destructive testing to large fractures extending across the full depth of the columns. At the time this damage was discovered, the structural steel industry and engineering profession had little understanding of the specific causes of this damage. The incidence of damage was sufficiently pervasive to cause worldwide concern by structural engineers and building officials with regards to the safety of these structures in seismically active regions. In response to these concerns, problem-focused studies of the seismic performance of welded steel and composite steel concrete moment connections mainly have been launched in the US, in Japan and in Europe. The amount of work performed has been enormous, in particular in the US, with the SAC joint venture. Research efforts in Europe and Japan have also reached unprecedented levels bearing on more specific aspects related to steel and composite moment frames (Europe) and to frames with tubular columns (Japan). After about 4 years of work, many practical outcomes are available and they have been or will soon be translated into design practice and codes. It is the case with the existing FEMA guidelines (FEMA,1997)and the AISC seismic provisions(AISC,1997) and with the soon revised part 1.3 of Eurocode 8 work(Eurocode 8, 1998). This research effort has been pushed by the need of an immediate ability to define design conditions avoiding brittle behaviour. It is quite natural that the seismic design context is the one where the concern on local ductility is the highest Mazzolani & Piluso(1996)and the research effort the most important. It is because a plastically deformed structure, still standing, is the ultimate limit state design objective under the design earthquake, while the plastic state in the plastic design of structures submitted to static loading, is a situation normally never reached under design loads and, if reached, with low requirement on ductility. However, the context of static loading, like the seismic one, requires correct evaluation of the ductility of sections. It is the essential subject of the six papers presented in this Session 5

“Local Ductility” of SDSS 99, which present studies aiming at the definition of better computation methods for the ductility of I shaped beams in bending or their web panel in shear, related to the wall slenderness and to the yield strength of the material. For this reason, the general report will focus mainly on the results and the practical implications of the “postNorthridge and Kobe” researches, stressing the aspects which appear as new or as revisions of old ones. All these developments can only be understood with reference to the “capacity design” of structures for earthquake resistance. It must be recalled that, in that concept, specific elements of the primary lateral force resisting system are chosen and suitably designed and detailed for energy dissipation under severe imposed deformations. The critical regions of these members, often termed plastic hinges or dissipative zones, are detailed for inelastic behaviour. All other structural elements are protected against actions that could cause failure, by providing them with strength greater than that corresponding to the development of the maximum feasible strength in the plastic hinge regions. The following features characterize the procedure: • Potential plastic hinge regions within the structure are clearly defined and designed to have dependable strengths. • Potentially brittle regions, or those components not suited for stable energy dissipation, are protected by ensuring that their strength exceeds the demands originating from the overstrength of the plastic hinges.

Figure 1:Principle of Capacity Design (from Paulay and Priestley, 1992) To highlight the concept of capacity design, the chain shown in Figure 1 is often considered. As the strength of a chain is the strength of its weakest link, one ductile link may be used to achieve ductility for the entire chain. The nominal tensile strength of the ductile link is subject to uncertainties of material strength and strain hardening effects at high strains. The other links are presumed to be brittle, but their failure can however be prevented if their strength is in excess of the real strength of the ductile weak link at the level of ductility envisaged. A correct application of the capacity design principle thus requires the knowledge of the material properties(resistances,…) both of the plastic zones and of the neighbour zones and the evaluation of the stresses and strains which must be sustained by the material of the plastic zones. It is clear now that pre-Northridge or Kobe design were characterised by weaknesses in the evaluations made either on the resistance side or on the action effect side. In the following, these weaknesses are summarised and the counter measures defined at present to obviate them are presented.

2. THE RESISTANCE SIDE: MATERIAL AND CONNECTORS. 2.1. Yield stress of base material. It is obvious that capacity design cannot be based on a nominal strength or a lower characteristic value of the yield strength without taking a chance to underdesign a

neighbouring connection. This fact was already recognised in the ECCS Design Recommendations, ECCS 1988, which indicate to consider an upper limit of yield stress fy max, and later by Eurocode 8(1998), which prescribes to consider “an appropriate estimation of the actual value of yield stress” and to indicate the allowable maximal yield strength in drawings. To our knowledge, the use of design values of yield stress higher than the nominal value was not prescribed in the US or in Japan before the Northridge and Kobe earthquakes. This has generated damageable effects on connections, which had to sustain beam bending moments higher than foreseen and failed in welds at the beam ends or in the flanges of the columns. For seismic design, codes must absolutely prescribe design values of yield strength higher than the nominal values, especially because recent studies have reported that the margin between the actual average yield strength and the specified yield strength has progressively increased over the years for some structural steels, even though the steel specification itself remained unchanged. For example, a few decades ago, yield strength of 255 to 270 Mpa (37 to 39 ksi) were typically reported for ASTM-A36 steel, whereas 20years later, the same steel grade revealed a substantial increase of the yield strength, with values ranging from 325 to 360 Mpa (47 to 52 ksi) and eventually some values in excess of 420 Mpa (60 ksi),FEMA 1997.To manage this problem, some North American structural shape producers are preparing an ASTM specification for a grade 50 steel having a specified yield strength of 345 Mpa (50 ksi), an upper limit on yield strength of 448 Mpa (65 ksi) and a specified maximum yield-totensile strength ratio of 0.85; it is already produced by some steel mills. Because of the uncertainties on the real value of the yield strength, very detailed prescriptions now deal with the problem of design yield strength for structural elements and for connections in the recent U.S. design recommendations FEMA 267 (1997) or AISC (1997). These prescriptions go further than previous European rules, by defining explicitly “expected yield strength” Fye to be used for purposes of computing plastic moment Mpl. These Fye have been established by statistical survey of web coupon tensile tests. Strain hardening is also accounted for, by a coefficient β ranging from 1.1 to 1.2, depending on the steel grade and the design plastic moment is equal to Mpl = β.Fye.Zb. It must be pointed out that, because of the scatter in yield stresses mentioned above for the low grade ASTM 36 steel (Fy nominal 36 ksi), the expected yield strength Fye of that steel ranges as high as 51 to 55 ksi, that means about 1.5 times the nominal value. 2.2. Through thickness resistance of column flanges. Through-thickness failures of column flanges has been observed both in buildings damaged by the Northridge Earthquake and in some test specimens. They are not well understood, because they can be a function of many factors: metallurgy of the steel(sensitivity to lamellar tearing), effect of axial loads, of loading rate, of tri-axial restraint, of local hardening and embrittlement in the welds heat affected zone, effect of stress concentration due to backing bars or defects at the root of the beam to column flange welds, effect of deformations caused by connection components like for instance the bending stresses and flange curvature induced by panel zone yielding. Given these many complex factors, the determination of a reliable and rational basis upon which to set permissible through-thickness stresses requires significant research, which is still under way. As provisory indications, FEMA 267 (1997)and AISC seismic provisions(1997) limit the through-thickness resistance: Fu=0.9 Fy ( that is 45 ksi for A36 and A572 steels).

2.3 Toughness of base and weld material. The existing data on cyclic behaviour of structural connections have essentially been obtained in cyclic quasi-static tests, though the earthquake effects are dynamic ones. Translated into strain rates ε’, it means approximately that most test results correspond to ε’=10-4/s, while real ε’ is around ε’=10-1/s. Material properties and notch effects are different in these two ranges of ε’ values and higher strain rates correspond to lower cyclic resistance of steel components, as was demonstrated by many tests, Bruneau et al, 1998. It has been established that tensile strength fu and yield strength fy increase at higher strain rate (at room temperature), but fy increases more than fu, so that fu/fy is reduced. It is known that high enough values of fu/fy are needed to avoid a failure in which only very high localised strains, meaning low ductility and early cracks, are developed. As the spreading of plasticity and the development of significant plastic hinge rotations are needed to realise a complete plastic mechanism in a structure, a material which exhibit adequate values of fu/fy in dynamic conditions is required. This characteristic can best be set forward by a dynamic notch test like the CHARPY V-notch, which, in fact, characterise the energy absorption capability of the material realised in the specimen by spreading of plasticity in dynamic conditions. Explicit requirements now define, in U.S. specifications like FEMA 267(1997) or AISC Seismic Provisions(1997), a minimal notch toughness for heavy structural shapes with flanges thicker than 1.5 inch (38mm). This value of toughness is 20 ft -lbs at 70 degrees F, which correspond to 27 J at 20°C. The restriction in the application field (tf>38mm) corresponds to the fact that this value of toughness is normally always realised in smaller sections than those with flanges thicker than 38 mm. The CHARPY tests must be conducted on specimens made of the weakest part of the section. For what concerns toughness, it is the intersection zone between flange and web, where the grain size of the material is higher because of the slow cooling realised there in the fabrication process. This U.S. prescribed toughness is similar to the minimum value of Eurocode 3 for structural steels. It corresponds to the European notch toughness quality “B”, the lowest steel quality with specified values of material properties; quality “A” means that no properties are required; “C” and “D” correspond to 27 J at 0 °C (32 F) and –20 °C (-4F) respectively. Requirements for weld filler metal are also defined in U.S. specifications. They are stronger than for steel shapes: 20ft-lbs at 20°F -, equivalent to 27 J at –7°C. In Europe, the requirement on weld toughness must be demonstrated on the weld itself, as realised in a test coupon made with all conditions similar to the real weld; the toughness of this weld material may not be less than the toughness of the base material. 2.4. Strain rate effect on bolted connections. In the European construction practice, bolting on site is preferred to welding and typical European design like the one of fig involve high strength prestressed bolt working in tension. Recent experiments realised quasi statically and dynamically, Plumier and Sanchez, 1999, does not show any negative effect of high strain rate on such connection design using 10.9 bolts, prestressed at the standard value 0.7 fuB.As.

3. THE ACTION EFFECT SIDE: STRESSES AND STRAIN IN THE CONNECTION ZONE. 3.1. Rotations at plastic hinges. Traditionally, structural engineers have calculated demand in moment frames by sizing the members for strength and drift using code forces (either equivalent static or reduced dynamic forces) and then “developing the strength of the members”. In the US, since 1988, “developing the strength of the members”.has been accomplished by designing connections corresponding to certain prescribed types. These connections had been tested in laboratories, generally with beams of relatively small depth, and had demonstrated ability to develop some plastic capacity. It was assumed that the prescribed connections would be strong enough so that the girder would yield (in bending), or the panel zone would yield (in shear) in a nearly perfectly plastic manner, producing the plastic rotations necessary to dissipate the energy of the earthquake. It is now known that the prescriptive connections were often incapable of behaving in this manner: during the 1994 Northridge and 1995 Kobe earthquakes, many moment-frame connections fractured with little evidence of plastic hinging of the beams or yielding of column panel zone. Certainly there are many reasons to be found on the resistance side (see paragraph 2) for this bad behaviour of connections which in principle had demonstrated experimentally a fair cyclic behaviour. But part of the explanation holds in the way the test results were interpreted. In the past, the stress in the interpretation of results was put on local ductility, ratio of plastic to elastic displacement capacity, and on comparison between local ductility and global ductility or behaviour factor. Little attention was paid to rotations in plastic hinges (expressed in radian, not in relative terms). It is now known that a realistic estimate of the interstorey drift demand for most structures and most earthquake is on the order of 1.5 to 2.5% of the storey height for structures designed to code allowable drift limits. This brings local yielding rotation demand of about 0.02 radian. Testing of the standard US pre-Northridge welded flange-bolted web connection as well as the standard Japanese fully welded H beam on tubular columns show that they are unable to reliability provide plastic rotations beyond about 0.005 radian. This is particularly true when the beam depth are high, because plastic strains required for a given rotation are increased with respect to a smaller section. Figure 2. The problem is similar when composite T sections are used.

Figure 2: Strains Increase with Section Depth and Shape

As a consequence of the explanations above, a connection plastic rotation demand of 0.03 radian has now been recommended as design criteria in the US recommendations, to provide a safety margin against the demands actually expected in most cases. 3.2. 3D stresses in welded zones. Weld shrinkage, if prevented, creates residual stresses. The bigger the weld throat and the longer the weld, the higher the residual stresses. At a beam column welded joint, weld shrinkage in the direction of the weld length is almost totally prevented, because the section offered by the beam and column flange to resist shrinkage force is high. This creates residual tension stresses parallel to the weld in and around the weld which can be as high as the yield stress of the weld material. If thick flanges or connections with cover plates are used, tension stresses perpendicular to the flanges surfaces may also be generated. Poor local ductility in the steel of the beam, close to the weld results from this situation, because yielding requires the development of slip planes which can only be achieved if the yield shear stress fy’ is exceeded. Considering the Tresca yielding criteria, for instance, fy’=fy/2 and shear stresses higher than fy’ are needed for plastic behaviour. It can be shown that triaxial or biaxial residual stresses can reduce shear stresses under the value needed to develop slip planes and create the conditions for brittle failure. See Figure 3, Bruneau et al, 1998 . This may explain in part the failures observed after the Northridge earthquake and justify a design option in which the plastic zone is moved away from the column face- see paragraph 4.1.

Figure 3: Triaxial Stresses in Unrestrained and Restrained Steel Elements(Bruneau, 1998) 3.3. Flexibility of column flange. Many damages observed after the Kobe (1995) earthquake have been explained by the difficulty to properly stiffen the SHS tubular section which are very widely used for columns in the Japanese constructional practice. As a uniform stress distribution in the beam flanges is

needed to achieve ductility in a plastic hinge formed close to the column, the design must either improve the stiffening of the column or choose solutions which, by moving the plastic hinge away from the column, reduce the stresses on the column flanges. 3.4. Real distribution of stresses and strains at beam ends. It is clear that the distribution of stresses in beams at some distance away from the column face agrees with that predicted by the classical beam theory, but it has been established recently by several researchers that the stress distribution changes drastically in the vicinity of the connection, Goel et al,1997, JSSC,1997. The area in the middle of the web near the shear tab is virtually devoid of stresses, while high stress concentrations occur in the flanges and flange welds. Figure 4.

Figure 4: Von Mises Distribution of Stresses in a Classical Moment Connection (Goel, 1997) The distribution of stresses is governed not only by imposed loading (the traction boundary condition), but also by the deformation constraints imposed by the connection details (the displacement boundary condition). Traditional welded steel moment connection details prevent free deformation of the top and bottom fibbers of the beam. Such boundary conditions constrain the warping of the beam cross section due to shear deformations. Stresses caused by boundary effects in small portions of structural elements are often not considered in practical structural design. However, for a wide flange section beam with a large web depth-tothickness ratio, boundary effect induce a substantial direct shear force in the beam flange welds, which are more strained than it is considered in the classical design considering that shear is transmitted by the web and bending by flanges. 3.5. Composite character of sections. Many structures are constructed with concrete floor slabs that are provided with positive shear attachment between the slab and the top flanges of the girders. The resulting composite action, by raising the position of the neutral axis of beams, increase the strains in the bottom flange and reduce the ductility of sections, as already mentioned at paragraph 3.1. The composite action also increases the effective strength of the girder significantly, particularly at sections where curvature of the girder places the top flange into compression. This effect

cannot be neglected in the design of systems classified as moment resisting steel frames, because it can result in a number of effects including the unintentional creation of weak column-strong beam and weak panel zone conditions, or a reduction of the effectiveness of reduced section or “dog-bone” type connection assemblies. All these effects need to be assessed, because they can influence the local and global ductility. 3.6. Panel zone deformations. If the joint panel zone of the column yields in shear before either beams or columns reach their plastic stage, the state of stress/strain in a beam welded connection beam will be affected by the panel zone distortion. Fractures of the welds connecting the beam to the column may be caused in part by the kinking occurring in column flanges when the panel zone distortions become large, see Figure 5, Krawinkler, 1997.

Figure 5:Local Effects of Panel Zone Distortion

4 RECENT DESIGN TRENDS AIMING AT AN IMPROVED LOCAL DUCTILITY. 4.1. Position of plastic hinges. A decisive step has been made in the recent US guidelines for the seismic design of moment frames FEMA 267 and AISC Seismic Provisions in the seek for local ductility. It is prescribed that beam-column connections should be designed with sufficient strength (through the use of cover plates, haunches, side-plates, etc.) to force development of the plastic hinge away from the column face. It is also mentioned that this condition may be attained through local weakening of the beam section at the desired location for plastic hinge formation. (Reduced beam section RBS or “dogbone” solution see paragraph 4.2.). This stringent recommendation results from the consideration of the factors influencing negatively the local ductility described at paragraph 3. All these factors are present when and only when plastic hinges are developed within the beams at the face of the column or when the column panel zone itself yields. If the column panel zone yields, the column flange deformation results in stresses on the beam flange to column flange joint, a condition which can contribute to brittle failure. If the plastic hinge forms in the beam, at the face of column flange material, this can result in large through thickness strain demand on the column flange material and large inelastic strain

demands on the surrounding heat affected zones. These conditions can also lead to brittle failure.

Figure 6: Desired Plastic Frame Behaviour (FEMA267,1997) Of course, past and present researches have demonstrated that there are conditions of material properties, design and detailing configurations that permit reliable performance of connections with yielding occurring at the column face. In particular, the two recent European research projects in which over 40 tests have been run in quasi static as well as in dynamic conditions on welded subassemblage of IPE 450 (450 mm depth) beams, have demonstrated that a ductile behaviour of the welded connections was achieved when the base material and the welding conditions were properly defined, as it is explained with some details at paragraph 4.3. However it is felt in the U.S. that these conditions are not yet mastered for the high beams used in the U.S. building practice and further research is going on. 4.2. Qualification of design. Because recent earthquakes have raised many questions on the effective local ductility of steel structures, U.S. design guidelines (FEMA 267, AISC Seismic Provisions), states that connection design must be based on experimental evidence coming from either tests reported in research or from documented tests performed for other projects that are demonstrated to reasonably match the project conditions or from tests that are conducted specifically for the project and are representative of the project. The requirements for representativity are high: the test specimens must be constructed using materials, sizes, configurations, processes and quality control and assurance methods that match as closely as practicable those of the real connections; For instance, beams with a tested yield strength that is more than 15 percent below Fye shall not be used in qualification testing. These requirements are understandable in the context of poor local ductility of steel structures in the recent earthquakes, but they are very strong ones. Few existing results do correspond to the requirements on the test specimens mentioned above or to the required testing procedure, in particular a plastic rotation of at least 0.03 radian.

4.3. Design of reduced beam sections. Since the initial development of the reduced beam section concept (Plumier, 1990), the idea has been studied into detail all over the world. Several alternative configurations of RBS connections have been tested. As indicated in Figure 7, these include constant section, tapered section, curved section, and drilled hole patterns. The tested specimens had reductions in the beam flange area on the the order of 35% to 45% and produced moments at the face of the column that resulted in stresses on the weld and column as large as 90 to 100 % of the expected material strength of the beam. It appears that some configurations are more desirable than others. In particular, the drilled hole section pattern is subject to tensile failure across the reduced net section of the flange through the drill holes. A few RBS tests utilising straight or tapered cuts have failed within the reduced section at plastic rotation demands less than 0.03 radian. The failure occurred at locations at which there was a change in direction of the cuts in the beam flange resulting in a notch effect. There have been no reported failures of RBS connection assemblies employing the circular curved flange cuts, but successful test have also been performed using the straight and tapered configurations. It is important that the pattern of any cuts made in the flange be proportioned so as to avoid sharp cut corners. All corners should be rounded to minimise notch effects and, in addition, cut edges should be ground in the direction of the flange length to have a surface roughness value less than or equal to 1000 as defined in ANSI/ASME B46.1. It has been observed that hinging which occurred at the reduced section was less prone to buckling of the flanges than in some of the other configurations which have been tested, due to the very compact nature of the flange in the RBS. On the contrary, the tendency for lateral-torsional buckling is significantly increased, suggesting the need for lateral bracing of the beam flanges near the reduced section. There has been a concern that the presence of a concrete slab at the beam top flange would tend to limit the effectiveness of the reduced section of that flange, particularly when loading places the top flange into compression. It is possible to mitigate this effect with proper detailing of the slab. Testing of RBS specimens with composite slabs has recently been successfully conducted, Trembley et all, 1997, on beams where the shear studs were omitted from the portion of the top flange having a reduced section. This design minimises the influence of the slab. In addition, a 1 inch wide gap was realised in the slab around the column to reduce the influence of the slab on the connection at the column face. Concerns have been raised by some engineers over the strength reduction inherent in the RBS. Clearly, code requirements for strength, considering gravity loads, wind, seismic and other loads must be met, considering that the resistance of the beams is that of the reduced section. For high seismic zones, where beam sizes are often governed by elastic stiffness considerations for drift control, it can be shown that RBS reduce the elastic stiffness of the structure by 5% only, when inelastic strength of RBS is in the range of 70% of that of full section. Practically, this means that a building designed using RBS will require little or no increase in size of the steel sections needed,Plumier,1997.

Figure 7: Reduced Beam Section Patterns Finally, it must be stated that the present U.S. design guidelines concerning RBS are still interim ones and contains some unclear options, because of the uncertainties about the resistances of welds. On the one hand, the prescribed allowable through-thickness stresses in the column flange are so low that they practically oblige the designers considering the RBS solution to reduction over 50%: through-thickness Fu is worth 32.4 ksi for A36 steel and 45 ksi for A50 steel, while the design yield strength of beam is 55 or more for A36 steel. On the other hand, all tests on RBS run until now, including the successful ones, considered reduction in the beam flange area on the order of only 35 to 45% and, considering the problem of drift control as well as the lateral buckling of beams and general economy of projects, it is certainly not advisable to go to higher reduction in area. 4.4. Some recent results on welding conditions allowing yielding of beam at column face. The problem of welded moment frame connection has been studied in two recent European research projects. In the STEELQUAKE project, Plumier et al 1998, 4 types of butt welds have been tested at the connection of an IPE450 beam to a HE300B column. Figure 9. The base material belonged to class C or D (27 J at 0°C or –20°C in CHARPY test); connections were made either by means of direct welding to the column or by using a welded end plate which was later bolted to the column. The various types of weld preparation are shown at fig .K preparation like in C4 was also used for the welding of end plates. All welds were made using FCAW technique: flux cored are welding with ATAL gas protection, currently called MIG semi automatic; welding material is basic (FLUXOFIL 31). Each test was run at constant amplitude of displacement in order to allow a direct interpretation of results in terms of low cycle fatigue, with a presentation of results in Wöhler curves. The main observations and conclusions are the following ones: - except with the C3 type of weld, which was a prototype with triangular backing bar, no brittle failure was observed and all failures took place in the beams - welds made with V preparation and a permanent steel backing bar gave scattered results; - welds made with V preparation and a temporary copper backing gave good results, similar to those obtained with a K preparation and welding from top and bottom;

-

the best results were obtained in the design using butt end plates, probably because some flexibility of the end plate contributed to the ductility, reducing the strains in the beam; tests run with low amplitudes of displacements generated failures close to the welds, without buckling and without strength degradation before a final drop in resistance tests run with large amplitudes of displacement generated buckling of the beam, with progressive strength degradation and failure in the buckles; the number of cycles to failure in case of low amplitudes of displacements can be as low as the one in case of high amplitudes, because of the two different failure modes observed in these two circumstances.

Figure 8: Beam Flange Welds Tested in the STEELQUAKE project. The STEELQUAKE tests were complemented in the RECOS project by considering the strain rate parameter on similar specimens, Plumier and Sanchez 1999. No brittle failure was observed, though the number of cycles or energy absorption was reduced. Globally, it can be concluded from these European tests that adequate requirements on base and weld material make it possible to achieve ductility with classical fully welded connections when European IPE beam sections up to 450 mm are used. However, the adequacy of requirements for higher beams with thicker flanges still has to be assessed. 4.5. Connections stronger than the beams. As a result of the actual U.S. design option for the position of plastic hinges, the design of connections stronger than the connected beam is promoted. Some types of such connection are classical, like those using haunches. Other types have been developed in a proprietary way, like the side plate connections or the bolted bracket connections. Three types, which are public and shown at Figure 9, are also a logical consequence of observations concerning the real distribution of stresses and strain at the beam ends (see paragraph 3.4): - the beam with tapered flange, successfully tested by SOTIROV et al,1999. - the flange rib connections - the cover plate connection.

Figure 9 : 3 Types of Connections Stronger than the Beams In the latter, which corresponds to a strong trend in the U.S. short cover plates are added to the top and bottom flanges of the beam with fillet welds adequate to transfer the cover plate forces to the beam flanges. The bottom flange cover plate is shop welded to the column flange and the beam bottom flange is field welded to the column flange with a common weld. The web connection may be either welded or high strength (slip critical) bolted. A variation of this concept which has been tested successfully ,FORELL and ELSESSER Engineers,1995, uses cover plates sized to take the full flange force, without direct welding of the beam flanges to the column. In this version of the detail, the cover plate provides a cross sectional area at the column face about 1.7 times that of the beam flange area. In the detail which has been tested, a welded shear tab is used, and is designed to resist a significant portion of the plastic bending strength of the beam web. One of the issues that must be faced by designers utilising coverplated connections is the sequence of operations used to attach the cover plate and the beam flange to the column. Proper control of pre-heat and heat input are important because improper procedures can result in brittle conditions in the heat-affected zone(see3.2.).When developing cover plated connection details, designers should attempt to minimise the total thickness of beam flange and cover plate, so as to reduce the total size of the butt weld. As a rough guideline, it is recommended that the thickness of the weld connecting the beam flange (eventually) and the cover plate should not exceed twice the thickness of the beam flange nor 100% of the thickness of the column flange. Another delicate issue at present in the design of connections stronger than the beams is the design of the beam web connection to the column flange. As explained in paragraph 3.4., some recent finite element studies of typical connections by GOEL, 1997,and others have suggested that even when the shear tab is welded, shear demands at the connections tend to be resisted by a diagonal tension type behaviour in the web that tends to results in much of the shear being resisted by the flanges. Investigation of these effects is continuing and, at present, design indications have deleted the former requirements for supplemental web welds on bolted shear connections, because these welds could contribute to the potential for shear tab failure at large induced rotations. It is in fact unclear whether it is desirable to develop as much bending strength in the web as possible. It has been observed in some laboratory tests that pre-mature slip of the bolted web

connection can result in large secondary flexural stresses in the beam flanges and the welded joints to the column flange. But if the flange connections fail first, welding of shear tabs to the beam web may promote tearing of the tab weld to the column flange or the tab itself through the bolt holes and total failure of the beam at its support. For that reason; there are also suggestions that fully welded joints for the web should be avoided and that web connections should incorporate horizontally slotted holes to limit the moment which can be developed in the shear tab, thereby protecting its ability to resist gravity loads on the beam in the event flexural connection failure. 4.6. Composite moment frames. Local ductility in composite beam sections can be obtained: - if early concrete crushing is avoided; - if plastic strains are developed either in the rebars of the slab or in the steel sections - in the latter case, if premature buckling of the bottom flange is obviated. There are two options to succeed in fulfilling these conditions: - either define the design conditions achieving the preceding goals; this is done in a recent European research development, ICONS Topic 4, 1999. - or decoupling concrete from steel at the location of potential plastic hinges, so that the problem becomes a better known problem of pure steel section; recent testing have shown that this required total disconnection of the slab in zone around the columns with a diameter of 6 to 8 times the depth of the steel section(same reference). 4.7. Partial strength connections. Partial strength connections offer many positive aspects for seismic design, in particular, the possibility to adapt seismic design resistance to demand, avoiding that dissipative elements, dimensioned for other criteria than seismic resistance, generates general over dimensions of the structure. Partial strength connections should have in the future a wide field of applications supported by standard design methods, especially for structures in which drift is not the dominating design criteria. At present, these is no standard design method and code guidelines for steel structures made with partial strength connections, but research work is active, both on the development of design methods and on the study of behaviour of typical connections, which generally have demonstrated a great potential of local ductility.

5. CONCLUSIONS. Because of the poor local ductility of many steel structures during the Northridge (1994) and Kobe (1995) earthquakes, an extraordinary research push has been made in the end of the 90’s. It has conducted to many new developments concerning requirements on material characteristics, testing requirements and procedure and connection design. The present state of the art summarises the results of this research effort which are known at present. Many questions are still open. One of them is certainly crucial and it concerns the application field of the most classical connection design: the direct welding of a beam to a column by means of butt welds, without any extra stiffening plate on the beam itself and without any weakening of the beam section. At present, this type of connection is not accepted in the USA without an experimental validation in the very conditions of the real structure to be constructed: material and size of the sections, welding material and process. This is an important change and

problem for the constructional practice and economy. For sure it is a first priority problem, which has to be solved by research during the next years, in order to define requirements on the material characteristics, fu/fy ratio, notch toughness, and on the welding process, type of joint, welding technology, welding parameters, in order to allow again a standard use of this classical type of connection.

References. FEMA-267A(1997). Federal Emergency Management Agency. Program to Reduce the Earthquake Hazards of Steel Moment Frames Structures. Interim Guidelines. FEMA Report Distribution Center. PO Box 2012. Jessup, MD 20794.USA. Fax 1 3013625335. FEMA-288 (1997) Federal Emergency Management Agency. Program to Reduce the Earthquake Hazards of Steel Moment Frames Structures. Background Reports : Metallurgy, Fracture Mechanics, Welding, Moment Connections and Frame Systems Behaviour. FEMA Report Distribution Center. PO Box 2012. Jessup, MD 20794. USA. Fax 1 3013625335. EUROCODE 8.(1998).Design Provisions for the Earthquake Resistance of Structures. CEN Comité Européen de Normalisation. Rue de Stassart, 36, B-1050, Brussels .Belgium AISC(American Institute for Steel Construction).(1997). Seismic Provisions for Structural Steel Buildings. ECCS(1988). European Recommendations for Steel Structures in Seismic Zones. Report N°54. ECCS Publications. Brussels. B JSSC (Japanese Society of Steel Construction). 1997. Kobe Earthquake Damage To Steel Moment Connections and Suggested Improvements. Technical Report N° 39. Mazzolani F.M. and Piluso V.(1996). Theory and Design of Seismic Resistant Steel Frames. E&FN Spon, London, UK Bruneau M., Uang C.M. and Whittaker A.(1998).Ductile Design of Steel Structures, McGraw-Hill, USA. Paulay T. and Priestley M.J.N(1992). Seismic Design of Reinforced Concrete and Masonry Buildings. John Wiley and Sons. USA. Plumier A. and Sanchez L.(1999). Technical Report N°3 .RECOS Research. University of Liege Civil Engineering Department. Plumier A., Agatino R., Castellani A., Castiglioni C.A. and Chesi C.(1998).Resistance of Steel Connections to Low Cycle Fatigue. Proceedings of the Eleventh European Conference on Earthquake Engineering. Balkema. NL Goel S.C., Stojadinovic B. and Lee K.(1997).A New Theory for Moment Connections. Proceedings of STESSA 97 Conference. Kyoto.E10/17.I

Krawinkler H. and Gupta A.(1997). Deformation and Ductility Demands in Steel Moment Frame Structures. SDSS ’97 Stability and Ductility of Steel Structures Conference, Nagoya, 1997, Japan Society of Steel Construction, J Tremblay R, Tchebotarev N. and Filiatrault A. Seismic (1997). Performance of RBS connections for steel moment resisting frames: influence of loading rate and floor slab. STESSA 1997, Kyoto.E10/17.I Plumier A.(1997). The Dogbone-Back to the Future. AISC Engineering Journal 2/1997 Vol 34 N°2, Sotirov P., Rangelov N., Ganchev O., Georgiev T., Milev J and Petkov, 1999. Improved Beam to Column Joints for Moment Resisting Frames-An Experimental Study. RECOS Progress Report. University of Sofia. BG. Forrel and Elsesser Engineers Inc, 1995.Lawrence Berkeley National Labs Steel Joint TestTechnical Brief. San Francisco. USA ICONS Topic 4-Draft of Specific Rules for Steel Concrete Composite Buildings. 1999.University of Liege Civil Engineering Department Report. B