EXPERIMENTAL ANALYSIS OF STEEL DISSIPATIVE BRACING SYSTEMS FOR SEISMIC UPGRADING

. Energy dissipating devices, such as metallic ductile dampers, could represent one reliable system for seismic performance upgrading of reinforced concrete (RC) structures. This paper illustrates the significant improvement to the seismic response of RC structures equipped with dissipative bracing systems, such as eccentric braces (EBs) and buckling restrained braces (BRBs). In fact, the results of experimental tests carried out on two similar two-storey one-bay RC structures, respectively equipped with EBs and BRBs, are described. Referring to EBs, 3 lateral loading tests have been performed. Each test is characterized by shear links with bolted end-plate connections. Different design criteria have been applied in the design of the connections. In the first test, capacity design criteria have not been considered. In the second test, a capacity design criterion has been applied, with a link shear over-strength factor equal to 1.5. In the third test, a design criterion similar to the one adopted for the second test has been implemented, but with a larger over-strength factor. In case of BRBs, two types of ‘only-steel’ braces have been tested: one type was made using two buckling-restraining rectangular tubes that are fully welded together with steel plates; the other type is detachable, being made again with two buckling-restraining rectangular tubes but joined together by means of bolted steel connections. In both cases, the internal yielding core was a rectangular steel plate. The experimental results of both bracing systems are encouraging about the possibility to use these devices for improving the seismic resistance of existing RC structures.


Introduction
Existing reinforced concrete (RC) frame buildings with non-ductile detailing represent a considerable hazard during earthquakes. This type of buildings suffered severe damages and were responsible for most of the loss of life during the major Italian seismic events such as the 1981 Irpinia earthquake. Improving the seismic response of this type of construction can be considered as one of the main concern for structural engineers. The work presented herein gets in this contest.
Among seismic performance upgrading methods, there are several options normally available, one of which is to employ energy dissipation devices, such as friction, viscoelastic and metallic dampers, etc.. Energy input by a strong earthquake is expected to be greatly dissipated by these devices, and if they are damaged they make the rehabilitation easy after the earthquake, since these devices are designed to be replaceable. In particular, this paper focuses its attention on removable steel eccentric braces and on buckling restrained braces as a seismic upgrading approach for protecting RC buildings from severe earthquake damage. In detail, the results of several lateral loading tests, up to collapse, performed on existing RC structures seismically retrofitted by steel eccentric bracings (EBs) and buckling restrained braces (BRBs) are presented and discussed. Tests have been carried out within the context of a wider experimental research activity, named the ILVA-IDEM project, having the purpose to evaluate several innovative technologies for the seismic retrofitting/upgrading of existing RC structures, such as ductile steel bracing, ductile shear panels, base isolation and use of C-FRP as external RC reinforcement. The experimental part of this study started from the exceptional opportunity to carry out experimental tests in inelastic range of response on a real RC building ( Figure  1a). The building, which is located in Bagnoli (Naples, Italy), was destined to demolition by competent Authority, within the dismantling process of the Italian ex-steel mill ILVA (or Italsider). In order to increase the potential number of specimens for testing different upgrading solutions, slabs were cut at the first and second floor, in such a way to divide the whole building into six separate structures to be analyzed, as shown in Figure 1b. Before cutting the slabs, external and partition walls, as well as non-structural elements, were removed in order to get bare RC structures. A deep and detailed description of the whole experimental activity can be found in [1,2].

Description of the RC structure and the test setup
The RC sub-structure is essentially constituted by four columns sustaining two floors. Columns have a square 300mm x 300mm cross-section. The structure of the two floors can be essentially described as made of Tsection beams going parallel in the transverse direction and supported by two longitudinal L-section beams. Column longitudinal steel rebars are in number of four, placed at the section corners and have a diameter of 12mm. Transverse stirrups have a diameter of 8mm and are spaced of about 200mm. Figure 2 shows the longitudinal section of the generic RC sub-structure.
The mechanical properties of materials were estimated both in-situ and in the laboratory, by means of tests on specimens, such as steel bars and concrete cores taken from the structure.  . In particular, this vertical steel beam (Figure 3c) was used for distributing the applied lateral force between the two stories of the structure to be tested. This arrangement reproduces an inverted triangular lateral load pattern which is often assumed in theoretical pushover studies. The strengthened structure was subjected to a cyclic loading history up to the development of a clear collapse mechanism.  During the test, floor displacements have been measured, by using a video-camera installed above each floor for measuring the floor lateral displacements ( Figure 4). This displacement-measuring device proved to give lateral displacements with the same precision of a topographic total station, which was used in a former test on a similar structure.

Description of the experimental tests on the RC structure equipped with eccentric braces
As it is well known, in case of EBs forces are transferred to the brace members through bending and shear forces developed in the ductile steel link. The link is designed to yield and dissipate energy while preventing buckling of the brace members. In case of RC frames, the concrete beams are incapable to perform as a ductile link for the steel bracing system that is inserted in the frame bays. In the light of this, it is impossible to adopt for RC frames the common inverted k-brace configuration (typically used in steel frames). Hence, the need to adopt a Y-inverted bracing configuration, with a vertical steel link, can be easily recognized. Besides, bolted connections at the link ends are required, what could have the advantage to permit replacement of the dissipative members (links) after a damaging earthquake.
As a consequence, this bracing system has been applied to the seismic retrofitting of one RC sub-structure, as mentioned above. The geometry of this bracing system is summarized in Figure 5. a) RC structure plan. b) RC structure elevation. Three experimental pushover tests have been carried out [1,2] on this structural unit. Each test showed different structural performance, because of different design criteria adopted for the design of link end-connections.

1. Test No.1 on EBs: general design data and experimental response
The first bracing system was designed neglecting capacity design criteria. The link and its connections to the RC slab and to the diagonal braces are shown in Figure 6. The normalised link length, which is defined as the ratio of the actual link length (0.25 m) to the limit value e s , was equal to e/e s = 0.92.

2. Test No.2 on EBs: general design data and experimental response
For the second test, link end-connections were strengthened using capacity design. Namely an ultimate shear strength of links equal to 1.5 times their yielding strength was assumed [3]. This implied that the end-plate thickness increased from 10 to 25mm.
According to [4,5], the ultimate bending moments transferred by links can be conservatively evaluated through static balance. Link end connections have been designed to resist these flexural actions. The normalized link length was equal to e/e s = 0.81.
The second test showed shear failure of bolts. In fact, as shown in Figure 9, the global response curves stop suddenly at a base shear value corresponding to the brittle failure of link-to-brace connections. Figure 10 shows that the plastic bending of end-plate connections was now completely avoided, while a moderate plastic engagement of links along with a strong plastic deformation concentrated as shear hinging of bolts at the link-to-brace joints was observed.

Test No.3 on EBs: general design data and experimental response
The third link was designed in order to increase the system ductility by forcing plastic deformation to be confined within links. In fact, the previous test revealed link over-strength larger than that expected; hence, as shown above, the brittle shear failure of bolts occurred. Because it was impossible to modify the geometry of link-to-brace joints, a steel built-up section (Figure 11) was now designed for the links, in order to have shear strength of connections at least 2 times larger than the average yielding strength of links. This led to the need of increasing bolt steel grade (compare Figure 6 and Figure  11). The flexural resistance of the built-up section was chosen to be similar to the one of HEA100 (which is the section used for test No.1 and No.2), while the shear resistance (consequently the web area) was lesser than in the previous cases. The normalized link length was equal to e/e s = 0.30.  Collapse was due again to the brittle shear failure of bolts of link-to-brace joints. However, the response curves ( Figure 12) show that the behaviour of the retrofitted structure was now characterized by larger energy dissipation capacity. As in test No.2, plastic bending of connection endplates is now avoided, while significant plastic shear deformation of links was observed ( Figure 13).

Description of the experimental tests on the RC structure equipped with buckling restrained braces
BRBs are special devices that solve the problem of the limited ductility of classic concentric bracing. In fact, the axial strength is de-coupled from flexural buckling resistance; the axial load is confined to the steel core (thus providing complete truss behaviour), while the buckling restraining mechanism resists overall brace buckling and restrains high-mode steel core buckling (rippling).
Several studies proposed a large number of different types of BRBs, but all of them based on the basic concept to use tubes for restraining lateral displacements while allowing axial deformations of the core. In the most classical form, the restraining tube is filled with concrete and an unbonding layer is placed at the contact surface between the core plates and the filling concrete, thus this version is called 'unbonded brace'.
'Only-steel' solutions have been also proposed, with two or more steel tubes in direct contact with the yielding steel plates. Contrary to the "unbonded", this type of BRBs can be designed to be detachable. This aspect implies that is possible to design these systems to be inspected, so that it is possible to control their condition after each seismic event and to allow an ordinary maintenance during the life-time. To do this the restraining tubes should be connected by bolted steel connections [6,7]. Moreover an 'only-steel' BRB is lighter than an 'unbonded' one; this implies a technical and economical advantage during the assembling.
These considerations led to study a special only-steel detachable BRB to be used for improving the seismic response of RC buildings. In particular two different types of this special device have been applied on one of the RC unit (see Section 2).
The diagonal braces were directed in alternate way, in order to evaluate the response of the studied braces in tension and in compression. In particular, the location of these braces is shown in Figure 14a-b.   Fig. 14. Configuration of the tested BRBs.
Two experimental pushover tests have been carried out [1] on this structural unit. Each BRB system has been tested under lateral cyclic loads as in the case of EBs. Figures 15 and 16 give the fundamental geometric properties of the first type of BRB tested. The yielding steel core is a rectangular plate (25mmx10mm), made of European S275 steel. The actual average yield stress of the core was measured to be 319MPa (i.e. 1.16 times the characteristic value).  The buckling-restraining action is given by two rectangular steel tubes (100x50x5), with a ratio between the Euler buckling load (P E ) of the two tubes and the actual yield force (P y ) of the internal steel core P E /P y = 2.1.

1. Test No.1 on BRBs: general design data and experimental response
The experimental evidence showed a good response of the brace when it is in tension, with the expected relative displacements developing between the internal yielding core and the restraining tubes (see figure 17a). But, the compression brace ductility was limited by the local buckling of the core, near the brace ends. This buckling produced strong flexural deformation of the closing plates, which were welded for joining the tubes at their ends (as shown in Figures 17b and 17c). This localization of damage ultimately led to the premature core fracture owing to the strong plastic strain developed at the transition section between the reduced core and the end tapering. Figure 17d illustrates plastic hinging at RC column end.

Fig. 17.
Damage pattern for the first BRB system.
The measured base shear vs. first and second story lateral displacement relationships are plotted in Figure  18a,b. At each floor two measures of lateral displacement were taken, approximately symmetric with respect to the loading axis. As it can be seen, the difference between the two displacements at each floor (hence the floor rotation) is small, with a maximum of about 15% of the average displacement in the inelastic range. The maximum firststory drift was 1.9% of the first-story height.
The global story ductility (µ) reached a maximum of about 4.75. In fact, the yielding value of the first story drift angle (which corresponded to yielding of BRBs at first story) was equal to about 0.004rad, hence µ = 0.019/0.004 = 4.75.

2. Test No.2 on BRBs: general design data and experimental response
Figures 7 and 8 illustrate the geometry of the second type of tested BRB. Two main changes were made with respect to test No. 1. First, the BRB internal core was now tapered in a more gradual manner (Figure 7). Second, the two restraining tubes were now joined together by means of bolted stiffened elements (Figure 8), allowing the BRB to be opened for inspection and monitoring at the end of the test    Figure 21a highlights the relative displacement between the internal core and the restraining tubes developed when the BRB was in tension. Figure 21b illustrates the local buckling failure of one end-plate during compression of one BRB at the first story. Figure 21c shows the local buckling of the internal core of one BRB placed at the first story: this phenomenon became very apparent at the large first-story drift reached during the test (5.6% of the story height). Finally, Figure 21d shows large flexural cracking occurring in RC columns at the first story in correspondence of the peak values of story drift.

Fig. 21.
Damage pattern for the second BRB system. Figure 22a,b shows the base shear vs. first-story and top-story lateral displacements. Also in this second test, difference between the tension and compression behaviour of the BRBs was within the expected range of behaviour, originating relatively small torsion of floors. Obviously, this difference became larger when local buckling of one end-plate affected the compression response of one BRB at the first story. Notwithstanding some localization of local buckling at one end of the internal steel core of the compressed BRB, the global story ductility (µ) was quite large, reaching a maximum of about 14. This value can be computed assuming the yielding value of the first story drift angle conventionally equal to 0.004rad, which corresponds approximately to a base shear equal to 40% of its maximum value, hence µ = 0.056/0.004 = 14.

Comparison of test results
Both the bracing system presented in the previous Sections demonstrated to be a reliable solution to improve the seismic performance of existing RC structure. In Figure 23 the lateral-load response of the all tests on the two tested bracing systems (EBs and BRBs) is compared in terms of envelope curve corresponding to the positive loading direction. Besides, the behaviour is also compared with the results of a previous pushover test, which was carried out on a bare RC structure very similar to the one tested with the bracing systems.
All tests showed a significant increase of lateral stiffness and strength respect of the one of the original unbraced RC structure. In particular, in case of EBs it was observed an increase of the lateral capacity from 5.65 to 8.34 times respect to the capacity of the original unbraced RC structure, while in case of BRBs from 4.08 to 4.95 times. The main cause of the larger values of the lateral strength achieved by EBs can be found in the shear over-strength exhibited by the tested steel links [8]. In fact, the maximum shear developed in links during test can be approximately estimated by taking one half of the value of the measured peak base shear, the latter reduced of 50kN, which is the value of the base shear force obtained from the capacity curve of the bare RC structure, a roof displacement equal to the displacement measured during the test in correspondence of the peak strength. The maximum peak shear force, corresponding to this computation, is equal to 5.42 times V y,link . Test results, briefly summarized here, highlight that links made with the European wide-flange hot-rolled profiles exhibit over-strength largely in excess of that implicit in modern design codes [9].
BRBs are characterized by lower over-strength capacity (at the most equal to the material axial overstrength), but they can provide for the structure a larger displacement capacity than EBs. In fact, referring to the studied cases, short shear links should develop shear deformation angles larger than 0.60 rad in order to provide the same displacement capacity of the tested BRB type-2. This large shear deformation is not reasonable, since no shear link is able to provide it. This implies that BRBs let to control stiffness, strength and ductility better than EBs. Comparison of response curves of tested bracing systems.

Conclusion
Among the possible solutions to retrofit an existing structure, bracing systems are a simple and effective retrofit system, especially when story drifts need to be limited. The idea is to design systems that are strong enough to resist the seismic forces and light enough to keep the existing structural elements far from needing further reinforcement. Furthermore, if these systems could be installed quickly and eliminate the need to disrupt the occupants of existing structures, they would be even more desirable (in the context of a hospital retrofit for example). In this sense both EBs and BRBs may be a viable solution for seismic retrofitting of RC structures. In fact, they provide high elastic stiffness, stable inelastic response and excellent ductility and energy dissipation capacity.
In case of eccentric braces, link end-connections exhibit a key role in determining the system ductility, especially if bolted connections are selected for removable links. Experimental test results clearly highlight this aspect, emphasizing large over-strength of short links with respect to the first yielding shear and the consequent danger of connection failure. In particular, it has been shown that short links, made with the European hot-rolled steel profiles, may exhibit over-strength largely in excess of that suggested by current code provisions. Namely, the ratio between the link ultimate shear strength and its yielding strength can be significantly larger than 1.5.
Based on some preliminary and approximate backanalysis, a value of about 5 has been estimated for the over-strength factor exhibited by the tested links.
Respect to EBs, BRBs revealed to provide a more complete structural performance, since they can improve not only the lateral stiffness and strength capacity but also the displacement capacity of the structure. In fact, test results on two different types of "only steel" BRBs showed good ductility of this system.
The efficiency of the first type of BRB (test No. 1) was impaired by the flexural failure of the end closing plates, which were unable to restrain the end portion of the brace core. This produced a strong flexural plastic engagement of the core at its ending portion. Hence, ductility of the system was quite limited, even if the strength and stiffness of the upgraded RC structure met the expected improvement.
The second type of BRB (test No. 2) showed instead large ductility, being able to adequately restrain the core from buckling, though some additional improvements are required in the design. In fact, the combined effect of sliding friction between the core and the restraining tubes and the small thickness (hence flexural rigidity) of the tube walls produced some localization of damage in the yielding core, hence some limitation in the expected system ductility. In addition, local buckling of one end plate was observed during this second test, even if the end plate satisfied a capacity design criterion. In fact, the local buckling Euler load (conservatively estimated by the assumption of beam-type behaviour and assuming a buckling length equal to the distance between the bolt centre-line and the starting section of the restraining tubes) can be computed to be 3.8 times larger than the maximum expected compression load. Hence, buckling of the end plate (occurred at just one location) may be attributed to: (i) strong local geometric imperfections, which grew up during cyclic loading; (ii) a flexible endrestraint, which produced an increase of the buckling length and also some coupling of lateral and torsion effects. Anyway, the maximum story-drift angle reached during test No. 2 (5.6% of the story height) is appreciably larger than the maximum values commonly applied in the past testing of BRBs. Therefore, results can be considered satisfactory and encouraging for the future