Concrete Repair, Rehabilitation and Retrofitting II – Alexander et al (eds) © 2009 Taylor & Francis Group, London, ISBN 978-0-415


3D analysis of seismic response of RC beam-column exterior joints before and after retrofit
R. Eligehausen, G. Genesio & J. Ožbolt
Universität Stuttgart, Stuttgart, Germany

S. Pampanin
University of Canterbury, Christchurch, New Zealand

ABSTRACT: This paper presents an application of microplane based Finite Element (FE) approach for threedimensional (3D) modeling of reinforced concrete beam-column joints under cyclic loading. Experimental tests have shown that the structural behavior of poorly detailed joints is decisive for the structural response of older frame building. Due to inadequate reinforcement ratio in the joint, poor bond properties of longitudinal reinforcement (plain round bars) and deficiencies in the anchorage of reinforcement (bars with end-hooks) a brittle failure mechanism is expected. Tests on poorly detailed beam-column exterior joints, primarily designed for gravity loads only, were conducted at the University of Canterbury (UC) in order to evaluate their seismic response. The numerical analyses presented in the paper are performed with the FE Code MASA, developed at the Universität Stuttgart and capable of 3D nonlinear analysis of quasi-brittle materials, like concrete, based on a microplane material model with relaxed kinematic constraint. In this contribution, the experimental results are used for calibration of the model with particular focus on joint strength, failure mode mechanisms and strength and stiffness degradation. The influence of different parameters is investigated with the intention to quantify their influence on the performance of the joint. In the final part of the paper a retrofit solution, proposed and experimentally evaluated at the UC, is also numerically simulated and discussed.



Reinforced concrete frame buildings designed before the introduction of modern seismic oriented codes in early 1970s, offer an inadequate response to lateral loads typical of seismic events. In this work the attention is focused on the behavior of exterior beamcolumn joint, since it is recognized that they are the most vulnerable part of moment resisting RC frames, due to their lack of a reliable joint shear transfer mechanism. This poor behavior is mainly due to: inadequate reinforcement detailing (lack of transverse reinforcement in the joint region), poor bond properties of the reinforcement (plain round bars) and deficiencies in the anchorage details (bars with end-hooks). The experimental results of quasi static tests on poorly detailed beam-column joints performed at UC are used for development and refinement of the finite element code MASA, developed at the Universität Stuttgart and capable of three-dimensional (3D) nonlinear analysis of concrete-like materials and reinforced concrete structures. The program is based on the microplane model with relaxed kinematic constraint (Ožbolt et al. 2001).

Following a brief description of the investigated test specimens, the microplane constitutive law for concrete and the other features of the employed FE model are described. The results of the numerical analysis are compared with the experimental results. Additionally, a study was carried out to investigate the influence of several parameters including the variation of axial load in the column and the influence of the modeling of the bond between concrete and reinforcement steel. Finally, the numerical simulation of the reversed cyclic behavior of the connection is compared with the experimental results. In the last part of the paper the simulation of a retrofit solution tested at UC is also presented. 2 EXPERIMENTAL TESTS

The experimental tests considered for the validation of the FE model include three exterior beam-column joints with plain round and deformed bars, before and after retrofit. Fig. 1 shows the test specimens considered for the numerical simulation. The specimen TDD1 (Fig. 1a) is characterized by deformed bars


Failure mechanism 1 2 3 1 2 3 1 2 3 Beam yielding Joint diagonal cracking Column yielding Joint diagonal cracking Beam yielding Column yielding Beam yielding Joint diagonal cracking Column yielding TDP2 In the FE code used in the analysis. ±4. as longitudinal reinforcement of the beam.8). the elements were taken as linear elastic (darker element in Fig. ±3. as shown in Table 2. ±2. ±0. Hexahedral elements with side length of approximately 25 mm were adopted for the mesh in the numerical simulations (Fig. To prevent a local failure of the concrete elements in the vicinity of the supports and at the point of application of axial and lateral loads. i. ±1. 3 Figure 1. The lateral load is considered as positive when acts opening the joint (increase of axial load in the column) and negative when closing it (decrease of axial load).0%. Geometry and reinforcement of the test specimens: a) Specimen TDD1 with deformed bars bent in the joint.2. The bond between longitudinal reinforcement and concrete was simulated using discrete bond elements.5%. Material properties of the test specimens. only one-half of the specimens was modeled. and in the specimen TDP2 (Fig. For transverse reinforcement. 2a). ±2. ±0. Specimen TDD1 Hierarchy of strength of the test specimens. a rigid connection between steel and concrete was assumed. 1983) is employed. smeared crack approach was used. More information about the experimental tests is available in Chen (2006). The analysis was performed in the framework of continuum mechanics. b) Specimen TDP2 and THR3 with smooth round bars and end-hooks (Chen 2006). 1b) plain round bars with end-hooks were employed. This assumption neglects the influence of the relative displacement between stirrups and concrete.0%. The discrete bond model was calibrated as explained in Section 3.7 Table 2.9 D10 324 457 191 TDP2 R6 408 482 191 R10 333 467 206 THR3 R6 352 436 224 26. i.0%. In the analyses. The retrofitted specimen THR3 had the same detailing of TDP2.e. The main material properties are summarized in Table 1. The specimens were expected to have different hierarchy of strength. In order to obtain results which are independent from the element size. FE Model (TDP2): a) Discreti-zation of concrete: b) Discretization of reinforcement (TDP2).e. 1142 .8 R10 347 474 219 Test Specimen Steel Bar size fy [MPa] fu [MPa] Es [GPa] fcc [MPa] Ec [GPa] Concrete 25.1%.0%. crack band approach (Bažant et al.5%. FINITE ELEMENT MODEL Table 1. The specimens were loaded under reversed cyclic lateral loading. where α = 1. 2).5%. THR3 bent in the joint.0 28. the microplane material model for concrete and 1D three-linear constitutive law for reinforcement steel were used. In the numerical simulation the material properties of concrete and reinforcement steel were chosen according to values presented in Table 1. To reproduce the asymmetric effects during an actual cyclic push-pull test on the prototype frame system. ±1.was varied during the experiments as a function of the lateral load (Nc = 75 kN ± αVc.2%. the axial load in the column a) b) Figure 2. The imposed loading history consisted of a series of two cycles at increasing top drift level of ±0. the vertical symmetry of the specimen was utilized. TDD1 R6 424 495 177 22.

More detail related to the used model can be found in Ožbolt et al. 1b) was calibrated on the pullout test performed by Fabbrocino et al. Basic assumptions of the bond model implemented in MASA (Lettow 2006).0 6. Bond parameters used in the numerical simulation (parameters defined in Fig. which depends on the state of stresses and strains in concrete and reinforcement. 5 the assumption for the bond-slip cyclic relationship for deformed bars is shown. 3) with a bond-slip relationship.5 8 0. Bond degradation is assumed to occur Figure 5.1 Microplane model τ-s-Relationship: CyclicLoading MonotonicLoading τ 1 The microplane model is a 3D macroscopic model in which the material is characterized by uniaxial relations between the stress and strain components on planes of various orientations called “microplanes”. after a certain slip due to the mechanical damage of the concrete-steel interface produced by the ribs of the reinforcement bars. Figure 6. Lettow 2006). More detail related to the bond model can be found in Lettow 2006. Figure 3.3. The bond-slip behavior of plain round bars of the test specimen TDP2 (Fig. Bond-slip monotonic relationship for plain round bars (Fabbrocino et al. 2002).1 1. 2001. kunload τf τm [MPa] [MPa] [N/mm] 0. In Fig. At each finite element integration point the microplanes can be imagined to represent damage planes or weak planes of the microstructure of the material. Bond-slip monotonic relation-ship for deformed bars (Lettow 2006). 1983). The macroscopic response is obtained by integrating contributions of all microplanes.0 Figure 4. 4). on type of loading and on geometry (Fig. Bar type R10 D10 k1. More detail can be found in Eligehausen et al. 6). (1983).5 1. 4). Bond-slip cyclic relationship for deformed bars (Eligehausen et al. The bond stress is in this case not the sum of a mechanical and a frictional components (see Fig. Table 3.5 4 100 20 ksec ss s3 [N/mm] [mm] [mm] 80 12 0. 2002 (Fig. 4). but only 1143 . 1983.2 Bond model 3 5 7 9 8 6 4 s 2 The discrete bond model implemented in MASA consists of a 1D nonlinear springs (Fig. It has been demonstrated that the model is able to correctly predict bond behavior of deformed steel bars for monotonic and cyclic loading (Eligehausen et al. 3.

The results of the monotonic analyses. The values of k. b2) End of the experimental test TDP2. 2006) was also numerically simulated. Fig. D10 (TDD1) according to pullout tests available in literature (Lettow 2006 and Fabbrocino et al. Furthermore. 10 (TDD1).175 with a top drift of 0. The cyclic behavior was assumed similar to the one adopted for deformed bars.73% and 0.2 and 0.1 Simulation of different failure modes As pointed out by several experimental studies in literature (e. are in a range between 0.of a frictional contributes and an initial adhesion. The localization of the damage in the concrete (cracking) is represented in a smeared way (dark zones in Figure 7. A retrofit solution developed in order to increase the structural performance of the test specimen TDP2 (Pampanin et al. performed varying the axial load coefficient (α) on the test specimen TDP2. the variation of axial load in the column influences the shear capacity of the joint panel. the ultimate load was reached not when the first cracks occurred.195 and 0. 1144 .2 * f ′c0. in contrary to what was observed in the numerical analyses. 10 (TDP2) and Fig. R10 (TDP2 and THR3) and deformed bars.g. In the comparison between experimental and numerical results it has to be considered that the monotonic curves obtained in the numerical simulation are compared with the envelope of experimental cyclic tests. pt). A typical flexural failure mechanism is shown in Fig.a2 and shear failure type of the joint panel is shown in Fig. a2) End of the experimental test TDD1. The cracking patterns and associated failure mechanism shown in Fig. 8 and Table 4). Furthermore. obtained varying the axial load.23.5. Pampanin et al. but at a higher drift level. b1) Shear cracks in the joint panel in the FE analysis of TDP2.t = − f a ± ⎜ a ⎟ + v 2 jn ⎝ 2⎠ 2 (1) where: fa = Axial compression stress acting in the column. No experimental data are available to evaluate the monotonic behavior of this exterior beam-column joint. 2003). Predicted and in experiments observed failure modes for: a1) FE analysis of TDD1 with formation of flexural hinge at the beam-column in-terface. calculated according to the Mohr’s circle theory. takes into account the influence of axial load in the column (Priestley 1997): ⎛ f ⎞ pc . 7a1. The predictions of the cyclic behavior of the beam-column joints TDP2 and TDD1 are compared with the hysteretic behavior and cracking patterns observed in the experimental tests. The upper limit for the formation of the first shear crack in the joint is given by: pt = k ⋅ f c′ (2) Central problem in the modeling of beam-column joints is the capability to reproduce properly the failure mechanisms occurred in the experimental tests. In the experimental test of the specimen TDP2 the factor k was equal to 0. Pampanin et al.b2. 4. The capability of MASA to predict different failure modes of beam-column joints is presented. 2002). 7). In the case of longitudinal plain round bars anchored with hooks in the joint. 4. seem to confirm the above assumption (Fig. respectively. 7 refer to the hysteretic behavior of that is shown in Fig. 7b1. (2003) proposed an upper limit for the formation of shear crack in the joint panel as 0. where: f ′c = cylindrical concrete compressive strength. vjn = Nominal shear stress in the joint. The formulation of the joint shear capacity in terms of principle tensile and compression stresses (pc.46% for the load direction with increase and decrease of axial load.2 Influence of axial load in the column 4 NUMERICAL ANALYSES In the following sections the main results of an extensive parametric study are presented. In Table 3 are given bond parameters for plain round bars. the influence of axial load in the column on the ultimate resistance of the joint and the importance of correct modeling of the bond-slip behavior of longitudinal bars are discussed.

41 21. the use of discrete bond model leads to a similar failure load. 1145 .6 Vc due to the stress that comes from the beam bars into the panel zone.8*Vc 10 20 30 40 50 60 70 80 20 15 10 5 0 0 0.Lateral displacement [mm] 0 30 SpecimenTDP2 25 N .6*Vc Exp.21 1.5 1 1.2 Vc Nc.85 0. Comparison of numerical results with and without discrete bond model. 9 the load-displacement curves from monotonic loading show that. On the other hand using a perfect bond assumption between steel and concrete the behavior of the joint is characterized by a ductile failure with the formation of a flexural hinge in the beam (see Fig. In Fig.0—1.20 0.6 Vc Nc.75 k [-] 0. Hysteretic behavior of test specimen TDD1: comparison between experimental and numerical results.8*Vc N + 1. 7b1). which occur in the analysis with discrete bond leads to a loss of resistance of the beam-column connection after a top drift of approximately 1%. Variation of axial load in the column in the test specimen TDP2 (the peak values correspond to first shear cracking in the joint panel).2 53.0 + 1.6 118. A strength degradation of the beam-column connection.0 + 3.2 132.3 Influence of bond modeling -20 -25 -4 -3 -2 -1 0 1 Numerical Experimental 2 3 4 The simulation of bond between concrete and reinforcement plays a central role in the modeling of beam-column connections. with a difference of approximately 10% of the ultimate capacity of the joint.0 + 5.8*Vc N + 3.22 0.5 4 Top drift [%] Figure 8.6*Vc N . characterized by a shear failure of the joint was obtained. In the experimental test Lateral displacement [mm] Perfect bond Discrete bond -80 25 20 -60 -40 -20 0 20 40 60 80 Specimen TDD1 Lateral force [kN] 15 10 5 0 -5 -10 -15 Top drift [%] Figure 9. A complex interaction between flexural response of the adjacent beam element and the joint shear transfer mechanism occurs Top drift [%] Figure 10.04 0.0% top drift in the numerical simulation.1.8 Vc Nc. Analysis of the results obtained varying the axial load in the column and determination of the factor k.3. However the ultimate load and displacement capacity are higher in the analysis without using the discrete bond model.N . 7a1).35 Nc [kN] 190.N + 1.22 0. In the case of monotonic loading in negative direction (decrease of axial force in the column) the same failure mode occurred regardless the bond modeling adopted. 10.97 22. The shear cracking of the joint (see Fig.0—3. However.4 42. In both analyses a failure mechanism. a different post-peak behavior is observed.1.5 2 2.84 1.52 16.23 0. due mainly to slippage and buckling of the longitudinal bars of the beam becomes relevant after 2.0%. 4. The specimen TDP2 (with plain round bars) was considered for this investigation. 4.6 Top drift [%] 1.8 Vc Nc. Table 4. as when using a perfect rigid bond between steel and concrete. in the case of positive loading (increment of axial force in the column). The comparison between experimental results and numerical analysis shows an acceptable correspondence of the behavior until a top drift of 2.22 Load Type Nc.8*Vc Exp.21 17. Vc [kN] 23.4 Cyclic analyses Lateral force [kN] Lateral displacement [mm] -80 25 20 15 10 5 0 -5 -10 -15 -20 -25 -4 -3 -2 -1 0 1 2 3 4 -60 -40 -20 0 20 40 60 80 Specimen TDP2 Lateral force [kN] The hysteretic behavior of the test specimen TDD1 is shown in Fig.5 3 3.

c) Cracking patter with Kd = 50. 7a2).5 2 Kd = 100. 4. the cracking pattern obtained in the numeric simulation (Fig. 13). strength degradation.b the cracking patterns indicate the formation of a flexural hinge in the beam. Figure 12. 11 the comparison between the experimental test and the numerical simulation of the test specimen TDP2 is shown. 12 shows the implementation of the metallic haunch and the corresponding flexural hinge. Fig. due to the same causes. (2006) the stiffness of the connection and the slip between the metallic diagonal plays a central role in the efficiency of this retrofit solution. while in Fig. slip 1 mm Exp. 2006). based on a diagonal metallic haunch. the simulation was able to reproduce with sufficient accuracy the main features of the cyclic behavior of the joint: (i) relevant strength degradation after 2. In the design model presented in Pampanin et al. Lateral displacement [mm] -80 30 25 -60 -40 -20 0 20 40 60 80 In order to modify the internal hierarchy of strength of the beam-column joint TDP2 (see Table 2) and to avoid the shear failure of the connection in the joint panel a retrofit solution. was developed at the UC (Chen 2006.000 kN/m and no slip. Evaluation of the influence of the stiffness (Kd) and slippage of the haunch connection: a) Cracking pattern with Kd = 100. The resulting cracking patterns are presented in Fig.0% top drift and (ii) increasing pinching at higher drift levels. envelope 2. a shear crack in the joint panel Lateral force [kN] 20 15 SpecimenTHR3 10 5 0 0 0. 1146 .000 kN/m and no slip.5 Analyses of retrofitted specimen a) b) c) Figure 13.000 kN/m.5 3 3. was observed during the 3.5 4 Top drift [%] Figure 14. 13a. 7b1. Pampanin et al. Consistently to the experimental behavior (Fig.000 kN/m Kd = 50. Comparison of monotonic behavior of the test specimen THR3 and the results of the numerical simulation using the parameters of Fig. Even if the ultimate capacity of the beam-column joint in the positive direction was overestimated by the numerical analysis (approximately 20%).Lateral displacement [mm] -80 25 20 Specimen TDP2 -60 -40 -20 0 20 40 60 80 Lateral force [kN] 15 10 5 0 -5 -10 -15 -20 -25 -4 -3 -2 -1 0 1 2 3 4 Numerical Experimental Top drift [%] Figure 11.0% drift cycle.5 1 1. In Fig. b) Cracking pattern with Kd = 50. Hysteretic behavior of test specimen TDP2: comparison between experimental and numerical results. 13c. In Fig. Haunch retrofit solution (Chen 2006).b2. 13. Numerical analyses well agree with the experimental results confirming the reliability of the design approach and the experimental observations (Fig.000 kN/m and 1 mm slip. due to the effect of the slippage.000 kN/m Kd = 50. 7a1) shows the formation of a flexural mechanism.

Z. Experimental behaviour straight and hooked smooth bars anchorages in existing r. G. 2003. B.. Lettow. The influence of the variation of axial load.. Ožbolt. IWB Universität Stuttgart. due the asymmetric effects during an actual cyclic push-pull test on the prototype frame system. The analyses were able to show the formation of the shear crack in the joint. 5 CONCLUSIONS The 3D numerical analyses of poor-detailed exterior beam-column joints under monotonic and cyclic lateral loading have been presented. UCB/EERC 83. Popov.c.. 2006. G. Chen. Pampanin.. Journal of Earthquake Engineering.23. Germany (in German).. 2006. 13a correspond to the experimental test THR3.J.P. & Chen. before and after retrofit. 12th European Conference on Earthquake Engineering. 1 No. Fabbrocino. 2683–2711. Elsevier Science Ldt. Christopoulus. Y. Dissertation. Vol. A refinement of the mesh in the joint panel could possibly improve the performance of the numerical model.. 13. & Bertero V. plain round bars. observed in the experimental tests. 1157–192. The envelope of the cyclic behavior of the retrofitted beam-column joint is compared in Fig. 14 with the monotonic analyses of the joint for the three cases presented in Fig. Carr A. G. buildings. the FE simulation overestimates the initial stiffness and strength capacity of the joint. & Kožar I. Eligehausen. when perfect bond is assumed. Master thesis. C. New Zealand.. fib International Symposium 2003.M.. The parameters of Fig. Megenes & G. J. 2002. Li. International Journal of Solid and Structures. S. Adopting a realistic bond model the cyclic behavior of the test specimens was simulated with sufficient accuracy regardless of the use of deformed or 1147 .governs the behavior of the connection. 2001. & Manfredi.-H.-H. In both cases more research work is required to improve the numerical determination of the ultimate resistance of the beam-column connection. T.V 1983. Crack band theory for fracture of concrete. R. A Variation of the bond assumption significantly influences the prediction of failure mechanism of the joint. 2001. Ein Verbundelement für nichtlineare Finite Elemente Analysen—Anwendung auf Übergreifungsstöße. 35:1739–1766. 2006. Christchurch. An unconservative assessment of the seismic response of a joint may be done. Microplane model for concrete with relaxed kinematic constraint. Local . Earthquake Engng Strct Dy. Mater Struct RILEM 1983. 12.. Verderame.. Priestley. Paper 393.N. Retrofit strategy of non-seismically designed frame systems. Athens. E. 1983. with cracking pattern that is shown in Fig. 1997. 2006.P.-H. S. Pampanin. which determine the subsequent degradation of the joint resistance. M. bond-stress relationships of deformed bars under generalized excitation. University of Canterbury. 38: pp. T. Greece.. Development and validation of a metallic haunch retrofit solution for existing under-designed rc frame buildings. REFERENCES Bažant. since it may lead to a ductile flexural mechanism rather than a brittle shear failure. However.93(16):155–77.. was confirmed by the numerical simulations. S. The simulations were able to reproduce correctly the general behavior of the test specimens with deformed and plain round bars. & Oh. Displacement-based seismic assessment of reinforced concrete buildings. Modelling of shear hinge mechanism in poorly detailed rc beamcolumn joints.