antique Engineering, Tent Word Conference. 1982 BaBems, Roterdam. ISBN 9054100605, Collapse simulation of steel frames with K Takanashi, K.~i,Ohi & ¥.Chen Institute of Industrial Science, University of Tokyo, Japan L.Meng local buckling Technical Research Institute, Fujita Corporation, Yokohama, Japan ABSTRACT: An inelastic portion of stee! beam-column is modeled as an assembly of. Simple hysteresis rule is proposed to simulate ther inelastic behaviors in consideration of ig elements, and a resistance deterio~ ration due to local buckling. After the model parameters are chosen so that they match with the monotonic test results, the validity of the model when it is used in the response analysis is checked throu the comparison withthe results ofthe on-line response tests, some of which simblate the complete collapse of steel frames. 1 INTRODUCTION ‘To establish theultimate safety of frames to earthquake loading, it is indispensable to carry out @ number of response simulation over various structural parameters {oan extensive set of earthquake motions. Constitutive ‘modeling used for such simulation is better to be es simple as possible, but should never neglect essential roperties of inclastic structural behaviors. Especially for stel frames, several inelastic bchaviors, such aS yielding, strain-hardening, Bauschinger’s effect, local buckling, P~A effect, and so on, shouldbe appropriate Jy considered in the model construction In this paper, an analytical model termed "malti— spring inelastic joint” (Ohi, 1992) is studied to simulate the behaviors of inelastic portions of steel beam= columns subject to uni-axial bending and axial load. ‘Similar modeling was already proposed for concrete members subject to bi-axial bending (Lay, 1984). A Simple hysteresis rule suitable to express the above ‘mentioned inelastic behaviors is also proposed herein ‘After model parameters are calibrated to the results of the monotonic loading tests, the validity ofthe model when iti. used in the response analysis is demonstrat ed through the comparison with the results of the on~ line pseudo-dynamic tests, which are performed on H-staped and square box specimens. Maulti-spring inelastic joint \ —aph) a 7 Elastic beam clement Figure 1 Simple beam-column model 2 A SIMPLIFIED BEAM-COLUMN MODEL A steel beam-columa i regarded as an asserbly of two Kinds of clemeats es show in Figure 1: one an elastic beam, andthe ater i an Snelasic oi Inthe following analysis, nodes are located ai the tips of these two elements, and geometrical non-linear or PA effect considered only fr th dplacements of these nodes by updating each clement coordinate ‘yam ne ena oles ‘ inelastic ot consi of fur axial springs and a shear panel as shown in Figuce 2, Inthe presence of ih sey tanamiaion among te springs and the panel each sping may cay el free varying a Brejoim axle, Sut the anda force at the center chosen asthe epresctative axial free ofeach ping. (© Shear panel (@) Axial springs Figure 2 Multi-spring inelastic joint 2.1 Hysteresis model for springs ‘Two different skelcton curves are used for the tension~ side and the compression-side behaviors of each ing, respectively. ‘The tension-side skeleton curve all Be similar to the uni-axial stress-strain ‘curve of steel material, and the compression-side skeleton curve shall be arranged in consideration of the ‘strength deterioration due to local buckling. Piccewise linear models are adopted herein for the both side skeleton curves as shown in Figure 3(2). ‘Two imaginary points termed ‘target points’ on the skeleton curves, one for cach side, are considered herein. These points are referred to determine unload- ing and reloading paths. Each target point is sc atthe clastic-limit point inthe initial state. When a loading atbeyond the elastc-limit is made along one sie skelo~ toa witha certuin amount of peste deformation, the target point ofthe loading side moves together wih the loading point, and atthe same time, the other Skeleton curve inluding the oter target point stall be Shifted tthe loading dieesion 35 much ae W times the Plastic deformation: Such # loading procedure on the celeron curve i stated in Figure 30). IF set to zero, neither hardening nor degrading occurs during cyclic reversals ithin the past peak amplitudes, If 0 iS setto one, the hysteresis includes o softening dus to the Bauscbinger’ effect. Actual Behaviors Of stea! rrembers are believed to fall onthe intermediate state between these two exteme sates. ‘Unioading and reloading paths ae modeled as por sions of the Ramberg Osgood funtion, denoted by 6 = Rey (P). The parameters included inthis function is emia to satity (D8, = Ro Py). where the point Gy. Fy) te lst teoadng polac ) Bee Regs SHeeogt bye gets betel” ton curve inthe unloading and reloading Gil) K,dR.g/aP (P,)=L, where K, is the int tie stiffness. These cOnditions are ilustrated in Figure ‘3(c). The tangent stiffness denoted by K’ dered trom these conditions as KEK K/EK + 6 KIC-P YEP] where K, = @ PGB) Gest hs Serio) * |, ae r (@ Initial skeleton curves (b) Shift of skeleton curve 22 Calibration of model parameters with monotonic test results Monotonic bending ets were cased out o [Stped and gute box beancoamge odes Stat alate Se shown in Figure © He Square bor sections with varous wide mete ‘lio were welded from sel plates inthe meee the widthticknss rato fo hl the Mange ono H-shaped section are 6, 8, and 10, and those for the ful moana wi of aguare box section re 38S fn db The cates of a0 anil Toad and the cont, Anil load with 0 or 20 percent ofthe Jill ack Were consicere. he yield ates and the maximum test eng needed to aang the tensions hsiaon cee Served om the preliminary tension et eal 40) Sts and 990 Mu in average, pect Aa oe omprssion-sde skeleton curt, iis apically ‘coogi thatthe maximum moment cpacky oF 6S ‘eamecoluma would be underestimated the esate fhe si clin compreson tts Wet we ay Sete This might be ested from the diferent sstress gradient in the beam-column tests and the stub folum teste Consequenisn. ther params thn Stomg moditey the ytd Shes he enol sense aad fe lengt ofthe telat oat assumed eed the clu dept, se adjusted by al andes thc the stmultdinootoic cures tally mich wk te caresponding test curves 7 (tasting Poiae) Figure 3. Hysteresis rule for axial springs =| a 0 200% 200 ACTUATOR L [Las Figure 4 Test setup Brat itn Nps sitet cave ia (© Unloading and reloading 2a x x? a 7 le | ni x [x Beto 10,1690» a) 8 [2 (a) Loading test _—_(b) Vibrational 7 = system analyzed in online tests Figure S. Test schema19 «omen 1a com wal al 7 i. a 12 r9 oom sonpstes in re esmn (computed) tre, om) nears tal aL eat tq osm ta como # ra gsm i (ested) H-shaped beam-columns Figure 6 Hysteresis loops under carthquake loading — tested and computed (parameters: =0.8,"=5) 83 sours el Bre wa moe al 2 wenene (computed) ested) Square box beam-colamss’ ai toro Figure 7 End moment vs. rotation under monotonic loading -—EEt— ca Figure 8 Analyzed beam-column In the loading tests, the yielding and the initiation of the local buckling were detected atthe both sides of the central loading point, but after tht, the local buckling. failures were brought up larger at only one side. Con= sidering this observation, two inelastic joints are in— serted at the center of the beam=column and no resist ‘ance deterioration is assumed for one of the inelast joints inthe following analyses, as shown in Figure 8. ‘The load-deflection curves observed in the monotonic loading tests and those simulated by the models are shown in Figure 7 3 COMPARISON WITH ON-LINE TEST RESULTS Oncline respons tests under eathauake loading were performed by use of the same test setup used in the ‘monotonic loading tests described in Section 2.2. The N-S component of El Centro 1940 was used as the input ground motion, the intensity of which was ad~ justed to attain two kinds of damage levels for ea ‘cases: one is moderate yielding for the case of no axial load, and the other is nearly in the complete collapse State forthe case of constant axial load. In these tests, half ofthe load applied at the center of the specimen during the tests was regarded as the restoring force of the single-degree-of-freedom vibrational system shown in Figure S. It should be noted thatthe deterio~ tation of th lateral resistance of the vibrational system is induced by not only local buckling failure but also the P-d effect due to the constant axial load applied, Fictitious mass was arranged from the elastic stiffness ‘measured prior to the on-line tests so thatthe natural period would be commonly 0.8 seconds, and ro ficti~ tious viscous damping effect was assumed. 2, Taare a tested and modeled real ‘Numerical simulations were carried out based on the ‘multi~spring joint models arranged in Section 2.2. Hysteresis loops observed in the on-line tests and the ‘numerical simulations are compared in Figure 6. The ‘numerical simulation does not always agree exactly ‘with the displacement history observed test, that is, the simulated direction i it permanent set occurs is sometimes opposite to the test result. Even in such a case, however, no fatal eors are recognized in the prediction of the global damage level and the peak response displacement. 4 CONCLUDING REMARKS Inelastic portions of steel beam-columns are modeled as an assembly composed of inelastic springs and ¢ shear panel. Also, a hysteresis rule suitable for the spring elements in such a modeling is proposed in consideration of strain-hardening, softening duc to the Bauschinger’s effect, and resistance deterioration due to local buckling failure as well. "After the model parameters were calibrated tothe re~ sults of the monotonic loading tests which were per- formed on H-shaped and square box beam-columns with various width-thickness ratios, response analyses ‘based on the inelastic joint model were carried out and ‘compared with the results of the on-line response tests under earthquake loading. Ttis demonstrated that the global responses are well predicted for the practical purpose by the present ‘model even in sucha final collapse state as the lateral resistance ofthe frame almost disappears. REFERENCES Lai, SS., G.T. Will & S. Otani 1984. Mode! for inelas~ tic bi-axial bending of concrete members, ASCE, ‘Vol.110, STII. Obi, K.'& K. Takanashi 1992, Multi~spring joint model for inelastic behaviors of steel members with local buckling, In ¥. Fukumoto & G. Lec (eds.), Sta~ Dility and ductility of stecl structures under cyclic loading, CRC Press: 215-224.