ACI STRUCTURAL JOURNA\ CHNICAL PAPER Title no. 95-855 Pivot Hysteresis Model for Reinforced Concrete Members by Robert K. Dowell, Frieder Seible, and Edward L. Wilson For nonlinear dynamic seisnic analysis of bridge stactres tobe proc ca, onty ee dominant nonlinear eharcterists ofthe structure shone Included. Based on capacts design principtes. the carmen seismic bridge ‘design palosophy be generaly to force al member nolonaries ato the ends of duce colunns. Thersfore, nonlinear characteristics ofthe cok ‘ans must be adegucely defined fv his paper a hysteresis made spre seated which accurately eapuures the nonlinear behavior of reinforced concrete members in terms ofa forcediplacenent, or nement-roatian, responce, Whot mates tic mos otactive, when compared to other hys teresis models, i hat unsymmatric sections Unansymnmetre cross-section _geemenry andlor tension reinforcement amounts ithe two loading dec Fons, acyclic vial load, and srenath degradation maybe Included. The ‘mel isbased on fw simple ues. Results bused om the proposed tyser ‘36 model show close agreement with varions experiments on renjorced ‘concrete member Keywords: bridge: damping: guviy; dynamic: Hysteresis nonlinear, prctcal design; enforced conte seismic, INTRODUCTION Recent bridge failures caused by moderate o strong earth- {quakes in Califormia and elsewhere have prompted a re- hewed Inerest in understanding the nonlinear seismic behavior of existing and new bridge structures, To accurate- ly mode! the nonlinear dynamic response of a bridge strue- ture subjected to a severe earthquake, an understanding of significant nonlinearities is required. Current bridge design philosophy dictates that inelastic behavior be strategically located at the ends of columns in the form of ductile plastic hinges. Based on capacity design principles, the adjacent su- perstructure and footings, as well as the column center por tion, are protected against inclastic action by providing ‘enough strength to force plastic hinges to form at the column ends. Therefore, a nonlinear dynamic time-history analysis of a typical bridge structure requires, among other things, that suitable nonlinear elements be used to accurately model the inelastic action expected at the column ends, In the past, elasto-plastic hysteresis rules were frequently used fo model the nonlinear behavior of columns. Although this model might prove reasonable for steel members, it is not a good representation of reinforced concrete. In 1970, ACI Structural Journal / September-October 1998 ‘Takeda et al. published a paper which included a set of gen- cral hysteresis rules for the response of reinforced concrete ‘members, commonly called the Takeda Degrading Stiffness Model.' In this landmark paper, the importance of correctly modeling the cyclic behavior of reinforced concrete was demonstrated by excellent comparisons to dynamic response fest results, It was shown that the constantly changing stiff ness of reinforced concrete, producing less damping per ey- cle than an elasto-plastic response, must be explicitly accounted for in the hysteresis rules to produce realistic non- linear dynamic results. In the years that followed this paper. changes were made to the basic model, resulting in the Mod. ified Takeda Model? The most notable difference is that the pre-cracked stiffness is often no longer included, resulting in an initial stiffness based! on the cracked section. Time-histo- Ty resulls using the cracked stiffness have been shown 10 agree very well with results using the original Takeda mod- cl. Some form of the Modified Takeda Model is used in sev- cral nonlinear programs.?# ‘The 1970 Takeda paper showed that a stiffness-degrading hysteresis model can adequately capture the nonlinear dy- namie response of a reinforced concrete cantilever column subjected to base excitation, Not included in the Takeda Model are (1) variable column axial loads, (2) unsymmetric sections, (3) shifting of the column point-of-contraflexure, (4) biaxial bending effects, and (5) sudden strength degrada- tion in non-ductile members and response modes, Since the Takeda Model was introduced many other hys- teresis models have been proposed, as discussed by Ghosh,? ‘with varying levels of sophistication and capabilites. To in- clude biaxial bending and varying axial loads, several hys- teresis models (4 or 5) are typically used at the critical AC Scan ar 98 No 5 Spend be 1 ‘ight 108, Arpercan Cones Totti ested nudge males fees untrained tom he sepab ropes, Fearn de fos il bpd ne ayAagst D9" ACY Smeal ewe re errTobe R Dowel ca war ngnar wi ANATECH Copan ere ofr tara ieee he Unies of Colon, San Diego mbt eee BS ‘gmetad aso bg engineer with Clr. Hr rch bare Id ona sei naan dein of refrced core cae. et mente eer Stl profesor of ct nein a the Uivesiy {§ Catforna, Son Dig, Heit a mambo nt ACEASCE Comer 536 Con hese Shall Degen Conan ad 1, Concee Bige Deseo ACT Cente) ergs Resi Conc sige and €27 Steg va non of Eng Corre Scores. Bévard Winton Profesor Een of Satara Engineering the Unies lf Cabfomta cs Betty wars he as ajay member 9618s been remorse for he deecpen feel Cnr rograns aaeet the Engineering Criteria Board of BCDC ard charmar of the Seismic Review Com re forte Berkley snp section representing composite (concrete and steel) nonlin- cearsprings.® ‘Some of the above-mentioned issues can be addressed by fiber or filament models, in which the member cross-section is divided into many longitudinal filaments representing the Tongitudinal steel, and confined and unconfined concrete CCyelic stress-strain rules are assigned to each material, al- Towing the moment-curvature hysteresis response 10 be found, including variable axial load and members with any combination of materials, loading, and section shape. The foree-displacement (or moment-rotation) response of the member is found by assuming a plastic hinge length, or by ‘monitoring the moment-curvature response at several sec tions along the member length, Recent efforts in fiber modeling, by Taucer eta? for ex- ample, show that good hysteresis results of reinforced con- crete columns and beams are possible, when reasonable hysteretic constitutive assumptions for concrete and rein- forcement filaments are used. The success of fiber modeling. techniques has led sorge researchers to conclude that this ap- proach is the most rfsonable for the dynamic noninear analysis of bridge structures. However, the computational ‘cast and storage requirements of even a single-degree-of- freedom fiber model can be prohibitive within the frame ‘work of 2 nonlinear time history analysis. Itis not reasonable to use this method at the ends of each column of an entire bridge structure for routine design and/or analysis. There fore, predefined member end force deformation hysteresis ‘models need to be developed which can accurately describe the inelastic response. Required parameters can be calibrated from the more advanced fiber model analysis and verified (where possible) from experimental results. In this paper, a hysteresis model is presented for force-dis placement or moment-rotation response of reinforced con crete members. Effects of a cyclic axial load, lack of section syrnmetry, and strength degradation are included. The use of, the more advanced fiber model for calibration of the hyster- sis parameters is demonstrated for circular columns and verified by experimental results. The primary advantage of the proposed hysteresis model, when compared to other models found in the literature! is its ability to capture the dominant nonlinear characteristics of very complicated member response with three simple nutes. PRIMARY PNOT POINT /) ily INCHING PIVOT at) aly INITIAL STIFFNESS LN i ‘pRaRY Iver POINT Fig. 1—Typical forcesdisplacement behavior of reinforced concrete column RESEARCH SIGNIFICANCE “This paper presents a new hysteresis model for the nonlin ear response of reinforced concrete members. The cyclic be- havior of the model follows three simple rules based on ‘geometry, Despite its simplicity, the Pivot Model gives z00d results as judged by comparisons to measured results, and re- sults from the more rational fiber model MODEL DEVELOPMENT Measured force-displacement hysteresis results of large- scale reinforced concrete members* consistently show that (1) unloading stiffness decreases as displacement ductility increases, and (2) following a nonlinear excursion in one di- rection, upon load reversal, the force-displacement path crosses the idealized initial stifTness lin prior to reaching the ‘idealized yicld foree (unlike elasto-plastic response). Exper- mental observations show that unloading, back to zero force. from any displacement level, is generally guided toward a single point in the force-displacement plane, on the idealized initial stiffness line (primary pivot point in Fig, 1). Thisis wot ‘ new idea: others have previously used the same approach for unloading. Also observed is that all force-displacement paths tend to cross the elastic loading line at approximatelyFig, 2—Strength envelope the same point (pinching pivot point in Fig. 1). The develop- ment of the proposed model starts with these observations and expands them to. hysteresis model for very complex sit- uations. Note that the force-deformation designation used in the Pivot Model description is completely general. Itcan rep- resent the moment-rotation response of 2 plastic hinge, force-displacement response of a column in single or double bending, or force-displacement response of an entire bridge frame. Experimental results shown in Fig, 1 are for specimen (C2 described tater. Description of rules ‘The Pivat Model is governed by a set of rules which de- pend on the properties of the member and the history of load= ing. Because the model aust represent all possible loading sequences and member response paths, the rules must be ‘complete and result in « unique force-deformation relation- ship. Figure 2 displays the strength envelope for monoto cally incteasing positive and negative louding, which is taken as the upper bound for cyclic loading, and may be de- termined as discussed in Ref. 10, The first and second branches of the strength envelope model clastic stiffness (cracked section) and strnin-hardening stiffness, respective- ly. The third branch represents strength degradation, such as from shear failure, confinement failure, or bar pollout. The ourth, and final, branch allows fora linearly decreasing re- sidual strength, As demonstrated in Fig, 2, the two leading dizections cun have different strength envelopes. In this way, members “which are unsymmetric, or have a cyclic axial load assoc ‘ed with frame action, may have diferent intial eracked stff- rnosses (based on frst yield ofthe reinforcement as discussed in Ret 10) and yield forces (nominal member eapacity) rep resented for the vo loading directions, It should be noted that these properties may be found from a simple push-over analysis of the frame (based on the plastic capacity of the columns) when a cyclic axial load is considered and no strength degradation is included. The analysis indicates ad- ditional axial loadin one loading direction, and reduced axial ACI Structural Journal / September-October 1998 Fig, 3—Quadrane definition load in the opposite loading direction, with corresponding changes tothe initial cracked stiffness and yield force of the ‘member. In this way, a cyclic axial load is treated ia the same fashion as an unsymmetric member, with the inital stiffness, yield force, and hysteretic parameters specified separately for the ewo directions of loading. Note that the cyclic axial Toad discussed in this paper is a consequence of frame action ‘which allows the force-deformation envelope to be defined using & unique relationship between deformation and axial Toad. The mor general variation in axial load associated With vertical accelerations is not included in the hysteretic response Prior to presenting rules which define unloading anc re- loading, it is necessary to define four quadrants, as shown in Fig. 3. Note that the four quadrants are defined by the hori- zontal axis and the elastic loading lines (not the vertical force axis), as indicated by the hatching in Fig. 3. This definition simplifies the model since different rules apply on either side of the elastic loading lines. It is also necessary to define pri- ‘mary pivot points P; through Pg. which control the amowat of softening expected with increasing displacement, and pinching pivot points PP and PP, which fix the degree of| pinching following a load reversal (Fig. 4). As will become ‘evident inthe rules section discussed later, the number des= iguation of the pivot points (Fig. 4) corresponds to dhe par- ticular quadrant they affect. For example, unloading, and Foakiing in quadrant Qy is guided by point P). From Fig. 4 it is clear thatthe pivot points lie on the elastic loading line ‘which corresponds to the quadrant they affect, This ensures thar the complete force-displacement plane is covered for ‘members with different initial (cracked) stiffnesses in the ‘wo directions of loading. ‘The response follows the strength envelope shown in Fig. 2 so long as no displacement reversal occurs. Once the yield ‘deformation has been exceeded (in either direction), subse- ‘quent strength envelope is developed requiring the introduc- tion of points 5; and S> which move along the strength envelope and are defined by the previous maximum dis placements (see Fig. 4). The modified strength envelope 609Fig. 4Pivor point designations (acting as the upper bound for fuuure eyelic load fined by lines joining points PP, to S,, and points PP> to S3 (Fig. 4), Although the pinching pivot points are initially fixed, they move toward the force displacement origin once strength degradation has occurred. The parameters which de- Fine these points are Bi gas Sd) Fiuax) By imax > dud where fy and By define the degree of pinching for a ductile flexural response prior to strength degradation, Displace- ments djya (Fig. 4) and dj, (Fig. 2) are the maximum dis- placement and sigengih degradation displacement, respectively, in the “Pairection of loading. ‘The following isa list of rules which define unloading and loading for the four quadrants shown in Fig. 3. Loading is de- fined as +Ad for quadrants Qy and Q. and -Ad for quadrants Q, and Qs, with Ad defined as a displacement increment. Untoading is defined as ~Ad for quadrants Q; and Qs, and ++Ad for quadrants Qy and Q5. For the force-displacement path to leave the current strength envelope, a displacement reversal is required. While viewing these rules it may help to look at Fig. 5 which displays the same rules graphically. A sample point (in the force-displacement plane) is placed in each quadrant, The arrows extending from these sample points represent potential loading or unloading paths in the force-displacement plane. RULES: Quadrant 1 + Rule 1) Unloading—move along a line toward point Py + Rule 2) Loading—move along a line away from point P, Quadrant 2 + (Rule 3) Unloading—move along a line away from point P 610 Alample 8 Qs Fig. 5—Graphical representation of rules + Rule 4) Loading—move along a line toward point PP Quadrant 3 + Rule 5) Unloading—move along a line toward point Ps + Rule 6) Loadi Ps move along line away from point Quadrant 4 + Rule 7) Unloading —move along a line away from point Py + Rule 8) Loading—move along a line toward point PP, By examining thelist of rules above, the number of rules ccan be condensed to a set of three simple rules. This is due to the logical numbering which was assigned between pivot points and quadrants, allowing the use of symmetry 0 re: duce the mumber of rules. The reduced set of rules is given below. CONDENSED SET OF RULES ‘Quadrants 1 and 3 + (Rule 1) Loading and unloading in quadrant Q, ts directed away from or toward point P,, respectively Quadrants 2 and 4 + (Rule 2) Loading in quadrant Q, is directed toward point PP, + ule 3) Unloading in quadrant Q, is directed away ‘rom point P, Modified Pivot Model ‘The hysteresis model presented thus far does not recognize 1 softened intial stiffness following a nonlinear excursion Inother words, ifthe force-displacement path passes through the origin, the initia! elastic stiffness is regained (until the modified strength envelope is encountered). However, ex: perimental and fiber model results indicate a softened initial stiffness following a large nonlinear excursion. This is also soon in dynamic tests where the natural period of the struc tare is lengthened atthe end ofthe experiment. ACI Structural Journal / September-October 1998OF yy OuFys (1 +7)) Fig. 6Modifed elastic loading tine To include this phenomenon within the framework of the Pivot Model, the elastic Toading lines are free to sotate as & function of the maximum deformation and specified param- ter 1) (Fig. 6). Therefore the quadrant definition is updated ‘whenever the previous maximum displacement is exceeded in either direction. Figure 6 demonstrates the rotation of the clastic loading line sacegle ‘quadrants Q; and Qy. The un: loading stiffness of the mafimuin displacement excursion in ‘quadtant Q; is guided toward point P, (unloading stiffness is same as original model at maximum displacement. This line continues until point P,* is reached at a force (1 + 1) times, larger than the force which defines point P,. A line extend- ing from point P,* through the origin defines the new sot ened elastic loading fine K*, and updated quadrants Qy and Q¢. Point P* is on the new elastic loading line at the same force evel as point Py, Point PP,* is also on the new elastic loading fine, but ata force defined by the intersection of the ‘modified strength envelope (line berween points PPg aud S,) and &* (Fig. 6). The hysteresis rules provided earlier are ‘used with the updated pivot points (indicated by an asterisk) ‘The points in quadrants Q; and Qy are affected because they border the modified elastic loading line, The same pro- cedure is followed if » nontinear displacement occuss in the opposite loading direction, reducing the elastic loading line between quadrants Q> and Qs, thus changing the behavioe in those quadrants, Note that with n=O, the model is identical to the original model. With a large value of n (say, 10) the “unloading stiffness for small eycles is approximately equal to the unloading stiffness at maxinnum displacement, To the authors’ knowledge this is the only hysteresis model capable of specifying the degree of initial stiffness los. PARAMETERS FOR CIRCULAR RC COLUMNS. WITHOUT STRENGTH DEGRADATION ‘The required hysteresis parameters for symmetric sections ‘with no axial load variation and without strength degradation are cand ( (Fig. 4). Results from multiple analyses of irew- lar reinforced concrete columns (Fig. 7) using a simplified fi ‘ber model," indicate that these parameters may be expressed sa function of the longitudinal steal ratio and axial load ra tio (ania! load divided by the unconfined concrete strength and gross section area - P/fedg). Since the charts were de- veloped with Grade 60 reinforcemeat, an effective longitudi nal steel ratio of 5 the actual ratio should be used for Grade 40 steel, The required bilinear envelope of the Pivot Model ‘cam be found ftom moment-curvature analysis, Chart verification ‘To examine the validity of the charts presented in Fig. 7, parameters cand f are measured from experimental results of citcular reinforced concrete column specimens C2, C4, ‘GI. and FA (as shown in Ref. 11). The parameters ate taken 2s the average values from push and pull directions at dis- placement ductility 6, evs found at the intersection ofthe ini- tial cracked stiffness and the softened unloading stiffness, \efined by a line from the force at displacement ductility 6 ‘hough the displacement at zer0 foree. Bis found following ‘load reversal at the intersection of the force-displacement path and th initial cracked stiftness. Columns C2 and C4 are shear-ritical columns"? which were retrofitted 1 behave flexurally, while GI isa flexural column'® which willbe is cussed in greater detail later. F4 ig also a flexural column.® ‘Table I displays relevant information for each specimen, in cluding measured hysteresis parameters o and B. By com paring these parameters with those obtained from Fig. 7 (and splayed in Table 1), the columa test indicate that the con tours detived from the fiber model are realistic. Measured «. perameters for Specimens C2, C4, and GI axe slightly lower than the chart values, while the measured cz parameter for Specimen Fé is slightly higher. inching characteristics, rep Table 1—Specimen details and measured and calculated hysteresis parameters Five Longing | tenga Aviallout |'Stel rato, | seetaies | a, | ocho cha Specimen _[atinyperent| “percent” | “perent™ | memired | (hig) | mestwea | (hig 7 G 5 a5 st 5 ec ce is 35; 16) [isp ae [os | 0 ai 15 2s 265 4 S__[ oss [033 5 Lae a5 ior [as | 18] 0s [ose ACI Structural Journal / Saptemiber-October 1998 en& g \ 2 2 } 3 3 os Jo88| jos z é as | 3 PS i /) YF? 3 me Bs. 2 j/ * 2 ° = 3 Ld Oe Longitudinal Steel Ratio (%) Longitudinal Steel Ratio (%) (a) (b) Fig. ~Contours for ysteress parameters for circlar reinforced concrete columns: (ao parameter, and () parameter [=a] 1 15 20 25 3 a6 4 Longitudinal steet ratio (%) Fig, 8 Ratio of post elastic stiffness to inital cracked Stiess “t" fr creular columns resented by Bare accurately captured for specimens C2 and. Gi, and are reasonably close for specimen F4. Specimen C4 showed more pinching chan that predicted from the fiber model. This is probably due to the high shear forces of this ‘column, which is known to cause pinching and reduction of B. Overall, measured parameters from the four specimens ‘match those found from the chart (developed by fiber model analyses) very well, In ation to the hysteresis parameters discussed above, the ratio of post-elastic stiffness to initial cracked stiffness “P's provided in Fig, 8 as a function ofthe axial load ratio (ALR) and longitudinal steet ratio. This figure was also de- veloped from fiber model analyses. MODEL VERIFICATION Loading in deformation control ‘To demonstrate the capabilities of the proposed model in {deformation contol, measured force-displacement hystere- 612 sis loops for circular reinforced concrete cantilever column specimen Gl (Fig. 9) are compared with results from the Piv- ot Model (modified, unloading stiffness of small displae- ments approximately equal to unloading stiffuess at saximum displacement with 1] = 10). Asa hench-mark, the Takeda model is included in some of the comparisons, as well as the more rigorous fiber model, Since itis not the pur pose ofthis paper to discuss the test specimen in detail, om a cursory explanation is provided. Further details ofthe test specimen can be found elsewhere? In the following figures the member force F is normalized to the nominal force Fy ‘fozce ratio) and the displacement is normalized tothe ideal- ized yield displacement (displacement ducility) permitting the hysteretic response of various members to be displayed in dimensionless form. ‘Column tess are typically conducted with increasing dis- placemenis in both loading directions, producing symmetric hysteresis loops, Although this loading provides valuable in- formation, such as energy dissipation characteristics for a siiven ductility level, as well as the expected monotonic re sponse, there are important hysteretic modeling, consider ations that remain unanswered. Under the influence of an ccarthquake, displacement cycles will differ arcaty from ar- ificially imposed, steadily increasing, symmetric loading cycles of the laboratory. For example, itis clear from typical reinforced concrete column ests that unloading stiffness softens as ductility is increased. What is not known from these experiments ishow the member stiffness, at small duc- tiliis, is affected by previous excursions to larger ductility levels. Also unanswered is the direction the force-displace- ‘ment path takes if cyeling occurs in one direction only. Specimen GI had its loading pattern intentionally altereé from the standard fully reversed cyclic testing procedures of the Inboratory'? to ensure that the Pivot Model responds ep- propriately for amore realistic loading. Figure 10 shows that portion ofthe loading history which contain these added dis- placement cyeles. With o-and B found carlir from the charts a 5 and .55 (Table 1), and ratio of post-elasti stiffness to ACI Structural Journal / September-October 199815 Experimental Repults 0 4 - Fiber Medel i i eae Tested ra Displacement Duetility @ Boa ty a3 2-7 EE Displacement Duet (b) Fig. 9Complete test of Specimen G1: (a) measured and fiber model results, and (b) Pivot Model vesults initial cracked stiffness "Y” of 1 (from Fig. 8), the Pivot Model responds similarly tothe test specimen (Fig. 9) Figure’ 11 compares measured unsymmetric force-tis- placement cycles, tothe response ofthe Pivot Model, Takeda ‘Model, and fiber model. Figures 11(b), (), und (f) compare the measured response of an additional eycle applied in the sume loading direction (corresponding 10 '¥; cycles 3 through 5 in Fig. 10), with results from the fiber, Pivot and ‘Takeda models. Figures 1 (a), (c), and (e) compare the mea- sured response of small displacement cycles following large displacement eycles (corresponding to '/ cycles 11 through 15 in Fig. 10), with results from the fiber, Pivot and Takeda models. Figure 11 demonstrates thatthe response of the proposed hysteresis model closely resembles the measured response, While the Takeda Mode! fjpas to point tothe previous max. ‘mum displacement too early, resulting ina response for un. balanced cycles which may he quite different fom that measured. The measured amount of energy dissipated (rela- tive energy in terins of ductility and force ratio) at each duc~ Lilly level tesiea (Fig. 12[3), is similar to that found from the Pivot Model and Modified Takeda Model (calibrated to ‘match the unloading stiffness af the Pivot Model at displace- ‘ment ductility 4), This shows that for the simple case of a symmetric member loaded with even displacement eycles, the Pivot Model behaves similarly tothe widely used Takeda Model, Figure 12(b) indicates the relative energy dissipated for the unbalanced loops (loops |, 2, and 3 correspond to '/ cycles 11 through 15, neglecting the small inser loop, 3 through 5, and 12 through 14, respectively, in Fig. 10). As ‘expected from Fig. 11, the energy dissipated from the Pivot Mode! is quite close to that measured, while the Takeda ‘Model generally underestimates the energy dissipated. This is especially true for the larger unbalanced displacement cy le loop 1 Dynamic loading Te-was shown in the previous section that large symmetric displacement cycles and unbalanced displacement eycles of ACI Structural Journal / September-October 1998 Displacement Ducthity a o& oN ho Fig. 10 Loading history for unbalanced eyeles tested reinforced concrete column are well represented by the proposed Pivot Hysteresis Model. However, to more re alistvally test dhe rules of a hysteresis model it should be subjected to an earthquake motion. This exercises the force displacement paths in somewhat random fashion, loading and unioading many times in each quadrant, As discussed previously, column specimen G1 was tested in deformation control. No dynamic loading ofthe specimen was conducted However, a nonlinear time-history analysis of the specimen, subjected tothe first 30 seconds of the 1940 El Cento earth- {quake (Imperial Valley Ivigation District, SOOE), was per- ormed with 2 percent viscous damping, The inertial mass is ‘adjusted so thatthe initial elastic period of the structure is fone second, In order to ensure reasonably large displacement ductility. the earthquake record is magnified by a factor of six over that required to produce the idealized yield force of the specimen (strength reduction factor Z of 6), ‘The same nonlinear analysis was conducted with the Pivot Mode! and the fiber model, The Modified Pivot Model was used, with unloading stiffness of small displacements ap- 61a1s Taperimental Rate ier Model wih “9 os Eon ——s | - “0 Z| -10 a0 yeaa oT ESTE ETS Wee eaa eas i PTGS OTS Displacement Ductility Displacement Ductility a) @) 1 —Bperimeniat Rebuts Gaperimental Reps ber Mode! ‘okeda Model tg | Biber Mode! 4 Al os 4 Af -10! i 5 eaa4 isd TESTOR STS SR BACT EIT SETS Displacement Ductility Displacement. Ductility (b) (e) 18 18 Expeginentel Rasuita {— caperimental = Filodet == Takeda Model 10 10 os os Sos 05 -10 “'iy-es-s-a eo eae OB 8 ges g-2-10 7 234567 0 Displacement Ductility placement Ductility © ® Fig. 11—Response of models compared to measured results for unbalanced displacement eles of Specimen GI: (a) Loops } and 3 fiber model and measured; (b) Loop 2 fiber model and measured; c) Loops 1 and 3 Pivot Model and measured; (2} Loop 2 Pivot Model and measured: (e) Loops 1 and 3 Takeda Model and measured: and (f) Loop 2 Takeda Model and measured proximately equal to the unloading stiffness at maximum displacement (7 = 10). This gave better results than the orig al Pivot Model following the maximum nonlinear excur- sion, As discussed previously, unlike « hysteresis model, the ea force-displacement response of a fiber model requires 20 rules. Iis reasonable, therefore, o expect that the dynamic response ofthe finer model will perform properly, anc that it may be used as a bench-mark forthe dynamic response of the ACI Structural Journal / September-October 1998ce t Soe | : snes : Ea fee a cep timber (a) (b) Fig. 12 Relative hysteresis energy: (a) synmetric displacement excles, and (b) unbal- anced displacement cycles —— Pivot Model Fiber Model Abmeormwee Displacement Duetiity Time (sec) (@) Displacement ductility timo-histry 1S - Pivot Model 1.0 - Fiber Modet 2 05 & 0.0 Ay -0.5 - ~1g5 h ne 10 15, 20°" 25" "30 Time (sec) (©) Force ratio time history Fig. 13—Dynamie response of Specimen GI subjected to the 1940 El Centro earthquake (constant axial toad) (a) displacement ductility uime-histors. and (b force rato time-Bistory 1s Pivot Model. Figure 13 shows that the response ofthe Pivot Model is very similar tothe fiber model behavior through the 10 entite 30 seconds of loading. The displacement ductility time-history (Fig, 13[a)) indicates good correlation in maxi- 8 ‘mum displacement and period of vibration. Als, force ratio Evo timerhistory reals show elose agreement between the two = models Fig. 131b). 08 ‘To demonstrate unsymmetsic behavior of the proposed hysteresis model under dyaamic response, specimen (313s, “10 investigated with a cyclic axial load. ‘The seismic axial load vaties in proportion to the ertical section moment with a SIESTA GT EG EEF constant of proportionality of 1.08 Hin (0.0315 Iiem), Note ‘Displacement’ Ductitity that this constant was chosen arbitrarily to demonstrate un symmetric member behavior associated with axial load vari- Fig. 14 Specimen GL with varying axial load ation which can be attbuted to frame action. The fiber ACI Structural Journal / Sentember-October 1998 615Pivot Model Fiber Model Displacement Ductility 10 15, 20 30 Time (sec) (@ Displacement ductility time-histry (©) Force ratio time history Pivot Model Fiber Model Time (sec) Fig. 15—Dynamic response of Specimen GI subjected tothe 1940 El Centro earthquake (varying axial load): (a) displacement ductility tme-histors, and (b) force eatiotime-history ‘model response of the specimen in deformation control is shown in Fig. 14, along with the corresponding Pivot Model response. Required parameters of the Pivot Model were ound from the fiber model results displayed in Fig. 14. In the following figures, nominal force and ductility one are based on dead foad only, allowing response variations due to the changing axial load to be quickly observed. Figure 14 “demonstrates that in deformation control the Pivot Model re- sponds similarly to the fiber model with the large axial load variations induced, As stated previously, a more thorough test of @ hysteresis model is in its dynamic response. Therefore, specimen G1 was subjected tothe first 30 seconds of the El Centro earth quake with the above discussed axial load variation, and ‘magnification from the prior nonlinear analyses. A notice: able difference between the response of the specimen with axial load variation and without axial Ioad variation is seen by viewing Fig, 13 and 15, The Pivat Model does a goo job ‘of mimicking the dynamic behavior ofthe fiber model (Fig, 15) for this unsymmetric column response. Figure 15(b) shows the importance of correctly modeling an unsymmettic 616 response, with the force on the tension side being significant- ly Jess than nominal, and on the compression side much sreater then nominal. This demonstrates that the proposed hysteresis model can quite accurately capture dynamic un- symmetric member behavior. Note that in this example a varying axial toad caused the unsymmetric behavior. [Lcould have been produced by other means, such as 2 different amount of reinforcement inthe two loading directions, oF an “unsymmetic seetion shape. OTHER POSSIBLE HYSTERESIS BEHAVIORS The following provides several examples of possible Pivot Model behavior for various members and response modes Unlike the previous examples, the Pivot Model parameters ane taken directly from the measured data. Therefore these ae not predictive models. Figures 16(a) and (b) show the xe- sponse ofa retrofitted outrigger bent, which performed in @ ductile maoner with no strength degradation, and an “as- buat” outrigger bent, which failed in joint shear in one iree- tion only." Figure 16(e) demonstrates a elessie column shear failure. with sudden oss in force capacity prior to ACI Structural Journal / September October 1998%0 orce (Kips) Displacement (in ) Displacement (in.) (b) Menwured Displacement (in.) () Fig, 16--Potetal Pivot Model behavior compared with various experiments: (a) retrofitted knee join (outrigger): (b) “as-built” knee joint (outrigger): and (c) shear column ductility one. In all cases the proposed hysteresis model responds similarly tothe measured results, CONCLUSIONS AA simple and efficient hysteresis mode! is proposed for force-displacement, or moment-rotation, response of rein- forced concrete members. Fora symmetric column, with no strength degradation, the Pivot Model responds similarly to the popular Modified Takeda Model, with improved results for unsymmetric displacement cycles. It was shown that add ced effects from an unsymmetric section, cyeic axial load as ACI Structural Journal / Seotembar-October 1998 sociated with frame action, and strength degradation may be included, Since strength envelope parameters, as well a un- Joadting and loading parameters, are specified separately for the two loading directions, the hysteresis response can mim= i very intricate behavior. Presently, the Pivot Mode does not include (1) continued strength degradation for muliple cycles to the same dis- placement level, (2) strength degradation in one loading di rection caused by sudden strength loss in the opposite loading direction, and (3) biaxial bending effects. These phe ‘nomend are curently under investigation for inclusion i fue ture versions of the Pivot Model. The unique quadrant approach of the Pivot Model, which s central ots hyster behavior, may sid in the development of a simple model ‘which includes biaxial bending, To define the Pivot Model parametcrs and investigate the above discussed phenomena, simplified fiber model is under development which closely ‘matches experimentally obtained results." CONVERSION FACTORS. Tie, Tip REFERENCES 1. Takeda, Ts Sozen, M.A and Nico, N,N, "Reinforced Concrete Response o Simulated Eahanakes." J. Strc. Engr. Dit, ASCE, V9, No. 12, 1970, 9p 2257-2573, 2 Car, A. J, RUAUMOKO User's Gude, Deparment of Civil Eg cerig, University of Camry, Christe, New Zaki, 195, 5. Sali, M, “Hysteresis Models foe Reinforced Cort,” J Siu ‘Engr Dr, ASCE. V 108, No.8, 1982, pp. 1077-1087 4. Powe, G.H, "Supplement to Comptes Program DRAIN-2D," Sup plement to Renee PRAIN.2D User's Gude, University of California, Ber ky, 1975, 5. Ghoeh, S. K., Earthquake Rexstont Concrite Sucre elie Response and Devign, ACLSP-127, Amstican Concrete, Detroit (991 6: ing, ¥. Sait, Mt, “Foor Spring Ema for Cycle espe RAC Colans* J Suet. nrs, Dis ASCE, V.116, NO 4 19%, pp ‘orga 17 Tava, Fs Spaone End Pippa, FC, “A Fiber Bem Com [lemeot for Semis Azuyas of Reinforsed Conese Stren” Rant No. UCRFERC-91/17, Colege of Enaiozing, Universi of Califia, Berkley, 1981, 8. Presley. M.J.N ar Sel, F, “Sekanic Assesment nd Retrofit of Bislges"Aepor No, SSRP-9IN3, Depaanas of Applies Mechanics aha Engincesing Seieaces, Univerny of California, San Diego, 1981 9. Kuanat, $1; Renhor, A M.sahd Path, ¥.J, "Aaa Model ing of loelasic Seine Response of RIC Stores” J Siac. Engrs Ds, ASCE, W116, No 4, 190, 9. 861017 10, Piste. M.J.No Sel, F: and Calvi, M, Seamie Design ond Revit of Bags. ohn Wiley and Sens, New Yor, 1956 UL Dowel, RK, *Nowliner Seismic Analysis and Design of Rela ford Conerete Brie Stoctbes” PAD dsseriaon, Uivernty of Cal fami, San Dag, 1998 2. reste, M1. N; Seible,F; Xiao, Yad Verna, R “Steel Iacket Rewriting of Renforced Conerete Bridge Onlurns fr Enhance Shear Suenglh-Pur 2: Test Reus and Comparson with They” ACT Siac tral Journal ¥. 91, Ne. 8, ep O. 19, pp. 53 13, Smith, G. “Suaepe Relocation of Plste Hinges i Bridge Col uns." Master's Thesis, University of Caifoen Sa Diego, 196, 1M. Ingham, J, “Sela Performance of Bhs Knee Joints” PhD, Dissertation, Univesity of Cafe, San Diego, 1995, 617,ACI Structural Research Award wae _ presented to: Paannd L. Wilson 2000 ‘Yor your co-authored paper, ‘Pivot Hysteresis Model for Reinforced Concrete Members,” September(October 1998 ACY Structuxal Journal, which describes the development of a simplified approach for madeling the nonlinear behavior of reinforced concrete bridge structtores subjected to earthquakes” american concrete institute President. international : Presentation Date. March 30, 2000 Mie