Engineering Structures 31 (2008) 3016-3033 Contents lists available at ScienceDirect Engineering Structures Journal homepa nv.elsevier.comilocatelengstruct A computer method for nonlinear inelastic analysis of 3D semi-rigid steel frameworks CG. Chiorean”* Facly of il Engineering, Techial Unversity of i Napoca, 15. Decovc Stet, RO-#00020, Ci-Nopoca, RamanaEngineering Structures 31 (2008) 016-3033 Contents lists available at ScienceDirect Engineering Structures journs! homepage: www.elsevier.comlocate/engstruct Acomputer method for nonlinear inelastic analysis of 3D s¢ mi-rigid steel frameworks i * CG. Chiorean* \ [4 Faculty of i Engineering, Techical University of Caj-Napoa, 15. aicovic Street, 40020, Cu-Napot, Rome i INFO ARTICLE ABSTRACT ‘Fre ison Received 9 Apri 2008, Received in eised form ‘june 2008, Accepted 3 August 2009 ‘Avalble online 15 Augus 2008 ‘This paper presents an efficient computer method for inelastic and large deflection analysis of steel space frames with non-linear flexible joint connections, based on the most refined type of second order inelastic analysis, the plastic zane analysis, The method employs modeling of structures with only one element per ‘member, which reduces the number of degree of freedom involved and the computational time. Gradual ‘yielding of cross-sections is modeled using the nonlinear inelastic force strain relationships, and then using the Mexibilit approach the elasto-plastic tangent stifess matrix and equivalent nodal loads vector ‘of 3-D beam-column element is developed. The method ensures also thatthe plastic bending moment is nowhere exceeded once a full platified section develops A zevo-length rotational spring element is sed fr incorporating the connection flexibility into the element tangent stfiness matrix and equivalent Keywords Piasic-zane analysis Semitgd space amenorks Large deletions nodal forces. The combined effects of material, geometric and connection behaviour nonlinearity sources ‘Advance analysis are simulated into an object oriented computer program automatically. Ths program was sed to study ‘the ultimate response of several ste! frames, which have been studied previously by other researchers. ‘The example of computations and the comparisons made have proved the robustness, accuracy and ime saving ofthe proposed analysis method, © 2008 Elsevier Ltd. All rights reserved, 1. Introduction Inthe spread of plasticity approach, the gradual spread of yield- ‘With the rapid advancement of computer technology, research works are currently in full swing to develop advanced nonlinear inelastic analysis methods and integrate them into the new and ‘more rational advanced analysis and design procedures. There cur rently exist several methods of large deflection elasto-plastic anal- ysis that calculate strength limit states of 3-D steel frames with rigid and semi-rigid connections (1-8). These methods, that use “tine elements” approach, are based on the degree of refinement in representing the plastic yielding effects and can be categorized in two main types, plastic hinge versus spread of plasticity ap- proaches. n the plastichinge approach, the effect of material yel@- ing is “lumped” into a dimensionless plastic hinge. Regions in the frame elements other than at the plastic hinges are assumed to behave elastically, and if the cross-section forces are less than cross-section plastic capacity, elastic behaviour is assumed. The plastic hinge approach eliminates the integration process on the cross section and permits the use of fewer elements for each member, and hence greatly reduces the computing effort. Unfor- tunately, as plastfication in the member is assumed to be concen- trated at the member end, the plastic hinge model is usually less accurate in formulating the member stiffness. Tel: +40264504967; ox: 407264 594967, El eddres:cosminchioreantmecanstliro (041-4206(5 se font matter © 2009 Elsevier Ld Alsight reserved oi 10.036)engstru.2008.08.003, ing is allowed throughout the volume of the members. There are ‘two main approaches that have been used to model the gradual plastfication of members in a second-order inelastic analysis. In both cases the member needs to be subdivided into several el- ements along its length to model the behaviour accurately [9,5 Cross sectional behaviourmay be described by moment-curvature- truss (M-2-P) relations derived for each cross-section in the analysis. Alternatively, the cross-section can be subdivided into el- ‘emental areas and the states of strain, stress and yield stress are monitored explicitly for each of the elemental areas during the analysis, In this case, the effects of residual stresses, geometric imperfections and material train hardening can be accurately in- cluded in the plastic-zone analysis. Hence, the plastic-zone ap- proach is considered to closely simulate the actual behaviour of member, but the computational effort is greatly enhanced and the method becomes prohibited computational inthe case of large scale frame structures. However, the rapid development of computer technology in recent years has enabled the plastic zone theory in which the spreading of plastic zones in members is taken into account to be ‘developed. A number of computer programs have been developed ‘on this theory in recent years by researchers. Unfortunately, the currently available methods for second-order spread of plasticity analysis and advanced analysis are not user friendly for practical applications. These methods ignore many important characteris- tics and requirements for practical design, consistency between(6 Chireon/ Engineering Sucre 31 (2009) 3016-3033 au the linear and nonlinear models due tothe need to use several el- ‘ements per member to model the distributed loads and spread of plasticity along the member length and the computational effort. ‘The approach presented in this paper is intended to over- come these inconveniences and represents an efficient computer ‘method for large displacement distributed plasticity analysis of 3D, semi-rigid steel frames fulfilling the practical and advanced analy sis requirements. The non-linear inelastic static analysis employed herein uses the accuracy of the most refined type of second order inelastic analysis, the spread-of-pastcty analysis and addresses its efficiency and modeling shortcomings through the use of only, ‘one element to model each physical member ofthe frame. Grad ual plastifcation through the cross-section is handled by fitting nonlinear equation to data forthe force-strain behaviour ofa unit length segment of the element. Because these relations depend on the characteristics of each cross-section inthe analysis, the gradual plastfication ofthe cross-section of each member is accounted for by smooth force-strain curves that are experimentally calibrated, ‘This way, inelastic behaviour in the member is modeled in terms of ‘member forces instead of the detailed level of stresses and strain, with favourable effects on the computational effort, The inelastic cross-section modelis then used to obtain flexibility coeficient for the full member by numerical integrations along its length. The ef- fects of material yielding along the member, onthe element sti ness degradation, are considered by axial and flexural rigidity, oF the element, depending on the sectional efforts, at each load incre- ment, The development of the element stifiness matrix is carried ‘out through the flexibility matrix of the element. The geometr cal nonlinear local effects are taken into account in analysis, for ‘each element, by the use of stability tifness functions ina beam- column approach, and updating at each load increment the length, flexural rigidity, and axial force of the element, following an ap- proach outlined in the next section A procedure is presented toen- force the force-point movement to remain on the plastic strength surface of a member, once a full plastified section develops. In this way, since the force-point movement remains on the plastic strength surface of a member, the yield criterions always satisfied after the full plastic strength of cross-sectionisreached, To perform, the nonlinear analysis of frame structures, in the majority of previ- ‘ous publications the loads are assumed to apply only at the nodes. Inthe present investigation, the loading due to the member lateral, Joads is transferred tothe nodes and included automatically in the analysis. Ths leads toa significant saving in imputing the member Joads, without the need to divide a member into several elements for simulation ofthese loads. The effect of semi-rigid connections is included inthe second-order elasto-plastic analysis using the zero- length rotational spring element approach. A procedure is presen- ted to asses this effect on element stiffness matrix and equivalent, nodal forces. The analytical model proposed in [10,11] is used to describe the non-linear behaviour of semi-rigid connections. Using an updated Lagrangian formulation, the global geometri- ‘al effects are considered updating the geometry of the structure ateach oad increment. Inthe present spread-ofplastcity analysis, the web plane vector approach proposed in 12] is effectively used to update the frames element coordinate. For normal building frameworks, the lateral response of the foot slab may be characterized by two translational and one rotational degrees of freedom located at the floor master node. In the present approach, the multi-freedom constraints, required by the rigid body floor model, are imposed by augmenting the beam- column element model, with the penalty elements. ‘The combined effects ofthese three nonlinearity sources are simulated into an object-oriented computer program automati- cally. This program was used to study the ultimate response of, several rigid and semi-rigid ste! frames thathas been studied pre- viously by other researchers [11,13,5.6]. The example computa- tions and the comparisons made have proved the effectiveness and ae Fa 8, Ban a ig. 1. Positive convention for free and displacement components of a beam cam element. time saving of the proposed method. The proposed software is pre- sented as an efficent, reliable tool ready to be implemented into design practice for advanced analysis and pushover analysis ofspa- tial frame structures 2, Mathematical formulation In this paper, the folowing assumptions are adopted in the formulation of analytical model: (1) Plane section remain plane after flexural deformation; warping and cross-section distortion are not considered; (2) flexural-torsional buckling do not acc (3) small strain but arbitrarily large displacements and rotations are considered; (4) nonlinearity is due to flexural joint flexibility, ‘material inelastcity, local and global geometrical change; (5) the connection element is of zero length. A frame member (Fig. 1)'s idealized as 2-noded 3D beam-column element with 12 degrees of freedom, Each ofthe element nodes has 6 DOF (3 displacements and 3 rotations). The elastic unloading effec is included in analysis, but hysteretic and softening effects associated with damage in building frames under severe loadings ae not taken into account. ‘The proposed approach is based on the most refined type of second ‘order inelastic analysis, the plastic zone analysis, and employs ‘modeling of structures with only one element per member, which reduces the number of degree of freedom involved and the com- putational time, 2.1. Elasto-plastic tangent siuess matrix and equivalent nodal loads 2.11. Elasto-plasticcros section analysis ‘The cross-section stifiness may be modeled by explicit inte- sration of stresses and strains over the cross-section area (eg. as micro model formulation) or through calibrated parametric equa- tions that represent force-generalized strain curvature response (eg. macro model formulation) inthe micro mode! formulation, the cross-sections ate subdivided into elemental areas and the states of strain and stress monitored explicitly foreach elemental area during the analysis, but the computational efforts are greatly enhanced [1,14]. As it was stated previously, in the presentelasto- plastic frame analysis approach, gradual plastification through the ‘cross-section subjected to combined action of axial force and bi- axial bending moments may be described by moment-curvature- {thrust (MN), and moment-axial deformation-thrust (M-e-N) analytical type curves that ate calibrated by numerical tests. The effect of axial forces on the plastic moments capacity of sec- tions is considered by standard strength interaction curves (5,15), ‘Two force-deformations type curves are taken into account in the present investigation, Ramberg-Osgood and modified Albermani force-strain curves, respectively 21.1.1. Ramberg-Osgood force-strain relationship. Gradual plasti- fication through the cross-section subjected to combined action of3018 (C6. Chore Engineering Sractres 31 (2008) 5016-3033 wie) Fig. 2. Ramberg-Osgood eatonships: (a Fore-train variation; b) lena xa) gies variation axial force and biaxial bending moments may be described by Ram- berg-Osgood moment-curvature-truss (M-~@ ~N), and moment- axial deformation-truss (M ~ e — N) type curves. In this case, the curvatures for each bending directions y and z and axial deforma- tions are expressed as (Fig, 2a): Py oS oe ane |"). te, 1 2-[ | Myo 5 alia [Mf]. (1b) zm | | IM ™ in which a and n are two shape parameters, M, is the full Plastification moment under combined bending and axial force for the given loading direction, N, is the full plastification axial force under combined biaxial bending moments, oy and so is the reence curvature and axl stain especie deine by 4 = #, foreach bending direction, &> = BE, Ep and Fag being the constant reference flexural and axial rigidity of the element respectively, Nand M axial force and bending moments with respect to strong (2) ot weak ais (y). 2.11.2. Proposed force-strain relationship. In order to take into account more explicitly the effect of residual stresses of the ‘cross-sections we propose another smooth force-strain curve to ‘model the graduak-plastification, The proposed relations modify the force-strain relations proposed by the Albermani 16] in order to simulate @ nonlinear variation of the cross-section axial and. flexural rigidity. The behaviour ofa cross-section subjected to the combined action of axial force and bending moments is expressed, ina set of equations relating the bending moments (axial forces) and the flexural (axial) rigidity for a certain value of axial force: Koy F< Fite ihn, ap), Fea SF Sf (2) ; «-fel-( where ‘Ky—represents the tangent axial or flexural rigidity about either the strong or weak flexural axis [Kg~represents the constant reference flexuralJaxial rigidity; Forepresents the cross-sectional generalized force (axial or bending moment with respect to either strong of weak cross- sectional axis); FyagyTepresents the generalized force (axial or bending moment) at the first yield; Fy-represents the full plastification force; [p-tepresents a strain-hardening parameter, for p hardening effect is neglected; represents a shape parameter (0 express the not Variations ofthe tangent flexural anal git oferosssecton or 1 we find the Albermani equations. ‘The shape parameters p and q are determined throughout ‘numerical calibrations. The effects of the combined action of axial force (N) and bi-axial bending moments (M,,M,) and as well the effect of residual stresses are taken into account in the expressions ofthe fist yielding Fj.) and full plastfcation(F,) generalized forces. For instance, the full plastfication bending moment is obtained from the full plasification surface equation fora given values of bending moments and axial force, whereas the first yielding bending moment is expressed as NM Mya) = (s 0-5 ) ee 8) ‘where oy represents the yield stress, o, represents the maximum value of residual stress, A the area of the cross-section, and Way represents the elastic section modulus ofthe crase-section with Tespect to strong or weak flexural axes. Similarly the first yielding arial force is expressed as Nye = (o- MMe ya “ Way Eq, (2) are plotted in Fig. 3. Note that the axial and flexural rigidity of the section reduces as the magnitude of the shape parameter « decreases. For a = 1, as is proposed by Albermani, a linear stifess degradation is considered. However, for steel cross- sections, where the effects of residual stresses are significantly ‘on the inelastic behaviour, a nonlinear stiffness degradation is required to represents adequately the partial plastifiation effects associated with bending and axial forces. Asitcan be seen in Fg. 3b, by varying the shape parameter a, smooth transitions from elastic to plastic behaviour can be modeled, the strain- Wasa Wa 2.1.1.3. Plastic interaction surfaces. The effect of axial forces on the plastic moments capacity of sections is considered by a standard strength interaction curves. Fr steel beam-columns with compact ‘wide-flange section the plastic interaction function proposed by Orbison [5] or AISC-LRFD [15] may be used. Orbison’s full Plastfication surface of cross-section is given by: P(N. My, M,) = 1.15p? em + mf + 3.67pm + 3p 4.65mi — (6)6. Chore /Egineering Sractures 31 (2009) 3016-5035, 2019 € Sy | yet c07 Pod) @ 0 ae 0 ‘Tangent rigidities ratio (axial/Mlexural) Fig. 3. Proposed foce-stain elaionships (a Force-strain variation; (b) Tangent (lexural or axa rigtes variation, AISC-LRFD's bilinear plastic interaction surface of the cross-section ‘may be expressed as: F(N.M.M,) p+ omy +$m—1=0, p> 2m, + 2m = fhtante sa 8 © fptmtm-1=0, p< 2m+2m where p= N/N? is ratio of the axial force to the axial yield strength, N? = Agy,Aisthe cross-section area, cy yield stress, my = M,/MM2, and m, = Mz/MQ, are ratios of the minor- and major- axis moments to the corresponding plastic moments, respectively, Myc) = Zpai\0y Zui) ate the plastic section moduli with respect to weak and strong axis, respectively, 1” = 0 represents, the full plastifcation of a cross-section, and in principle, a state of forces characterized by /" > 0 isnot allowed, 2.1.2 Flexibility based derivations. Member analysis ‘The procedure for the derivation of the instantaneous defor- rmational-stifness relations ofan element taking into account only, the elasto-plastic material behaviour using the force-strain rela- tionships is briefly shown. The response in torsion and shear is, assumed to be linearly elastic, consequently the nonlinear equa- tions for the force-strain relationship associated with torsion and shear are omitted in the following discussion for simplicity. The elasto-plastic behaviour of cross-section is modeled by the above mentioned force-strain relationships in conjunction with either Orbison’s (Eq, (5)) or AISC-LRFD (Eq. (6) failure surfaces. ‘The spread of inelastic zones within an element is captured con- sidering the variable section flexural and axial rigidity Ele) (€) and EA @) depending on the bending moments My, (€) and axial force 'N. When Ramberg-Osgood force-strain relationships are assumed to model the gradual plasification of cross-section, the instanta- neous flexural ad axial stiffness ofthe section is expressed by the equations: Mz) (E) = Ely) ) - Ay) (6) ” aN @) =EA@)- de) where Mya) (E) is the sectional bending moments about either strong (z) of weak ()) axes and dy) (€) and « (E) the curvatures, and axial deformation respectively nthe section é = x/Lofthe de- Formed element. From Eq, (1) the variable flexural rigidity ly) (@) and axial rigidity EA (E) can be expressed, for strong and weak axis, as (Fig. 2b): My yo) an (8) & or Ea fe 6) W) where he correction factors) () ate: 1 bro = bean |e joe 10) &e) 7 t+e-n-fats| in which Mp) (€) is the full plastification moment under com- ined bending and axial force forthe given loading direction, N () isthe full plasification axial force under combined biaxial bend ing moments. It is important to note that, Np 2) is not constant ‘long the member length because its obtained from the full plas- tification surface (Eg. (5) or (6) for agiven values of bending mo- ‘mentsM, (6) and M (6) respectively. Inthe same way Myy (6) and M2 (€) can be obtained ‘When the proposed stiffness degradation relations Eq (2) are used, the axial and flexural rigidities can be expressed asin Ea, (9), with the fllowing expressions forthe correction factors 1, My 6) < Myatses [ 7 ( Mya) @) = Myits y a-p fr = Maye) @ = Mery : Mysyey © My) @) = My) @) P,_ Mya) 6) > Mey @) 1, NS Niaa N= Nya) \" LO = [ - Cc nat) q ») Nyon @) 2N =) p> Ne) Fig. 4 shows the deformed shape of a 3D beam-column element in Tocal system attached to the initially straight center lin. The (1)020 CG. Chiren Engineering Sucures 31 (2008) 3016-5033 @ Fig 4 Beam-column element with rigid modes removes. element, with the rigid body modes removed, has six degrees of freedom (natural coordinates). They are two bending moments or Totations at each node, one axial force or displacement and one torsional moment or axial rotation, The element is subjected to distributed loads with linear variations. In this case, assuming a third degree parabola forthe bending moment variations along the >beam-column element, the element force fields bending moments and shear forces) are directly obtained as functions of the nodal element forces and distributed loads. With positive convention as shown in the Fig the intemal forces, about each principal axis(y or2),are given by the following expressions: Myo) (6) = My & ~ 1) + Myer ee EE hye 6 + ty 6-2)] dM yz) (E) _ Miyizy ++ Myce) (12) Fn 6) = Mig — Men Man ’ 1 (0-2) or Non-linear analysis by the stiffness method requires incremental loading, ie. the inelastic behaviour is approximated by a series of elastic analysis. The element incremental flexibility matrix 4, which relates the end displacements to the actions As, can be derived directiy from energetic principles. Assuming elastic behaviour within a load increment, and no coupling of axial and flexural responses at the section level, the increment ofthe strain energy AW can be wiritten as follows, including the additional shear and torsional deformations, Fig. 4: LP LP MO, bP Me, wok f Mae t L I sh mm*tal Bet taf He” Lp om LP PO 1p Re +h f cate [BO st PEO ay idas**ilaatil ae in which the bending moments and shear forces are given by the 54, (12) Using the second theorem of Castigliano, the relationship be- tween incremental deformations and efforts can be readily calcu- lated and partitioned as follows: aaw aN aaw uy | My Ny po ay | | aaw My} | By 8| | 2M | _ fiom Sx J. My] 4 155] ray On aaw 343) faysxay Me ia 8) | Mg Me] |e 64 | aaw ae OMy aaw OM, or in condensed form Au, = fy» As, +8, (15) ‘where f, represents the incremental flexibility matrix ofthe beam- column element without rigid body modes, and in which the ma- tices f, (i = 1, 2) have the following expressions in Box I: and 8, is a term resulting from loading actions: _ PPEGHV [aor 6 +) - dyin € -2)] oe [ GE) - daph (& = 3) = den [E — D? - + [ 2Glyn @) a fa : = [ ESM 649d = Bap wh SEiye) ©) * gaeanl 8 rank [E ah Wye @) ‘The integrals of the flexural coefficients ae calculated numerically using Eg (9) to express flexural and axial rigidity El, (), Els €), EA). To determine the bending moments My) each element, at each load increment, is divided ina few stations and the bend- ‘ng moments in each station are computed through a second order transfer matrix as functions of the nodal element forces (1), Thus, the response of a beam-column element is typically computed as the weighted sum of the response of a discrete ‘number of cross-sections, The cross-sections ae located at control points whose number and location depends on the numerical Integration scheme. In this work, the Gauss-Lobatto rule for element quadrature is adopted. Though this rule has a lower order ofaccuracy than customary Gauss-Legendre rule, ithas integration points at each end of the element, where the plastic deformation is important, and hence performs better in detecting yielding. To simplify the expressions ofthe flexibility coeficents the following. “cortection coefficients ae introduced: ial Tey =f Rd cum = [ape we [Toye “7 ie eae 2 "E0-8) 7b bn ® a 5a ®) which are numerically calculated by the Gauss-Lobatto rule. To produce the deformational-siffness relation, the Eq, (15) is, inverted, resulting the following deformational-stifiness equation: As, =I - Au, 4, (18) where the vector q, is the equivalent load vector, whereas I, represents the instantaneous element tifiness matrix of the beam column element without rigid body modes, determined by matrix inversion of the flexural matrix f,: roa" [Ape [bee go] 0 BY [ee tapes, 19) a =k5,6. Choro / gern Socures 31 (208) 3016-2022 2m L ° ° ° ae) i GW yl fae £G-0) 1-1 5 "eE-) 01 0 L a+ f f eased L me *tih BO bP 7 oy "EG Dep eel ats nf ° eo ae “h ‘hae ae | Lf Saal T@-1F +f ot bh, B®,” "Ly ab bh, @,” "Lh, A ® bh Teé-¥ is eel lf ae hp HO IhAG bh fo th AG ° Box - where Ik; and ky have the following expressions: where EAg ae ape 0 2 Mey = 39g [de (14e52)¢19 1 ~ BF Co) 2 =| 0 Be, Hep, (208) + diac (18256542) ~ 7Byem9)] ey 2Elg AE Pe 0 =a Ty. Myc = 5g [ax (7Brereae ~ 16909) 4, Bay g + tu BBaeoe) ~ Uso] ob, 2 ah and in which the correction coefficients used in the above equa- =|, Shy 0 (200) tions are given by: tegen? oo eae f £E-WE+ De ba ®) 7 inwhich the correction factors 2, eB, y are: 15 EE —17@—2) (25) =-B f fE-) Ey il 3(cm +22) on 2 Ley neds ayy = ee a(a- BO Tar hase en wm =— ee EO Da (con — 59) some) § y 2 be Geo ey yy cy = 2 f EE=VE-Day Ja * besye Sy + Gey) 7 Syea . Bye) = Aer taye) ~ Sys For common cases of uniformly distributed transversal loads Sy) = Cy) + Gaye + Gay and where the coefficients t, and t, account for transverse shear deformation effects in a beam-coumn element: pw th 128, 0 yE ae ‘where GA, and GA, are the constant shear rigidity along local y- axis and z-axis respectively. The resulting correction factors are sub unity and reflect the change in member rigidity as plastic zone develops along the member length. In the same way, a5 for curvature and axial stain, similar nonlinear equations for the {orce-strain relationships associated with torsion and shear may be used. As aforementioned, however, this approach is limited to consideration of three-dimensional plastic behaviour for biaxial bending and axial force, only. ‘The elasto-plastic equivalent nodal forces, transferred to the nodes, from the member loads, will not be constant during the analysis, and will be dependent on the variable flexural rig along the member according with the process of gradual formation, of plastic zones. For the case of linearly varying distributed loads, the equivalent nodal forces q, of the beam-column element without rigid body modes is computed according to Eq, (19) and can be written as: (22) = [0 my my me me off 2) (ave) = ay) = dye) the above relations become: aeypl? | 9(2eiyi) +121) C5) 12 Ze 2 chy Zx) mye) d 26) —— BR + Cyn) yen — 2A Cy ve Ze Be inwhic: aa f' G28) af = ca Toe oma pee a Since gradual platification through the cross-section subjected to combined action of axial force and biaxial bending moments is described by simplified force-strain relationships, the residual stress effects and explicit distribution of plastic zones through cross-sections cannot be accurately captured. Altematively, the cross-sections can be subdivided into elemental areas and the states of strain and stress monitored explicitly for each elemental area during the analysis. In this case the variable flexural and axial rigidity in any sections, are computed through basic equilibrium, compatibility and material nonlinear constitutive equations — 6, by an iterative process [14 In this case, the effects of residual stresses, shape geometry, and material strain hardening can be3022 6. Chorean Engineering tracts 31 (2009) 3016-3033 luded in the analysis but, the computational effort is greatly enhanced [14] ‘The resulting element stfiness matrix is a6 x 6 matrix. Injecting the rigid body modes, the stiffness matrix is pre- and ‘post multiplied by a transformation matrix to result inthe required 12 x 12 matrix. A similar treatment is applied for elasto-plastic equivalent nodal load vector following a detailed procedure given inSection 25.1 22, The second-order effects on element tangent siness mate ‘The geometrical nonlinear effects for each element are taken into account in the present analysis in a beam column approach, by the useof the stability stiffness functions and updating a each load increment the length, axial force and the flexural rigidity about of each principal axes of the element. This method minimizes the ‘modeling and solution time as generally oly one or two elements, are needed per member. The element stiffness matrix coefficients, kg are computed through the correction factors gy as: kg = kya 1,F=1,6 (28) ‘where ky are the stiffness coefficients of the elasto-plastic stifiness mattix of a 3D element with rigid modes removed. Explicitly, for compression, the correction factors gy for ky matrix (see also Eq. (21)) are: (29) where ¢(2) and s (v2) are the stability functions accounting for the effect of axial force on flexural stiffness, Hy and Hp are the stability functions accounting forthe effect of flexure on axial stiffness (17) and vjz = Ly/lc. N being the axial compression force in the element and EI?” and EI are the average flexural rigidity ofthe element: jo- [| Oar Hm Elo i; te: [ hat 300 the plastic hinge approach is ‘modeled, and for a = 1 and n = 20-50 the analysis corresponds to the plastic zone model, in which gradual yielding of cross- sections and strain-hardening effects are modeled, whereas using the proposed force-strain analytical curve plastic zone approach is ‘modeled when the parameters a = 2+ 3 and p = 0.001 are used. Inthe present approach, one element has been used to model each column and beam in all computational examples. 4.1, Example 1: Two-story plane frame with semi-rigid connections The accuracy of the proposed method in handling the effect cof nonlinear behaviour of semi-rigid connections is illustrated ‘clearly inthis example. The Liew’s two-story frame [11], member and material properties, and applied loading are shown in Fig. 11. Connections for the lower-story girder are C-3/4 having the following characteristics: Ry = 107, 548 kip-injrad, My = 173 kip-in and n_ = 0.80 and connections for the upper- story girder is C-1/2 having the following characteristics: Rp 205, 924 Kip-injrad, M, = 814 kip-in and n= 1.57. It was analyzed by Liew [10,11] using a refined plastic hinge analysis, ‘method implemented in the computer program PHINGE, Liew used 4 elements for each beam and the distributed Joad has been ‘modeled a5 concentrated loads. Inthe proposed approach a single ‘ “ gecraviy ' a * Be Pree ie vera i a oF ‘locc0 0.0510 0.0520 0.0580 0.90%0 0.0580 0.0060 0.0070, ‘ater espacement in] Fig. 12, Lateral oad-spacements traces, ‘element with seven integration points has been used to model the beam, The initial geometrical imperfections of the frame have been. ‘modeled directly inthe analysis. Gradual yielding of cross-sections ‘is modeled using the proposed force-strain relationship with the following parameters: ce = 2, p = 0.0001 and maximum residual stresses o, = 0.33a,. AISC-LRFD plastic interaction surface has been used. Comparable load-displacement curves for the top and bottom floors are shown in Fig. 12. As it can be seen, the proposed analysis results ae in close agreement with the refined plastic hinge analysis. The inelastic limit points are practically the same. Bening momen diagrams, Rexurl gies (1~ 2) dstbution, an load-defection curves for top flor care out by the NEFCAD analysis considering different types of connections, are shown in Figs. 13-15. The bending moment values are compared with those abjained by the refined plastic hinge analysis Asi can be seen the results agree fay well 42. Example 2: Two story space frame ‘The Vogel's two-story spatial frame [13], member and material properties, and applied loading ae shown in ig. 16. The accuracy of the proposed method in handling the effect of distributed Plasticity effect along the member length is clearly shown in this example, All he frame members are from American profiles WI 43 with Young's madulus £ = 210 x 10° Nem? and shear modulus G 80 x 10° N/cm*. The yield strength of all6, ChioreanEngnerng Structures 31 (2009) 3016-3033 3029 14x Gay as aint pie. 18 y 000 i Serer ie = ie — serene inet 8 2 ow cI + 3, fom 00. ‘0.0000 0.0010 0.0020 0.0080 0,0040 010080 0.9080 0.0070 0.6080 Lateral aisplacement Fig. 1, Load-deection curves for dierent connections. waste tee} ey HB is |“ a [NEFCAD, Semisgi joims (non nea) (NEFCAD, Seri joins (linear) ENEECAD, Rigi joins) {PHINGE} Fig 14, fending moment digramsat factored gravity fad (1.09) members is 6, = 24830 N/cm?, In the rigorous plastic zone solution [13] the members of the frame are divided in a several fiber-clements and the material nonlinearity is considered by the stress-strain relationship of the fiber. In the proposed approach <— 2In a Fig. 16. Twostory spacetime a single element with eleven integration points has been used to model the elements and gradual yielding of cross-sections has been modeled using the proposed force-strain relationship with ‘the following parameters: « = 2, p = 0.0001 and maximum residual stresses 0, = 0.330,. The effect of axial forces on the plastic moments capacity of sections is considered by Orbison’s plastic interaction surface. Comparable load~ 2 0 © Displacoment em) Fig. 24, Load-dsplacement cuveat¥ direction of node, Semi-rigd joints effects. The frame is analyzed for the combination of gravity loads = 4.8 kN/m? and wind loads = 0.96 KN/m?, acting in the y- direction. In Liew’s et al. [8] study, rigid floor diaphragm action is assumed in the global analysis, Liew employed one plastic- hinge beam-column element to model each steel column and four elements for each beam, and the limit load of the frame is reached at the load ratio of 1.031, In Ngo-Huu's etal. [4] study, the fiber plastic hinge concept has been used to predict the the second-order inelastic behaviour of the frame, and the inelastic limit point reported is 1.003. Inthe rigorous plastic zone solution of Jiang et al (6] each member of the frame is divided in an eight fiber-elements and the material nonlinearity is considered by the elastic-perfectly plastic stress-strain relationship of the fiber. The inelastic limit point reported in this case is 1.00. In present analysis one element with seven integration points (NG = 7) has been used to model each column and bearn and Ramberg-Osgood and proposed force strain relationships in conjunction with the Otbison's plastic interaction surface has been used, For n = 300 and a = 2 and p = 0.0001 respectively, the limiting strength of the frame is reached at load ratio of 1.005, wihereas, a load ratio ‘of 1.062 is obtained by the proposed analysis method for n = 30 and @ = 2and p = 0.01, respectively. The load-deflection ‘curves of nodes A and B at the top of the frame calculated by the previous researchers and the results obtained from the new Lo zg Wy o ae i NY ane rf wioreo oy A in Cre a wine? IN wiastes it I AY is ty = 203.6580 73.60 M wieaa (Column section) oe NC yf Mt an witasis Lk i fh EL fy Ww 14si59 in Ke wine ‘ ig. 25, Twenty story space frame: (a) Plan view: (b) etspective view. ‘method are shown in Fig. 26. Fig 27 shows the variation of the flexural rigidities (1 — $£) along the member lengths, a collapse, for the case when gradual plastification through the cross-sections hhas been modeled using the proposed force-strain relationships with the following parameters a = 2 and p = 0.0001. Running on a Pentium Il personal computer at 500 MHz the NEFCAD analysis ‘was performed in approximately 6 min which is almost 200 times shorter than the time taken to complete the same analysis with the fiber-element analysis computer program of fiang etal. [6].6. Choren / Engineering Series 3 (2008) 3016-3083 3033 © Per elonet aaj. Sergeta eroxn ramon Oxon nen —neFoxD renee ona 80 4 Mstenege anaes, Lene errhp toot Appa load factor an atet peat 0 700 160 20 Top stor detection nthe ¥-Sretion om) Fig. 26, Load displacement races Zar AAD AAAAA, LA ee Fig. 27. Fesural ites bution at ultimate load factor. 5. Conclusions A reliable and robust nonlinear inelastic analysis method for semi-rigid space frames has been developed. The proposed ‘model is based on the closed-form force-strain relationships which can approximately account for the gradual plastifica- tion of cross-section. The proposed analysis can practically ac- count forall key factors influencing steel space frame behaviour: gradual and distributed yielding associated with biaxial bend- ing and axial force, shear deformations, local and global sec- ond order effects, rigid diaphragm action, and nonlinear be- haviour of semi-rigid connections, with computational efficiency, and the necessary degree of accuracy. The proposed method em- ploys modelling of structures with only one element per phys- ical member which reduces the number of degree of freedom involved and the computational time. Furthermore, distributed Toads acting along the member length can be directly input into the analysis without the need to divide a member into several elements for simulation of these loads leading to a consistency in the linear and the nonlinear structural models. ‘The model has been verified by comparing the predicted re- sults with the established results available ftom the literature, ‘The studies show that the proposed analysis compares very well to exact plastic zone solution with much less computa- tional effort than a fiber-element analysis, Future work is en- visaged, considering the application of the proposed method to non-linear inelastic analysis of 3D reinforced concrete frameworks. Acknowledgement The writer gratefully acknowledges the support ftom Romanian National Authority for Scientific Research (ANCS and CNCSIS-Grant PNIL-IDEI No. 193/2008) fr this study. The writer would also like to thank Prof. GM, Barsan for his constructive comments and suggestions References [1] iorean CG, Barsan GM. Large defection dstibued plastic anasis of 3 ‘ee fametorks Comput @ Structures 2005-83(19-205 1555-71 [2] Kim SE Kim ¥, Choi SH: Nonlinear anajss of -D ste ames. hin Walled Struct 200: 38:445-61, [5] Kim SE, Choi SH, Practical advanced analysis fr semi space Frames. Jncernat Solids Sutures 2001;38(80-51)9111-3. [4] Nao-Huc, Kim, On Jk Nonlinear analyss of ste ames using ie plastic ‘ing concept. Eg Strut 20072(4)648-57 15) Orbison JG, McGuire W. Abel Vel surface applications in nontiner steel frame ali. Comput Metnods App Mech Engrg 1982:3%1)557-73, 16] Jang XM, Cen Hi Liew JR. Spread of pasty analysis of thre dimensional ‘tee ames J Consruce Stee Res 200238:183-212. (0) Lev. }¥m Chen Shanmugam NE, Chen WF. Iproved nonlinear plastic hinge analysis of space ramesrutues Eg Struct 2000 2: 1224-38 {8] Lew J¥m. Chen H. Shanmugam NE inelastic analysis of steel ames with ‘composite beams) strut Eng, ASCE 2001:127(2 184-202. {a] Attia MR Deierlein Co, Mecuize W. Spread of psi: Qas-pastic hinge approach.) Strt Br. ASCE 1994 20(3)-2451-73. [Ho] Lew i¥R White Di Chen WA. Lime states design of semi-rigid frames using advanced analysis Part Conection madeing and clsseation, Jans fas 1908326:1-77 [11] lew White Di, Chen WA, Ln states design of semi-rigid frames sing ‘vanced analysis: al Analysed design jConsract es 1993:28:29-57, [12] Basan, ELNimierM.Largedeiectlon spatial bucking thin-walled beams and ames. Engrg Mech, ASCE 157366 1259-81 [13] Voge U, Malet DH. EinfluderSchabweehhet be de Taglastbeechnung umber Systeme, Der Saban 198782717. [14 Baran GM, chicrean Co. Infuence of residual sess on the carying- capacity of See! famed sutures In: Dubin, vany Meters Namen Investigation Stabity and Duciiy of Sie! Structures Elsevier 198. panr-2 115] AISC-LRFD, American Istiute of Stee Construction, Specification for Structural Stel Buildings, lS, Chicago, 1994. [16] AlbermaniF, Kiporacal 8 Eisto-plstic lage deformation anlyiof thin ‘wale structres. Engrg Struct 1990; 12(1)28-36, [17] Ekhande Se, Sevappalam NM, Madogula KS, Stabily functions for tzee- mensional beamcolurs. Struct Eg, ASCE 1989;115(2)467-79, [18] Lui EM, Chen WE. fects of jit extbity on che behevir of sel frames Comput Structures 1987:265) 719-22. [19] Chu KH, Rampesreer RH Lage gelectionbucling of space frames. Strut Dis, ASCE 1972:ST12:2701-22 20] Yang YB, Yau JO, Lew L. Recent developments in geomexiclly nomtinear and postbutkling analysis of famed sutures. Appl Mech Rev 2003:584) 1, [21] Felipga' CA. Introduction to finite elements method, Deparment of ‘Aerospace Engineering Sciences, Univ. of Colerado at Souder, 2008, (itp: fe colorado sdulenneciglcsiourses EMA). [22] Lam We, rie CT. Arc-length metho for passing ln ois in sutra caleulatin.} Sct Eng. ASCE 1982;118(1) 169-85.