Sie sind auf Seite 1von 9
ISURFACE INTERFACE MODEL FOR ANALYSIS or Masonry STRUCTURES Mui’ By Paulo B. Lourengo' and Jan G. Rots* Agsraact: The performance of an interface lastoplastic constitutive model for the analysis of unreinforced ‘masonry structures is evaluated, Both masonry components are discretized aiming ata rational unit-joint model able to describe cracking, slip, and crushing ofthe material. The model is formulated in the spint of softening plasticity for tension, shear and compression, with consistent treatment of the intersections defined by these ‘modes, The numerical implementation is based on modern algorithmic concepts such as local and global Newton- Raphson methods, implicit integration ofthe rate equations and consistent tangent stiffness matrices. The pa- rameters necessary to define the model are derived from microexperiments in units, joints, and small masonry samples, The model i used to analyze masonry shear-walls and is capable of predicting the experimental collapse Toad and behaviour accurately. Delailed comparisons between experimental and numerical results permit a clear understanding of the walls structural behavior, flow of internal forces and redistribution of stresses both in the ‘re- and post peak regime, INTRODUCTION ‘Masonry is a composite material made of units, eg, clay bricks or concrete blocks, and mortar. The large number of influence factors, such a6 anisotropy of units, dimension of ‘units, joint width, material properties of the units and mortar, arrangement of bed as well as head joints and quality of work= ‘manship, make the simulation of masonry structures extremely difficult. Moreover, an accurate masonry description needs complete set of experimental data Recently, the masonry research community began to show interest in sophisticated numerical models as'an opposition to the prevailing tradition of rules of thumb or empirical for- ‘mulae, These were, in general, acquired with years of practice ‘and experimental esting. The present article follows naturally from a rational attempt to achieve a more fundamental insight in the behavior of masonry and to permit more competitive tse of masonry. Here, attention will be given to micromodels, ‘wherein the (wo masonry components are modelled separately. Interface elements are used as potential crack, slip, or crushing planes. A new interface cap model, recently developed by the ‘writers see Lourengo et al. (1994), in. modern plasticity con- cepts, is able to capture all masonry failure mechanisms. The ‘model includes a tension cutoff for Mode I failure, a Coulomb fiction envelope for Mode Ul failure and a cap model for com- pressive failure. In addition, interface elements are considered to model potential cracks in the units. The implementation of the model is briefly reviewed here and includes proper han- dling of the nonsmooth corners, use of a local Newton-Raph- son fo solve the implicit Euler backward retum mapping and consistent tangent moduli. Such set of procedures leads to a robust and accurate numerical algorithm that preserves qua- dratic convergence. The parameters nevessary 10 define the ‘model are derived from carefully displacement controlled mi- ‘roexperiments, The main difference between the present stal- egy and the research carried out by other authors, see Lour- fengo and Rots (1993) for references and Lofti and Shing (1994), lies in the assumption that all the inelastic phenomena ‘occur in the interface elements, This assumption leads to a robust type of modelling, capable of following the complete Toad path of a structure until total degradation of stiffness. To the Knowledge of the authors most previous numerical mi- croanalyses (and also most of the experimental results) are limited to the structural prepeak regime. Computations beyond the limit load down toa possibly lower residual load are, how- ever, needed to assess the safety of the structure CAP MODEL FOR MASONRY ‘An accurate model for masonry has to include the basic types of mechanisms that characterize the material: (a) Crack- ing in the joints; (b) sliding along a bed or head joint at low values of normal stress; (c) cracking of the masonry units in direct tension; (@) diagonal tension cracking of masonry units at values of normal stress sufficient to develop friction in Joints; and (e) splitting of units in tension as a result of mortar lilatancy at high values of normal stress (sce Fig. 1). It is Lirtin vane @ "Rex. Engr, Dei Univ, of Technol. Fas. of Civ, Engg, Def, The [Netherands: presently at bept. of Ch. Engre, School of Engrg, Uni. of Minho, Azurém. 4860 Guimaris, Poaugal 2c Res Fellow, Dollt Uni. of Technol/TNO Build and Conse: Res, PO. Box 49,2600 AA Delf, The Netherlands ‘Note Associate Editor: Robert ¥. Liane. Disession open until De- camber 1, 1997. To extend the closing date one mont, a writen request ‘ust be fled with the ASCE Manager of Jouals. Te manuscit for this paper wat submited for review and pone polation on March 1995. Tis paper Is part ofthe Journal of Engineering Mechanles, Nei, 123, No.7, July, 1897. ASCE, ISSN 0738-93901970007-0600— 68188.00 + 850 pes page. Paper No. 1040 {660/ SOURNAL OF ENGINEERING MECHANICS / JULY 1997 dbibidad bbdbeddd hott ~ TT wa rat FIG. 1. Fallure Mec racking; (b) Joint Si Diagonal Tension Crack: nisms of Masonry: (8) Joint Tension ) Unt Direct Tension Crack; (2) Unit 3) Masonry Crushing clear from the described phenomena that (a,b) are joint mech- anisms, (¢) is a brick mechanism and (d, e) are combined ‘mechanisms involving bricks and joints. Atlempis to use in- terfaces for the modelling of masonry were carried out in the last decade with reasonably simple models and without in- cluding all the above mechanisms, see Lourengo and Rots (1993) for references, ‘The question remains of how to consider all phenomena in the model. The approach followed here is to concentrate all the damage in the relatively weak joints and, if necessary, in ‘potential pure tension cracks in the units placed vertically in the middle of each unit (see Fig. 2). The joint interface yield surface has then to include all the mechanisms referred above ‘except uniaxial tensile cracking of the unit. Inclusion of the first two mechanisms (ensile and shear failure of the joint) has been pursued before but the cap model here set forth is novel. By limiting the compression/shear stress combinations the compressive damage can be included in the model as well ‘as the combined mortar shear failure and unit diagonal tension failure, The new interface cap model to be developed is used ‘ata micro-level, Remarkably, experiments cartied out on, €-., ‘shear-walls (Mann and Muller 1982) result in similar macro- level yield surfaces (see Fig. 3). “The formulation of isoparametric interface elements has ‘evolved rather far and the reader is referred to Hohberg (1992) for an exhaustive discussion, One of the most important issues is the selection of an appropriate integration scheme, ¢.8., 4 | _ ef | he F Potential crack —_f Interface elements in the unit Goins) ero thickness 2 byt he Continuum elements (units) FIG. 2. Suggested Modeling Strategy [Units (u} Which Are Ex: panded in Both Directions by Mortar Thickness, Are Modeled ‘with Continuum Elemente; Mortar Joints (m) and Potente! ‘Cracks n Unita Are Modeled with Zero-Thickness Interface Ele- ments] lumped or Lobatto scheme, as Gauss integration was reported to lead to oscillations of stresses when high stifTnesses are used, see Rots (1988). An interface element allows discont nuities in the displacement field and its behaviour is described in terms of a relation between the tractions t and relative dis- placements w across the interface. In the present paper such ‘quantities will be denoted as a, generalized stress, and e, gen- eralized strain. In this case the elastic constitutive relation be- tween stresses and strains is given as usual by o=De a ag (ky ky), @ = (a, 2)" and © = (a 4)", Where m and s = normal and shear components, re- spectively. The terms inthe elastic sifness matrix can be ob- tained from the propertios of both matonry components and the thickness of the joint as For a 2D configuration D a a ert WG. GD ° where E, and £. = Young's moduli; G, and Gx = shear moduli, fespectively, for unit and mortar and fy = thickness of joint ‘The stiffness values obtained from these formulae do not cor- respond to a penalty approach, which means that overlap of neighboring units subjected to compression will become visi- ble already in the elastic regime. This feature is, however, intrinsic to the interface elements formulation and is indepen- dent from the values of normal sifness, even if itis clear that ‘the amount of penetration will be higher with decreasing in- terface stiffness. The interface model includes a compressive cap where the complete inelastic behaviour of masonry in ‘compression is lumped. This is a phenomenological represen- tation of masonry erushing because the failure process in com- pression is, in reality, explained by the microstructure of units ‘and mortar and the interaction between them, In the model the failure mechanism is represented in such a way thatthe global stress-displacement diagram is captured and corresponds to ‘one unit Iterally imploding over the other. As can be gathered from some confusing statements in the literature, i takes some time to get used to this zero or “negative'’ volume (in com pressed slate) of interface elements (Hohberg 1992). MULTISURFACE PLASTICITY ‘The elastic domain is bounded by a composite yield surface that includes tension, shear and compression failure with soft ‘ening (sce Fig. 4). For the implementation ofthis model single Surface and multisurface plasticity theory is reviewed below in modern concepts, including consistent tangent operators and correct handling of the corners. For multisurface plasticity the form of the elastic domain is defined by each yield function f; < 0. Loading/unloading can be conveniently established in standard Kuhn- Tucker form by means of the conditions AZOKS0 and AK=0 @ Grashing “Tensile faire ‘Shear fale ofthe ofthe ofthe sr panel = pesicins ai ets “Kestraetce || * FIG. 3, Fallure Surface for Shear Walls (Mann and Miller 1982) FIG. 4, Proposed Interface Cap Model JOURNAL OF ENGINEERING MECHANICS / JULY 1997 /661 where A, = plastic multiplier rate. Here it will be assumed that the yield functions are of the form Ko.) = Po) + Vie) ® Where scalar x = amount of hardening or softening: and @, “Y, = generic functions. The usual elastoplatic equations for single surface plasticity hold: the total strain rate € is decom- pposed into an elastic component é* and a plastic component enete © the elastic strain rate is related to the stress rate by the elastic ‘constitutive matrix D as o=De © tnd the assumption of nonassociated plasticity yields mes enik o where g = plastic potential. The scalar « introduced before reads, in case of strain hardening/softening Re V@e ® For any comer of the proposed model two yield surfaces are active and the previous equations must be appropriately stated for multisurface plasticity. According to Koiter's generaliza- tion (Koiter 1953), (7) reads Pee eee © and, assuming a form of explicit coupling for the hardening softening parameters, we obtain Rem ek + Gok and Em Gok #1) where fr: and {sy are coupling terms. Integration of Elastoplastic Equations ‘The return mapping algorithm is strain driven and basically consists of two steps, the calculation of the elastic trial stress, also called the elastic predictor, and the return mapping to the yield surface, ie, the plastic corrector. We will use the implicit Buler backward integration scheme due to its unconditional stability (Ortiz and Popov 1988) and accuracy (de Borst and Feenstra 1990). For single surface plasticity, manipulation of ()-(7) for finite increments and application of an implicit Euler backward integration scheme gives the update of the stress values as r+ BO ye: = 8, + Dde iy) = 0, + Dideyn: ~ AF ay ‘This equation can be recast as ct 28 som — an Ze | a2 with oft =o, + DAe,,,. The subscript m + 1 means that all the quantities are evaluated at the final stage. For simplicity it will be, in general, dropped in the derivatives. (12) repre- fents a nonlinear system of equations with the stress tensor ‘components and one scalar (usually AA,..) as unknowns. En- forcing the satisfaction of the yield condition atthe final con- verged stage results in the necessary additional equation a “The constitutive nonlinear equations [(12) and (13)} atthe in- tegration point level can be solved locally with a regular New- (52/ JOURNAL OF ENGINEERING MECHANICS / JULY 1997 ton-Raphson method. In most cases, and for the yield functions used in the present paper, (12) can be solved in order to obtain explicitly the updated stress value as a function of the updated plastic multiplier aes = Guss(Bes) ay Furthermore, inserting (7) in (8) yields for finite increments Arses = Beers: Mba) as, Substitution of these two equations in the yield function, ef (13), leads to a nonlinear equation in one variable, namely Breer! foer(Bdens) = 0. This constitutive equation is solved here witha local Newton method. The derivative of f(D) ‘with respect to AA,.1, which is needed inthis procedure, reads after some manipulation xe max ‘where the modified yield surface gradient and the hardening ‘modulus h are given by a oe a6) a le ae aan |,., For multsurface plasticity a similar procedure applies and (12) reads now as) We shall equally assume that this equation can be solved ex plicily as rrs(Brnets Bhan) ay) (10) yields for fini HEM. Mdraets Mages EMO AMseete Mane) 20) Substitution of these two equations in the active yield func- tions leads to a nonlinear system of equations with a set of scalar unknowns, namely AX,.., and Adress Finwi(Adsarts Manet) #05 fare Dinrts Adan) =O (21) ‘This nonlinear constitutive system of equations is solved by a local Newton-Raphson iterative method. The Jacobian neces- sary for the Newton-Raphson scheme reads 1 2e 4 hoe "aby * ax BBs 7 e a, “The problem that remains is how to determine the set of active yield functions, see Simo et al. (1988). In the present article @ wial-and-error procedure to solve the return mapping ‘is used. It will be assumed that the set of initial active yield functions is the one defined by Simo et al. (1988) (f?™ = 0) If, after the retum mapping is completed, any AN}, < 0 oF fier > 0 is found, the number of active yield surfaces is ad- jisted accordingly and the return mapping is restarted. During this process one restart may be required, or, rarely and only for larger increments, more restarts may be required before the correct number of active yield surfaces is obtained. Evaluation of Tangent Operator ‘A tangent operator consistent with the integration algorithm is needed to obtain quadratic convergence and a robust global Newton-Raphson method. For single surface plasticity, differ- entiation of the update equations ((12) and (15)} and the con- sistency condition (df... = 0) yield after algebraic manipula- tion the consistent tangent operator D* as, see Lourengo etal (1994) 1 og ae ie pra —_ roa vee a ao where the modifed compliance mattis His given by Se BaD + a 2 es ‘The above expression for D” is particularly interesting as it clearly shows that @ nonsymmetric tangent operator will be oblained even in the case of associated plasticity if the hard- ‘ening parameter update is nota linear function of the plastic ‘multiplier update, ie. if Avg # eMAye, Where cis aconstant. ‘Similarly, for mulisurface plasticity, differentiation of the Update equations (18) and (20) and the consistent conditions (Gres =0 and snes = 0) yield after algebraic manipulation the consistent tangent operator B® as, see Lourengo etl (1998) pre] eee vee IU WH 26) hee the moe comptiane matic H eds now nD" 4 Bde, 2 + Ahaee Ht Wed" + dhe Zi 6 then an the mac U ren =) 3828 we [22] = the mauie V reads veo a o snd he mate E ead a seat ma] oy FORMULATION OF MODEL, ‘Cap models originated in the field of soil mechanics. The introduction of a spherical cap for the Drucker-Prager model was firstly made by Drucker etal, (1957) to describe plastic ‘compaction and to enhance the behaviour in hydrostatic com pression. Since then various models were suggested, e.., the well-known Cam-clay model (Roscoe and Burland 1968). Re- ccently the numerical algorithm has been revised by Hofstetter et al, (1993) with the use of unconditionally stable closest point projection return mappings, tangent operators consistent ‘with the integration algorithm and proper handling of the cor- ners. Previous cap models have been, in general, limited to associated plasticity and hardening/softening of the cap while the other yield surfaces remain in ideal plasticity. FFor the application envisaged here the behavior found ex- perimental leads to a more complex model. Most masonry Joints have extremely low dilatancy and the mode! must be Formulated in the wontext of nonassociated plasty. Also softening behavior should be included for all modes of the composite yield surface. The model presented inthis article is ‘of general application but the Formulation is shown inthe char- acteristic interface (6, 1)-space. The reader can assume the usual (h,, VJ,)-space formulation for regular continuum ele- ments Tension Mode For the tension mode, exponential softening on the tensile strength is assumed according to the mode I experiments by van der Pluijm (1992) (see Fig. 5). In this figure, the shaded area represents the envelope of three tests. The yield function reads. Be. «) =o She) on where the yield value f reads =foen (Sx) tetoew (Be om where fo = tensile strength of joint or, more precisely, of unit- ‘mortar interface (Which is the weakest link); and G} = mode 1 fracture energy. An associated flow and a strain softening hypothesis are considered, Assuming that only the normal plastic relative displacement controls the softening behavior, () yields ay Shear Mode For the Coulomb friction mode, the yield function reads For 3) = [2] + ota (43) — of) es Based on microshear experiments by van der Pluijm (1993), the yield values and tan ¢ are assumed = a a : oS ee | | on | a — re ‘mental Results trom van der Pluljm (1982) (f,= 0.30 MPa; @ = ‘Oo12 Nem) JOURNAL OF ENGINEERING MECHANICS / JULY 1967 /683 = DMPA 20 Sa] = Experiments | — Numerical 1LoMPs 0s __o= 05 MPa 2 -2.1MPa | 00, —_- 00 a4 06 0810 ‘Shear stip =, fmm) FIG. 6. Sheer Behavior of Present Model versus Experimental Fesults by van der Pluljm (1999) for Different Confinement Lov- els (c,= 0.87 MPa; tan Gy=1.07;tan 6,=0.73; G; «0.058 0-130 (Wimp) Where ¢) = intial cohesion of joint; tan dy = initial friction angle; tan ¢ = residual friction angle; and G? = mode Tl frac- ture energy. Note that exponential sofiening is assumed forthe cohesion (see Fig. 6) and that, for simplicity, the softening of the friction angle is taken proportional to the softening of the cohesion. A nonassociated plastic potential g: galt tong on ‘where a dilatancy angle Y and a strain softening hypothesis are considered. In the computational implementation of the ‘model, the dilatancy angle is considered as a function of the plastic relative shear displacement and the normal confining pressure. Under increasing values of these two quantities, the dilatancy angle tends to zero. This is physically realistic due to the micro granular structure of mortar and is also confirmed in the experiments, see van der Pluijm (1993). Assuming that only the shear plastic relative displacement controls the soft- ening behavior, (8) yields he co) Permanent Coupling of Tension/Shear Mode ‘The lesion and Colom friton mates ate coupled with isovopicsofeing. This mean thatthe peconge St sen ingon the caesion nase toe the tame a the tele Seng and or exponen softening (3) and 8) mat Be Zh 0 ‘These relations can be directly obtained from (32) and (38). It is noted that (39) according to Kuhn-Tucker conditions, ef. {@), indicate that the shear and tension modes represent, in reality, only one discontinuous yield surface and not two. The ‘wo scalars shown as a measure of softening, x, and xa, ep- resent a single scalar. In the computational code however, the ‘measure of shear softening and tensile softening are updated independently to posterior check which type of softening has ‘occurred, During postprocessing of the results, this is relevant for a good understanding of the structural behaviour, Cap Mode For the cap mode, an ellipsoid interface model is used. The yield condition is now given by filer, 6) = Cua? + Cy + Car PO) (40) 1664/ JOURNAL OF ENGINEERING MECHANICS / JULY 1997 where Cou Cun and C, = set of material parameters; and o = yield valve. For the hardening/softening behavior the law shown in Fig. 7 was considered. Note that the curved diagram leads to a C' continuous o ~ € relationship (see Fig. 8). The approach shown in Fig. 7 can be used for interface elements because a direct relation between stresses and displacements is established. The energy under the curve can be related to a “compressive fracture energy.” Using matrix notation, (40) can be rewritten as Ke, x) = V20"Pe + p'e ~ 3°06) ap where P = diag {2Cm 2C,) and p = (Cy, 0)". ‘An associated flow and a strain hardening/softening hypoth- esis are considered. This yields, upon substitution in (8) & = hViPo + po + p) (42) Comers ‘The theory for multisurface plasticity is described previous section. Only additional remarks about the comers Will be detailed here. “ sta epee(a ga | 58 Experiments | — Numerical oo! ‘800 “0002 000s 0005 .008 Masonry compressive stain =e FIG. 8. Model versus Experimental Compressive Behavior [Compressive Strength = 20.6 Ma; Inelastic Parameters trom [Atkingon and Yan (1990) hie aa RF Meanie FIG, 8. Softening ot Tension/Shear Mode For the tension/shear corer (39) yields a stronger penali- zation than adjacent single surface modes as shown in Fig. 9. ‘This figure shows three different possible ial stress states close t0 the corner and the respective return mapping. It is further assumed that the yield surface at stage n + 1 is the same for the three cases (this is clearly wrong but yields a ‘more legible picture), the stifiness matrix is equal to the iden tity matrix and the dilatancy equals zero. Then, for the shear ‘and tension modes, the amount of softening is measured by the distance AAway = [o"™" — Oey,| In the comer, acoording to (39), the two values will be added representing @ nonac- ccepiable penalty. For this reason a quadratic combination is assumed, reading laur + (GBs): V GF ho GP fe Ties) ay V(G4s) + 00 «) Note tat thee equations yield atonal terms forthe Jao- tian ued forth bal Neston Raphson method, cf (2) For the sheacap comer the yield sures ae assumed to Be om ‘ounle, NUMERICAL EXAMPLES ‘Tests on masonry shear walls were carried out by Ratimak- ers and Vermeltfoort (1992) and Vermeltfoort and Raijmmakers (1993). Two types of walls were considered, cither with an opening in the center, hereby designated by “hollow wall,” ‘or a complete panel, hereby designated by “solid wall.” These tess are particularly suitable for the appraisal of the model Se nf s | / ar TABLE 1. Properties for Potential he he te @ a @ 8 @ Toxig | ox 20 O08 ‘Wm | ___ Nia? Nim’ Nin, “ @ @. 6. 765709 ais es | 36 50:36 Ni a ‘Nin ‘Nim? performance as quite comprehensive experimental data was recorded, Moreover, the parameters necessary to characterize the material model are available from companion microexper- ‘The specimen consists of a pier with a widtivheight ratio of| fone (990 X 1000 mm’), built up with 18 courses (16 active courses and 2 courses clamped in stee! beams) of Joosten solid clay bricks (dimensions 204 98 X 50 mm’) and 10-mm~ thick mortar (:2:9, cement:lime:sand by volume). The piers were subjected to a vertical uniformly distributed load p before 1 horizontal load was monotonically increased under top dis- placement control d unt failure (ee Fig. 10) For the numerical analyses, bricks are represented by plane stress continuuin elements (8-noded) while line interface ele- ments (6-noded) are adopted for joints and, if necessary, for the potential vertical cracks in the middle of the brick. Bach brick is modeled with 4 2 elements. For the joints, the ‘composite yield function described is adopted and, for the po- tential cracks in the bricks, a simple Mode I cracking model ‘with exponential tensile softening and immediate drop to zero of the shear stress after initiation of the crack is assumed. The ‘material data are obtained from the micro tension, compression and shear tests given before and are given in Tables 1-3. ‘When more than one value is given in the same column this ‘means that different values are used for the walls with an in- itial vertical load p of, respectively, 0.30, 1.21, and 2.12 Nim’, according to the microtests in samples obtained from each wall, The hardening/softening law for the cap is the same as defined for the hollow walls. The hardening/softening law for the cap mode is defined by the set (0, x}, = {(Ja/3, 0.0); (Cw 0.09): (ful2, 049); (fl, +2)). Hollow Walls ‘Two tests were sucessfully carried out with an initial ver tical load p of 0.30 N/mm’. An opening placed in the middle ofthe wal defines two small, relatively weak piers and forees the compressive strut to spread through both sides ofthe open- ing leading to the ultimate cracking patterns shown in Fig. 11 For the numerical simulation potential racks inthe bricks axe not modelled asthe experimental results indicated, in gene eral, no visible cracks. Mierocracking is likely to oecur but it ‘was assumed irelevant for this analysis. “The comparison between the numerical and experimental Joad-displacement diagrams is shown in Fig. 12 The results agree wel. The calculated failure load is equal to the failue Toad of one of the tests whereas the other test features a col: lapse load 20% lower than the calculated value. A good im- pression about the behavior of the model is also obtained be- Sause the calculated softening sitfaess of the structure agrees well with the softening sifness of one ofthe test. This in- dicates thatthe same failure mechanism is predicted. It is how ‘ever recognized that the behavior obtained numerically is more ‘rile than the experimental observations. Globally, the analysis captres wel the experimental bchav- four of the walls, as illustrated in Fig. 13 (compare with Fig. 11, In his figure the word “"damage" is used forthe equiv lent plastic strain of each mode of the cap model. Note that the cap “damage” is ony shown in the softening range. Ini- tally, two diagonal cracks arise from the comers ofthe open TABLE 3._Inelastc Properties for Joints “Tension ‘Shear cap, ie ai ten é. | tang, | tang a te Gm» T Gn | Gy oH) @ w | o | © a @, ® | oo | ai 025; 0.16, aoe ‘01s | 075 | 00 [0.128 0080; oas0] 105. 11.5 10 | 90 | 00 Neon | Nam leet |e ‘Nm ‘Nan 2 | ne JOURNAL OF ENGINEERING MECHANICS / JULY 1997 /685 0.30 Nn p<0.m Nea iy ill 4 « » FIG. 11, Experimental Fallure Patterns for Hollow Walle [Raijmakers (1992); Vermeltfoort (1982)), Hezntal force RN Horizotal displacement d{msn] 12, Load-Displacement Diagrams for Hollow Wall ing. These cracks are accompanied by less evident horizontal ‘racks in the top and bottom of the small piers as well as one horizontal crack inthe bottom of the wall (See Fig. 13a). Under increasing deformation, the diagonal cracks that arose intially ‘cannot progress to the compressed toes and two additional diagonal cracks start to open (see Fig. 13b). These become ‘gradually predominant and, when the compressed toes atthe bottom and top of the wall begin to crush, the previous di- ‘agonal cracks become inactive, Finally, a collapse mechanism is formed with the failure of the small piers in bending (crush ing of the compressed toes and extended horizontal tension cracks) and the wall behaves similarly to four rigid blocks connected by the hinges shown in Fig. 13c. ‘Remarkably itis possible to trace the complete path of the test without numerical difficulties, Full quadratic convergence was found during the entire prepeak and postpeak regimes. [Note also that the snap throughs found in the numerical anal- ysis are converged states due t0 the opening of new diagonal Cracks and closing of previous diagonal cracks. These snaps ‘were traced with arc-length control over opening, shearing or crushing of selected interfaces. It is noted that, under the same loading conditions, the re- sponses of the two tests, in terms of load-displacement dia- rams, are different (see Fig. 12), and correspond to different crack patterns (see Fig. 11). Itis believed that this is due to the combination of the scatter in the material properties and the small number of units that constitute the wall. It seems that different crack locations can be triggered for close failure loads, ie, within the observed 20% range. The lower collapse load and more ductile behavior observed for one test corre- spond to an early sliding failure mode, involving the part of the wall above the opening, Fig. 11 (righ). This is in oppo- sition with the bending failure mode (with opening ofthe top- Fight comer) of the same part ofthe wall featured by the other test, Fig. 11 (lef), A possible approach to reproduce an ob- served failure mode is to penalize or strengthen a certain num- (666/ JOURNAL OF ENGINEERING MECHANICS / JULY 1997 TOTAL DISPLACEMENTS Damacte = Sear Tension cn ae = | Ex] =| ee eo » ° Horizontal Displacement d(Hol- (@)d=20mm; (e) d=25.0mm ber of joints in order to obtain exactly the crack pattern ob- served in a specific test but such inverse fiting is outside the scope of the present paper. Mesh insensitivity ‘The mesh objectivity of the composite interface model is checked here by an additional numerical analysis in. which teach brick is modelled with 8 X-4 elements. This leads to 4 times as many continuum elements and 2 times as many in- terface elements. As shown in Fig. 14, the results are almost identical, regarding the collapse load and the entire load-dis- placement diagram used to characterize the behavior of the structure. The interface cap model can be therefore considered to be mech insensitive. This is another advantage of the inter- face approach when compared to crack band type models where mesh objectivity is a debated issue, Solid Walls ‘Tests were carried out for different initial vertical loads p of 0.30, 1.21 and 2.12 N/mm. The crack patterns forthe dif- ferent walls tested are shown in Fig. 15. The behavior of the ‘walls with a low initial vertical load is characterized by init 4x2 elements per brick — 8xéelemensperbrick | 00 100 7200 300 Horizontal displacement df] FIG. 14. Loed-Dieplacement Diagram for Two Ditferont Mesh Diecretizations: oso fae} ‘0S orzo placement ft] FIG. 16. Load-Displacement Diagrams for Solld Walls, TOTAL PISPLACEMENTS. DAMAGE, Sen Tenion Cop Experimental Failure Patterns for Sol Wa S0KN; (8) P=T20KN; (e) P= 210 KN horizontal tension cracks that develop at the bottom and top ‘of the wall, The collapse forall the walls is however charac- terized by a diagonal shear crack including cracks in the bricks and by crushing of the compressed toes. From the failure patterns it is evident that cracks in the bricks need to be modeled. This is confirmed in Lourengo and Rots (1993). In that paper, the potential cracks in the bricks ‘were not included, which fed to an overestimation of the col- lapse load and a response much stiffer than observed in the experiments. The comparison between numerical and experi- ‘mental load-displacement diagrams is shown in Fig. 16. The experimen behavior is reproduced satisfactorily and the col- Tapse load can be estimated within a 139 range of the ex: perimental values. The sudden load drops are due to cracking ina single integration point of the potential cracks in a brick and opening of each complete crack across one brick (note ‘that, when a potential vertical crack is almost complete across fone brick, relutively high shear stresses are set (0 zero in a single step. In the former case a mesh refinement is likely to ‘smooth the curve but, in the latter, a snap through or snap back will, as in case of the hollow walls, most certainly be found, Here, these points were sporadically “jumped over”” in the analysis by the use of a secant stiffness instead ofa tangent stiffness, only for the discrete cracking criterion in the bricks. FIG. 17. Numerical Results st Horizontal Displacement (Bolla Walls): (a) d= 2.0 mm; (o) d= 4.0mm For the interface cap model, the consistent tangent operator ‘was used throughout the analysis. “The wall with an inital vertical load of 0.30 N/mm’ was chosen for further discussion because two specimens were built. The behavior of the wall is well captured by the model 4s illustrated in Fig. 17. Initially, two horizontal tension cracks develop at the bottom and the top of the wall. A stepped di- ‘agonal crack through head and bed joints immediately follows (Gee Fig. 17a). This crack starts in the middle of the wall and Js accompanied by initiation of cracks in the bricks. Under increasing deformation, the crack progresses in the direction of the supports and, finally, a collapse mechanism is formed with crushing of the compressed toes and a complete diagonal crack through joints and bricks (see Fig. 17b) CONCLUSIONS ‘The present paper contains some computational results of a research program aiming at development of rational and robust humerical techaiques for the analysis of masonry structures. At this first stage a micromodeling strategy, in which the units and joints are separately discretized, is considered, JOURNAL OF ENGINEERING MECHANICS / JULY 1997 /687 ‘An interface cap model that includes all the possible failure ‘mechanisms of masonry structures has been developed. The composite yield surface is implemented in an accurate and robust algorithm. The fully implicit Euler backward return ‘mapping is solved with a Newton-Raphson technique for all ‘modes of the cap model, including the corners. Tangent op- erators, consistent with the integration algorithm are derived forall the modes of the cap model. Isotropic coupling is as- sumed between tensile softening and decohesion. Application of the model to experiments on shear-walls shows good agree- tment and provides additional knowledge about the behavior of masonry structures. The model is able to reproduce the ‘complete path of the structures until total degradation without hhumerical difficulties. Its further noted that within the toler- tance adopted for the satisfaction of the return mapping (107 Of the initial violation of the yield criteria), no inaccurate ppoints are found during the large scale analyses. The global convergence for the analyses using exclusively the interface ‘cap model detailed in this article also behaves extremely well (ypically 3 to 4 global iterations are needed to reach an energy norm of 10". Finaly, itis shown that the modeling strategy adopted is mesh insensitive. ‘The writers believe that the importance of numerical mod- cling for supporting safe and economic designs of masonry structures is already evident from the present article. For a thorough interpretation of the behavior ofthe shear walls pre- sented herein the reader is referred to Lourengo and Rots (1994), ACKNOWLEDGMENTS “The models reported in the present paper have been implemented in the DIANA fie element code, which #4 repistered vademark of TNO Building and Construction Research, The esearch x soported fnacilly by the Netherlands. Technology Foundation (STW) under grant DCT 5323052 andthe Netherlands Bick industries via CUR commitees C96 fd A33 on Stuctral Masonry. APPENDIX. REFERENCES ‘Atkinson, R. HL, and Yan, G. G. (1990). “Resus of a saisial study ‘of masonry defrmabiliy." Masonry Soc J, (), 81-90 de Bort and Peonse, PH. (1990), "Stoic in anlsowople plasticity with reference tthe Til een Int J. Mumer Methods Enare. 23, 315-336, Drucker, D.C, Gibson, RE. and Henke, D. J (1957. Soil mechanics ‘and work hardening theorca of plasty.” Tens, ASCE, 122,338 Bae 1660/ JOURNAL OF ENGINEERING MECHANICS / JULY 1967 Hotter, G. Simo, J... and Taylor, RL, (1993). "A modified cap ‘model; eldest point solution algorithms.” Comp. & Strut, 4602), 203-218 ohberg, 1-M. (1992). “A jin element forthe nonlinear dynamic an- ysis of arch dams.” Rep. No. 186, Inst of Stoct. Engrg. ETH Zurich, Svzeiand Koites, WT. (1983), "Sires-sunin relations, uniqueness and variational probloms for elapse materials witha singular yield ntsc.” ‘Quarterly Appl Mash, 11, 380-354 Loft, HR. and Shing, FB. (1984). “Interface model apie to fracture (of masonty structures" J. Sut. Engrg, ASCE, 1201), 63-80. Loong, PB. abd Row, J. (1983). "Oa the tse of ico modeling forthe analysis of masonry shear- wall” Computer methods In siruct tural masonry—2, G.N. Pande and J. Middleton, eds, Book & Jout- nal Int. Swansea, United Kingdom, 14-26 ‘Lourngo, PB. and Rots, J. G. (1984). “Understanding the behaviour ‘of shear walls w numerical review.” Proc 10th Int. Brictblock Ma lonry Conf, N.G. Shive and A. Huize, eds, Calgary, Canada, ‘Loareago, P.B, Rots, J.G. and Blaaswendraat, J. (1994). “Implemen- tation ofan itertie cap model fr the snalysis of masony sie: tures" Computational modeling of Concrete stractres H. Mang. N Bicani, and Re Borst, eds, Pnetidge Press, Swansea, United King om, 133-134, ‘Minn, Wand Miller, H. (1982). “Falla of shear stessed masonry — fan ealaged theory, teste Sn aplication o shear walle” Proc, Brith Ceramic So. 30, 223-235. (mz, M, and Popov. P. (1988) “Accuracy and stability of integration lgortims for elsto-plaric.consttaive relations” Int J. Numer Methods Engr. 21, 1961-1576 Raljmakers, TM. 1, and Vermelfoon, A. T. (1992). "Deformation con- tolled meso shear tests on masonry piers." Rep. B-92-1156, TNO- BOUW/TU Bindhoven, Bull and Const: Res, Eindhoven, The Net. estan in Dutch). Roscoe, KH. and Burland, J.B. (1968) "On the generalized sress- ‘tain behaviour of wet iy." Engineering plasty. J. Heyinan an FLA. Lee, eds, Cambridge University Press, London, Eagland, 535-08. Ross, J, (1988). “Computations modelling of concrete fracture.” Dis ‘eration, Delt Unis. of Techaology, Delft, The Netherlands. ‘Simo, J. C, Kennedy, J. G.. and Govindee, 8. (1988), "Non-smooth useurface platiity and viscoplastic Loading/unlosding cond. ions and numerical aigothme" Dn J Numer Methods Engr. 26, isl -2188, ‘an der Plu, R. (1992). "Material properties and is components under tension and seas” Proc, 6th Can. Masonry Symp. Saskatoon, Can- sa van der Plijm,R, (1993), “Shear behaviour of bed joints.” Proc, 6k ‘North Am. Masonry Con. Pilaelphia, Pa 125~136. Vermaltoar, A. Raljmaker, TM. J, and Tonssen, H. J. M, (1992) “Shear ets on masonry walls." Proc. th North Am. Masonry Cont. PPiladlpi, Pa 1183-1193,

Das könnte Ihnen auch gefallen