Available online at www.sciencedirect.

com

Construction and Building

Construction and Building Materials 23 (2009) 4053

MATERIALS
www.elsevier.com/locate/conbuildmat

Review

An equivalent frame model for seismic analysis of masonry and reinforced concrete buildings
Y. Belmouden *, P. Lestuzzi
Ecole Polytechnique Federale de Lausanne, ENAC-IS-IMAC, EPFL, Station 18, CH-1015 Lausanne, Switzerland Received 13 August 2007; received in revised form 29 October 2007; accepted 29 October 2007 Available online 19 December 2007

Abstract In this paper a novel equivalent planar-frame model with openings is presented. The model deals with seismic analysis using the Pushover method for masonry and reinforced concrete buildings. Each wall with opening can be decomposed into parallel structural walls made of an assemblage of piers and a portion of spandrels. As formulated, the structural model undergoes inelastic exural as well as inelastic shear deformations. The mathematical model is based on the smeared cracks and distributed plasticity approach. Both zero moment location shifting in piers and spandrels can be evaluated. The constitutive laws are modeled as bilinear curves in exure and in shear. A biaxial interaction rule for both axial forcebending moment and axial forceshear force are considered. The model can support any shape of failure criteria. An event-to-event strategy is used to solve the nonlinear problem. Two applications are used to show the ability of the model to study both reinforced concrete and unreinforced masonry structures. Relevant ndings are compared to analytical results from experimental, simplied models and nite element models such as Drain3DX and ETABS nite element package. 2007 Elsevier Ltd. All rights reserved.
Keywords: Seismic evaluation; Unreinforced masonry; Reinforced concrete; Structural wall; Equivalent frame

Contents 1. 2. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A model for structural walls with openings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1. Description and hypotheses of the structural model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2. Formulation of a structural wall model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A nonlinear analysis of framed structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Pushover analysis of a RC building . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1. Description of the structural model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2. Application. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Pushover analysis of an URM building . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1. General assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2. Application. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Appendix 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 41 41 42 44 44 44 45 49 49 49 52 52 53 53

3. 4.

5.

6.

Corresponding author. Tel.: +41 21 693 6382; fax: +41 21 693 47 48. E-mail addresses: youssef.belmouden@ep.ch (Y. Belmouden), pierino.lestuzzi@ep.ch (P. Lestuzzi).

0950-0618/$ - see front matter 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.conbuildmat.2007.10.023

Y. Belmouden, P. Lestuzzi / Construction and Building Materials 23 (2009) 4053

41

1. Introduction Earthquakes are considered to be the major cause of structural failure of buildings in Europe. Despite their rarity and moderate intensity, earthquakes in the interior of northwest and central Europe have the potential to cause extensive damage and associated nancial losses, due to the vulnerability of the local building stock. The mitigation of earthquake hazard involves the collaboration of many specialists with dierent tasks. One of these topics is structural engineering providing and advancing the knowledge for earthquake resistant construction. However, a problem arises for existing buildings analysis. In this context, in the few last decades, technical advances have been made in seismic engineering and particularly in the seismic vulnerability assessment of existing buildings. The vulnerability assessment focuses on the study of the extent of damage for dierent earthquake scenarios. In almost all countries, the majority of the building stock is classied as existing buildings. This is why extensive assessment of such structures is motivated since they have been generally designed to resist gravity loads. Nevertheless, the seismic vulnerability of existing buildings designed against wind loads, is found to be very low. This paper makes a contribution to the seismic vulnerability assessment of existing buildings through the development of a simplied analytical model. The need for such models is always motivated by rst, the large amount of structures that should be analyzed in a very short time and second, the search for optimal solutions for structural retrotting. For vulnerability assessment purposes, the analysis of a large number of existing buildings requires relatively simple approaches that are capable of representing their essential characteristics. The models should be able to evaluate the ultimate strength, maximum displacements and the failure modes. Dierent models are developed based on analytical and nite element approaches [1]. The analytical models are found to be very simple to use and require lesser amount of data. However they are very limited, particularly for large building analysis in terms of structural behavior (coupling eect, distribution of the nonlinearity, modes of failures prediction). The performed analysis show that they are conservative and are not able to represent all features of such buildings [2]. On the other side, nite element approach is a powerful tool for seismic analysis but it is time consuming and requires a large amount of data. Moreover, rened models based on either discrete or continuum approaches suer from the strong mesh-dependency and require numerous parameters that may not be directly extractable from structural analysis. Hence, these models are very sensitive to the parameter calibration that aects closely the reliability of the results and the analysis stability (lack of convergence, ip-op occurrence, sudden load falling, and so on). With such methods it is not possible to treat a stock of buildings. Thus, these methods are cumbersome due to the high analytical skills required for

their numerical implementation and they are restricted only to practitioners with a high level of knowledge. A widely used model for structural analysis is the linear (beam-column element) nite element or the equivalent frame models. Despite of some limitations in the equivalent frame model, it is very attractive in comparison to complex nite element models [1,35]. Moreover, they have shown satisfactory results particularly for RC structures. In this context, the proposed model is based on beam-column element and distributed of nonlinearity approaches. It is adapted to analytical methods without use of nite element method. In this paper, the developed model deals with the seismic vulnerability assessment of existing multistoried buildings. 2. A model for structural walls with openings 2.1. Description and hypotheses of the structural model The mathematical model can represent solid walls, frame structural elements (made in beams and columns), coupled walls and perforated walls (or framed walls) [6]. The model can represent dierent openings. However, the vertical axis should lie through all vertical piers elements as well as for the horizontal axes that should lies through all spandrels. The structural model consists of an assemblage of vertical plane walls with openings that form a single perforated wall. Each structural wall is made of pier elements with or without rigid osets and a portion of spandrels such that there are two kinds of individual walls: exterior walls and interior walls (Fig. 1). The length of these parts of spandrels is equal to the zero moment length, and can be updated at each step depending on the bending moments at the spandrel ends. In the equivalent frame models that are based on nite element method, nonlinear exural springs (lumped plasticity) are inserted into the model at the ends of the piers and/ or spandrel elements. These elements are dened in terms

hi
F
lsp

hp
Opening Rigid zone Deformable part

Li
Edge wall Intermediate wall Edge wall

Fig. 1. A schematic representation of equivalent frame model for planar walls with openings.

42

Y. Belmouden, P. Lestuzzi / Construction and Building Materials 23 (2009) 4053

of momentrotation laws. Translational shear springs are added at each pier and spandrel at mid-points. These springs are expressed in terms of shear forcedisplacement laws. However, the occurrence of yielding is unlikely along spandrel spans and piers heights. For that reason, nonlinearity should be distributed along the clear pier height and clear spandrel length. Thus, the proposed model is based on the spread nonlinearity approach. Each pier and spandrel can be discretized into a series of slices [7] while cross-sections are considered as homogeneous. The structural element behavior is monitored at the center of the slices [7] while bending moments are evaluated at slice ends. The mechanical model undergoes exural as well as shear deformation. In the current formulation, the model only considers a biaxial interaction between axial forces bending moments (NM) and axial forcesshear forces (NV) only. The so-called shifting of the primary curve technique is used in a simple manner [8]. The axial force is evaluated in a simple manner based on initial axial forces plus vertical shear forces produced in spandrels at joints. A triaxial interaction rule, (NMV), is not currently considered. At present, only interaction curves that represents bending moment or shear force interaction with regards to a compressive axial force are considered. The major features of this model are summarized as follows: 1. All previous attempts to use simplied models based on static equilibrium method, always consider a constant zero moment location [914], and others. The wall formulation herein permits the capture of the coupling eect in elevation due to the nonlinearity distribution in both piers and spandrels. Thus, the zero moment location in both piers and spandrels can be mitigated during the nonlinear analysis. 2. In the current development, the variation of the axial vertical loads are considered for piers only and they are based on an over-simplied approach. The axial loads on piers are updated based on the initial axial forces at each storey plus the shear forces developed on spandrel ends. 3. The nonlinearity is treated using a smeared plasticity approach [7]. Thus, the piers and spandrels are discretized into nite homogenized slices [15]. Variable sections can be specied over either spandrels or piers. In pier elements the axial forces can increase or decrease. In that case, the pier slices can shift either from elasticto-plastic or from plastic-to-elastic state depending on the axial force distribution. 4. The model can take into account both exural and shear behavior in the inelastic range. The interaction eect can be dened by using experimental and phenomenological models. These equations are considered as failure criteria that can be dened by points and linear segments. The non linear constitutive model for both exural and shear behavior is considered as a bilinear envelop curve with a very small post-yield stiness to avoid

numerical problems. The exural behavior is modeled as a moment-curvature law that is based on an equilibrium statement in a cross-section. 5. The present formulation deals with a Pushover analysis. It is based on the well-known event-to-event strategy. A simplied algorithm for systems with interaction eect is presented through an equilibrium correction at each step of calculation. The analysis is performed by a force-controlled technique. The change of sign in a structural element is permitted only in the elastic range. In the inelastic range, this leads to stoppage of the analysis. 6. The structural wall is a planar structure (two-dimensional). However, the sum of all capacity curves, on the basis of the equal top displacement assumption, permits to analyze an entire building and to develop capacity curves. 2.2. Formulation of a structural wall model The structural walls, composed in an assemblage of piers and spandrels, are modeled as an equivalent frame structure. Rigid osets can be added at the top and/or the bottom of the piers, or the left side and/or right side of spandrels. The storey rotations and lateral displacements are calculated so that exural and shear deformations can be considered. It should be noted that only in-plane deformations and rotations of the entire walls are considered. In the following a general method for structural analysis of a multistory building is given. The wall base is considered as xed. The storey momentlateral force formulation of a structural wall element is expressed by: fM bs g K Frame fP g fP g LPT LTF fF g 2 B1 C 1 6A B 6 2 2 6 6 0 A3 6 1 K Frame 6 . . . 6. . 6. 6 40 0 0 0 P n Dn T n1 En T n 1 2 3

0 C2 B3 . . . 0 0

0 0 C3 . . . 0 0

... ... ... .. . ... ...

0 0 0 . . . AN 1 0

0 0 0 . . . BN 1 AN

0 7 0 7 7 7 0 7 7 . . 7 . 7 7 C N 1 5 BN 3 4

{Mbs} represents the base storey bending moments vector, [KFrame] is the equivalent frame stiness matrix, {P} is the reduced shear forces vector, [LPT] is the reduced storey shear forcesstorey shear force transformation matrix, [LTF] is the storey shear forceapplied lateral force transformation matrix in absence of vertical distributed loads, {F} is the lateral load pattern vector. The equivalent frame matrix and the reduced shear forces vector are dened by the expressions An, Bn, Cn, Dn, and En (with n = 1, N; N is the number of storeys) (Eqs. (3), and (4)) such as An k sp eq;n1
1

Y. Belmouden, P. Lestuzzi / Construction and Building Materials 23 (2009) 4053

43

2 Bn sp sp p k eq;n1 k eq;n k n C n k sp 1 eq;n hn1 Dn sp k eq;n1 E n hn 1 1 p sp k n k eq;n

! 6 7 8 bflx hn n cflx hn n d shr n 9

mNslices X m1 mNslices X m1

Z Z Z

gm

gm1 gm

1 nndn K v nm n2 dn K v nm dn K d nm

18 19 20

gm1 gm gm1

1 hn

mNslices X m1

The equivalent stiness of a spandrel for an individual structural wall i at level n (Fig. 2) is: For an interior wall: ! 1 1 sp 10 k eq;n cspandrel aspandrel n i i1 For a left side exterior wall: k sp aspandrel n i eq;n k sp cspandrel n i1 eq;n
1 1

11

For a right side exterior wall: 12

Herein, the expressions (17)(20) are formulated for pier elements; hn is the storey height at level n, n is a normalized variable. gm1 and gm are the left side and the right side coordinates of a slice respectively; Nslices is the number of slices in a wall element, nm = 0.5(gm1 + gm) represents the slice centre at which the slice behavior is monitored. The same expressions are also used for the calculation of the nonlinearity redistribution coecients in the spandrels with respect to their zero moment lengths L0R and L0L (Fig. 2). The redistribution coecients are then discretized such that a structural element is decomposed into nite slices [7]. The inter-storey rotations between level n and n-1, for a given single wall, omitting the index i for clarity, are given by: hn hn1 bpier apier M n1;n bpier cpier M n;n1 n n n n 21 hn hn1 M sp n;tot k sp eq;n M sp n1;tot k sp eq;n1 22

The equivalent stiness of a pier for an individual structural wall element at level n is: k p cpier bpier n n n
1

13

The superscripts p and sp mean pier and spandrel elements respectively. The superscripts x and shr mean exure and shear respectively. The nonlinearity redistribution coecients are dened as follows: an aflx d shr n n bn cn bflx d shr n n cflx d shr n n
mNslices X m1

14 15 16 Z
gm gm1

aflx hn n

1 n dn K v nm

17

M sp and M sp n;tot n1;tot represent the total bending moment developed in spandrels at the nth and n 1th level respectively (Fig. 2). k sp and k sp eq;n1 represent the equivalent stieq;n ness of spandrels at the nth and n 1th level respectively (Fig. 2) for an individual wall i. The lateral story-displacements for each individual wall, in a given section yj, are given as follows: Z yj Z yj vy j hj1 y j vny j ndn dndn 23
0 0

L0 L ,i ,n

L0 R ,i ,n

L0 L ,i +1,n
sp M R ,i , n

Level n

M n ,n +1
Tn
sp M L ,i , n

M n ,n 1

The term level means the centre-line (or the neutral axis) of the spandrels between two adjacent storeys that form the pier-to-spandrel joints. The wall curvatures and shear deformations are expressed as follows: vn Mn=K v n 24 25

L 0 L ,i ,n 1 M n 1,n 2

L 0 L ,i +1,n 1 L 0 R ,i ,n 1
sp M R ,i ,n 1

hn

dn T n=K d n

Level n-1

M n 1,n

Tn 1
Level n-2

sp M L ,i ,n 1

L 0 L ,i +1,n 2 M n 2 , n 3 L0 R ,i ,n 2
Wall i+1

hn 1

M n 2,n 1
sp M L ,i , n 2

sp M R ,i , n 2

L0 L ,i ,n 2
Wall i

Span i

M(n) is the bending moment and T(n) is the shear force in a slice, Kv(n) and Kd(n) are the exural and shear stiness of a slice, respectively. The curvatures are assumed linear over each slice. Moreover, the equilibrium equation at a rigid joint is formulated as follows: X X M spandrels 26 M piers The pier-to-spandrel joint equilibrium for an individual wall i (Fig. 2) is expressed as follows:

Fig. 2. Moment diagrams in an equivalent frame model and decomposition into simplied individual wall element. L0L,i,n and L0R,i,n are the left side and right side zero moments lengths for the wall i at level n.

44

Y. Belmouden, P. Lestuzzi / Construction and Building Materials 23 (2009) 4053

M sp M n;n1 M n;n1 n;tot M sp n;tot M sp L;i;n M sp R;i1;n For a left side exterior wall i:
sp M sp R;i1;n k eq;R;i1;n hn

27 28
Equilibrium correction

S1
Compression

Slice elastic before loading (state k-1) kSlice yields after loading (state k)

( ( S 2k ) = S 2 yk )

29 30

S1(k )
S1(k ) k 1

T sp R;i1;n

M sp R;i1;n =L0R;i1;n

For a right side exterior wall i: M sp k sp hn L;i;n eq;L;i;n T sp M sp =L0L;i;n L;i;n L;i;n 31 32
( S 2 yk 1)

S1(k 1)
( S 2k )
( ( S 2k ) = S 2 yk ) ( S 2k 1)

L0R,i,n and L0L,i,n dene the zero moment location for the ith spandrel element at the nth level, M sp and M sp L;i;n R;i1;n are their corresponding spandrel ends bending moments in the span i (Fig. 2). The initial values for zero moment locations are calculated according to the relative stiness between spandrels and the adjacent piers. For an interior wall, the expressions from Eqs. (29) to (32) are considered simultaneously. 3. A nonlinear analysis of framed structures In the following, a simplied approach of the Pushover method based on the event-to-event strategy is presented (Table 1). In this method, the event factor for the entire frame subjected to the predened lateral load pattern is calculated such as ak = Min(ak , i = 1, Nwall), Nwall is the i number of parallel individual walls. A wall event factor, ak , is calculated for the nominal event plus the tolerance i [7]. A wall event factor is extracted from the lowest slice

+ S 2 y(k )

+ S 2 y(k 1) S 2

Tension

Fig. 3. General representation of an interaction process between S1 and S2 parameters.

event factors. The occurrence of an event depends on the yield surfaces function and yielding limit states for all slices in exure and in shear (see Figs. 3 and 4). 4. Pushover analysis of a RC building 4.1. Description of the structural model A three-dimensional multistoried building made in RC structural walls is studied (Fig. 5). The structure was modeled on both Drain3DX [16] using a ber beam element (type 15) and ETABS [17] using a point hinge beam ele-

Table 1 Flow chart of the proposed Pushover analysis for the equivalent planar-frame model Step 1 Step 2 Step 3 Structural modeling: Frame geometry, lateral loads pattern {F}, initial compressive axial loads on piers N, failure criterion (NM) and (NV). Slices state initialization: Yielding bending moments and yielding shear forces, elastic stinesses, event overshoot tolerances, initial values for zero moment locations L0Land L0R. For each individual structural wall i = 1, Nwall: (ak = 1) 3-1 Apply a load increment ak{DF} 3-2 Having Lspan;k1 , Lspan;k1 ,Nk1 and M k1 , and T k1 y y 0L 0R 3-3 Calculate an, bn and cn 3-4 Form [KFrame]1, and {DP} 3-5 Solve Eq. (1) 3-6 Calculate {DMbs} and {DT} 3-7 Extract the bending moments at the top of all storeys 3-8 Calculate the total bending in spandrels DM sp , for n = 1,N (Eq. (26)) n;tot 3-9 Calculate Dhn Dhn1, for n = 1,N (Eqs. (21) and (22)) sp sp sp 3-10 Calculate DM sp R;i1;n , DM L;i;n , DT R;i1;n and DT L;i;n , for n = 1,N (Eqs. (29)(32)) 3-11 Calculate the lateral displacementsDv(Eq. (23)) 3-12 Calculate the event factor for the entire frame (parallel walls) ak 3-13 Adjust and update the applied lateral force, the storey rotations, the lateral displacements, the bending moments and shear forces, S, and their corresponding deformations, D, (slice curvatures and slice distortions) in both piers and spandrels: Sk = Sk1 + akDSk, Dk = Dk1 + akDDk 3-14 Update axial loads on piers and spandrels by force equilibrium:Nk 3-15 Equilibrium correction using Nk: determine corrected yielding states M k and T k for the next iteration (Figs. 3 and 4) y y 3-16 With respect to the current yielding limits (Fig. 3), update the shear and/or exural stiness of the yielded slices (Fig. 4): Determine the updated slice stinesses 3-17 New estimation of Lspan;k using spandrel ends bending moments 0 3-18 Check for structural stability: Quit the analysis if the structure becomes unstable (signicant loss of stiness), or if a given pier is subjected to a tension axial load. Otherwise perform the next iteration, k + 1 (go to step 3-1) Plot the capacity curve for the entire structure

Step 4

Y. Belmouden, P. Lestuzzi / Construction and Building Materials 23 (2009) 4053

45

Bending moment
PL

M >M
PL

k 1 y

PL PL EL PL EL
k M > My k M M y 1 k M y 1 k M > My k M < M y 1 k My k M My k M < M y 1

k M = M y 1 + M tol

step k 1

EL EL EL EL EL
k M = M y 1 + M tol

Equilibrium correction

EL EL EL
k M My

step k

PL EL
k M > My

PL
k M > M y 1

PL
k M > My

Curvature For : N and M

PL
k M > M y 1

PL
k M > My

PL

PL

Fig. 4. Bending moment redistribution and the equilibrium correction in the case of a compressive axial force loading.

Fig. 6. Plan view of a RC wall section with ber discretization on Drain3DX model (dimensions in m).

top and bottom storeys. In the equivalent frame model (EFM), the nonlinear behavior for each slice is dened by a moment-curvature relationship in compression only. The analysis performed on Drain3DX was force-controlled. On the other hand, the analysis performed on ETABS was displacement controlled in the presence of a given lateral load pattern. The adopted load pattern represents the distribution of inertia forces corresponding to the rst mode of vibration. 4.2. Application The building studied herein is a six storey torsionally balanced reinforced concrete structure with a total height equal to 6.0 3.4 m and a total oor area equal to

Fig. 5. Plan view of the RC structure.

ment. The ber beam is a nonlinear nite element model in exure but linear elastic in shear. The behavior of bers is dened by a stressstrain relationship for both steel and concrete materials. The use of bers to model cross-sections accounts rationally for axial forcebiaxial bending moments (Fig. 6). A detailed description of the element is given in the reference including related capabilities, assumptions and limitations [16]. In this application, an elastic perfectly plastic model was adopted for steel bers. However, a parabolicrectangular stress block was adopted for concrete material. The oor was modeled as a grid system using the elastic linear beam element type 17 [16]. A validation of the nite element type 15 against experimental results can be found in the reference [2]. On the other side, ETABS provides a exural point hinge nite element model (PHFE) called P-M2-M3. This model considers an interaction between two-way moment curves and axial forces. Since the structural model behaves in the in-plane direction (Z-direction, Fig. 5), the point hinge model performs with a biaxial interaction rule. The hinges are located at the ends of all beam elements, at

Fig. 7. A view of the structural model developed on ETABS.

46

Y. Belmouden, P. Lestuzzi / Construction and Building Materials 23 (2009) 4053

Moment

30.0 18.0 m2 (Fig. 7). The oor is of a grid-type with RC slab. In the case studied, the spandrels consist of beam elements representing a grid oor with underbeams. For estimating the stiness of the oors with underbeams, the eective width of the oor slab is calculated according to the rules suggested by Bachmann and Dazio [18]. The second moment of inertia of the oor section is equal to 0.0262 m4 [11]. The oor load carried by each wall is equal to 188 kN, 300 kN and 900 kN for walls 1, 2 and 3, respectively. The resulting normal forces acting on each oor level are summarized as follows: the axial forces acting on walls A1, A2, and A3 at the 6th oor are equal to 211 kN, 324 kN, and 924 kN, respectively, while for lower levels, these forces are equal to 235 kN, 347 kN, and 947 kN, respectively. Three cases are investigated to study the axial force redistribution and the axial forcebending moment interaction rule. They are: (1) rigid-oor-type structure with 100% of the oor stiness, IFloor, (2) semi rigidoor-type with 50% IFloor and (3) exible oor-type with 10% IFloor. For each oor-type model, two cases were studied for the EFM and four cases for the PHFE model on ETABS. The case studies are dened as follows: 1. PHFE M1 and M2: Bilinear and elasticplastic momentrotation law respectively, without (NM) interaction, 2. PHFE M3 and M4: Bilinear and elasticplastic momentrotation law respectively, including (NM) interaction, 3. EFM 1 and EFM2: Without and with (NM) interaction, respectively. In this simulation, the spandrels represent the oor beams that were elastic linear. Thus, the zero moment lengths were chosen to be at the middle of the span length, without rigid osets, and kept constant during the analysis. The material properties were dened by the tensile strength of concrete, the compressive strength of concrete, the steel maximum strength, the concrete Young modulus and the steel elastic modulus that are equal to 5 MPa, 45 MPa, 500 MPa, 37,500 MPa, and 210,000 MPa, respectively. Additional details on the material properties are found in [11,19]. For the EFM, the walls at each storey are discretized into 20 slices over a storey height. The EFM necessitates moment-curvature laws that are considered idealized for both elastic perfectly plastic and bilinear curves (Fig. 8). For the RC wall section, the mechanical properties (Fig. 8) are summarized as follows: The yielding moment, My, the ultimate moment, Mu, the rst yield curvature, /0y , the nominal curvature for a bilinear idealization, /y, and the ultimate curvature, /u, are equal to 3034 kN m, 4786 kN m, 1.648 m1, 2.60 m1, and 28.5 m1, respectively. On ETABS, only yielding moments should be specied for the exural hinges since the yielding rotations are calculated by the program. The axial deformations of the

Mu
Elastic plastic

My

r Bilinea

y = y'

Mu My

y' y

Curvature

Fig. 8. Bilinear idealization of the moment-curvature law.

beams are neglected on both ETABS and Drain3DX models. With regards to the biaxial failure criteria used herein (Fig. 9), the hinge yielding in both PHFE model and slices of the EFM depends closely on the axial load. For low axial loads, the yielding of the sections is delayed, while for high axial loads the yielding is anticipated. These mechanisms are closely related to the axial load redistribution, the oor stiness and wall coupling. Hence, the global response of the structure is aected (elastic stiness, structural displacement, damage occurrence). Figs. 1012 display capacity curves for the three oortype models analyzed by Drain3DX, ETABS and the proposed EFM. The (NM) interaction eect increases with the total base shear. The (NM) interaction has small eect in the rst stage of the analysis. As the oor stiness increases, the force redistribution capacity of the structure increases, the normal forces increase, and then the eect of (NM) interaction becomes signicant. When axial force is still small, the (NM) interaction is negligible. In other

5000 0 -5000 -10000

Compression Tension

Axial force (kN)

-15000 -20000 -25000 -30000 0 1000 2000 3000 4000 5000 6000 7000 8000

Bending moment (kNm)

Fig. 9. Axial forcebending moment interaction law for RC wall sections dened by linear segments and extracted from ETABS according to the Eurocode design code [17].

Y. Belmouden, P. Lestuzzi / Construction and Building Materials 23 (2009) 4053

30000
Fiber model 100%Ipl Fiber model 50%Ipl Fiber model 10%Ipl

47

25000

Total base shear (kN)

20000
PHFE model M1

15000

PHFE model M2

50%IFloor
10000

PHFE model M3

PHFE model M4

10%IFloor
EFM1

5000
EFM2

0 0 0.05 0.1 0.15 0.20 0.25 0.3

Roof lateral displacement (m)

Fig. 10. Capacity curves for rigid-oor-type model.

30000

Fiber model 100% Ipl Fiber model 50% Ipl Fiber model 10% Ipl

25000

Total base shear (kN)

20000

100%IFloor
15000

PHFE model M1

PHFE model M2

PHFE model M3

10000

10%IFloor
5000

PHFE model M4

EFM1

EFM2

0 0 0.05 0.1 0.15 0.20 0.25 0.3

Roof lateral displacement (m)

Fig. 11. Capacity curves for semi rigid-oor-type model.

words, the (NM) interaction rule has no eect for exible oor-type structures (Fig. 12). This application tends to demonstrate the ability of the EFM, in comparison to ETABSs results, to reproduce the interaction between the oor stiness, the structural wall coupling, the force redistribution, and the failure criteria on the global response of the building. The order of occurrence of the plastic hinges can be obtained for all the steps of the displacement control from a pushover analysis. The collapse sequence at ultimate state is presented in the Figs. 13 and 14 [20]. The comparison should be drawn in term of general behavior of the structure, not in terms of exact location of the hinges. Note that,

the EFM is compared only to the elastic perfectly plastic PHFE model. In the rst steps of analysis, the diagram of moments (elastic diagram), correspond to the relative stiness between beam-oors and walls. However, as the nonlinearity grows, the wall element stiness decreases while the beam-oor stiness remains elastic linear. This means that the shape of the moment diagram tends gradually to a frame-type moment diagram. When a plastic hinge forms at a given wall base in the EFM, the bending moment remains constant at this slice. Generally, as the load increases, the zero moment location shifts to the mid-storey height as explained above. The plastic hinge presented

48
18000

Y. Belmouden, P. Lestuzzi / Construction and Building Materials 23 (2009) 4053

100%IFloor
16000

Fiber model 100% Ipl Fiber model 50% Ipl Fiber model 10% Ipl

50%IFloor
14000

Total base shear (kN)

12000
PHFE model M1

10000
PHFE model M2

8000
PHFE model M3

6000
PHFE model M4

4000
EFM1

2000
EFM2

0 0 0.05 0.1 0.15 0.20 0.25 0.3

Roof lateral displacement (m)

Fig. 12. Capacity curves for exible oor-type model.

Fig. 13. Plastic hinges sequence for rigid-type oor and with no (NM) interaction. Right: PHFE model at Droof = 0.1348 m case M2. Left: EFM1 case for Droof = 0.149 m.

Fig. 14. Plastic hinges sequence for rigid-type oor and with (NM) interaction. Right: PHFE model at Droof = 0.2094 m case M4. Left: EFM2 case for Droof = 0.209 m.

Y. Belmouden, P. Lestuzzi / Construction and Building Materials 23 (2009) 4053

49

 a1 N b 1 N 2 f N ; V MIN V c1 N  p V a2 b2 N V a3 1 b3 N 6 0

34

Two constants a and b are required for exure failure criteria, while nine constants ai, bi and ci (for i = 1, 3) are required for shear failure criteria [2123]. N is the axial compressive load acting on a pier element. The same failure criteria can be found in many other procedures for masonry assessment [1214, 24] and others). These equations deal with elastic perfectly plastic models in terms of momentrotation and shear forcedisplacement laws. In this study, the behavior of the spandrel is assumed to be elastic linear both in exure and shear. 5.2. Application A full-scale two-storey unreinforced masonry tested at the Pavia University was chosen for model validation
Fig. 15. Geometry of the model building.
6.435 5.7525 5.5069 5.07

1.365

herein corresponds to the step for which the EFM achieves its ultimate state under force control. For an elastic perfectly plastic moment rotation model without (NM) interaction, the EFM behaves as a point hinge model with lumped plasticity. The use of force interaction permits modeling the eect of the axial redistribution on the yielding capacity of the structure. Figs. 13 and 14 show the dierence in the number of hinges and the yielded slices when the (NM) interaction was activated. The results extracted from capacity curves and structural damage assessment point out that the use of force redistribution in simplied models is necessary in particular for existing building analysis more than for structural design. 5. Pushover analysis of an URM building 5.1. General assumptions The proposed model can be used also for URM structures modeling. The URM piers and spandrels are subdivided into a series of slices. The slices represent an homogeneous bricks and mortar one-phase material. As known, the masonry material is a weak isotropic material with very limited ductility. Thus, the softening behavior is very burdensome for computation and causes failure of convergence particularly when the analysis is force-controlled. The post-peak behavior with softening is beyond the scope of this model. The yield criteria considered are expressed for exure (Eq. (33)) and for shear behavior (Eq. (34)) according to the Magenes model [2123] as follows: f N ; M M aN bN 6 0
2

2.7625

1.235 2.1088

3.835 3.398

2.99 2.5314 2.145

1.69

2.5314 0.94 0.94

P1

P2

P3

Fig. 16. Elevation view of the wall D and geometry (in m). Exterior walls length and axial loads on the bottom and top levels are equal to 1.15 m, 56 kN, 26.9 kN, respectively. Interior wall length and axial loads on the bottom and top level are equal to 1.82 m, 133 kN, and 64.5 kN, respectively.

1000 900 800

Axial force (kN)

700 600 500 400 300 200 100 0 0 100 200 300 400 500 600
Exterior walls Interior walls

Bending moment (kN)

33

Fig. 17. Flexural criteria for rocking mode of failure in piers.

50

Y. Belmouden, P. Lestuzzi / Construction and Building Materials 23 (2009) 4053

[2123]. This structure has been extensively studied in literature. A remarkable feature of this structure, is that the axial load in piers varies during the experimental test. The variation of the axial load in the considered structure is exploited to study the sensitivity of the model to the axial force variation on piers. The structural model is subjected to increasing lateral forces that are applied at the oor levels, keeping a 1:1 ratio between the force at the rst and the second oor. In this application the door wall D (Figs. 15 and 16) was chosen because of no ange eect is considered. The elastic properties of the structure used in the model are summarized as follows:

500 450 400 350

Mode 3
Vshear mode1 Vshear mode2 Vshear mode3 Minimum Shear

300 250 200

Mode 2
150 100 50

The maximum compressive strength of a masonry prism orthogonal to the mortar bed, fm, is equal to 7.9 MPa [23], the shear modulus (Ge = 90fm) is equal to the eective value. For full data see Ref. [23]. In the current application, The (NM) interaction is shown in Fig. 17. However, the (NV) interaction was not activated. In fact, it was found that axial forces in the second storey are conned to the rst failure mode domain of validity (Fig. 18) since the variation in axial forces for this storey is very low. The second and third modes of failures are not activated. However, despite the variation in axial forces in the rst storey, they are conned to the second failure mode domain of occurrence during the analysis (Fig. 18). Hence, an elastic perfectly plastic model without (NV) interaction is used. The use of rigid osets is a crucial issue in equivalent frame modeling. The dimensions of rigid osets in piers are calculated based on an empirical approach proposed by Dolce [25]. In this study, full rigid osets are considered. The capacity curves (total base shear versus top lateral displacement) are developed for dierent cases (Table 2, Appendix 1). In the light of the obtained results, the following recommendations are made: 1. The eect of the axial forcebending moment, (NM), interaction is showed by the cases 1 and 2. As displayed in Fig. 19, as axial compressive load increases, exural strength of the piers also increases with regards to the failure criteria (Fig. 17). 2. The nonlinear eect of shear mechanism is illustrated by cases 3 and 4 in the absence of rigid osets, and by cases 7, 8 and 9 in the presence of rigid osets (Fig. 20). As expected, the contribution of shear

Axial force (kN)

Mode 1
0

50

100

150

200

250

300

Shear force (kN)

Fig. 18. Shear failure criteria for piers only. Mode 1: (cs) shear failure along bed-joints at the end section cracked in exure; Mode 2: (ws) diagonal cracking at the centre of the panel due to mortar joint failure; Mode 3: (b) diagonal cracking at the centre of the panel due to brick failure.

160
Experiment

140

Case 3 Case 2 Case 1, 4 and 6 Case 5

120

Total base shear (kN)

100

80

Experiment Case 1

60

Case 2 Case 3

40

Case 4 Case 5

20

Case 6

0 0 2 4 6 8 10 12 14 16 18

Roof lateral displacement (mm)

Fig. 19. Capacity curves of the wall D with no rigid osets.

Y. Belmouden, P. Lestuzzi / Construction and Building Materials 23 (2009) 4053

180
Case 8

51

160
Experiment Case 12

140

Total base shear (kN)

120

100 80

Experiment Case 7 Case 8 Case 9

60

Case 10 Case 11 Case 12

40

Case 13

20

0 0 2 4 6 8 10 12 14 16 18

Roof lateral displacement (mm)

Fig. 20. Capacity curves of the wall D with rigid osets.

F F F S S

S:

Shear cracks

S S S

F:

Flexural cracks

Fig. 21. Crack patterns from the experimental test of the URM building (at failure state (top displacement equal to 24 mm)). Fig. 22. Crack patterns from numerical results at 17 mm.

mechanism tends to decrease the capacity of the structure due to the occurrence of shear damage. This feature is successfully captured by the simplied model. 3. As displayed in Figs. 19 and 20, the rigid osets have a signicant eect on the global response not only on stiness, but also on strength capacity of the structure [6,15,22]. This is expected as the horizontal element stiness closely aects the contribution of the frame mechanism to structural response (cases 10, 11 and 12). The capacity curves obtained from EFM (case 12) versus PHFE model (case 13) are satisfactory. 4. In cases 12 and 13, the two capacity curves are close to a certain extent in spite of the smeared approach in the EFM. Both cases 56, and 1213, show the com-

parison of the modeling performance, including shear eect and (NM) interaction rule and using either the EFM and the PHFE model with or without rigid zones. Due to the coupling eect, the resulting crack patterns displayed by the numerical simulation are dierent on the left and right sides (Figs. 21 and 22). On the other hand, the crack pattern predicted by the EFM is symmetrical (Fig. 23). The shear cracks on spandrels were not obtained since the spandrels were modeled as elastic linear. It is clear that the numerical results should be more accurate, in comparison to the EFM, since both the axial deformation and axial force redistribution were not considered in the EFM. The axial force and deformation are crucial issues when using failure criteria for the plastic hinge formation and

52
c

Y. Belmouden, P. Lestuzzi / Construction and Building Materials 23 (2009) 4053

c

F c F F S

Fig. 23. Crack patterns from analytical results (on the EFM) F: exural crack in one slice, S: shear crack over a pier.

damage occurrence. Moreover, in comparison to the test crack pattern, the numerical and analytical results were extracted at 17 mm while the tests represent a crack pattern at 24 mm. Also, the shear failure was neglected in the lintels in both models. These are the reasons for the lack of match between the experimental, the analytical and the numerical crack patterns. 6. Conclusions This paper presents a simplied formulation of an equivalent frame model. The model permits to consider many relevant features of structural behavior such as structural wall coupling, zero moment location shifting, axial forcebending moment interaction, axial forceshear force interaction, and failure modes prediction. However, in the case of URM buildings, it is well known that smeared
Table 2 Case studies for both EFM and PHFE models Case 1 2 3 4 5 6 7 8 9 10 11 12 13 Model type PHFE PHFE EFM PHFE PHFE EFM EFM EFM EFM EFM EFM EFM PHFE Rigid zone in pier Rigid zone in spandrel (with 2Em) (with 10Em) (with 10Em)

crack approach suers from a few limitations. The smeared crack model is enable to represent eectively the rocking and bed joint sliding mode of failures. For the development of capacity curves, the obtained results from the proposed model show good agreement with experiment and numerical results (Figs. 19 and 20). The model has proven its capability to satisfactorily predict the maximum strength. The calculated maximum strengths, in particular for the masonry structure (in the range of 9%), could be judged as good results since the model is based on simplied approaches in comparison to nite element models. However, the post-peak behavior with softening is not yet obtained since the model is force-controlled. Care should be taken when modeling dual buildings as framewall structures in particular with respect to the initial zero moment lengths assumption. In all cases, obtained results should be considered from an engineering point of view as is generally done for all simplied existing models. It is evident that the failure mode identication is a challenging task even if nite element models are used. This feature is sensitive to various analysis parameters such as the modeling of shear mechanisms, the lateral load pattern and force redistribution capabilities. The proposed model works well for RC structure. However, it requires further improvements for URM structural modeling (displacement control, variation of axial force in spandrels, multilinear models with softening). Finally, the proposed model is formulated in order to extract capacity curves with damage identication. It can be implemented readily using any programming platform. The model can be used to assess URM structures, RC structures as well as dual structures that are commonly adopted in many countries. Acknowledgements This work is a part of a project dealing with seismic vulnerability assessment of unreinforced masonry existing buildings in Switzerland. The project was supported by

(NM) failure criteria

V shear eect

Maximum strength ratio (*) (%) 20.1 15.5 10.7 21.7 22.4 20.5 9.9 +18 9.9 9.0 8.5 7.1 9.3

(with (with (with (with (with (with (with

2Em) 2Em) 4Em) 10Em) 2Em) 10Em) 10Em)

Legend: PHFE: point hinge nite element model, EFM: equivalent frame model, () option considered, () option not considered, (*) the maximum strength ratio = analytical/experimental maximum strengths %, Em is the masonry Young Modulus.

Y. Belmouden, P. Lestuzzi / Construction and Building Materials 23 (2009) 4053

53

the Federal Oce for the Environment, FOEN (Department of the Environment, Transport, Energy and Communications). This nancial support is gratefully acknowledged. Appendix 1 See Table 2. References
[1] Tzamtzis AD, Asteris PG. FE analysis of complex discontinuous and jointed structural systems: part 1: presentation of the method A state-of-the-art review. Electron J Struct Eng 2004:1. [2] Belmouden Y, Lestuzzi P. Seismic vulnerability assessment of existing buildings in Switzerland. Applied Computing and Mechanics Labo ratory. Structural Engineering Institute. Ecole Polytechnique Federale de Lausanne. Research Report No 6, April 2006. [3] Roca P, Molins C, Mar AR. Strength capacity of masonry wall structures by the equivalent frame method. J Struct Eng 2005;131(10):160110. [4] Salonikios T, Karakostas C, Lekidis V, Anthoine A. Comparative inelastic pushover analysis of masonry frames. Eng Struct J 2005;25:151523. [5] Kappos A, Penelis G, Drakopoulos C. Evaluation of simplied models for lateral load analysis of unreinforced masonry buildings. J Struc Eng, ASCE 2002;128(7):8907. [6] ATC-40. Seismic evaluation and retrot of concrete buildings, vol. 1. Redwood City, California: Applied Technology Council; 1996. [7] Belmouden Y, Lestuzzi P. Analytical model for predicting nonlinear reversed cyclic behaviour of reinforced concrete structural walls. Eng Struc 2007;29(7):126376. [8] ElMandooh K, Ghobarah A. Flexural and shear hysteretic behavior of reinforced concrete columns with variable axial load. Eng Struc 2003;25:135367. [9] Nuray Aydinoglu M. An incremental response spectrum analysis procedure based on inelastic spectral displacements for multi-mode seismic performance evaluation. Bull Earthquake Eng 2003;1:336. [10] Kilar V, Fajfar P. Simple pushover analysis of asymmetric buildings. Earthquake Eng Struc Dyn 1997;26:23349. [11] Lang K. Seismic vulnerability of existing buildings. PhD thesis. Institute of Structural Engineering, Zurich, Switzerland, Swiss Federal Institute of Technology, February 2002.

[12] FEMA 356. Prestandard and commentary for the seismic rehabilitation of buildings. Washington (DC): Federal Emergency Management Agency; 2000. [13] FEMA 306. Evaluation of earthquake damaged concrete and masonry wall buildings, basic procedures manual. Washington (DC): The Partnership for response and recovery; 1999. [14] Paquette J, Bruneau M. Pseudo-dynamic testing of unreinforced masonry building with exible diaphragm and comparison with existing procedures. Construct Build Mater 2006;20:2208. [15] Penelis GRG. An ecient approach for pushover analysis of unreinforced masonry (URM) structures. J Earthquake Eng 2006;10(3):35979. [16] Prakash V, Powell GH, Campbell S. DRAIN3DX: Base program description and user guide, UCB/SEMM-1994/07. Berkeley: Department of Civil Engineering, University of California; August 1994. [17] ETABS. Integrated building design software, computers and structures, Inc. Berkeley, California, USA, Version 9, November 2005. [18] Bachmann H, Dazio A. A deformation-based seismic design procedure for structural wall buildings. In: Proceedings of the international workshop on seismic design methodologies for the next generation of codes, Bled/Slovenia, A.A. Balkema: Rotterdam; 24 27 June 1997. [19] Dazio A. Entwurf und Bemessung von Tragwandgebauden unter Erdbebeneinwirkung. Institut fur Baustatik und Konstruktion, ETH Zurich, Bericht Nr. 254. Birkhauser Verlag: Basel; 2000 [in German]. [20] Wilkinson SM, Hiley RA. A non-linear response history model for the seismic analysis of high-rise framed buildings. Comput Struct 2006;84:31829. [21] Magenes G. A method for pushover analysis in seismic assessment of masonry buildings. Paper 186612WCEE; 2000. [22] Magenes G, Della Fontana A. Simplied non-linear seismic analysis of masonry buildings. In: Proceedings of the fth international masonry conference, London; 1315th, 1998. [23] Magenes G, Calvi GM. In-plane response of brick masonry walls. Earthquake Eng Struc Dyn 1997;26:1091112. [24] Tomazevic M. Seismic resistance verication of buildings: following the new trends, seismic methodologies for the next generation of codes. In: Fajfar P, Krawinkler H, editors. Balkema: Rotterdam; 1997. [25] Dolce M. Schematizzazione e modellazione per azioni nel piano delle pareti (Models for in-plane loading of masonry walls). Corso sul consolidamento degli edici in muratura in zona sismica, Ordine degli Ingegneri, Potenza; 1989.