10th International Conference on Fracture Mechanics of Concrete and Concrete StructuresFraMCoS-X

G. Pijaudier-Cabot, P. Grassl and C. La Borderie (Eds)

RETROFITTING UNREINFORCED MASONRY BUILDINGS WITH ASTRAIN-HARDENING CEMENT-BASED COMPOSITE TO ENHANCE

SEISMIC RESISTANCE

MARISKA KOTZE∗, GERT C. VAN ROOYEN† GIDEON P.A.G. VAN ZIJL‡

∗University of StellenboschStellenbosch, South Africa

e-mail: 18179142@sun.ac.za

†University of StellenboschStellenbosch, South Africa

e-mail: gcvr@sun.ac.za

‡University of StellenboschStellenbosch, South Africae-mail: gvanzijl@sun.ac.za

Key words: SHCC, Masonry, Seismic

Abstract. In South Africa there are many low-rise, unreinforced, load-bearing masonry (ULM) build-ings in seismic zones due to shortcomings in South Africa′s design codes. These buildings pose a greatrisk to their occupants in the case of seismic activity . A possible solution is to retrofit these build-ings with a strain-hardening cement-based composite (SHCC) overlay. It is postulated that the strainhardening characteristic of the SHCC overlay may sufficiently increase the ductility of such buildingsto prevent collapse during seismic activity. This paper discusses the effects SHCC have on the dam-aged behaviour of masonry walls under in-plane seismic action. These effects were evaluated usingexperimental pushover tests with different retrofitting techniques. From the tests, it was concludedthat adding an SHCC overlay increased the in-plane shear resistance of a masonry wall. The ductilitywas improved by introducing debonding strips between the masonry wall and SHCC overlay. TheSHCC retrofitting scheme was numerically optimised using non-linear finite element (FE) analyses.Masonry was modelled using a smeared crack model with a Rankine-Hill material to allow for ten-sile softening and compression hardening followed by softening as well as an Engineering-Masonrymodel to allow for multi-directional loading. The interface between the masonry wall and SHCCoverlay was modelled using a Coulomb friction model. The SHCC overlay was modelled using aRankine-Rankine model with compression and tension hardening.

1 INTRODUCTION

There were no seismic design specificationsin the South African design codes before 1989,therefore there are many ULM buildings thatwere constructed in seismic areas [1].Thesebuildings need to be retrofitted in order to with-stand seismic events should they occur.

SHCC was proposed by [2] as a possiblebonded overlay retrofitting solution to increasethe ductility and shear resistance of ULM build-ings to better resist seismic action. De Beer[2] developed a sprayable SHCC mix to ap-ply on masonry walls. He conducted exten-sive experimental research, consisting of small

1

Mariska Kotze, Gert C. van Rooyen and Gideon P.A.G. van Zijl

masonry triplets tests with and without SHCCoverlay and larger shear dominant beam spec-imens, also with and without SHCC overlay.The aim of these tests were to obtain the modelparameters, namely the compression strengthof mortar and individual bricks, shear strengthof brick-mortar bond, SHCC-brick shear bondstrength, SHCC overlay shear strength, tensilestrength and Young’s modulus.

The ductility factor (µ), as seen in Equation1 was used to determine if the SHCC overlayimproved the overall ductility of the masonry.

µ =δuδy

(1)

δu is defined as the in-plane shearing deforma-tion when the post peak shearing resistance hasreduced by 80%, and δy is the yield shearingdeformation at the intersection of the elasic-plastic bilinear approximation of the load defor-mation response as defined by Mahmood andIngham [3]. The ductility factor of the beamspecimens, with and without SHCC overlay,tested by De Beer [2] were on average 1.8 witha CoV of 0.187. It was therefore concluded bythat although the SHCC increased the shear re-sistance of the beam specimens, the ductility re-mained unaffected, and further research had tobe conducted on how to improve the ductility ofthe SHCC overlay.

Research conducted by De Jager [4] con-sisted of setting up FE models for the SHCCretrofitted masonry walls to further test and de-velop debonding strips as a way to increase theductility of the SHCC overlay. Promising re-sults were obtained that are discussed in de-tail in section 5. An energy contribution factor(ECF), as seen in in Equation 2, was developedto numerically illustrate the improved energydissipation of the SHCC overlay with debond-ing strips in comparison to a fully bonded over-lay.

ECF =δsu − δsyδru − δry

(2)

δs referes to the wall with debonding strips, andδr refers to the wall without debonding srips.

The subscript y refers to the shearing displace-ment at a preselected shear force in the elasticregion, and the subscript u refers to the dis-placement at the same force in the plastic re-gion.

The experimental push-over tests conductedby [2] and [4] consisted of a 1150x935mm ma-sonry wall, as illustrated in Figure 1, that waseither single- (110 mm) or double-leaf (220mm). The size of the test wall was selectedto facilitate shear failure, as this is the fail-ure that typically occurs during seismic events.It is also the typical size of the wall betweenopenings in a building. It was supported at thetop and bottom with a 100x400 mm concretebeam with 175 mm overhang on either side. A345x345x89 H steel beam was placed om topof the concrete beam and the loads of the upperstories were applied to the steel beam throughthree rows of steel rods, two on each side, eachwith a cross-sectional area of 380mm2. Thedisplacement was directly applied to the mid-dle of the steel beam. The shear resistance ofthe wall was determined by displacing the steelbeam a total of 20mm at a rate of 2mm/min andmeasuring the reaction force at the point of dis-placement. The boundary conditions were fixedat the bottom, and out of plane displacementswere constrained. The top of the wall was fixedto the concrete beam by using strong mortar.

Figure 1: Basic setup of full scale shear wall test [2]

2

Mariska Kotze, Gert C. van Rooyen and Gideon P.A.G. van Zijl

2 REPRESENTATIVE BUILDINGThe aim of this study is to ultimately model

a typical full scale building, as seen in Figure 2,with and without SHCC overlay.

Figure 2: Typical low income residential building in CapeTown/Stellenbosch

A simplified modal analysis will be doneto determine the displacement of the buildingdue to an earthquake, using a quasi-static ap-proach. The bare building will be comparedto the retrofitted building to determine the op-timum configuration of the SHCC overlay.

3 NUMERICAL MODEL METHODOL-OGY

All the elements shown in Figure 1 weremodelled as 3D membranes except the steelrods that were modelled as 3D line trusses.

The finite element model was solved usinga non-linear phased analysis. Firstly the ownweight was applied to the model, followed bythe pressure in the rods representing the weightof the upper stories, and lastly the incremen-tal displacement was applied to the middle ofthe steel beam, in 0.2 mm increments. Theforce of the middle rod pair in the experimen-tal setup, done by De Beer [2], was kept con-stant by means of springs, while the forces inthe rods at the end increased as the horizontaldisplacement increased. Typical forces in therods are illustrated in Figure 3 over time, asthe displacement was applied in 2mm/min in-

crements. These forces were measured by DeBeer [2] in his experimental tests.

0 2 4 6 80

50

100

Time [min]

Rod

forc

e[k

N]

Middle rod

Second rod

Third rod

Figure 3: Rod forces

Masonry can be modelled either at macrolevel, that simulates the global behaviour of thestructure, or meso level, where the behaviour ofthe masonry structure is simulated in more de-tail, however this is a computationally intensiveapproach. The macro level approach was cho-sen as the aim is to obtain a global representa-tion of the model. The 15 mm SHCC overlaywas modelled using a rotating smeared crack-ing model, with parameters as illustrated in Ta-ble 1. With the high tensile fracture energy, thepseudo-plastic post-cracking tensile behaviourof SHCC is simulated.

Table 1: Rotating smeared crack material model parame-ters for SHCC

Symbol Value DescriptionE 15000N/mm2 Young’s modulusν 0.2 Poisson’s ratioρ 2100 kg/m3 Mass densityft 2.0 N/mm2 Tensile strengthGft 10 N/mm Fracture energy

To model masonry in Diana 10.2 there aretwo different material models namely, Engi-neering Masonry and Rankine-Hill. The Engi-neering Masonry material model is a smearedfailure model that can simulate cyclic loading ofmasonry structures. The Rankine-Hill material

3

Mariska Kotze, Gert C. van Rooyen and Gideon P.A.G. van Zijl

model is an anisotropic, multi-surface plasticitymodel that allows for tensile softening and com-pression hardening followed by softening. Itshould be noted that unloading in the Rankine-Hill model is elastic, and does not take crackclosure into account in unloading [5]. Boththese material models were used and compared.The parameters for the Rankine-Hill and Engi-neering Masonry material models are listed inTable 2 and 3 respectively. It should be notedthat the values were obtained by starting withexperimental values and adjusting them untila reasonably good correlation between the nu-merical model and the experimental tests wereobtained. A double-leaf wall (220 mm) wasused in this study, as the main focus is on loadbearing masonry walls which are typically 220mm wide.

Table 2: Rankine-Hill model parameters for masonry

Symbol Value DescriptionE 1800N/mm2 Young’s modulusν 0.22 Poisson’s ratioρ 2100 kg/m3 Mass densityftx 0.35 N/mm2 Tensile strength

(x)fty 0.22 N/mm2 Tensile strength

(y)fcux 20 N/mm2 Compression

strength (x)fcuy 20 N/mm2 Compression

strength (y)Gtx 0.15 N/mm Fracture energy

in tension (x)Gty 0.1 N/mm Fracture energy

in tension (y)Gcx 2 N/mm Fracture energy

in compression(x)

Gcy 2 N/mm Fracture energyin compression(y)

εu 0.001 Ultimate com-pressive strain

The interface between the masonry wall andSHCC overlay was modelled using a Coulomb

friction model, with cohesive strength of c = 2.3MPa and friction angle of φ = 1.08 rad. Theseparameters were determined by De Beer [2]from the standard triplet test. The debondingstrips discussed in section 5 were modelled as aweak interface between the SHCC overlay andthe masonry. The steel beam and the concretebeam were modelled as linear elastic material.

Table 3: Engineering Masonry model parameters for ma-sonry

Symbol Value DescriptionEx 1440N/mm2 Young’s modulus

(x)Ey 850N/mm2 Young’s modulus

(y)Gxy 350N/mm2 Shear modulusν 0.22 Poisson’s ratioρ 2100 kg/m3 Mass densityfty 0.15 N/mm2 Tensile strength

(y)ftr 0 N/mm2 Residual tensile

strengthfc 15 N/mm2 Compression

strengthGt 0.1 N/mm Fracture energy

in tensionGc 0.4 N/mm Fracture energy

in compressionGs 0.5 N/mm Fracture energy

in shearα 0.8rad Diagonal crack

anglec 2.3 N/mm2 Cohesion

3.1 MeshTo determine the appropriate element type

and mesh size, a mesh dependency analysiswas done. Quadratic quadrilateral and quadratictriangular elements were considered with nor-mal and high numerical integration respectively.The mesh dependencies are illustrated in Figure4 and Figure 5. The figures illustrate shear forceagainst mesh size, with the shear force taken atdifferent displacement points.

It was found that quadrilateral elements witha 2x2 integration scheme does not follow any

4

Mariska Kotze, Gert C. van Rooyen and Gideon P.A.G. van Zijl

significant trend, and therefore it does not de-liver reliable results. Using a 3x3 integra-tion scheme yields consistent results up to9.5mm displacement for mesh sizes smallerthan 30mm, as shown in Figure 4. Displace-ments larger than 9.5mm are however still meshdependent and unreliable.

101 101.2 101.4 101.6 101.8 102

160

200

240

Mesh element size [mm]

Shea

rfor

ce[k

N]

7.5 mm

9.5 mm

12 mm

20 mm

Figure 4: Quadrilateral element with 3x3 integration

The mesh dependency for quadratic triangu-lar elements was done only with high integra-tion. There is no difference between reduced,normal and high integration for quadratic trian-gular elements in Diana 10.2, as all these in-tegration schemes use 3 integration points [6].As shown in Figure 5, fairly consistent resultswere obtained for all displacements if the ele-ment size is 25 mm or less.

101 101.2 101.4 101.6 101.8 102

200

240

Mesh element size [mm]

Shea

rfor

ce[k

N]

7.5 mm

9.5 mm

12 mm

20 mm

Figure 5: Triangular element with high integration

Consequently, a triangular mesh size of

25mm with higher order integration was se-lected for the models, as illustrated in Figure6. The element is a six-node isoparametric tri-angular three dimensional plane stress element,referred to as a CT18GM element in Diana. Thetriangular mesh was also preferred above thequadrilateral due to the orientation of the meshelements having an effect on where shear fail-ure occurs. It was found by van Zijl et. al [7]that element sides are preferred directions forcomputational localization, therefore the trian-gular mesh will provide a better representationof the diagonal shear. The crack band width isdefined as the square root of the finite elementarea to ensure objective fracture energy dissipa-tion in crack formation through a single line offinite elements.

Figure 6: 25mm Triangular mesh of the shear wall model

4 EXPERIMENTAL AND NUMERICALRESULTS

Figure 7 illustrates the shear force againstdisplacement of two walls tested by De Beer [2]in comparison to the shear forces obtained fromthe non-linear finite element models, using bothRankine-Hill and Engineering-Masonry mate-rial models for the masonry. It can be seen thatboth the numerical models provide a good rep-resentation of the experimental tests, and reachsimilar maximum shear forces. The Engineer-ing Masonry model appears to approximate theascending branch better than the Rankine-Hillmodel.

5

Mariska Kotze, Gert C. van Rooyen and Gideon P.A.G. van Zijl

0 2 4 6 8 100

50

100

Displacement [mm]

Shea

rfor

ce[k

N]

Experimental

220-0-Rankine-Hill

220-0-Engineering Masonry

Figure 7: Shear force of masonry wall without SHCC

A 15 mm SHCC overlay was selected since a10 mm overlay resulted in premature failure anda 20 mm overlay results in a brittle failure withdelamination, as found by De Beer [2]. A 15mm SHCC overlay was applied to the walls andtested. The shear forces obtained for three ex-perimental tests, as well as the Rankine-Hill nu-merical model are shown in Figure 8. The shearforce for the two bare wall specimens are alsoincluded in the figure to illustrate the increasein shear resistance that the SHCC provides. TheSHCC overlay significantly increases the shearresistance of the masonry wall but the ductilityhas not been improved. De Jager [4] proposeddebonding strips as a means to improve the duc-tility of the masonry walls.

0 5 10 15 200

50

100

150

200

250

Displacement [mm]

Shea

rfor

ce[k

N]

Experimental with SHCC

Rankine-Hill

Experimental without SHCC

Figure 8: Shear force of masonry wall with SHCC

5 DEBONDING STRIPSThe aim of the debonding strips is to prevent

reflective cracking from the masonry wall to theSHCC overlay, cracking in the masonry thatreflects straight through to the SHCC overlay.As stress is transferred from the masonry wallto the SHCC overlay through the bonded ar-eas, multiple cracks form in the debonded areasand the stress is distributed more evenly overthe SHCC overlay [8]. Figure 9 illustrates thedebonding strips applied to the wall before theSHCC overlay is applied. Two diffrent layoutsof debonding strips were applied to the masonrywall, (i) 75 mm width strips with a spacing of150 mm, and (ii) 100 mm width strips with aspacing of 200 mm. Five walls were tested, onerepresentative bare wall and two walls of eachlayout of debonding strips.

Figure 9: Debonding strips on full scale tests [4]

The shear force against displacement for thewalls with debonding strips are illustrated inFigure 10, along with a representative bare walland the numerical results. It can be seen thatthe debonding strips can potentially improve theductility of the masonry walls significantly. Theductility factor of these walls is 2.4 on averagewith a 0.15 CoV. It should be noted that one ofthe four tests with strips failed prematurely dueto delamination in the bottom right-hand cor-ner of the wall. The delamination occurred dueto excessive compression in the SHCC overlay.This was prevented in the final test reported inTable 4 by cutting a 25 mm gap in the SHCC

6

Mariska Kotze, Gert C. van Rooyen and Gideon P.A.G. van Zijl

overlay at the top and bottom of the wall toprevent the concrete beam from inducing addi-tional compressive forces in the SHCC overlay.

Figure 10: Shear force of masonry with SHCC overlayand debonding strips

The ductility factors for the numerical andexperimental specimens are reported in Table4. The labeling of the specimens is accord-ing to the masonry wall thickness, SHCC over-lay thickness, debonding strip width, debondingstrip spacing, and in the case of multiple tests ofthe same specimen, the number of the test.

Table 4: Ductility Factors

Specimen δy δu µ Fp

(mm) (mm) (kN)N. - Numerical220-0 2.6 9.1 3.5 131.3220-15 2.9 7.3 2.6 259.9220-15:75-150 4.4 16.8 3.6 260.0E. - Experimental220-0 3.1 6.8 2.2 129.2220-15 8.8 13.2 1.5 241.7220-15:75-150:1 6.8 16.8 2.5 218.9220-15:75-150:2 8.8 17.6 2.0 243.8220-15:100-200 8.8 23.5 2.7 257.0

The energy contribution factors for the spec-imens are presented in Table 5 for the givenforces Fi (kN). ECF is an expression of energyand consists of the contributions of cracks in the

masonry and SHCC overlay, as well as shearslip in the interface between the masonry andthe SHCC.

Table 5: ECF values at different forces

Fi 75-150 (avg.) 100-200N. E. E.

175 1.54 1.31 2.28200 2.35 1.32 3.10225 2.31 1.94 3.36

6 CONCLUSIONComputational models to analyse ULM

walls with SHCC overlay have been devel-oped and verified by laboratory experiments.Fully bonded SHCC improve the shear resis-tance of the masonry walls and debonding stripsimproved the ductility of the masonry walls,by preventing reflective cracking and improv-ing the stress distribution throughout the SHCCoverlay. The following conclusions are drawn:

• The in-plane shearing resistance of 220mm ULM walls are increased by an av-erage of 100% by the application of a 15mm SHCC overlay.

• The ductility of a masonry wall is not sig-nificantly increased by the fully bondedSHCC overlay.

• Diagonal debonding strips can signif-icantly increase the ductility of theretrofitted masonry wall, provided thatdelamination can be prevented, withoutcompromising the shear resistance gainedby the SHCC overlay.

In addition to the modelling of an entire build-ing retrofitted with SHCC, different debondingconfigurations will be investigated, comprisingof but not limited to (a) strips in both direc-tions, (b) horizontal strips, (c) vertical strips, (d)debonding surface in the middle of the wall and(e) different strip widths and spacing.

7

Mariska Kotze, Gert C. van Rooyen and Gideon P.A.G. van Zijl

REFERENCES[1] SABS 0160. The structural use of masonry,

part 1: Unreinforced masonry walling.South African Bureau of Standards, SouthArica, 1989.

[2] Leon Rhoode De Beer. Developingand testing a sprayable overlay of StrainHardening Cement-based Composite forretrofitting of unreinforced load bearingmasonry walls. (October), 2016.

[3] Hamid Mahmood and Jason M. Ingham.Diagonal Compression Testing of FRP-Retrofitted Unreinforced Clay Brick Ma-sonry Wallettes. Journal of Composites forConstruction, 15(5):810–820, 2011.

[4] Dirk Jacobus Adriaan de Jager. Assess-ment of SHCC overlay retrofitting of unre-

inforced load bearing masonry for seismicresistance. (August), 2018.

[5] DIANA FEA BV. DIANA Theory manual.(March), 2006.

[6] DIANA FEA BV. User ’ s Manual ElementLibrary.

[7] G. P. A. G. van Zijl, R de Borst, andJG Rots. A numerical model for the timede-pendent cracking of cementitious materials.International Journal for Numerical Meth-ods in Engineering, 52(7):637–654, 2001.

[8] Yum Mook Lim and Victor C. Li. Durablerepair of aged infrastructures using trap-ping mechanism of engineered cementi-tious composites. Cement and ConcreteComposites, 19(4):373–385, 1997.

8