Shaking Table Tests on the Large Scale Model of Mustafa Pasha Mosque Without and with FRP

Shaking table tests on the large scale model of Mustafa Pasha Mosque without and with FRP

L. Krstevska, Lj. Tashkov & K. Gramatikov University “St. Cyril and Methodius”, Skopje, Republic of Macedonia R. Landolfo, O. Mammana, F. Portioli & F.M. Mazzolani University of Naples “Federico II”, Naples, Italy

ABSTRACT: Shaking table tests were carried out on a large scale model of Mustafa Pasha Mosque in Skopje within the activities of Sixth Framework Program PROHITECH – “Earthquake protection of historical buildings by reversible mixed technologies”. The main aim of experimental investigation was to evaluate the seismic stability of the monument after applying a reversible technology for strengthening. Beside the experimental shaking table testing of the model, three-dimensional finite element analyses were carried out using the general purpose software package ANSYS. The numerical models were calibrated on the basis of mechanical properties of masonry obtained from compression and shear tests on wall specimens built with same technique used for the original prototype. The comparison between experimental and numerical results allowed the reliability of the implemented models to be evaluated. Finally the effectiveness of strengthening and the evolution of crack patterns observed during each phase of the testing programme has been analyzed on the basis of the obtained results.

1 INTRODUCTION strengthening schemes based on the calculated tensile stress distributions. Shaking table testing of the large scale model of the To assess the seismic capacity of the structure Mustafa Pasha mosque in Skopje, (15th century) was and the effectiveness of FRP reinforcement, nonlin- performed in the Laboratory of the Institute of Earth- ear pushover analyses were carried out by means of quake Engineering and Engineering Seismology in the implemented finite element model. In the follow- Skopje in the period November-December 2006. The ing, a detailed description of the implemented models main objective of the tests was to investigate experi- and of the corresponding obtained results is also mentally the effectiveness of the proposed reversible reported. technology for strengthening of this type of historical monuments. The performed testing was a part of the activities within the Sixth Framework Programme 2 EXPERIMENTAL INVESTIGATION PROHITECH – “Earthquake Protection of Historical Buildings by Reversible Mixed Technologies”. The model of the mosque was constructed to a length In order to predict the response of the large scale scale of 1:6 according to the “gravity forces neglected” model during the shaking table tests performed at the modeling principle, using the same materials as those IZIIS Laboratory in Skopje, a finite element model of the prototype: stone, bricks and lime mortar (Gra- was implemented. matikov et al. 2007). The experimental testing was The main aim of numerical investigations was to performed on the bi-axial seismic shaking table at the evaluate the peak ground accelerations to be assigned IZIIS’ Laboratory. Following the main objective, the to the shaking table during the test phases of exper- testing was performed in three phases: imental programme that were devoted to the assessment of the seismic strength of both the original and 1) Testing of the original model for low intensity level, strengthened structure. In particular, it was important to provoke small damage; to control the damage on the original Minaret and 2) Testing of repaired model and strengthened minaret on the Mosque in order to prevent collapse before with FRP, until total collapse of the minaret; the application of FRP strengthening. Moreover, the 3) Testingof strengthened model with FRP and carbon numerical models were used to determine suitable FRP fiber bars until collapse.

383 Figure 2. Damage to the minaret (horizontal crack) and of the mosque after phase 1.

Figure 1. Dimensions and instrumentation of the large scale model of Mustafa Pasha Mosque.

To investigate models’ response during seismic action different types of transducers were placed at characteristic points – accelerometers, linear poten- tiometers and LVDTs,as shown in Figure 1 (Krstevska et al. 2007).

2.1 Phase 1 – Testing of the original model Before starting the seismic testing, the dynamic char- acteristics of the model were defined by means of ambient vibration method as well as by low intensity random excitation within the frequency range Figure 3. Characteristic response time histories at the top of 0.1–50 Hz. The accelerogram of the Montenegro of the model for input intensity 10%g, original model. earthquake of 1979 (Petrovac N-S component) was selected as a representative earthquake excitation. The damaged model is presented in Figure 2, while The earthquake time histories were scaled (com- some representative response parameters for input pressed) 6 times, according to the similitude require- intensity of 10% g are given in Figure 3. ments and several tests were performed with intensity of 0.01–0.10 g. 2.2 Strengthening of the minaret Under input intensity of 2% g, the first horizontal crack appeared at the base of the minaret. In the next After performing the seismic tests on the original tests performed under intensities of up to 10% g, dam- model of the mosque, the damaged model was repaired age occurred on the mosque, too. The reason of this by crack injection and the minaret was strengthened was the frequency content of the applied excitation, by application of CFRP wrap, cut to have the corre- which was close to the natural frequencies of both the sponding width and applied upon a layer of epoxy glue minaret and the mosque. (Fig. 4).The obtained strips with a width of 15 cm were

384 Figure 5. Collapse of the upper part of the minaret.

Figure 4. Repaired model and strengthened minaret after phase 1 testing. placed on four sides along the length of the minaret in vertical direction for the purpose of its stiffening. To confine the structure, strips with a width of 10 cm were placed at four levels along the height of the minaret in horizontal direction, while a strip with a width of 20 cm was placed at its base.

2.3 Phase 2 – Testing of repaired model and strengthened minaret Before the seismic tests, the dominant frequencies of the model were checked by random excitation. Figure 6. Damaged model after phase 2 testing. For the minaret dominant frequency was 4.7 Hz, while for the mosque two frequencies were dominat- ing: f = 7.4 Hz and f = 9.6 Hz. The characteristic response parameters during the During this testing phase, 11 tests were performed seismic tests are given in Table 1, while the represen- with input acceleration between 0.2 g to 1.5 g.The first tative time histories for the final test with an input cracks on the minaret occurred under input intensity of intensity of 1.5g are given in Figure 7. 0.34 g, which indicated that the applied strengthening enabled stiffening of the minaret and increasing of its 2.4 Strengthening of the model bending resistance. The initial cracks in the mosque appeared under input intensity 0.42 g. After the final test of the model in phase 2 the During the next tests, the cracks developed and at minaret was removed and the mosque rebuild on input level of 0.49 g, the upper part of the minaret damaged parts. Then the model was strengthened. totally collapsed (Fig. 5). The main adopted principle in strengthening of the After reaching the max input acceleration of 1.5 g, model was that the methodology to be applied be the mosque model was heavily damaged and the testing reversible and invisible. Hence, the cracks in the dam- was stopped. aged model were not repaired by injection, but a There were cracks in the walls and also in the dome concept was adopted that the model be strengthened (Fig. 6). to the conditions after the preceding tests.

385 Table 1. Performed seismic tests, phase 2.

Inp. Top Top Input Top acc acc. acc. displ. displ. Test No (g) CH1 CH3 (mm) (dome)

1 0.2 0.65 0.4 2.24 2.3 2 0.28 0.82 0.6 3.3 3.3 3 0.34 1.7 0.7 4.0 4.0 4 0.42 – 1.0 4.7 5.0 5 0.49 – 0.88 5.48 6.4 6 0.53 – 0.9 6.1 7.5 7 0.58 – 1.0 7.0 8.9 8 0.65 – 0.96 9.1 12.4 9 1.05 – 1.1 11.4 14.7 10 1.4 – 0.7 13.9 18.8 11 1.5 – 0.54 15.3 22 Figure 8. The strengthened model ready for testing, phase 3.

course was formed whereby the tensile resistance of the wall was improved and synchronous behaviour of the bearing walls was achieved. • Formation of a horizontal belt course around the tambour by applying a CFRP wrap with a width of 10 cm. • Formation of a horizontal belt course at the base of the dome by use of a CFRP wrap with a width of 50 cm. The formation of these horizontal belt courses enabled better integrity of the tambour and the dome base and prevented “opening” of the dome which was the most common reason for occurrence and prolongation of cracks in the bearing walls. The strengthened model is presented on Figure 8.

2.5 Phase 3 – Testing of strengthened mosque model Before the seismic tests, the dominant frequencies of the model were checked. The resonant frequency of 9.2 Hz was obtained and it was compared to the frequency of 8.6 Hz measured after the last test of testing in phase 2. This pointed out that strengthening had increased the resonant frequency of the model for about 8%, meaning Figure 7. Response during the test with intensity 1.5 g, that the stiffness of the model hadn’t been recovered phase 2. completely compared to the initial state. During this phase, 24 tests were performed The strengthening consisted of incorporation of under input acceleration between 0.15 g to 1.5 g. horizontal belt courses for the purpose of increasing The accelerogram of the Petrovac earthquake (N-S the integrity of the structure at those levels and pro- component) was scaled by 6 (compressed) in the first viding as better as possible synchronous behaviour of 15 tests. During the tests performed under input inten- the bearing walls: sities between 0.15 g and 0.40 g. the model’sbehaviour • Incorporation of carbon rods in two longitudinal was stable, without provocation of large cracks. In the mortar joints around the four walls at two levels: the next 6 tests carried out under input acceleration of level above the openings and at the top of the bear- 0.60–0.80 g, the dome experienced sliding at a visible ing walls, immediately below the tambour. With the horizontal crack at its base. To provoke more intensive incorporation of these carbon rods, a horizontal belt response of the model, in the next tests, a time scaling

386 Table 2. Performed seismic tests, phase 3.

Input Scal. Input acc Top acc. displac. Top displ. factor (g) (dome) (mm) (dome)

6 0.14 0.30 1.5 1.5 0.18 0.35 2.0 2.0 0.25 0.42 3.0 3.0 0.29 0.50 3.5 3.5 0.35 0.59 4.0 4.3 0.38 0.65 4.7 5.0 0.40 0.75 – – 0.42 0.85 6.8 7.7 0.67 1.60 7.2 8.5 0.87 1.70 10.6 14.0 0.82 1.55 10.0 14.4 Figure 9. Damage of the model after phase 3 tests accom- 0.90 1.70 11.0 16.0 plishment. 0.80 1.3 12.6 16.7 3 0.20 0.60 7.8 13.0 factor of 3 was used, producing an input acceleration 0.46 0.93 15.0 22.0 of 0.46–1.5 g. In this series of tests, many new cracks 1.20 1.10 25.0 26.8 1.5 1.0 30.0 40.0 appeared in the walls as well as in the dome, decreas- = 2 0.15 0.40 8.4 12.0 ing the dominant frequency of the model to f 4.4 Hz. 0.75 0.70 27.0 35.0 This value was more than twice lower compared to the 1.00 0.80 45.0 52.0 initially measured frequency of 9.2 Hz, thus indicating 1 0.35 0.53 58.0 75.0 a pre-collapse state of the model. The next two tests were performed by a scaling factor of 2, under an input acceleration 0.75–1.0 g. analyses were performed for the evaluation of struc- Progressive cracks appeared, but still without collapse. tural capacity, both for original and reinforced model The final test was performed by a scaling factor of (Landolfo et al. 2007). 1, under an input acceleration of 0.35 g. Heavy damage The numerical model was generated by importing occurred in many places on the dome and around the in the FE program a three dimensional solid model of openings as well as big cracks and inclination of one the Mosque that was implemented in a computer aided corner of the model, including damage to the strength- design system. ening FRP belt at the lower level of one of the walls, In order to properly evaluate the structural inter- after which the testing was stopped. actions among the different parts, the implemented The damaged model is presented in Fig. 9.The char- geometrical model reproduces all the main elements acteristic response parameters during the seismic tests of the building accurately, including the openings and are given in Table 2. The representative time histories the pendentives connecting the walls with the dome. in the final tests are given in Fig. 10. Two separate FE models were implemented for the As a common conclusion after the performed exper- Minaret and the Mosque, respectively (Fig. 11). In the imental testing of the mosque model in phase 3, it last case, the symmetry of the model along the vertical might be said that the model’s behaviour was evi- plane parallel to the direction of the input displacement dently different in respect to that of the original model. was considered, in order to save CPU time for solving Under tests of moderate intensity, the existing cracks non-linear equations. were activated but during the subsequent more inten- To model the reinforcement, the outer surfaces of sive tests, the failure mechanism was transferred to the the Mosque and Minaret were divided into different lower zone of the bearing walls, in the direction of the areas corresponding both to the FRP sheets and bars. excitation, where typical diagonal cracks occurred due The properties of overlapping areas were set up in to shear stress. order to match the meshes corresponding to masonry and FRP reinforcement elements. The whole masonry structure was discretized with tetrahedral 3D solid ele- 3 FE ANALYSIS ments considering a mesh size of 100.0 mm in the case of Mosque model. 3.1 Types of analysis, geometric modeling and SOLID45 elements were used for masonry and meshing SHELL181 elements were employed to model the FRP The finite element analysis of the large scale model sheets. The shells corresponding to the FRP reinforce- was carried out using the ANSYS software. Push-over ment were directly overlapped to the masonry bricks

387 Table 3. Masonry material properties assumed in numerical simulations of Minaret and Mosque.

γ Ecϕδ [kN/m3] [MPa] ν [MPa] [◦][◦]

Minaret 19.00 16000 0.25 0.13 60.57 60.57 Mosque 19.00 680 0.25 0.09 69.13 69.13

and no interface elements were considered. The parts of the Mosque model reinforced with bars were mod- elled with elastic solid elements through the whole thickness of the shear walls with Young’s modulus equal to that of FRP material.

3.2 Material modeling The material properties used for the two finite element models are reported inTable 3. In particular, the elastic parameters are referred to the values that have been calibrated on the basis of first random vibration tests performed on the original undamaged structure. The strength properties were determined on the basis of both compression and shear experimental tests carried out on masonry wall samples. As far as structural masonry, the elastic-perfectly plastic Drucker-Prager material model was considered. In the case of Mosque, cohesion c and angle of inter- nal friction ϕ were calibrated in order to obtain values of 0.05 MPa and 1.0 MPa for tensile (ft ) and compres- sive strengths (fc). For Minaret, the considered values for ft and fc were 0.1 and 1.0 MPa. In order to assign an associative flow rule for plastic strains, the assumed Figure 10. Response of the model during the last test with dilatancy angle δ was equal to φ. With regard to com- intensity 0.35 g, scaling factor 1, phase 3. posites, an elastic material model was considered. In particular, the adopted Young’s modulus for sheets is equal to 240 GPa and the considered equivalent thickness is 1.0 mm, according to the nominal mechanical properties provided by the manufacturer.

3.3 Load modeling and boundary conditions With regard to the load modelling approach implemented in pushover analyses for the evaluation of seismic capacity, a uniform and a linear acceleration distribution along the horizontal direction was applied to the FE models of Mosque and Minaret, respectively. As far as the boundary conditions are concerned, full restraints were assumed at the base of the structure in the performed analyses.

3.4 Results of the non-linear pushover numerical analyses Figure 11. The FE model of the Mosque and Minaret The assessment of the ultimate seismic strength of implemented for non-linear pushover analysis. both the original and reinforced large scale model was

388 Figure 12. Distribution of first principal plastic strains on Figure 13. Distribution of first principal plastic strains on the original large scale model of the Minaret at collapse load. the reinfroced large scale model of the Minaret at collapse load. carried out by means of nonlinear pushover analyses performed on the implemented FE model. On the basis of obtained numerical results, the evolution of damage distribution till collapse load on the investigated prototype has to be ascribed to the attainment of tensile or shear strength, while the com- pressive resistance is never exceeded. The collapse loads corresponding to the original and reinforced prototype were determined by checking the attainment of the maximum plastic strain calculated for masonry on the basis of the FE models calibrated against shear tests. With regard to the Minaret, the results of numerical analysis show that the reinforcement increases the ultimate strength in terms of peak acceleration at the top from the value of 0.3 g corresponding to the original model to 1.2 g and the collapse mechanism shifts to the upper part without strengthening (Figs 12, 13). Figure 14. Distribution of first principal plastic strains on As far as the Mosque is concerned and accord- the original large scale model of the Mosque at collapse load. ing to the implemented numerical model, the original prototype collapses with a mixed pier/spandrel mech- out plane horizontal loads. In the second step, the dam- anism of the vertical bearing structures, that is typical age extends to the upper spandrels, between the second of weakly coupled perforated walls (Fig 14). On the and third row of opening in the shear walls parallel to basis of the numerical results, the evolution of collapse the direction of seismic loads. In particular, it develops mechanism can be divided in different phases.Accord- up to the dome, among the openings in the support- ing to the numerical model, the first diagonal tension ing polygonal drum and the ones in the shear walls. cracks occur in the shear walls, namely in the spandrels Finally, the seismic strength of the structure is attained between the first and second row of openings from the when the central wall at the base of the Mosque and basement. In this phase, the damage also develops at the lateral piers collapse for shear and bending mecha- the base of the walls perpendicular to the direction of nisms, respectively. The study of collapse mechanism, ground motion, owing to bending stresses induced by obtained from numerical analysis on both original and

389 1.8

1.6 original minaret strengthened minaret 1.4 Exp. Exp. FE FE 1.2

0.8 Top Acc. [g] 0.6

0.4

0.2

0 0 1 2 3 4 5 6 7 8 Relative Displ. [mm]

Figure 15. Distribution of first principal plastic strains on Figure 16. Comparison between experimental and numer- the reinforced large scale model of the Mosque at collapse ical response for the original and strengthened Minaret load. model.

1.8 reinforced Mosque, allowed the effectiveness of FRP reinforcement to be analyzed. The wraps around the 1.6 dome and the top of shear walls fully prevent the prop- 1.4 agation of cracks from the bottom part to the drum, as shown in Figure 15. 1.2 The role played by FRP bars in the shear walls is to 1 stiffen and strengthen the spandrels forming a sort of reinforced masonry beams at different levels able to 0.8 distribute the seismic action among the piers. Top Acc. [g] 0.6 According to the numerical model of the reinforced 0.4 original mosque strengthened mosque Mosque, in this case the collapse mechanism turns Exp. Exp. from a mixed into a weak piers/strong spandrel type. 0.2 FE FE 0 0 1 2 3 4 5 6 7 Relative Displ. [mm] 4 COMPARISON BETWEEN EXPERIMENTAL AND NUMERICAL RESULTS Figure 17. Comparison between experimental and numerical response for the original and strengthened Mosque The results of FE model were compared with experi- model. mental tests. The distribution of first principal plastic strains predicted by the numerical model fits well the crack patterns observed on the large scale model of The comparison between experimental and numer- the Mosque during shaking table tests. As shown in ical responses in terms of acceleration and relative Figures 6 and 9, in the case of the original Mosque displacement at the top is shown in Figures 16 and 17. cracks formed on the spandrels between the open- Also in this case, the numerical and experimental ings up to the tambour, while in the reinforced model results are in good agreement. cracks on the bearing walls were observed at the base In the case of Minaret, it should be noted that the of the structure, according to the predicted damage experimental value of 0.2 g on the original model cor- pattern. responds to first observed cracks and to the attainment With this regard, it should be noted, however, that of first plastic strains in the numerical model. sliding of the dome at tambour opening level has been However, during the testing the original Minaret did observed before the collapse of the shear walls. not collapse for values of peak acceleration at the top This collapse mechanisms is not predicted by up to 0.7 g and the structure exhibited a rigid block numerical results and can be ascribed to a different behaviour with rocking. quality of masonry at the base of the dome from other Also in the case of the strengthened Minaret the parts of the Mosque. experimental peak accelerations at the top are higher

390 than numerical ones. Considering that a non-linear in terms of ultimate seismic capacity and collapse static FE analysis was performed, this discrepancy can mechanisms. Further studies are in progress, mainly be ascribed to cyclic loading induced by seismic exci- devoted to refine the calibration of FE model by imple- tations and to the rigid block-type behaviour of the menting non linear time history analysis with other Minaret observed during testing. types of material models for masonry, such as smeared crack.The calibrated non-linear numerical models will represent a reliable tool both for the design optimiza- 5 CONCLUSIONS tion of FRPs and for the evaluation of the effectiveness of other types of strengthening on the original Mosque. In this study, a numerical and experimental investigation on the seismic strength of the masonry large scale model of the Mustafa Pasha Mosque reinforced with ACKNOWLEDGMENTS FRP has been presented. The experimental testing of Mustafa pasha large It is significantly acknowledged the financial support scale model was performed to investigate the seismic of the European Commission (grant No. INCO-CT- stability of the monument after applying a reversible 2002-509119), for funding the research project PRO- strengthening methodology, both for the minaret and HITECH (Earthquake PROtection of HIstorical Build- for the mosque. ings by Reversible MixedTECHnologies), which is the The strengthening applied to the minaret enabled main framework of the experimental and numerical stiffening and increasing of its bending resistance. activity presented in this paper. The mosque model’s behaviour after strengthening was evidently different in respect to that of the original model. Under tests of moderate intensity, the existing REFERENCES cracks were activated but during the subsequent more intensive tests, the failure mechanism was transferred Gramatikov, K., Taskov, Lj., Krstevska, L. 2007. Final report on “Shaking table testing of Mustafa-Pasha to the lower zone of the bearing walls, in the direc- mosque model”. FP6-PROHITECH Project, Workpack- tion of the excitation, where typical diagonal cracks age 7-Experimental analysis. University “Sts.Cyril and occurred due to shear stress. Methodius”, Skopje, Macedonia. In order to support the experimental test set-up, to Krstevska, L., Taskov, Lj., Gramatikov, K., Landolfo, R., design the FRP reinforcements and to analyze their Mammana, O., Portioli, F., Mazzolani, F.M. 2007. Exper- effects on the prototype, a finite element analysis was imental and numerical investigations on the Mustafa performed. Pasha Mosque large scale model. In Proceedings of Cost In particular, the implemented finite element model Action C26Workshop“Urban habitat constructions under allowed the evolution of partial and global collapse catastrophic events”, Prague 30–31.3.2007, 158–169. Landolfo, R., Portioli, F., Mammana, O., Mazzolani, F.M. mechanisms observed during the tests to be analyzed. 2007. Finite element and limit analysis of the large scale The results of numerical investigation have been com- model of Mustafa Pasha Mosque in Skopje strengthened pared with the experimental outcome. Generally, a with FRP. In Proceedings of the First Asia-Pacific Con- good agreement between the behaviour predicted by ference on FRP in Structures (APFIS 2007). Hong Kong, numerical models and test results was observed, both China, 12–14 Dec. 2007.

391