PERFORMANCE BASED SEISMIC ASSESSMENT OF UNREINFORCED MASONRY BUILDINGS IN NEW YORK CITY

Transcription

1 1NCEE Tenth U.S. National Conference on Earthquake Engineering Frontiers of Earthquake Engineering July 21-25, 214 Anchorage, Alaska PERFORMANCE BASED SEISMIC ASSESSMENT OF UNREINFORCED MASONRY BUILDINGS IN NEW YORK CITY J. Aleman 1 and G. Mosqueda 2 ABSTRACT A performance-based seismic assessment was conducted for an archetype unreinforced masonry (URM) building in New York City (NYC) with an emphasis on out-of-plane behavior. The probabilistic framework and new performance definitions provided by the ATC-58 project were used to conduct the assessment. Reliable macro-models for in-plane walls, wood diaphragms and out-of-plane walls were developed and validated using experimental test data available in the literature. These macro-models were used to develop out-of-plane URM fragility curves, a building-specific collapse fragility function, and estimate the seismic response of the building when subjected to moderate intensity ground shaking. Sample results are presented from the performance calculations that illustrate the probable loss and repair time for the archetype building. 1 Structural Engineer, Arup, Los Angeles, CA 966. Former Research Assistant, Department of Civil, Structural, and Environmental Engineering, University at Buffalo, The State University of New York, Buffalo, NY Associate Professor, Department of Structural Engineering, University of California San Diego, CA 9293 Aleman J, Mosqueda G. Performance based assessment of unreinforced masonry buildings in New York City. Proceedings of the 1 th National Conference in Earthquake Engineering, Earthquake Engineering Research Institute, Anchorage, AK, 214.

2 Performance Based Seismic Assessment of Unreinforced Masonry Buildings in New York City J. Aleman 1 and G. Mosqueda 2 ABSTRACT A performance-based seismic assessment was conducted for an archetype unreinforced masonry (URM) building in New York City (NYC) with an emphasis on out-of-plane behavior. The probabilistic framework and new performance definitions provided by the ATC-58 project were used to conduct the assessment. Reliable macro-models for in-plane walls, wood diaphragms and out-of-plane walls were developed and validated using experimental test data available in the literature. These macro-models were used to develop out-of-plane URM fragility curves, a building-specific collapse fragility function, and estimate the seismic response of the building when subjected to moderate intensity ground shaking. Sample results are presented from the performance calculations that illustrate the probable loss and repair time for the archetype building. Introduction Studies regarding the regional seismicity of New York City (NYC) indicate that earthquakes of magnitude greater than or equal to 5 have a 2-4% probability of occurring in a 5 year period [1]. Considering that a large population (19. million) is living in the region s old unreinforced masonry (URM) buildings that are not seismically designed, it is probable that even a moderate earthquake would have critical consequences on public safety and the economy of this area. The high vulnerability of URM buildings to out-of-plane damage and collapse has been observed in past earthquakes. Since the URM building stock in NYC shows evidence of missing joist anchorage, rotted wood joists and significant change of occupancy [2], the first stage of this study focused on estimating the performance of the buildings considering only this failure mode. A comprehensive study currently under way at the University at Buffalo will address additional modes of collapse and expand the fragility functions provided by the ATC-58 project [3]. This paper describes the implementation of this methodology, the numerical models developed to complete it, and presents preliminary results of the seismic performance of the archetype building. More detailed information can be found in [4]. Performance-Based Assessment Framework 1 Structural Engineer, Arup, Los Angeles, CA 966. Former Research Assistant, Department of Civil, Structural, and Environmental Engineering, University at Buffalo, The State University of New York, Buffalo, NY Associate Professor, Department of Structural Engineering, University of California San Diego, CA 9293 Aleman J, Mosqueda G. Performance based assessment of unreinforced masonry buildings in New York City. Proceedings of the 1 th National Conference in Earthquake Engineering, Earthquake Engineering Research Institute, Anchorage, AK, 214.

3 The ATC-58 framework explicitly accounts for all uncertainties involved in the process of estimating the seismic performance of a building. Uncertainties in the seismic hazard, construction materials, modelling and response of the building, damage states and its consequences are considered. The seismic performance is expressed in terms of deaths and serious injuries, repair and replacement costs, and business interruption time. It can be implemented in five steps, namely 1) assemble building performance model, 2) define earthquake hazards, 3) analyze the building response 4) estimate collapse fragility functions and 5) performance calculations. Fig. 1 illustrates the ATC-58 methodology and each step is briefly described below. Step 1: A building performance model is an organized description of building assets (i.e., structural system, non-structural components and occupancy); vulnerability of each of these assets; relevant demand parameters they will be subjected to during ground shaking; all potential damage they will experience due to these demands and the consequences of this damage. Step 2: Earthquake hazard is defined according to the assessment type considered. The methodology enables intensity-, scenario- and time-based assessments. An intensity-based assessment estimates the probable losses in a building subjected to a particular ground shaking intensity. Scenario-based assessment provides estimated losses for an earthquake of specific magnitude and location. Lastly, time-based assessment estimates probable annual losses considering all possible earthquakes and seismic sources. Step 3 and 4: Analyzing the building response involves computing demand parameters such as drift, floor accelerations, floor velocity, and residual drift. Nonlinear response-history analysis is usually utilized for calculating demand parameters and developing building-specific collapse fragility functions, which are required to determine potential casualties. Because the seismic response of the building is obtained with a limited suite of ground motions, additional performance calculations require generating simulated demands. This process is usually completed using Monte-Carlo analysis. Step 5: The simulated demands are used to determine whether collapse occurs, estimate damage, and compute casualties, repair costs and repair time. This process is usually completed using the Performance Assessment Calculation Tool (PACT), developed by the ATC-58 project. Figure 1. Performance-based assessment methodology (after ATC-58). Building Performance Model

4 Statistical information about the current building stock in New York City was obtained from a database developed by the NYCEM Project [1]. In this database, unreinforced masonry buildings accounted for 8% of structures in the area. The average length, width, number of stories and occupancy of the URM stock are illustrated in Fig. 2 Those average values of length and width were expected, since the layout of Manhattan closely follows the 1811 Commissioner s report that created the characteristic grid of avenues and streets in the city [5]. In establishing other construction details, the characteristics of common Row Houses were used. Wall thickness, diaphragm configuration, connections between wood joists and walls, windows and door sizes were obtained from relevant architectural and historical books. Limited information exists on the population models of the buildings and the default functions provided in the PACT software for residential and commercial buildings were utilized. The total replacement cost of the building performance model was calculated to be about US $75,. This value is similar to that estimated for URM buildings in Los Angeles [6]. A replacement time of 18 days was assumed. After past earthquakes, most owners have preferred to replace URM buildings with a reinforced concrete or steel frame structure; therefore the replacement cost and replacement time could be significantly larger. More details on the building performance model are presented in [4]. Figure 2. Building performance model for NYC (Dimensions in m). The Normative Quantity Estimation tool provided by the ATC-58 was utilized to define typical quantities, vulnerability and distribution of damageable components and contents for commercial and residential buildings. A user defined fragility and consequence function for out-of-plane URM walls was constructed using incremental dynamic analysis. In defining this function, two damage states were considered, namely, 1) onset of flexural cracking requiring cosmetic repairs (SD1) and 2) significant cracking along the height of the wall that may lead to partial or total collapse of the wall (SD2). Associated consequence functions were estimated from [7] where the cost of implementing several retrofit techniques is provided. Fig. 3 shows the 3D numerical model used in the simulations, and resulting preliminary fragility and consequence functions. An experimental setup that will be used to validate the numerical model is also shown.

5 Probability of Exceedance SD_1 SD_ Floor Acceleration (g) c) Out-of-plane Fragility Function 16 a) 3D Numerical Model Repair Cost ($) SD_1 SD_2 b) Experimental Model Repair Time (days) Wall Quantities 8 d) Repair Cost Consequence Function SD_1 SD_ Wall Quantities e) Repair Time Consequence Function Figure 3. a) Numerical model, b) Experimental model, c) Preliminary out-of-plane fragility function, d) Repair cost function, and e) Repair time function. Seismic Hazard For intensity-based assessment, ground shaking was represented by the 2% probability of exceedance in 5 years Uniform Hazard Response Spectrum (UHRS) for a site Class A located in New York City. The UHRS was obtained from the hazard tool provided by the USGS website. Table 1 presents a set of 1 ground motions that were selected and amplitude-scaled to match the UHRS. The spectral shape of this set is consistent with geologic characteristic of the Central and Eastern United States region, as shown in Fig. 4. The geomean spectrum of each ground motion pair was constructed over the period range of.17 seconds to 1.7 seconds, since the fundamental period T, of the performance building model was estimated equal to.85 seconds. The amplitude of each pair of ground motion was scaled by the ratio of Sa(T) obtained from the target spectrum to that of the geomean spectrum for the pair. More details on defining the seismic hazard for NYC are discussed in [4].

6 Table 1. Selected ground motions for intensity-based assessment. Earthquake Station Name Mw R (km) PGA (g) Gaza (US) Ball Mountain Dam Gaza (US) North Hartland Dam Gaza (US) North Springfield Dam Gaza (US) Union Village Dam Nahanni (CA) 698 Site Nahanni (CA) 699 Site Val-Des-Bois (CA) Orio Virginia (US) Corbin Virginia (US) Charlottesville Virginia (US) Reston Fire Station Spectral acceleration (g) Geomean response spectrum of selected ground motions Median response spectrum Target Uniform Hazard Spectrum Period (s) Figure 4. Target and median response spectrum of selected ground motions. Building Analytical Model A building analytical model is required to predict the response of the structure subjected to seismic loads. However, modelling older multi-story URM buildings poses a significant challenge due to the highly nonlinear behaviour of bricks, mortars and their interaction with other structural components. In this research, a reliable macro-model for masonry buildings supported by experimental testing was developed. This model accounts for the complex interaction between in-plane walls, wood diaphragms and out-of-plane walls. More details on modelling of components and their interaction can be found in [4]. In-plane walls were modelled using a modification of the Equivalent Frame method [8]. This method uses beam elements combined with nonlinear springs to simulate the seismic behaviour of individual piers. It accounts for rocking, sliding and shear failure modes. Rotational springs at top and bottom of the beam element simulate rocking whereas a shear spring at the middle captures the failure in shear. Currently, the method works only for 2D applications but ongoing research will incorporate a yield surface to account for bidirectional behaviour. The new method was validated using experiments from the literature [9, 1].

7 Wood diaphragms with straight-sheathing were modelled using a nonlinear finite element model with reduced degrees of freedom. The model is based in the nail-couple method suggested by ATC-7 [11] to estimate the in-plane shear strength of straight-sheathed diaphragms. Upon demonstration that the horizontal mechanical contact and friction between wood boards are negligible, the straight sheathing system was replaced with an equivalent Bernoulli beam element located at the center of the diaphragm. Nonlinear rotational springs were proposed to simulate the nail coupling at the connection between wood boards and joists. Comparison with refined finite element models and key experimental results from the literature [12] indicates that the MDOF model accurately predict the seismic response of this type of diaphragms at a fraction of the computational time. Out-of-plane walls were modelled using finite elements. A rigid shell element connected to rotational springs at assumed crack location captures the behaviour of the wall subjected to oneway bending. The moment-curvature of the nonlinear springs was obtained from simple rigid body motion equations and validated with one-way test data [13]. Refined finite element analyses are being conducted to simulate the cyclic behaviour of out-of-plane walls in two-way bending and to propose simplified macro-models. In developing the building analytical model, it is assumed that the out-of-plane walls are perfectly attached to wood diaphragms. Soil-structure interaction was ignored. Dead and live loads were applied to the structure as suggested by the ASCE-7 [14] for commercial buildings. Mean values of the material properties were estimated from experimental results of small specimens, in-situ flat jack testing on masonry buildings and previous research on URM walls. A numerical model of the walls and diaphragms was implemented in SAP2 as shown in Fig. 5. Nonlinear response-history analyses using the set of 1 ground motions were conducted using Newmark direct integration method. A damping ratio of 2% of critical damping was assumed for all modes of vibration. Collapse Fragility Analysis The URM analytical building model was used to establish a building-specific collapse fragility function. In developing the collapse fragility functions for the building, currently only the out-ofplane collapse of the walls was considered. Since the model does not explicitly captures the collapse of the walls, a collapse limit state is used to estimate this non-simulated collapse mode. For walls with aspect ratios h/t larger than 1, as in this case, the ABK methodology [15] specifies maximum velocities of the diaphragm of 25 in/s. Herein, collapse is assumed to occur when the maximum velocity of the diaphragm has exceeded this threshold. Following ATC-58 guidelines to estimate collapse fragility functions, each ground motion of the set previously identified was scaled up or down to generate 1 intensity levels. Then, nonlinear response-history analyses were performed using each intensity level. The conditional probability of collapse can be obtained by dividing the number of analyses for which collapse is predicted by the total number of analyses performed at that intensity. The conditional probability of collapse is then plotted as a function of intensity and a lognormal distribution is fitted to this data.

8 The resultant Incremental Dynamic Analysis (IDA) curves, the collapse fragility function with the median value of the spectral acceleration at the building s effective first mode period, Sa(T) and dispersion β are shown in Fig. 6 (a) and Fig. 6 (b), respectively. Figure 5. 3D Analytical Model of the URM Building in SAP2. (a) 1 (b) 1 Median Sa (T), g Diaphragm Velocity (in/s) Probability of Collapse Fit Fragility Median Sa=.43 Dispersion, β=1. Analyses Sa(T),g Figure 6. Collapse fragility analysis: a) IDA curves b) Building-specific fragility function for out-of-plane collapse. Building Response Nonlinear response history analysis was used to obtain peak drifts, peak absolute accelerations and peak absolute velocities for.1g ground shaking intensity. Since the geomean spectral shape of the scaled motions matches reasonably well with that of the target spectrum in the period range Tmin, to Tmax, the set of 1 records was expected to provide a good prediction of median response [3]. The reliability of the nonlinear URM model was checked by first conducting response spectrum analysis, nonlinear static lateral and gravity analysis and nonlinear analysis with and without P-Delta effects. The hysteretic response was verified for selected inplane, diaphragms and out-of-plane elements. To account for uncertainties in building definition and construction quality, a dispersion factor βc of.4 was assumed. Likewise, a dispersion factor βq of.4 was also assumed for model quality and completeness. Residual drifts were not currently considered in this study. A extended discussion on the building response can be found in [4].

9 Peak transient drift ratios and peak floor accelerations are presented in Table 2 and Table 3. Note that those parameters were estimated only for one direction of analysis. Two of the ground motions predicted collapse and therefore were removed from the simulation, as recommended in [3] Table 2. Estimated values of peak floor accelerations (g). Floor/Story EQ1 EQ2 EQ3 EQ4 EQ5 EQ6 EQ7 EQ8 Roof (g) Floor 5 (g) Floor 4 (g) Floor 3 (g) Floor 2 (g) Floor 1 (g) Table 3. Estimated values of peak story drift ratios (rad). Floor/Story EQ1 EQ2 EQ3 EQ4 EQ5 EQ6 EQ7 EQ8 Floor5-Roof E Floor E E Floor E E E-5 Floor E E Floor Performance Calculations and Results PACT uses the median drifts, accelerations and velocities to make damage state assessments for structural and non-structural components specified in the building performance model. This process requires fragility and consequence functions for each component considered in the building performance model. Before conducting performance calculations, a large number of simulated demands or demand distributions are generated accounting for variability of each random parameters and its consequence on the seismic response. Each set of demands, associated damage and consequences is called a realization. The performance assessment starts by determining the number of people present in the building when the earthquake occurs. This information will be used to determine the number of casualties if collapse occurs. PACT will then estimate for each realization if collapse has occurred. If so, the collapse mode of the structure must also be determined. Herein, the collapse mode is out-ofplane walls failure at any story of the building. The numbers of casualties are then estimated based on the number of people present in the building and the fraction of floor area at each floor subjected to collapse, in this case 1%. Finally, repair cost and repair time are set equal to the building replacement values, namely $75, and 18 days. If PACT determines that collapse has not occurred, simulated demands are used together with fragility functions to generate damages states for each structural and non-structural component in

10 the building. Those damage states are then used together with consequence functions of repair cost, casualties and repair time to estimate the associated losses for each realization. This process is repeated for each realization. Repair cost and repair time function are given in Fig. 7 and Fig. 8, respectively. These preliminary results indicate that if an earthquake of.1g intensity happens in NYC, the median repair cost for the archetype building will be US$75, which is mainly due to out-of-plane damage of the walls, as seen in Fig. 7 (b). This is 1% of the building s total replacement cost. Likewise, the mean repair time will be 123 days, as most of the damage will occur in upper stories, as implied from Fig. 8(b). Note that for this earthquake intensity, PACT did not predict deaths or serious injuries.. P (Total Repair Cost<= $C) (a) Lognormal Fitted Curve Binned Values $ C (U.S. Dollars) Cost (U.S. Dollars) (b) Performance Groups 1. Out-of-Plane Walls 2. URM Chimneys 3. Wall Partition type 1 4. Stairs 5. Wall Partition type 2 6. Ceiling 7. Piping 8. HVAC 9. Distribution Panel Performance Groups Figure 7. Repair cost results: a) Loss curve for cost associated with Sa=.1g b) Disaggregation of losses among performance groups for a 5% of probability of exceedance of the total repair cost. P (Total Repair Time<=Days) (a) Lognormal Fitted Curve Binned Values Time (days) (b) Floors Performance Group Out-of-Plane Walls Time (Days) Figure 8. Repair time results: a) Repair time probability b) Contribution of each performance group to the total repair time for a 5% of probability of exceedance of the total repair cost. Conclusions and Upcoming Work The preliminary results presented in this paper indicate that even a small intensity earthquake will have significant cost and business impacts in NYC. Ongoing data processing of a shake table test of two unreinforced masonry walls will provide additional data to properly model outof-plane walls and diaphragm connections. The validated numerical model will be used to improve fragility functions, estimates of repair cost and consequence functions.

11 Acknowledgements Financial support for the study described in this paper was provided by MCEER, the State of New York and ARUP. The Structural Engineers Association of New York (SEAONY) is collaborating with MCEER on this research project and their support is gratefully acknowledged. References 1. Tantala, M., et al., Earthquake Risks and Mitigation in the New York/New Jersey/ Connecticut Region, 23, NYCEM Summary Report Eschenasy, D., Cases of Failure of Unreinforced Brick Walls Due to Out-Of-Plane Loads, Structure Magazine, 211, May 211. p Applied Technology Council, Seismic Performance Assessment of Buildings Volume 1- Methodology, Aleman, J., Performance-Based Seismic Assessment of Unreinforced Masonry Buildings in New York City, PhD Thesis, 214, University at Buffalo, Buffalo, NY. 5. Friedman, D., Historical Building Construction: Design, Materials, and Technology. 1st ed, W.W. Norton, 1995, New York. 6. USGS, The Shake Out Scenario-Unreinforced Masonry (URM) Buildings, 28, Pasadena CA. 7. McMonies, W.W., Portland's Unreinforced Masonry Apartment Buildings: A Threatened Species? Center for Real Estate Quarterly Journal, 21. 4(3): p Calvi, G.M. and G. Magenes. Seismic evaluation and rehabilitation of masonry buildings, Proceedings of the U.S. - Italian workshop on seismic evaluation and retrofit, 1997, Buffalo, New York. 9. Manzouri, T., et al., Repair and Retrofit of Unreinforced Masonry Structures. Earthquake Spectra, (4): p Yi, T., et al., Lateral Load Tests on a Two-Story Unreinforced Masonry Building. Journal of Structural Engineering, (5): p ATC, Guidelines for the Design of Horizontal Wood Diaphragms. H.J. Brunnier Associates, Peralta, D.F., J.M. Bracci, and M.B.D. Hueste, Seismic behavior of wood diaphragms in pre-195s unreinforced masonry buildings. Journal of Structural Engineering, (12): p Griffith, M., et al., Experimental Investigation of Unreinforced Brick Masonry Walls in Flexure. Journal of Structural Engineering, (3): p ASCE/SEI 7-1, Minimum Design Loads for Buildings and Others Structures, 21, Reston, Virginia, American Society of Civil Engineers. 15. ABK, Methodology for Mitigation of Seismic Hazards in Existing Unreinforced Masonry Buildings : The Methodology, 1984, El Segundo, CA