ISSARS: An integrated software environment for ... - Semantic Scholar

10 downloads 2165 Views 3MB Size Report
Aug 29, 2012 - ... the efficiency of earthquake ground motion selection by accounting for ..... of the finite element program SAP2000 was used to provide access.
Advances in Engineering Software 58 (2013) 70–85

Contents lists available at SciVerse ScienceDirect

Advances in Engineering Software journal homepage: www.elsevier.com/locate/advengsoft

ISSARS: An integrated software environment for structure-specific earthquake ground motion selection Evangelos I. Katsanos, Anastasios G. Sextos ⇑ Department of Civil Engineering, Aristotle University Thessaloniki, Greece

a r t i c l e

i n f o

Article history: Received 29 August 2012 Received in revised form 19 December 2012 Accepted 17 January 2013

Keywords: Software advancements API functions Earthquake engineering Strong ground motion selection and scaling Time history analysis Structural response

a b s t r a c t Current practice enables the design and assessment of structures in earthquake prone areas by performing time history analysis with the use of appropriately selected strong ground motions. This study presents a Matlab-based software environment, which is integrated with a finite element analysis package, and aims to improve the efficiency of earthquake ground motion selection by accounting for the variability of critical structural response quantities. This additional selection criterion, which is tailored to the specific structure studied, leads to more reliable estimates of the mean structural response quantities used in design, while fulfils the criteria already prescribed by the European and US seismic codes and guidelines. To demonstrate the applicability of the software environment developed, an existing irregular, multi-storey, reinforced concrete building is studied for a wide range of seismic scenarios. The results highlight the applicability of the software developed and the benefits of applying a structurespecific criterion in the process of selecting suites of earthquake motions for the seismic design and assessment. Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction The evolution in computational power and the parallel processing capabilities of modern engineering software make nowadays the use of complicated structural analysis methods an attractive alternative for the design and assessment of structures. In contrast to the past, when the elastic static or response spectrum analysis was almost exclusively used for the seismic design of structures, the state-of-practice has progressively moved toward dynamicelastic, nonlinear-static (i.e., single mode or multi-modal ‘‘pushover’’ analysis), and even nonlinear time history analysis. The latter, capturing more efficiently the hierarchy of failure mechanisms, the energy dissipation, the force redistribution among the structural members and contact issues (such as gap, impact, sliding and uplift) is preferable in cases of significant material or geometrical nonlinearities and as such, is used for the design of seismically isolated buildings and bridges or the assessment of existing structures with various degrees of damage. Elastic time history analysis is also extensively used, primarily for structures whose response is dominated by higher or closely spaced modes (mostly tall and irregular buildings and towers), or structures of high importance that are typically designed to remain elastic even ⇑ Corresponding author. Address: Department of Civil Engineering, Aristotle University Thessaloniki, Division of Structural Engineering, 54124 Thessaloniki, Greece. E-mail address: [email protected] (A.G. Sextos). URL: http://www.asextos.net (A.G. Sextos). 0965-9978/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.advengsoft.2013.01.003

for long return-period earthquake intensities (i.e., industrial facilities, power plants, dams, critical administrative buildings, etc.). In all cases, the main task of the design procedures is to achieve more predictable and reliable levels of safety and operability against different levels of seismic intensity [1], a framework known as performance-based design and assessment. Despite the above major advances made in terms of structural analysis, the reliability of the analysis output and the subsequent structural performance prediction strongly depend on the decisions made for the selection of the seismic input used as ground excitation. Research has shown that among all possible sources of uncertainty stemming from structural and soil material properties, the modeling approximations, the design and analysis assumptions as well as the earthquake-induced ground motion, the latter has by far the highest effect on the variability observed in the structural response [2– 4]. Therefore, the selection of a ‘‘reliable’’ suite of earthquake ground motions constitutes an important prerequisite for the reliability of the structural analysis procedure as a whole. Nowadays, typically, the selection of earthquake records in most seismic codes and guidelines worldwide is primarily based on implicit parameters such as the earthquake magnitude, M, and the source-to-site distance, R. These parameters are defined by deterministic seismic hazard analysis, SHA, or by disaggregating a probabilistic site-specific SHA [5] and are used as the preliminary criteria for selecting an initial suite of eligible earthquake motions. Soil conditions at the site of the structure, the seismotectonic environment and other parameters (for instance, source mechanism,

E.I. Katsanos, A.G. Sextos / Advances in Engineering Software 58 (2013) 70–85

path of seismic waves and duration of the strong-motion) are used often to further sift the dataset of eligible records. However, the concurrent application of all the above parameters may significantly reduce the number of the eligible records. Thus, relaxation of these criteria may be inevitable to ensure an adequate set of motions for conducting time history analyses. The above limitations of the concurrent application of multiple criteria has also lead various researchers to investigate the relative impact of magnitude- and distance-based selection, the first found to be a more influential record selection parameter [6,7]. Another significant selection criterion is the strong motion intensity and the parameter used for its quantification. Peak ground acceleration, PGA, and the spectral acceleration at the fundamental period of the structure, Sa (T1), are typical examples of widely used intensity measures, IM (e.g., [3]). More elaborate IMs have also been proposed, involving the spectral shape and some structural characteristics. Such IMs are expected to result into a more accurate prediction of the seismic demand [8–10]. However, none of them is yet explicitly implemented in seismic code ground motion selection procedures. On the other hand, it is also common to envision convergence of the response spectra of the selected acceleration time series with a target spectrum [11–13]. This target spectrum can be specified by (a) a seismic scenario determined from a ground motion prediction relationship (e.g., [14]), (b) a seismic hazard assessment for the site of interest, (c) a conditional mean spectrum [15] or (d) the seismic code provisions. Along these lines, progress has been made for the quantification and improvement of envisioned spectrum convergence (e.g., [16,17]) as well as for controlling the standard deviation of the earthquake spectra themselves [18]. Several alternative or more advanced methods have been proposed for enhancing the reliability of the earthquake records selection and scaling process (as reviewed in [19]). Nevertheless, again, their main findings have not yet been reflected on the present state of practice and the corresponding drafting of seismic codes. In contrast, a somewhat simplified framework is only prescribed, without major differences among codes worldwide. 2. Motivation and objectives Recent research has shown that the above rather oversimplified prescriptions of seismic codes and guidelines for selecting and scaling suites of motions has adverse implications for both assessment and design of structures. More precisely: (a) the limited number of (commonly seven) records required by most guidelines and seismic codes for conducting time history analyses, undermines in advance the computation of a adequately stable estimate of the elastic or inelastic structural response [3,12,13,20]. (b) in most cases, the above variability in structural response cannot be reduced by a more accurate specification of the range of variation for the magnitude M or the source-to-site distance R, because structural response and M–R pairs have been found only partially correlated [3,8,21]. Few exceptions are structures sensitive to higher modes [3] or structures assessed on the basis of cumulative damage measures [22]. (c) the quality of spectrum convergence, which is the main selection and scaling criterion according to the seismic codes, is not explicitly ensured in a quantified manner. More specifically, the process only guarantees that the average response spectrum of the ground motions selected simply exceeds the ordinates of the target one, without considering the aleatory variability of the records or imposing an upper bound for the resulting mean spectrum of the seed records. This often leads to either unreliable estimates of structural response or overconservative design values [23].

71

(d) enforcement of spectral matching to periods as long as twice the fundamental period T1 of a building, as prescribed for instance in Eurocode 8-Part 1, is made irrespectively of the structural characteristics and ductility. As a result, spectral matching is enforced in the long period range, where it is very unlikely that a low-to-moderate ductility structure will ever respond. On the other hand, this scaling at the long period range, may substantially increase the spectral ordinates of the mean spectrum of the records selected in more critical periods of vibration (i.e., close or lower than T1) [23]. (e) even in case the designer is aware of the above issues, it is impossible to overcome them and achieve near optimal spectral matching without a specialized computational tool. Currently, useful algorithms and software have been made available for selecting and scaling seismic records [17,18,24–26] in all cases though, earthquake record selection is solely related to the fundamental period of the structure, without considering other structural or response parameters. Given the above limitation, an improved computational scheme has been developed and is presented herein for structure-specific selection and scaling of seismic motions aiming to accommodate a web-based selection of earthquake records for buildings and bridges according to the European and US standards. Particularly, the proposed framework retrieves records from the PEER-NGA strong-motion database (PEER-NGA, Ó2011, The Regents of the University of California, available in http://peer.berkeley.edu/ peer_ground_motion_database) and forms suites that comply with specific criteria (M, R, soil type) and the spectral matching requirements of Eurocode 8 [27] (Part1 for buildings and Part2 for bridges) or NEHRP Recommended Seismic Provisions for New Buildings and Other Structures, abbreviated in the following as FEMA P-750 [28]. The software proceeds, optionally, in filtering out, from the subset of eligible suites of motions that satisfy the code criteria and present the best matching with the target spectrum, those extreme cases that are associated with unrealistically high variability in response quantities, specifically for the structure studied. This is achieved by using the Application Programming Interface (API) functions of a widely used finite element software (SAP2000, CSI [29]) to conduct time history analyses of the structure in the computer background during runtime and quantify the induced variability in critical structural response quantities. In this way, the ground motion suite, which is eventually selected from a long list of code-compliant suites, has the following desirable features: (a) satisfies all the preliminary criteria (magnitude, source-tosite distance, intensity, soil class) and all the spectral matching provisions of current seismic codes in Europe and the US whichever is each time applicable, (b) is characterized by near-optimal average spectral matching to the target spectrum (as opposed to the minimum spectral matching criteria prescribed by EC8 and FEMA-750), (c) can be easily and automatically formed by more than seven records (which is the minimum number of records typically prescribed in the codes), thus improving the reliability of the mean structural response computed. (d) ensures reasonable (i.e., not excessive) structural response variability in the most critical structural elements (i.e., typically, though not compulsorily, bending moments at the base of shear walls or lateral storey displacements). Particularly, the user can (optionally) set the desirable level of structural response variability, by assigning a threshold confidence level for the standard error of the response quantities, and thus filter-out some of the, otherwise eligible,

72

E.I. Katsanos, A.G. Sextos / Advances in Engineering Software 58 (2013) 70–85

Table 1 Earthquake records selection and spectral matching criteria prescribed in the seismic codes and guidelines studied here.

a

Seismic codes and Guidelines

Selection criteria

Ensemble spectrum

Spectral matching period range

Lower bound

EC8 Part1 EC8 Part2 FEMA P-750

Seismotectonic features, soil type Source mechanism, M, R Source mechanism, M, R

Mean of individual spectra Mean of SRSS spectra Mean of SRSS spectra

0.2T1–2.0T1 0.2T1–1.5T1 0.2T1–1.5T1

0.90Satargeta 1.30Satargeta 1.17Satargeta

Satarget represents the spectral acceleration values defined by the target-code spectrum.

suites of ground motions that lead to highly scattered response results. Along these lines, the proposed computational framework provides a robust and fully automated procedure for record selection and structural analysis that proceeds beyond the minimum selection criteria imposed by modern seismic codes whilst simultaneously satisfying all relevant code provisions. Moreover, given its generic architecture, this framework is applicable independently of the structure studied (i.e., bridge or building), the complexity of the finite element model, the regularity of the configuration and the consideration of soil compliance. In the following, the principles of the software environment developed are presented together with a detailed demonstration for the challenging case of an existing, irregular-in-plan, multi-storey, R/C building, initially designed according to Eurocode 8.

3. Seismic code provisions for record selection and scaling The philosophy behind ground motion selection does not vary in principle amongst most modern seismic codes and is typically based in preliminary seismotectonic and soil compatibility criteria followed by rules for spectral matching between the mean response spectrum of the seed motions and the target spectrum of the seismic code after appropriate scaling. The ground motion to be used in (elastic and inelastic) time history analyses, is described by two independent orthogonal components while the terms and conditions for incorporating the vertical component of ground motion are also provided. Particularly for buildings, EC8-Part1 [27] permits the representation of seismic motion required for time histrory analysis by artificial, simulated or recorded accelerograms, depending on the nature of the problem studied and the information available. FEMA P-750 [28] and EC8-Part2 [27] for bridges though, permit the use of simulated ground motions only in cases of inadequate number of recorded accelerograms, the latter being selected on the basis of magnitude, fault distance, and source mechanisms consistent with those that define the design seismic action. It is only EC8-Part1 that explicitly imposes compatibility of the eligible records with the soil category at the location of the building studied. Regarding the spectral matching procedure, the minimum convergence criterion implies that the average response spectrum exceeds the lower bound of the design spectrum (Table 1) for the period range prescribed in each code. FEMA P-750 guidelines for spectral matching are almost identical to those in EC8-Part2 for bridges. Particularly, the square root of the sum of squares (SRSS spectra) of the 5%-damped elastic spectra ordinates of each component of the horizontal motions selected, are first determined and then the average of the individual SRSS spectra is computed to be compared with the target spectrum multiplied by 1.3 and 1.17 respectively (as comparatively summarized in Table 1). Spectral matching is imposed within the period range 0.2T1 up to 1.5T1, where T1 is the fundamental period of the structure. On the other hand, EC8-Part1 for buildings, requires that the mean 5%-damped elastic spectrum, calculated from all the response spectra of the individual records, has to exceed 90% of the target spectrum

ordinates, while the period range for the spectral matching is wider (that is, 0.2T1–2.0T1). One major difference among the above codes and guidelines is that FEMA P-750 permits individual scaling of each record by a different factor to facilitate spectral convergence, while EC8-Part2 prescribes explicitly the use of a unique scaling factor, uniformly applied to all the selected records of a given suite. On the contrary, EC8-Part1 does not provide specific guidelines regarding the scaling of seismic records in order to establish the required compliance with the design spectrum. Post-processing of the structural analysis results Eurocode 8 and FEMA P-750 is identical: when seven or more different records are selected, the average of the response obtained from all analyses is considered as the design value; when the selected suite consists of three to six records only, then the design value is defined as the maximum response observed. 4. Software environment for ground motion selection and structural analysis 4.1. System architecture and interaction with finite element software The proposed software environment for structure-specific earthquake record selection ISSARS (Integrated System for Structural Analysis and Record Selection)1 is organized into four main modules dedicated to (1) the initial selection of ground motion records, (2) the optimization and visualization of spectral matching, (3) performance of structural analysis and (4) further optional sifting of the eligible suites of motions based on the desired structural response confidence level. The algorithm was developed in Matlab [30] environment and the Application Programming Interface (API) of the finite element program SAP2000 was used to provide access during run-time, between the third-party application (i.e., Matlab in this case) and the analysis software itself. Other scripting options are also provided by the API through alternative integrated development environments (IDEs) such as Visual Basic.NET, Visual Basic for Applications (VBA), C#, C++ and Fortran. This API was activated within the Matlab script (i.e., .m file), thus making it feasible to gain access to an extensive collection of objects and functions and control the operation of SAP2000 entirely in the background, completely overriding the standard point-and-click procedure. It is noted that the above process provided access in real time both for pre- and post-processing purposes, as seen in Fig. 1. The overall system architecture is illustrated in Fig. 2 and is described in detail in the following sections. 4.2. Source of strong motion records The software environment developed, utilizes, during runtime and directly through the web, the Next Generation Attenuation Strong-Motion database, PEER-NGA [31] (available at http:// peer.berkeley.edu/peer_ground_motion_database) to search automatically, for suitable, recorded earthquake strong motions. Currently, the PEER-NGA dataset consists of 3551 publicly available, three-component ground motions (that is, of about 10,650 1

Available as a freeware at http://www.asextos.net/issars.html.

73

E.I. Katsanos, A.G. Sextos / Advances in Engineering Software 58 (2013) 70–85

Fig. 1. Application binding and typical data flow for coupling the Matlab GUI and the finite element program SAP2000 through the API.

individual seismic records), which have been recorded during 173 shallow crustal earthquakes from active tectonic regions worldwide. The majority of the embedded seismic events have occurred in California, while their magnitude ranges from 4.2 to 7.9 and their epicentral distance is in the range 0.20–600 km (Fig. 3). Apart from the magnitude and distance, the strong motion database provides basic information about the seismic source, including the date and time of the event, the location of the hypocenter, the mechanism of the fault, the seismotectonic environment and others. Detailed data about 1600 recording stations is also provided (i.e., site characterizations, surface geology, shallow subsurface conditions, and the location of the instrument).

4.3. Initial earthquake record selection module The earthquake magnitude, M, the epicentral distance, R, of the seismic events, the soil classification, S, of the site where the motion was recorded and the peak ground acceleration, PGA, of the strong-motion record, constitute the preliminary search criteria of the software developed. Based on these criteria, an initial list of eligible pairs of seismic records (containing all recorded components) is automatically formed by records retrieved online. The recorded accelerograms are used as the seed motions to form the alternative, eligible suites of records required for structural analysis. It is recalled that the total number of suites Ntot.suites of m pairs of seismic records that can be formed out of a larger group of k pairs, can be calculated by the following factorial formula of the binomial coefficient:

Ntot:suites ¼



k m

 ¼

k! m!ðk  mÞ!

ð1Þ

A typical value for the number of records m per suite is seven, since this is the minimum required by both Eurocode 8 [27] and FEMA P-750 [28] to permit the use of the mean of the response quantities for design purposes. Another reason for this popular design option is that the formation of suites with more than seven records requires a disproportionally large number of eligible records, k, in order to achieve an acceptable fitting of the average spectrum of the m individual records to the target spectrum [18]. On the other hand, a sample of seven response quantities is already small to derive reliable mean (and eventually design) values as well as to compute their standard error. For this reason, the software leaves this decision to the designer, but taking advantage of its automation ability, suggests the formation of suites with more than seven records.

Having defined the preliminary criteria, all suites of records are scaled by a factor sfavg. in order to ensure that their average spectral values Saavg.(Ti) exceed the minimum (Table 1) spectral ordinates Satarget(Ti) of the target spectrum within the prescribed period range:

 Sfav g: ¼

 min

1 Saav g: ðT i Þ ; Satarget ðT i Þ

i ¼ 1—N

ð2Þ

where Ti is the sample period and N is the size of the sample within which the prescribed period range is discretized. Next, all suites are ranked according to their ‘‘goodness-of-fit’’ to the target spectrum, as quantified by the normalized root-mean-square-error, d, between the scaled average and the target spectrum:

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u 2 N  u1 X sfav g  Saav g: ðT i Þ  Satarget ðT i Þ d¼t  N i¼1 Satarget ðT i Þ

ð3Þ

A wide variety of similar expressions has also been used in the literature (e.g., [32]). Further discussion on the efficiency of various spectral matching indicators can be found elsewhere [13,16– 18,26,33]. It is notable that despite the extensive computations required, the performance of the ISSARS software is very satisfactory; less than 180 s are required using an 8 GB RAM 1.60 GHz quad core processor for sifting approximately 10,000 records of the PEERNGA database on the basis of the preliminary criteria, retrieving online those which meet the criteria, calculating and scaling all individual, SRSS and mean spectra of the motions selected, and eventually ranking 480,700 suites of seven records (in case for instance, 25 eligible pairs of records). This computational efficiency, which is further improved when faster processors and network connection are used, permits the concurrent use of the algorithm for alternative sets of the preliminary criteria and quickly leads to near optimal convergence with the target spectrum (as seen in Fig. 4). 4.4. Spectral matching visualization module (optional) The eligible suites of records formed can be visually assessed by calling the plot module. For each one of the suites formed, ISSARS provides a figure illustrating: (a) the target spectrum, (b) the codeimposed lower bound of the design spectrum (necessary for the spectral matching procedure, according to Table 1), (c) the individual 5% damped response spectra of all records forming the suite, (d) the average spectrum of the suite, and (e) the scaled average spectrum. This illustration (Fig. 4) enables a visual assessment of

74

E.I. Katsanos, A.G. Sextos / Advances in Engineering Software 58 (2013) 70–85

Fig. 2. Flowchart of structure-specific earthquake record selection procedure proposed herein.

E.I. Katsanos, A.G. Sextos / Advances in Engineering Software 58 (2013) 70–85

75

joints). Next, mean values of the seismic demand and their corresponding standard errors are calculated for each structural member through the Matlab GUI environment. The only prerequisite for this procedure is to develop in advance the finite element model in the form of a SAP2000 (.sdb) file. The variation of structural response quantities is also available for visualization through histograms provided by the software GUI. This interactive computational procedure can be applied regardless of the structural type (i.e., buildings, bridges, silos [34], etc.), the material used (i.e., reinforced concrete or steel), the presence of base-isolation [35], the number of degrees of freedom of the finite element model, the potential for geometrical or material nonlinearities, as well as the consideration of soil flexibility. The designer has only to prepare the finite element model of the structure in advance and the structure-specific earthquake record selection runs effortlessly based on the seismic scenario defined in terms of magnitude, distance (far- and near-field conditions [36]), intensity, soil class and the seismic code applicable at the site of interest. Fig. 3. Magnitude, epicentral distance, and peak ground acceleration distribution of strong-motion records in the PEER-NGA Database (distribution computed based on the data available online).

the ‘‘goodness-of-fit’’ and may reveal period windows of codecompliant but less successful fit that would be otherwise suppressed by d value of equation (3) used to quantify the average spectral matching along the entire period range (0.2T1 up to 1.5T1 or 2.0T1).

4.5. Structural analysis module (optional) Having ranked all the eligible suites of records formed, the Application Programming Interface (API) of the finite element software SAP2000 (CSI [29]) is utilized directly by the Matlab [30] script. Through this connection, it is feasible to control SAP2000 without any intervention of the user, automatically run time history analyses in the background with the records of the selected suite and, post-process the structural response results (maxima in time of forces and displacements monitored at all frames and

4.6. Module to establish a desired confidence level of structural response (optional) As already mentioned, seismic codes provisions for selecting and scaling earthquake ground motions for the purposes of time history analysis aim to induce on average a level of seismic demand which is compatible to the target, uniform hazard spectrum prescribed, within a given period range. Further, it is desirable to achieve reliable enough (i.e., of low standard error) structural response estimates, namely response results that are as close as possible to those that could have been predicted if the structure was analyzed under a large sample (population) of ground motions, corresponding to the same seismic design scenario [20]. There are two main reasons that this is desirable. First, given that all the code-conformed procedures for both design and assessment of structures are based solely on the average response produced by the (linear elastic or inelastic) time history analysis using the selected ground motions, it is imperative to preserve the reliability particularly of the mean (design) value. Second, especially in case of inelastic analyses, the response variability should be kept reasonably low, otherwise, the overall rational for ‘averaging’ the

Fig. 4. Sample plot of individual, mean and target spectra computed for a suite of seven pairs of records.

76

E.I. Katsanos, A.G. Sextos / Advances in Engineering Software 58 (2013) 70–85

action effects is not always meaningful, since the demand parameters will be most likely derived partially from elastic and partially from inelastic analysis [23]. Along these lines, the purpose of this specific fourth module, is to introduce a threshold value, set by the designer, for the minimum acceptable variability of structural response, in order to filter-out suites of ground motions associated with extreme variability in structural response and prioritize others, which, apart from fulfilling the criteria set by the code they provide a more reliable estimate of the mean structural response. To quantify the above objective, the standard error of an estimate, SEE, is used. It is recalled, that SEE is a measure of the accuracy of a result which is obtained from a sample of data when the entire population cannot be used [37]:

r

SEE ¼ pffiffiffi n

ð4Þ

In the above, r is the population standard deviation of the underlying random variable and n is the sample size. In most of practical cases, the standard deviation of the population is unknown. Therefore, if the sample size is sufficiently large (i.e., n P 30), it is reasonable to assume that r can be replaced by the corresponding standard deviation of the sample, s, which is a point estimator of r. In such a case, the uncertainty associated with the unknown value of r is ignored. On the contrary, if the sample size is small (i.e., n < 30), the uncertainty introduced by the lack of definite knowledge of the true value of r cannot be ignored and thus, the equation for the standard error of an estimate is reformed as follows:

s SEE ¼ pffiffiffi  tðCL; df Þ n

ð5Þ

where the value of t-factor depends on the confidence level, CL, that is assigned to predict the response estimate, and df is the degrees of freedom for the two-sided Student-t Probability Density Function. For the case of the ground motion selection problem studied, assuming a suite of records with seven seismic motions, the t-factor is 1.943 for six degrees of freedom and 90% confidence level. Thus the standard error, defined as percentage of the estimated median of, say, the roof displacement with a sample standard deviation of s = 0.35 (i.e., the standard deviation of the log roof displacement), is approximately equal to 25%. To interpret this result, it is worth emphasizing that if one were to construct many response-samples of the same size drawn from the same population and the standard deviation of the samples remained constant, 90% of the ±25% confidence intervals on the median response the true (unknown) median level would be approximately included. Rather, the sample-median roof displacement is estimated with a confidence band of about ±25%. It is interesting to note that the median response (or more precisely, the geometric mean of the response values) used for determining the confidence intervals, is preferred over the commonly used arithmetic mean (i.e., the term describing the central tendency of a data set). This is because the median is more consistent with the log-normally distributed response quantities (e.g. [38]). However, the consequences of this issue are not significant if a consistent definition of the average measure is made for both scaling

4,60

0 B5 45/5

C1 O60

W4 45/150

1,75

0,45

2,45

8,70

5,00

W6

W6 25/485

12,00

0

/1 W6 35

Elevation well

B1 25/60

0 B3 25/6

25/210

W6 25/210

2,10

B10 45/50 W2 45/180

1,30

S3 d=15

2,40

1,80

4,65

65

0,55

S5 d=15

B4 25/6

0,55

3,30

W1 45/180

B2 45/50

S1 d=16

55 W3 40/2

B9 45/50

4,75

0,60

5,10

1,35 1,80

11,75

S2 d=18

2,55

Longitudinal direction

C2 O60

3,50

5 B6 55/5

B8

45 /50

4,10

S4 d=18

0 B7 45/5

85 40/2 W5

2,00

W4 45/75

2,85

2,8 5

4,30

4,85 7,20

4,60 4,70

13,65

Transverse direction Fig. 5. Plan view of the typical storey of the case study R/C building.

77

E.I. Katsanos, A.G. Sextos / Advances in Engineering Software 58 (2013) 70–85 Table 2 Modes of vibration, periods, and modal participating mass ratios of the building studied. Mode

Period (s)

UX

UY

UZ

Sum UX

Sum UY

RX

RY

RZ

Sum RX

Sum RY

Sum RZ

1 2 3 5 6

0.621 0.480 0.303 0.177 0.125

58.13 3.24 11.23 11.27 1.03

2.97 68.49 0.55 0.91 12.85

0 0 0 0 0

58.13 61.37 72.60 83.88 84.90

2.97 71.46 72.01 72.92 85.77

3.81 94.37 0.98 0.00 0.41

77.90 4.74 16.68 0.25 0.03

14.19 0.03 60.46 1.37 0.05

3.81 98.19 99.17 99.17 99.57

77.90 82.64 99.32 99.57 99.61

14.19 14.22 74.68 76.05 76.11

Fig. 6. The 3-dimensional finite element model of the case study building and the illustration of the first three mode shapes (T1 = 0.621 s, T2 = 0.480 s and T3 = 0.303 s).

Table 3 Preliminary searching criteria and the algorithm output based on the four alternate seismic scenarios considered. Seismic scenario

Preliminary searching criteria Magnitude (M)

A B C D a

Algorithm output

Distance (R)

6.5 6 M 6 7.5 5.5 6 M 6 6.5 6.5 6 M 6 7.5 6.5 6 M 6 7.5

PGA (g)

20 6 R 6 50 20 6 R 6 50 0 6 R 6 20 20 6 R 6 50

PGA P 0.16 PGA P 0.16 PGA P 0.16 PGA P 0.16

Site classa

C C C D

Pairs of seismic records Eligible

Used in analyses

36 21 32 54

28 20 28 28

Number of suites

1184,040 77,520 1184,040 1184,040

According to NEHRP soil classification: Site class C (760 m/s 6 vs 6 360 m/s), Site class D (360 m/s 6 vs 6 180 m/s).

Table 4 The selected suites of ground motions to be used for the time history analyses. Suites of recordsa

a

sfavg.

d

Suites of recordsa

sfavg.

d

Seismic scenario A (6.5 6 M 6 7.5, 20 6 R 6 50, PGA P 0.16 g, Site class C) 7, 10, 11, 13, 19, 23, 26 0.670 0.075 1, 7, 10, 11, 13, 19, 36 0.691 0.084 1, 11, 13, 14, 19, 23, 26 0.692 0.087 2, 7, 12, 17, 19, 21, 23 0.868 0.401 1, 3, 4, 11, 17, 19, 31 1.060 0.507 5, 8, 10, 14, 15, 19, 34 0.941 0.603 3, 4, 9, 14,17, 22, 35 1.152 0.702 2, 6, 8, 11, 12, 21, 31 1.406 0.822 4, 6, 9, 10, 23, 34, 35 1.304 1.002 3, 4, 16, 20, 27, 31, 34 4.196 3.546

Seismic scenario B (5.5 6 M 6 6.5, 20 6 R 6 50, PGA P 0.16 g, Site class C) 2, 9, 11, 13, 16, 17, 18 0.868 0.085 2, 3, 9, 13, 16, 17, 18 0.765 0.087 2, 3, 9, 10, 13, 17, 18 0.757 0.093 2, 3, 4, 6, 10, 13, 15 0.879 0.279 1, 3, 6, 10, 11, 12, 17 1.118 0.372 1, 2, 3, 4, 9, 14, 16 1.296 0.464 1, 2, 5, 12, 17, 18, 19 1.023 0.555 4, 5, 9, 11, 15, 18, 20 1.252 0.661 1, 9, 12, 14, 16, 17, 20 1.823 0.822 7, 8, 9, 12, 15, 19, 20 4.186 2.883

Seismic scenario C (6.5 6 M 6 7.5, 0 6 R 6 20, PGA P 0.16 g, Site class C) 1, 10, 22, 23, 25, 26, 30 0.464 0.086 1, 5, 22, 23, 25, 27, 29 0.415 0.090 1, 10, 20, 22, 23, 25, 30 0.457 0.091 1, 10, 11,13, 14, 26, 28 0.485 0.255 1, 3, 8, 23, 25, 30, 32 0.513 0.305 1, 9, 17, 25, 16,28, 29 0.515 0.348 2, 5, 8, 9, 21, 26, 30 0.515 0.393 12, 14, 18, 20, 22, 23, 28 0.707 0.444 2, 4, 6, 8, 18, 29, 32 0.648 0.518 2, 6, 7, 12, 24, 26, 32 1.064 1.060

Seismic scenario D (6.5 6 M 6 7.5, 20 6 R 6 50, PGA P 0.16 g, 8, 10, 11, 15, 16, 28, 46 0.832 3, 8, 10, 11, 15, 28, 46 0.853 3, 8, 11, 15, 16, 28, 46 0.902 1, 3, 15, 16, 29, 51, 52 0.822 3, 11, 17, 18, 20, 38, 52 1.020 2, 7, 11, 15, 27, 29, 52 0.858 8, 15, 20, 27, 29, 51, 54 0.893 2, 12, 15, 16, 18, 20, 54 1.441 1, 7, 8, 12, 18, 34, 53 1.410 7, 18, 21, 34, 37, 38, 53 1.887

The records are numbered according to Table A1 seismic motions identification.

Site class D) 0.061 0.067 0.071 0.292 0.355 0.409 0.464 0.528 0.619 1.395

78

E.I. Katsanos, A.G. Sextos / Advances in Engineering Software 58 (2013) 70–85

the accelerograms (through the spectral matching procedure) and measuring the structural response [20]. Based on the above, the ISSARS software developed performs a refined statistical evaluation of the structural response estimates based on the time history analyses results of the structural model studied. Particularly, the response quantities calculated for an eligible suite of motions (typically one of those presenting good average convergence with the target spectrum and thus exhibiting low value of d), are statistically processed. Then, an acceptanceor-rejection status is assigned to the selected suite of records depending on whether a user-defined level of uncertainty (as expressed by SEE, typically ±15%) on the median response estimates is fulfilled. In case of rejecting the suite of earthquake motions tested due to low level of induced accuracy to the median response estimates, the user can easily test alternative, code-compliant, suites of records until a desirable suite is found. One critical view to the above rational might be that the intention to control structural response variability by setting a threshold value for the standard error of response and thus, being able to exclude suites of strong motion records, which lead to highly scattered response estimates, suppress the inherent stochastic nature of earthquake itself and the resulting potentially extreme variability in the structural response. Here, one issue is that, the stochastic nature of earthquake excitation (i.e., the uncertainty of the input motion) is already incorporated in the definition of the uniform hazard, target spectrum either prescribed by the code or derived by site-specific probabilistic seismic hazard analysis, both

providing spectral accelerations with a given probability of exceedance. It is also notable that the alternative code-prescribed method for time history analysis with the use of artificially generated motions that closely match the design spectrum, clearly leads to structural response with negligible variability. That been said, it is deemed useful to compile suites of ground motions by optionally controlling the structural response variability within non-excessive limits, while satisfying all code-prescribed selection criteria.

5. Application example To illustrate the applicability and effectiveness of the proposed software environment and to investigate the influence that the code-based records selection procedure exert on the structural response, an existing, irregular-in-plan, multi-storey, reinforced concrete (R/C) building, in the city of Thessaloniki, Greece, was adopted as the case study. Different seismic scenarios were considered and a large number of records suites were formed and ranked based on spectral convergence requirements. The selected seismic motions were used for the bidirectional earthquake excitation of the building, modeled in SAP2000 as a means to take advantage of the ISSARS capabilities. The structural response, assumed to remain linear elastic, was calculated through linear time history analysis. An overview of this case study, as well as the analysis results derived are presented in the following sections.

Fig. 7. Initial earthquake record selection module for the case studied.

E.I. Katsanos, A.G. Sextos / Advances in Engineering Software 58 (2013) 70–85

5.1. Structural configuration and modeling aspects of the case study building The current case study is an existing, irregular-in-plan, sevenstorey (including pilotis and basement) R/C building with a dualresisting structural system consisting of four shear walls of approximately equal dimensions and a reinforced concrete core. The plan view of the building is illustrated in Fig. 5. The total area of the typical floor plan is about 135.0 m2 and the storey height is 3.0 m. The concrete class used was C20/25 (i.e., compressive strength fc0 ¼ 20 N=mm2 ) and the yield strength of the steel reinforcing bars (St. III) is equal to fy0 ¼ 500 N=mm2 . The building was designed and constructed in 2008 according to the latest version of the Greek Seismic Code (EAK [39]), currently in parallel enforcement with the Eurocodes. In principle, it can be considered that both codes lead to comparable levels of seismic demand and ductility. The design ground acceleration, ag, was 0.16 g as corresponding to the seismic hazard category I for the site of interest (also prescribed in the National Annex of EC8). The building was founded on stiff soil formations, classified as type B according to EC8 [27]. Frame elements were used for beams and columns, while the shear walls and the slabs were modeled using shell elements. Diaphragms were applied at each storey level. Young’s Moduli equal to 29,000 N/mm2 and 200,000 N/mm2 were considered for concrete and steel materials respectively. The dynamic characteristics of the particular structure, determined from eigenvalue

79

analysis, are listed in Table 2, while Fig. 6 presents a schematic illustration of the first three mode shapes. It is noted that the particular building has been selected deliberately on the basis of its complex structural behavior and significant contribution of higher and torsional modes. 5.2. Seismic scenarios for time history analysis Four alternate scenarios were developed for different earthquake magnitude, M, source-to-site distance, R, peak ground acceleration, PGA, and soil conditions, S; the latter also used as the preliminary records selection criteria (as summarized in Table 3). It is noted that values higher than 0.16 g were set for the PGA criterion as a means to prevent selection of ground motions with significantly low acceleration ordinates that would require in turn excessive scaling factors during the spectral matching process. The EC8 5% damped, elastic spectrum was defined for soil category B (800 m/s 6 vs 6 360 m/s) and the reference peak ground acceleration value was taken equal to 0.16 g for the reason described previously. Seven pairs of strong motions were used to form the required suites to comply with the state-of-practice and highlight the implications of this decision. The mean spectrum of each suite, calculated by averaging the 14 accelerograms (7 pairs of 2 components each) included in each set, was scaled to fit to the 90% of the design spectrum within the code-imposed range of periods (i.e., 0.12–1.24 s).

Fig. 8. Post-analysis visualization of the structural response variability.

80

E.I. Katsanos, A.G. Sextos / Advances in Engineering Software 58 (2013) 70–85

Based on the preliminary criteria, the eligible pairs of records were selected by ISSARS for each seismic scenario to be used in the foreseen time history analyses. Description of the seismic records selected, corresponding to the four seismic scenarios, is given in Appendix (Table A1). Except from the second scenario (B) where only 21 eligible pairs were spotted, 28 different pairs of horizontal components of strong motions were selected as the seed motions for defining the eligible suites. Thus, seismic scenarios A, C and D resulted to 1184,040 different suites of motions (containing seven pairs of records each). Because of the enormous computational cost related to the performance of time history analyses for the entire array of the records combinations derived, the case study building was excited using 10 selected suites of motions from each of the four seismic scenarios adopted, with various normalized root-mean-squareerror, d, values (covering very good to very poor spectral matching). Along these lines, 280 bidirectional time history analyses were performed automatically by the software. The scaling factors sfavg. applied to the individual motions selected and the corresponding spectral matching indicator d for the suites used to excite the building are summarized in Table 4. Snapshots of the initial earthquake record selection module and the post-processing of the time history analysis results using the software developed are presented in Figs. 7 and 8 respectively. 5.3. Effect of ground motion record selection on the predicted structural response Having established the four seismic scenarios and formed in total 3,629,640 different suites of eligible earthquake records, the time history analyses results were used to quantify the impact of the EC8-based spectral matching procedure to the response

estimates and the variability derived. Fig. 9 shows the variation of the average bending moments, normalized to the lower value of their sample, as derived from seven time history analyses with a given suite of motions characterized by a specific value of d, for ten particular beams, columns and shear walls at the ground floor of the building. In general, it is observed that, the higher the values of d, (i.e., the less successful the spectral matching), the higher the prediction of the average structural response. This trend can be attributed to the lack of an upper bound in EC8 spectral matching provisions for the spectral accelerations of the records used, in contrast to the explicitly prescribed lower bound (i.e. the requirement that spectral ordinates of the average spectrum exceed 90% of the target spectral accelerations), as already shown in Table 1. Fig. 9 also shows that the use of records, characterized by high values of d (i.e., d > 1.0, corresponding to weak spectral matching but still permissible according to the letter of the code), results in significant overdesign (or overconservative post-design assessment) of the structural system. In other words, it is seen that selecting and scaling earthquake records without an appropriate tool to automate the procedure and form all eligible suites and to ensure low values of d, may lead to unnecessarily high predicted forces that far exceed the ones predicted by either response spectrum or time history analysis using artificial accelerograms. On the other hand, it is interesting to notice that despite the fact that the increase of average response quantities with d is anticipated as a general trend, it can by no means be considered a rule. In fact, there are some cases of eligible suites of ground motions with high (inferior) values of d, which lead to lower average response quantities compared to suites characterized by lower (superior) values of d. Fig. 10 depicts this non-monotonic

Fig. 9. Normalized average bending moments of 10 structural members of the ground floor against the spectral convergence measure (d) of the ground motion suites studied (x-axis is nonlinear to facilitate visualization).

E.I. Katsanos, A.G. Sextos / Advances in Engineering Software 58 (2013) 70–85

81

Fig. 10. Bending moments (in kNm) of critical structural members of the ground floor derived from the chosen suites of records (ranked in terms of d) for seismic scenario C. Error bars indicate the average response ± one standard deviation.

relationship between the average structural response quantities (this time in terms of their absolute values) and d, for four shear walls (W1, W2, W3 and W4) and the third seismic scenario. It is seen that for instance, the third suite of motions which is characterized by a significantly low (i.e., third best) value of d (i.e., d = 0.091), produces average bending moments that are higher than those derived using suites of motions with significantly higher values of d (for example, suites of motions with d = 0.255, d = 0.348 and d = 0.444). The same observation is made when the coefficient of variation, CV, of the bending moments of the ten characteristic structural members is plotted with the spectral matching indicator d (Fig. 11). It is seen that for certain seismic scenarios, (i.e., C), a specific suite of records with significantly low value of d (i.e., d = 0.091), is again associated with very high bending moment variability (i.e., CV > 0.80) for all structural members; an extreme dispersion which is not reached for other suites characterized by significantly inferior spectral matching. This observation stems from the fact that, as already mentioned, the spectral convergence measure d is computed along the entire (and actually wide) code-imposed range of periods, hence potentially suppresses both some local extremes of the average spectrum of the records and their intra-suite scatter at specific periods. In other words, a good average matching of the target spectrum (through the use of d) by no means guarantees an equally close spectral matching at specific periods of potential

interest, especially for complex or irregular buildings where a large number of modes of vibration contributes to the overall dynamic response.

5.4. Benefits of using ISSARS for selecting ground motions To avoid the above undesirable implications, an acceptable confidence bandwidth for the estimated median response is set in the ISSARS software equal to ±15% (the user can easily define alternative values for uncertainty intervals, see Refs. [3,20] for more details). According to this threshold and in case the structure is designed for seismic scenario C, the suite of motion with a significantly low value of d = 0.091, is rejected because the corresponding standard error (SEE, as percentage of the median estimate of the bending moment of shear wall W5) is approximately equal to ±30% for a 95% confidence level. All other suites remain eligible (as they satisfy all Eurocode 8 prescribed criteria) unless a different confidence level is set by the designer. As an example, an alternative suite of motions, characterized by a less successful spectral matching (d = 0.444), leads to an acceptable median response estimate corresponding to a level of SEE equal to approximately 13%. This preview of structural response would not be feasible without the development of the particular software developed and the improved capability offered to the designer to conduct structurespecific earthquake record selection, in a way that satisfies to the

82

E.I. Katsanos, A.G. Sextos / Advances in Engineering Software 58 (2013) 70–85

Fig. 11. Coefficient of variation (CV) of the bending moments of 10 characteristic structural members at the ground floor against the spectral convergence measure (d values) of the records suites studied (x-axis is nonlinear to facilitate visualization).

maximum all seismic code criteria and the desirable objectives presented in Section 4.6. 6. Conclusions An integrated software environment is described herein that couples the earthquake record selection procedure with the structural analysis by considering the reliability of the predicted structural response as a final criterion for selecting suitable suites of motions to be used for the time history analyses of structures. A threshold confidence level for the standard error of the response quantities, set optionally by the designer to ensure reasonable response variability, can be utilized to reject suites with highly scattered results for critical structural members. At the same time, alternative suites of motions are prioritized that both fulfill every code-compliant record selection criterion and provide stable enough estimates of mean structural response. Overall, the software developed presents a number of improvements with respect to the current state-of-the-art and practice methods while permits, automatic, online, fully parameterized earthquake record selection by easily forming suites with more than seven records (which is the minimum number of records typically prescribed in the codes). Furthermore, a near-optimal average spectral matching to the target spectrum is also implemented as opposed to the minimum spectral matching criteria prescribed by Eurocode 8 and NEHRP. Additionally, its application for the case of an existing and quite demanding (highly irregular), multistory reinforced concrete

structure has revealed a number of problematic implications of the present seismic code provisions and the process of selecting earthquake records based on the normalized rootmean-square-error, d, between the scaled average of the individual spectra of the records comprising a suite and the code-prescribed target spectrum. It is believed that the coupling between earthquake ground motion selection and structural analysis is a feasible and useful approach that fulfills the code requirements while increasing the reliability of the response estimates. Acknowledgements This research was partially supported by the 7th Framework Programme of the European Commission, under the PIRSES-GA-2009-247567-EXCHANGE-SSI (EXperimental & Computational Hybrid Assessment Network for Ground-Motion Excited Soil-Structure Interaction Systems) Grant. This support is gratefully appreciated. The authors would also like to acknowledge the contribution of George Manolis, Professor at Aristotle University of Thessaloniki, Greece and Amr Elnashai, Professor at the University of Illinois at Urbana-Champaign, United States for their valuable feedback during the development of the software.

Appendix A. Summary of the ground motions selected See Table A1.

83

E.I. Katsanos, A.G. Sextos / Advances in Engineering Software 58 (2013) 70–85

Table A1 Description of the ground motions adopted for each seismic scenario. M is the earthquake moment magnitude; R is the epicentral distance (expressed in kilometers); PGA is the peak ground acceleration of the records (expressed as a fraction of g). No.

R

PGA

Used in analysis

Seismic scenario A: 6.5 6 M 6 7.5, 20 6 R 6 50 (km), PGA P 0.16 g, Site class C (760 6 vs 6 360 m/s) 1 KC 21.07.1952 Taft Lincoln School 7.36 2 SFE 09.02.1971 Castaic-Old Ridge R. 6.61 3 SFE 09.02.1971 Lake Hughes#12 6.61 4 SFE 09.02.1971 Santa Anita Dam 6.61 5 FR 06.05.1976 Tolmezzo 6.50 6 TAB 16.09.1978 Dayhook 7.35 7 IV 15.10.1979 Cerro Prieto 6.53 8 IR 23.11.1980 Brienza 6.90 9 NZ 02.03.1987 Matahina Dam 6.60 10 LP 18.10.1989 Anderson Dam 6.93 11 LP 18.10.1989 Coyote Lake Dam 6.93 12 LP 18.10.1989 Gilro-Gavilan Coll. 6.93 13 LP 18.10.1989 Hollister-South & Pine 6.93 14 LP 18.10.1989 San Jose-S.T. Hills 6.93 15 LP 18.10.1989 Saratoga-Aloha Ave 6.93 16 CM 25.04.1992 Shelter Cove Airport 7.01 17 LAN 28.06.1992 Lucerne 7.28 18 NOR 17.01.1994 Big Tujunga, Angeles 6.69 19 NOR 17.01.1994 Castaic – Old Ridge R. 6.69 20 NOR 17.01.1994 Glendale – Las Palmas 6.69 21 NOR 17.01.1994 LA-City Terrace 6.69 22 NOR 17.01.1994 LA-Cypress Ave 6.69 23 NOR 17.01.1994 LA-Fletcher Dr 6.69 24 NOR 17.01.1994 LA-Temple & Hope 6.69 25 NOR 17.01.1994 LA-Univ. Hospital 6.69 26 NOR 17.01.1994 La Crescenta – N.Y. 6.69 27 NOR 17.01.1994 Lake Hughes#12A 6.69 28 NOR 17.01.1994 Lake Hughes#9 6.69 29 NOR 17.01.1994 Manhattan Beach 6.69 30 NOR 17.01.1994 Moorpark-Fire Station 6.69 31 NOR 17.01.1994 Pasadena-N. Sierra 6.69 32 NOR 17.01.1994 Point Mugu-Lag Peak 6.69 33 NOR 17.01.1994 San Gabriel-E Grand 6.69 34 DZC 12.11.1999 Lamont 375 7.14 35 MAN 20.06.1990 Abbar 7.37 36 HM 16.10.1999 Hector 7.13

Earthquake

43.49 25.36 20.04 45.86 20.23 20.63 24.82 46.16 24.23 26.57 30.78 28.98 48.24 20.13 27.23 36.28 44.02 31.55 40.68 29.72 39.15 33.25 30.27 32.72 36.47 27.83 40.65 44.77 38.69 31.45 44.01 48.28 44.32 24.05 40.43 26.53

0.173 0.299 0.330 0.169 0.346 0.351 0.176 0.214 0.293 0.238 0.295 0.334 0.279 0.283 0.382 0.195 0.721 0.200 0.490 0.256 0.267 0.206 0.207 0.165 0.349 0.173 0.215 0.169 0.166 0.229 0.234 0.175 0.209 0.737 0.505 0.306

Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes No Yes Yes Yes Yes Yes No Yes No Yes No No No Yes No No Yes Yes Yes

Seismic scenario B: 5.5 6 M 6 6.5, 20 6 R 6 50 (km), PGA P 0.16 g, Site class C (760 6 vs 6 360 m/s) 1 P 28.06.1966 Temblor pre-1969 6.19 2 FR 06.05.1976 Tolmezzo 6.50 3 VIC 09.06.1980 Cerro Prieto 6.33 4 COA 02.05.1983 Parkfield-Fault Z.15 6.36 5 MH 24.04.1984 Coyote Lake Dam 6.19 6 MH 24.04.1984 Gilroy Array#6 6.19 7 PS 08.07.1986 San Jacinto-Soboba 6.06 8 WN 01.10.1987 Brea Dam 5.99 9 WN 01.10.1987 Glendale – Las Palmas 5.99 10 WN 01.10.1987 LB-Rancho Los Cerr. 5.99 11 WN 01.10.1987 N Hollywood-C.C. 5.99 12 WN 01.10.1987 Orange Co. Reservoir 5.99 13 CH 20.09.1999 CHY080 6.20 14 CH 20.09.1999 TCU067 6.20 15 CH 20.09.1999 TCU071 6.20 16 CH 20.09.1999 TCU075 6.20 17 CH 20.09.1999 TCU076 6.20 18 CH 20.09.1999 TCU122 6.20 19 CH 22.09.1999 CHY024 6.20 20 CH 22.09.1999 TCU129 6.20 21 CH 25.09.1999 TCU129 6.30

40.26 20.23 33.73 37.97 24.55 36.34 33.53 22.72 21.73 25.52 34.48 22.16 29.48 33.97 20.51 26.00 20.80 24.47 48.60 40.98 33.15

0.293 0.346 0.572 0.166 0.965 0.281 0.231 0.231 0.233 0.190 0.171 0.196 0.334 0.177 0.274 0.181 0.336 0.190 0.246 0.391 0.257

Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes No

12.82 6.80 9.01 7.17 18.46 16.51 16.34 12.56 10.36 4.51 13.67 16.27

0.644 1.056 0.526 0.498 0.783 0.342 0.457 0.517 1.345 0.624 0.249 0.510

Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Seismic scenario C: 6.5 6 M 6 7.5, 0 6 R 6 20 (km), 1 GAZ 17.05.1976 2 NAH 23.12.1985 3 LP 18.10.1989 4 LP 18.10.1989 5 LP 18.10.1989 6 LP 18.10.1989 7 LP 18.10.1989 8 LP 18.10.1989 9 CM 25.04.1992 10 CM 25.04.1992 11 LAN 28.06.1992 12 NOR 17.01.1994

Recording station

M

PGA P 0.16 g, Site class C (760 6 vs 6 360 m/s) Karakyr 6.80 Site 1 6.76 BRAN 6.93 Corralitos 6.93 LGPC 6.93 UCSC 6.93 UCSC Lick Observ. 6.93 WAHO 6.93 Cape Mendocino 7.01 Petrolia 7.01 Joshua Tree 7.28 Bev.H.-12520 Mulhol 6.69

(continued on next page)

84

E.I. Katsanos, A.G. Sextos / Advances in Engineering Software 58 (2013) 70–85

Table A1 (continued) No.

Earthquake

Recording station

M

R

PGA

Used in analysis

13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32

NOR 17.01.1994 NOR 17.01.1994 NOR 17.01.1994 NOR 17.01.1994 NOR 17.01.1994 NOR 17.01.1994 NOR 17.01.1994 NOR 17.01.1994 NOR 17.01.1994 NOR 17.01.1994 NOR 17.01.1994 NOR 17.01.1994 NOR 17.01.1994 NOR 17.01.1994 NOR 17.01.1994 NOR 17.01.1994 NOR 17.01.1994 NOR 17.01.1994 NOR 17.01.1994 KOB 16.01.1995

Jensen Filt. Plant Jensen Filt. Plant Gen. LA-Brent. VA Hosp. LA – Chalon Rd LA-Sepulv. VA Hosp. LA – UCLA Grounds LA-Wad. VA Hosp. N LA-Wad. VA Hosp. S LA 00 LA Dam N Hollywood-C.C. Pac. Palisades-Sunset Pacoima Kagel Can. Santa Susana Ground Simi Valley-Kath. Rd Stone Canyon Sylmar-Conv. Sta. E Sylmar-Olive View Topanga – Fire Sta Nishi-Akashi

6.69 6.69 6.69 6.69 6.69 6.69 6.69 6.69 6.69 6.69 6.69 6.69 6.69 6.69 6.69 6.69 6.69 6.69 6.69 6.90

12.97 13.00 17.95 14.92 8.48 18.62 19.55 19.55 14.41 11.79 13.12 18.22 19.28 14.66 12.18 14.41 13.60 16.77 14.19 8.70

0.764 0.765 0.179 0.215 0.803 0.391 0.265 0.339 0.319 0.453 0.279 0.332 0.348 0.253 0.745 0.339 0.647 0.701 0.259 0.486

Yes Yes No No Yes Yes No Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes No Yes

30.79 39.49 43.15 33.73 29.07 26.31 29.44 27.13 27.80 27.47 27.64 28.09 27.23 48.62 35.83 29.41 36.19 30.89 28.11 29.77 31.40 32.37 39.88 47.90 45.10 42.13 22.64 21.29 26.49 47.48 21.78 41.01 28.20 25.44 20.22 23.61 27.29 39.39 33.77 25.52 20.27 21.55 25.65 22.45 46.73 38.79 43.58 24.20 45.97 38.61 38.60 41.27 29.27 47.97

0.186 0.210 0.193 0.285 0.216 0.207 0.375 0.374 0.448 0.427 0.420 0.538 0.431 0.166 0.293 0.191 0.207 0.172 0.260 0.353 0.462 0.304 0.312 0.231 0.263 0.212 0.424 0.161 0.436 0.200 0.198 0.173 0.204 0.369 0.223 0.335 0.370 0.467 0.273 0.454 0.698 0.385 0.505 0.591 0.188 0.331 0.196 0.267 0.229 0.211 0.707 0.766 0.210 0.194

Yes Yes Yes No No No Yes Yes Yes Yes Yes Yes No No Yes Yes Yes Yes No Yes Yes No No No No No Yes Yes Yes No No No No Yes No Yes Yes Yes No No Yes No No No No Yes No No No No Yes Yes Yes Yes

Seismic scenario D: 6.5 6 M 6 7.5, 20 6 R 6 50 (km), PGA P 0.16 g, Site class D (360 6 vs 6 180 m/s) 1 NCA 21.12.1954 Ferndale City Hall 6.50 2 SFN 09.02.1971 LA-Hollywood Stor 6.61 3 IV 15.10.1979 Brawley Airport 6.53 4 IV 15.10.1979 Delta 6.53 5 IV 15.10.1979 EC County Center FF 6.53 6 IV 15.10.1979 El Centro Array#10 6.53 7 IV 15.10.1979 El Centro Array#11 6.53 8 IV 15.10.1979 El Centro Array#4 6.53 9 IV 15.10.1979 El Centro Array#5 6.53 10 IV 15.10.1979 El Centro Array#6 6.53 11 IV 15.10.1979 El Centro Array#7 6.53 12 IV 15.10.1979 El Centro Array#8 6.53 13 IV 15.10.1979 El Centro Diff. Array 6.53 14 IV 15.10.1979 Parachute Test Site 6.53 15 SH 24.11.1987 El Centro Imp. Co. C. 6.54 16 SH 24.11.1987 Wildlife Liquef. Array 6.54 17 SP 07.12.1988 Gukasian 6.77 18 LP 18.10.1989 Coyote Lake Dam 6.93 19 LP 18.10.1989 Gilroy-Historic Bldg. 6.93 20 LP 18.10.1989 Gilroy Array#2 6.93 21 LP 18.10.1989 Gilroy Array#3 6.93 22 LP 18.10.1989 Gilroy Array#4 6.93 23 LP 18.10.1989 Gilroy Array#7 6.93 24 LP 18.10.1989 Hollister City Hall 6.93 25 LP 18.10.1989 Hollister Diff. Array 6.93 26 LP 18.10.1989 Sunnyvale-Colton Ave 6.93 27 CM 25.04.1992 Rio Dell Overpass-FF 7.01 28 LAN 28.06.1992 Morongo Valley 7.28 29 NOR 17.11.1994 Canyon Country-W 6.69 30 NOR 17.11.1994 Downey-Co Maint 6.69 31 NOR 17.11.1994 Hollywood-Will. Ave 6.69 32 NOR 17.11.1994 LA-116th St School 6.69 33 NOR 17.11.1994 LA-Baldwin Hills 6.69 34 NOR 17.11.1994 LA-Centinela St 6.69 35 NOR 17.11.1994 LA-Century City 6.69 36 NOR 17.11.1994 LA-Hollywood Stor 6.69 37 NOR 17.11.1994 LA-N Westmoreland 6.69 38 NOR 17.11.1994 LA-Obregon Park 6.69 39 NOR 17.11.1994 LA-S Grand Ave 6.69 40 NOR 17.11.1994 LA-Saturn St 6.69 41 NOR 17.11.1994 Newhall-Fire Sta 6.69 42 NOR 17.11.1994 Newhall-W.P.Can. Rd. 6.69 43 NOR 17.11.1994 Pardee – SCE 6.69 44 NOR 17.11.1994 Santa Monica C.H. 6.69 45 KOB 16.01.1995 Abeno 6.90 46 KOB 16.01.1995 Amagasaki 6.90 47 KOB 16.01.1995 Fukushima 6.90 48 KOB 16.01.1995 Kakogawa 6.90 49 KOB 16.01.1995 Shin-Osaka 6.90 50 KOB 16.01.1995 Tadoka 6.90 51 KOB 16.01.1995 Takarazuka 6.90 52 DZC 12.11.1999 Bolu 7.14 53 DZC 12.11.1999 Lamont 1062 7.14 54 HM 16.10.1999 Amboy 7.13

E.I. Katsanos, A.G. Sextos / Advances in Engineering Software 58 (2013) 70–85

References [1] Lagaros ND, Papadrakakis M. Neural network based prediction schemes of the non-linear seismic response of 3D buildings. Adv Eng Softw 2012;44(1): 92–115. [2] Elnashai AS, McClure DC. Effect of modeling assumptions and input motion characteristics on seismic design parameters of RC bridge piers. Earthq Eng Struct Dyn 1996;25(5):435–63. [3] Shome N, Cornell CA, Bazzurro P, Carballo JE. Earthquakes, records and nonlinear responses. Earthq Spectra 1998;14(3):469–500. [4] Padgett J, Desroches R. Sensitivity of seismic response and fragility to parameter responses. J Struct Eng 2007;133(12):1710–8. [5] Bazzurro P, Cornell CA. Disaggregation of seismic hazard. Bull Seismol Soc Am 1999;89:501–20. [6] Stewart JP, Chiou SJ, Bray RW, Graves P, Somerville G, Abrahamson NA. Ground motion evaluation procedures for performance-based design. PEER report 2001/2009. Berkeley (CA): Pacific Earthquake Engineering Research Center, University of California; 2001. [7] Bommer JJ, Acevedo GD. The use of real earthquake accelerograms as input to dynamic analysis. J Earthq Eng 2004;8(1):43–91. [8] Baker J, Cornell CA. A vector-valued ground motion intensity measure consisting of spectral acceleration and epsilon. Earthq Eng Struct Dyn 2005;34:1193–217. [9] Luco N, Cornell CA. Structure-specific scalar intensity measures for near-source and ordinary earthquake ground motions. Earthq Spectra 2007;23(2):357–92. [10] Tothong P, Luco N. Probabilistic seismic demand analysis using advanced ground motion intensity measures. Earthq Eng Struct Eng 2007;36:1837–60. [11] Malhotra PK. Strong-motion records for site-specific analysis. Earthq Spectra 2003;19(3):557–78. [12] Shantz TJ. Selection and scaling of earthquake records for nonlinear dynamic analysis of first mode dominate bridge structures. In: Proceedings, 8th U.S. national conference on earthquake engineering, paper No. 1012, San Francisco, USA, 2006. [13] Beyer K, Bommer JJ. Selection and scaling of real accelerograms for bidirectional loading: a review of current practice and code provisions. J Earthq Eng 2007;10(1):13–45. [14] Abrahamson NA, Silva WJ. Empirical response spectral attenuation relations for shallow crustal earthquakes. Seismol Res Lett 1997;68:94–127. [15] Baker J. Conditional mean spectrum: tool for ground motion selection. J Struct Eng 2011;137(3):322–31. [16] Naeim F, Alimoradi A, Pezeshk S. Selection and scaling of ground motion time histories for structural design using genetic algorithms. Earthq Spectra 2004;20(2):413–26. [17] Youngs RR, Power MS, Wang G, Makdisi FI, Chin CC. Design ground motion library (DGML) – Tool for selecting time history records for specific engineering applications. In: SMIP Seminar on Utilization of Strong Motion data; 2007. [18] Kottke A, Rathje EM. A semi-automated procedure for selecting and scaling recorded earthquake motions for dynamic analysis. Earthq Spectra 2008;24(4):911–32. [19] Katsanos EI, Sextos AG, Manolis GD. Selection of earthquake ground motion records: A state-of-the-art review from a structural engineering perspective. Soil Dyn Earthq Eng 2010;30(4):157–69. [20] Hancock J, Bommer JJ, Stafford PJ. Numbers of scaled and matched accelerograms required for inelastic dynamic analyses. Earthq Eng Struct Dyn 2008;37:1585–607.

85

[21] Krawinkler H, Medina R, Alavi B. Seismic drift and ductility demands and their dependence on ground motions. Eng Struct 2003;25:637–53. [22] Bazzurro P, Cornell CA. Seismic hazard analysis of nonlinear structures. I: Methodology. J Struct Eng 1994;120(11):3320–44. [23] Sextos AG, Katsanos EI, Manolis GD. EC8-based earthquake record selection procedure evaluation: Validation study based on observed damage of an irregular R/C building. Soil Dyn Earthq Eng 2011;31:583–97. [24] Iervolino I, Galasso C, Cosenza E. REXEL: computer aided record selection for code-based seismic structural analysis. Bull Earthq Eng 2010;8:339–62. [25] Dias J, Castro JM, Romão X, Gonçalves M, Lopes JC. SelEQ: a web-based application for the selection of earthquake ground motions for structural analysis. In: Proceedings of the 14th European conference of earthquake engineering, Ohrid, F.Y.R.O.M.; 2010. [26] Jayaram N, Lin T, Baker J. A computationally efficient ground-motion selection algorithm for matching a target response spectrum mean and variance. Earthq Spectra 2011;27:797–815. [27] Eurocode 8. EC8: Design provisions of structures for earthquake resistance. Part 1: General rules, seismic actions and rules for buildings. Part 2: Bridges, Final Drafts pr EN1998-1 and -2, European Committee for Standardization (CEN), Brussels, Belgium; 2004. [28] FEMA P-750. NEHRP recommended seismic provisions for new buildings and other structures. Washington, DC (USA): Building Seismic Safety Council; 2009. [29] CSI, Computers and Structures. SAP2000: Integrated software for structural analysis and design, ver.14, Berkeley, California (USA); 2009. [30] Mathworks, Inc. MATLAB: The language of technical computing, v.R2010a. Natick, Massachusetts, USA. [31] Chiou B, Darragh R, Gregor N, Silva W. NGA project strong motion database. Earthq Spectra 2008;24(1):23–44. [32] Ambraseys NN, Douglas J, Rinaldis D, Berge-Thierry C, Suhaloc P, Costa G, et al. Dissemination of European strong-motion data, vol. 2, CD-ROM collection. UK: Engineering and Physical Sciences Research Council; 2004. [33] Buratti N, Stafford PJ, Bommer JJ. Earthquake accelerogram selection and scaling procedures for estimating the distribution of drift response. J Struct Eng 2011;137(3):345–57. [34] Silvestri S, Gasparini G, Trombetti T, Foti D. On the evaluation of the horizontal forces produced by grain-like inside silos during earthquakes. Bull Earthq Eng 2012. http://dx.doi.org/10.1007/s10518-012-9367-6. [35] Tirca LD, Foti D, Diaferio M. Response of middle-rise steel frames with and without passive dampers to near-field ground motions. Eng Struct 2003;25(2):169–79. [36] Tirca LD. Behavior of MRFs subjected to near-field earthquakes. In: Mazzolani FM, Tremblay R, editors. 3rd International conference STESSA 2000, Montreal; 2000. p. 635–43.Alavi B, Krawinkler H. Consideration of near-fault ground motion effects in seismic design. In: Proceedings of the 12th world conference on earthquake engineering, Auckland, New Zealand; 2000. [37] Benjamin JR, Cornell CA. Probability, statistics, and decision for civil engineers. New York: McGraw-Hill; 1970. [38] Cornell CA, Jalayer F, Hamburger FO, Foutch DA. Probabilistic basis for 2000 SAC federal emergency management agency steel moment frame guidelines. J Struct Eng 2002;25(4):371–400. [39] EAK. Hellenic Antiseismic Code. Athens (Greece): Earthquake Planning and Protection Organization, Ministry of Public Works; 2002.