Sie sind auf Seite 1von 27

THE STRUCTURAL DESIGN OF TALL BUILDINGS,

Vol. 5, 1-27 (1996)

SEISMIC DEMAND EVALUATION FOR A 4-STORY STEEL FRAME STRUCTURE DAMAGED IN THE NORTHRIDGE EARTHQUAKE
H. KRAWINKLER AND A. AL-ALI Department of Civil Engineering, Stanford University, Stanford, C A 94305-4020, U . S . A .

SUMMARY This paper summarizes results and conclusions from a case study concerned with the prediction of seismic demands and correlation of these demands with connection fractures discovered after the Northridge earthquake. Two adjacent buildings are used for this purpose. One is a 4-story building that experienced many connection fractures in a N-S perimeter frame. The other is a 2-story building that did not exhibit visible connection fractures. The discussion focuses on analytical modeling issues and the interpretation of analytical results obtained from eight series of static (pushover) and dynamic (time history and spectral) analyses. The analytical models are different in each analysis series, ranging from a simple elastic centerline analysis model to inelastic models that incorporate the contributions of the floor slab to the lateral strength and stiffness of moment resisting and simple frames. Two of the models are preliminary attempts to model the post-fracture behavior of one of the frame structures.

1. INTRODUCTION
The 17 January 1994 Northridge earthquake has exposed significant problems in steel moment resisting frame structures with welded connections. In more than 100 buildings, fractures were detected in and around fuii penetration welds connecting beam flanges to column flanges. Even though these fractures did not lead to collapses of buildings, they raise concerns about the seismic safety of steel frame structures and may necessitate a performance evaluation of existing structures. Fundamental questions concern to what extent analytical predictions of seismic demand can be used to assess connection performance and what consequences for the safety of the structural system may result from fractures connections. Structures damaged in the Northridge earthquake provide an apportunity to address these questions. This paper summarizes one of the several case studies of a coordinated program in which a number of damaged structures were analysed in detail. The case study discussed focuses on the following specific objectives: (1) to predict elastic strength and inelastic deformation demands for representative ground motions; (2) to evaluate the correlation between predicted seismic demands and observed connection fractures; (3) to assess the sensitivity of demand predictions to different assumptions made in the analytical model; (4) to provide a preliminary assessment of the consequences of connection fractures; ( 5 ) to assist in identifying areas needing future study.

CCC 1062-8002/96/010001-27 0 1996 by John Wiley & Sons, Ltd.

Received July 1995

H. KRAWINKLER A N D A . AL-ALI

Two adjacent buildings are used for this purpose. One of them, a 4-story building with steel moment resisting frames (MRFs) at the perimeter, experienced many connection fractures in one of the N-S perimeter frames during the Northridge earthquake. The other building, a 2-story steel MRF structure, did not exhibit visible connection fractures. The emphasis in this paper is on the demand predictions for the 4-story MRF in which connection fractures were observed. Eight series of static (pushover) and dynamic (time history and spectral) analyses were performed, utilizing four recorded ground motions, an equal hazard spectrum, and nine simulated records that were generated to represent the ground motions of the Northridge earthquake at the site of the buildings. The analytical models are different in all eight analysis series, ranging from a simple elastic centerline analysis model to inelastic models that incorporate the contributions of the floor slab to the moment resisting and simple frames. Two of the models are preliminary attempts to model the post-fracture behavior of one of the frame structures. This paper summarizes properties of the 4-story building and provides a brief description of the observed connection damage. Then it focuses on analytical modeling issues for the structure and its elements. The element models discussed are approximate, but they serve the purpose of providing information on the sensitivity of the predicted results. Much of the paper focuses on the evaluation of analytical results, the sensitivity of results to modeling assumptions and the relevance of the results for an assessment of the likelihood of connection fracture.

2. PROPERTIES OF 4-STORY CASE STUDY BUILDING A plan view of the 4-story building, which is rectangular with plan dimensions of 11 1 ft x 63 ft, is shown in Figure 1. The structural system consists of complete perimeter MRFs and interior simple frames with the beam and column sections shown in Figures 2 and 3. The MRF at the

tN

Figure 1. Plan view of 4-story building

STEEL FRAME SEISMIC DEMAND EVALUATION


+Story

Bldg.,

MRF on Line D (NS)

4-Story Bldg.,

Simple Frame on Llne C (NS)

(a)

(b)

Figure 2. Sections of M R F on line D (a) and simple frame on line C (b). (N-S direction, 4-story building)

$-Story Bldg., HRF on Llne 1 (EW)

4-Story Bldg.,
W16X31

Slmplo Frame on Line 2 (EW)


W16X26 W18X46 W16X26

W18X35

W18X35 rn

W21X57

W2lX50

W18X35

i
5
W21X62 W21X57

W18X35

W2lXSO

W18X35

W18X35

W2lX50
b ?

B
,
/'
/

5
MV'

__,
/

T
400'

f-- CB'
I

/'

Figure 3. Sections of M R F on line 1 (a) and simple frame on line 2 (b). (E-W direction, 4-story building)

east side is located 8 ft from the end of the building in order to avoid foundation interference with an adjacent building. The columns of the MRFs are embedded into grade beams and anchored to the top of the pile cap and may be considered as fixed at the base. The foundation system comprises drilled concrete piers with pile caps, grade beams and tie beams. Typical drilled piers are 2 4 diameter and 34ft long. One to three piers are provided at column locations. A 26" x 28" concrete grade beam is used at the ground floor with reinforcing bars welded to the columns. The columns

H. KRAWINKLER A N D A. AL-ALI

of the simple frames are connected to individual column footings with base plates and anchor bolts and are considered as pinned at the ends. All columns are made of A-572 grade 50 steel. All beams and girders are made of A-36 steel. The girder-to-column connections utilize full penetration groove welds at girder flanges and shear tab plates with A325 bolts at the girder web. A36 continuity plates, a,, less than girder flange thickness, are used opposite girder flanges with full penetration groove welds to the column. Shear doubler plates are not needed in any of the joints. The floor construction is steel framing with a 6p thick slab ( 3 p light weight concrete on 3 composite metal deck) at floors and roofs. The exterior wall system is thin set brick veneer supported on a metal stud wall. The building was designed according to the 1988 UBC. The structural system qualified as a regular structure with an R, of 12 as defined in the UBC. Roofs were designed for 20 psf live load, and floors were designed for 80 psf live load plus 20 psf for partitions. At the time of the earthquake, the building had not yet been occupied. Thus, almost no live loads and only a fraction of the design partition load of 20 psf were present.

3. DAMAGE DESCRIPTION In the 4-story building, a total of 14 bottom flange fractures and two top flange fractures in beam-flange to column-flange connections and one shear-tab to column-flange connection failure were detected. The top and bottom beam-flange to column-flange connection fractures may typically be described as partial or complete separation between the weld and the column flange, or fractures within the weld material. There were many additional cases in which cracks were detected through ultrasonic inspection, but the severity of these cracks was difficult to judge. All but one of the severely fractured beam flange connections were in the N-S moment frame on Line D (see Figure 1). The locations of these fractured connections are shown in Figure 4. The black portion of the fracture symbol indicates the location(s) of the fracture (bottom flange, or both flanges). No fracture was observed in the N-S moment frame on Line A, and only one fracture was observed in the E-W frame on Line 1. Thus it appears that the N-S frame on Line D was subjected to much more severe deformations than the other moment frames.

Figure 4. Location of fractured connections in N-S MRF on line D

STEEL FRAME SEISMIC DEMAND EVALUATION

4. ANALYTICAL MODELING ISSUES


Two- and three-dimensional elastic analyses and two-dimensional inelastic static (pushover) and dynamic (time history) analyses were performed. In all analyses, the following dead loads were applied as inital conditions. Uniformly distributed floors load Uniformly distributed roof load
= 79 = 74

psf (including 10 psf partitions) psf (including 6 psf for penthouse)

Weight of exterior walls = 16 psf over exterior surface The loads were distributed to the beams and girders according to well established load path assumptions. No live loads were applied, since the building was not occupied at the time of the earthquake.
4.1. Computer models of 4-story structure

An elastic model for the complete structure was formulated for three-dimensional analysis, which was needed to evaluate torsional effects caused by asymmetry in the N-S direction. No inelastic three-dimensional analyses were performed. The great majortiy of the analyses were two-dimensional, using the computer program DRAIN-2DX. For this purpose, the structure is modeled as a series of two-dimensional frames linked together at the floor levels by rigid links. Two frames each are used in the N-S and E-W directions, one representing an exterior perimeter MRF and one representing an interior simple frame, as shown in Figure 5. In the N-S direction the frame on line D is chosen as the MRF, since it has a larger tributory floor area than the frame on Line A and is expected to be subjected to larger seismic demands. Torsional effects are accounted for in an approximate manner by assigning seismic masses equal to tributary masses. The accuracy of this assumption will be discussed later. The incorporation of the simple frames is necessary in order to model their contributions to strength and stiffness. Even if all simple beam connections are modeled as hinges, the simple frames contribute to strength and stiffness because the columns that are part of the orthogonal MRFs are fixed at the base. Moreover, in several of the analysis models the contributions of the simple connections to strength and stiffness are also modeled. The gravity loads are distributed to the individual frames, including the simple frames, according to tributary areas. This is necessary in order to transfer P-delta effects from the

MRF on Line D

Simple Frame on Line C

Figure 5. Computer model for two-dimensional analysis (N-S direction)

H. KRAWINKLER A N D A. AL-ALI

COLUMN JOINT. SPRING

I I
BEAM

SPRING

Figure 6. Element modeling at beam-to-column joint

interior of the structure through the diaphragms (rigid links) to the perimeter MRFs, while still maintaining realistic axial loads in the MRF columns. The total seismically effective weight for the structures includes all uniformly distributed dead loads, the exterior walls, and a 10 psf partition load on each floor. The total seismically effective weights for the computer models are as follows:

N-S:

W, = W, = W 4
Koof =
F o t

356 k 312 k 1380 k (58% of total weight because of asymmetry)

E-W:

W, = W, = W 4 = 308 k Koof = 273 k 1197 k (half of total weight) Fo, =

The modeling of elements at a beam-column joint is illustrated in Figure 6. Columns are modeled as standard beam-column elements with the AISC Plastic Design M-P interaction diagram. Since only in very few cases did the axial beam column load exceed O.15Py,the column strength was usually equal to M,. Beams are modeled as linear elastic elements with a very large strength. Plastic hinge rotations at the ends of the beams are modeled by means of rotational springs which are given a very large elastic stiffness and a strength equal to the bending strength of the beam at the connection. This modeling facilitates modifications of beam strength properties at the connection and the incorporation of a consistent strain hardening stiffness (as a fraction of 6EI/L of the beam). Joint panel zone shear deformations are modeled by means of a rotational spring connected to scissors rotating about the center of the joint. This model will be dicussed in Section 4.2. 4.2. Modeling of element strength and stifness properties Various elastic and inelastic models are used in the analyses to investigate the effects of different element strength and stiffness representations on the predicted seismic demands. The properties of the beam elements, beam springs, and joint elements are varied using the element models discussed next. The models discussed are approximate, but are deemed adequate for the purpose of the study.

Table I. Strength and stiffness properties of beams in 4-story building W18 x 40 W24 x 68 W24 x 76 W24 x 94 W24 x 104

2 P

M R F beam properties

F, (ksi) M, (k in) M , (k in) MG (k in) M,,,,(k in) M,:,,, (k in) Ib(in") I,,, (in") L = 20' I,,, (in") L = 30' I,,, (in") L = 40'
50.5 3959 4833 6798 765 2566 612 1136 47.3 8372 9918 13255 1530 4460 1830 2967 3019 3110 47.3 12014 14527 18341 1530 4546 2700 4161 4293 4652 4788 47.3 13770 13770 16087 1530 4509 3100 47.3 9460 11386 14882 1530 4488 2100 3318 3378 3481

I
F, (ksi) M, (k in) M,, (k in)
50.5 2232 510 1527 301 661 790 50.5 2727 510 1694 375 50.5 3358 765 2500 5 10 983 986

.$

2
1426 2219 2446

Simple beam properties


I,,, (in") L = 20' I,,, (in") L = 3 0 I,,, (in*) L = 40'

< Ib(in") (k in)

50.5 4580 765 2583 712

5 05 4772 765 2859 843 1539

50.5 5555 765 2877 984 1743

50.5 6515 756 2902 1170

47.3 6811 765 2894 1330

8 Z

H. KRAWINKLER A N D A. AL-ALI

4.2.1. Contribution offroor slab to stzflness of beam. The variable moment of inertia resulting from composite beam action under lateral loading cannot be included directly in a DRAIN-2DX analysis. An effective constant moment of inertia is used to estimate the effect of the slab contribution to the stiffness of the structure. Approximations should be acceptable considering the many uncertainties in estimating structure stiffness due to non-structural contributions. Many different equations can be found in the literature to estimate the effective width B of the slab for stiffness contribution under lateral loading (e.g. Reference 2). The following simple approach is used. which compromises between several recommendations: (for interior span beams) B = b + 0.2L

B =b

+ 0.1L + min [O.lL, overhang]

(for exterior span beams)

where b is the flange width of the steel beam and L is the span of the beam. Using this definition of effective width, the positive (slab in compression) and negative (slab in tension and assumed cracked) moments of inertia are computed and averaged, This average is taken as the effective moment of inertia. A pilot study did show that the beam end rotations do not differ by much whether this average I is used or a linear variation of I is assumed. for the beam elements are listed in Table I. The effective moments of inertia (Ieff) 4.2.2. Contribution of froor slab to strength of beam at welded connection. A simple ultimate strength model is used to estimate the strength of the beam spring, which represents the bending strength of the beam at the beam-to-column connection. The spring stiffness is assumed to be very large, i.e. the connection is treated as a rigid connection. The following estimated ultimate forces are assumed to act at the centroids of individual stress blocks (see Figure 7(a)). Ffb= 1.5FYx A,, Beam bottom flange: F, = min [cap. in shear, cap. in bearing] At each bolt location in web: Beam top flange: Slab in compression: Slabin tension:
F,, = 1-5Fyx A,,

F,

= 2fc x

t,, x b,,

4 = A,

xfy

Figure 7. Connection strength models: (a) moment connection; (b) simple (shear) and fractured connections

STEEL FRAME SEISMIC DEMAND EVALUATION

where A,, is the area of the beam flange, A , is the reinforcement in the effective width of the slab, b,, is the width of the column flange, t,, is the thickness of the floor slab, F, is the yield strength of the beam steel, andf; is the compressive strength of the slab concrete. There are many assumptions involved in these equations, the most important being the following: (a) The beam flange can develop a stress of 1.5Fy without fracture at the weld. (b) The slab concrete bearing against the column flange can develop a stress of 2fc'. (c) In the web connection, failure modes other than bolt shear and bearing are not critical. (d) The contribution of partial welding of the shear tab is negligible. These assumptions will result in reasonable upper bounds of stress resultants, and correspondingly in upper bound estimates of the bending strength at the connection. In this case study, the force in the slab in tension is small in all cases since only a 6 x 6-W1.4 x W1.4 wire mesh ( A , = 0-029in2 ft - ') is provided as reinforcement. Thus, F, is ignored in the strength calculations. Accepting these assumptions, the positive and negative bending strengths at the connection ( M ; , M L ) can be computed from force couples as follows (see Figure 7(a)): Slab in compression:

+ F, x e2 Ffbx e , + F, x e2
Ffb

x e,

(ifF,,, < F,) (ifF, > F,)

where F, = I F b , e , is the distance from the beam bottom flange to the centroid of the compression forces (assumed to be 1.5" above the top flange) and e2 is the distance from the center of the bolt group to the centroid of the compression forces Slab in tension: where d , is the depth of the beam, Mpb is the plastic moment capacity of the beam and This strength model is used in one of the inelastic analysis models (Model 2), whose purpose it is to incorporate an upper bound on beam strength in order to estimate an upper bound on joint distortion demands and a lower bound on beam plastic rotation demands. Computed values for the positive and negative strengths ( M ; and M ; ) are listed in Table I.

I(Fbi x di)is the moment capacity of the bolt group (Mweb in Figure 7).

4.2.3. Estimates o f strength and stifness of simple beam connections. For simple connections, which contain only shear tabs and no flange connections, the same strength model is used as for welded connections, but no flange contributions are considered (see Figure 7(b)). Thus, the following bending strengths are estimated.
Slab in compression:
ML,, = min (F,,

C Fb)x e

( e = d,/2
di)

+ t,,/2)

(3)

Slab in tension:
= Mweb =

C (Fbi

(4)

Computed values for the positive and negative strengths (M;,, and M,;,,) for the simple beam connections used in the 4-story building are listed in Table I. The strength model discussed here is similar to that given in the recommendations of Deierlein and Leon.3 In regard to stiffness, these recommendations are followed literally. It is assumed

10

H. KRAWINKLER A N D A. AL-ALI

that the positive strength (MZ,,,,) is attained at a rotation of 002, and that the strength increases linearly to this value. The same elastic stiffness is maintained in the negative direction. The strain hardening ratio is assumed as zero, resulting in a bilinear moment-rotation relationship at the beam end springs and bilinear hysteresis loops. The latter result may be unrealistic for simple connections, but needs to be accepted in this case study because of the limitations of DRAIN-2DX.
4.2.4. Estimate o f beam strength at fractured connections. Two series of inelastic analyses are performed in which the post-fracture behavior of connections is modeled approximately. In these analyses it is assumed that weld fractures occur very early and that the pre-fracture stage can be ignored. In order to model connection behavior after fracture it is assumed that both the beam top and bottom flanges have fractured and that the fractured surfaces never establish contact again. With this assumption, the connection behaviour approaches that of a simple connection with the contribution of the floor slab. Thus, the previously discussed strength and stiffness model for simple connections can also be employed for fractured moment resisting connections. Computed values for the positive and negative post-fracture bending strengths (ML,,, and M;,,,) for beams in MRFs of the 4-story building are listed in Table I.

4.2.5. Modeling of shear behavior of joint panel zones. In order to model the shear strength and stiffness of joint panel zones, two nodes are defined at the center of the joint shown in Figure 6. Translational displacements of the nodes are identical, but different rotations are permitted in the elements connected to the nodes. These rotations are linked by a rotational spring element with appropriate moment-rotation characteristics. The dimensions of the joint are taken into account by defining rigid bodies equal to the depth of the column and depth of the beam (set of scissors type rigid bodies), representing infinite bending stiffness of the joint. Thus, only the shear distortion of the joint is taken into account, which is represented by the relative rotation of the rigid bodies, i.e. by the rotation of the joint spring. Rotational stiffnesses and moment strength properties are assigned to this spring to represent the shear stiffness and strength of the panel zone. The shear strength and stiffness properties of joint panel zones are modeled as proposed by K r a ~ i n k l e r .In ~ this approach the panel zone shear force demand V is computed from the unbalanced beam moments at the column faces ( M i + Mi) and the column shear force V , by the equation

where d b is the depth of the beam. A reasonable representation of the shear force-shear distortion relationship of a joint panel zone is trilinear, defined by a yield shear force V , and the corresponding yield distortion y y , an ultimate shear force V,, which is associated with a distortion of 4yy, and a strain hardening stiffness for distortions exceeding 47,. The quantities defining this relationship have been given by Krawinkler4 as follows:

V , = 0,55Fyd,t

(6)

yy=J3G
where F, is the panel zone yield strength, t is the panel zone thickness, and d,, b,, and tCfare

F Y

STEEL FRAME SEISMIC DEMAND EVALUATION

11

the depth, flange width and flange thickness of the column section. The properties of the joint spring representing the shear-force/shear-distortion behavior can be obtained by recognizing that the rotation of the joint spring 0 is equal to the average angle of shear distortion of the joint y and by converting the shear force quantities to moments at the center of the joint, which can be done by means of a geometric transf~rmation.~

4.3. Yield strengths o f beam and column materials


The yield strength of the structural material is based on estimates of mean values rather than nominal values. Based on statistical information given in Reference 5, the mean yield strength of the A36 beams is between 50.5 and 47.3 ksi (Group 1 and Group 2 shapes), and the mean yield strength of the A572 Grade 50 columns is between 57-5 and 57.3 ksi (Group 2 and Group 3 shapes). All strength values listed in Table I are based on these yield strengths.
4.4. Important indicators of likelihood of fractures

Disregarding the critical issues of residual stresses and workmanship, which cannot be assessed from a global structural analysis, the likelihood of fractures is expected to increase with the stress and deformation level at and around the beam-flange to column-flange welds. Fractures may occur when the global stress level in critical regions is still elastic. In this case the demands on beam moments, column moments and joint shear forces should be good indicators of the likelihood of fractures. These demands are best described in terms of strength demand/capacity (D/C) ratios, using the global yield capacity of members as the normalizing parameter (i.e. M , of beams, M,, of columns and V, of joint panel zones). Once these D/C ratios exceed unity, elements of the structure have yielded and redistribution occurs, and the D/C ratios become questionable indicators. Nevertheless, they are the only rational indicators if only an elastic analysis is performed. One of the objectives of this case study is to assess the value of the elastic D/C ratios for structures that are expected to undergo severe inelastic deformations. For structures that behave inelastically in earthquakes, the magnitude of plastic deformations is believed to be a better indicator. Global yielding around beam-to-column connections may occur at beam ends, in the column above or below the joint and in the joint panel zone. Measures of global demands are now the rotations at plastic hinges in beams and columns, and the plastic distortions of joint panel zones. Much emphasis is placed in this case study on an assessment of these indicators. Interstory drift is another global indicator. It does not provide direct information on plastic rotations and distortions, but it can be used to estimate these quantities, provided a reference yield value of drift can be estimated. Since it is often feasible to estimate story drifts from the total lateral displacement of the structure, and the latter can be related with reasonable accuracy to the spectral displacement, the interstory drift is considered an attractive indicator. Thus, the purpose of the case study analysis is to predict these three indicators (D/C ratios, plastic rotation and distortion demands, and story drifts) and assess their relevance in the determination of the likelihood of fractures.
4.5. Analytical models and types of analysis

Eight models are employed in this case study. The models are of increasing complexity and are intended to investigate the effects of various modeling assumptions on the prediction of those seismic demands on which an assessment of the likelihood of weld fractures may be based.

12

H. KRAWINKLER A N D A. AL-ALI

The eight analytical models are as follows: (1) Elastic Benchmark Model, using a bare steel frame model and centerline dimensions of beams and columns. (2) Elastic Model, considering the joints as elastic but deformable elements and incorporating the contribution of the floor slab to the beam stiffness. (3) Inelastic Benchmark Model, considering inelastic deformations in beams, columns and joint panel zones, but no contribution of the floor slab. Zero strain hardening is assumed in beams and columns. (4) Inelastic Model I, considering inelastic deformations in beams, columns and joint panel zones, and incorporating the contribution of the floor slab to the beam stiffness; 3% strain hardening in beams and columns is assumed in this and all following models. (5) Inelastic Model I with Elastic Joints, same as Model 1 but assuming that the joint panel zones have infinite shear strength. (6) Inelastic Model 2, considering inelastic deformation in beams, columns and joint panel zones, and incorporating the contribution of the floor slab to the beam stiffness and strength in the MRFs and simple frames. (7) Inelastic Model 3-1, similar to Model 2, but modeling connection fractures at the joints in which fractures have been observed after the Northridge earthquake (see Figure 4). (8) Inelastic Model 3-47, similar to Model 2, but modeling connection fractures at all joints of the MRF. Models 3-1 and 3-2 are utilized only for the N-S direction of the 4-story building, using the MRF in which almost all the fractures have occurred (MRF on Line D). The different models are employed to perform time history analyses, using ground motion records discussed in the next section. The elastic models are used also to perform modal analysis. In all dynamic analyses discussed in this paper, 2% damping is assumed. For all inelastic models a pushover analysis is carried out in which the UBC load pattern (lateral loads proportional to story mass times distance from the base) is used to perform an incremental static analysis. The pushover analysis execution is displacement controlled, using small displacement increments up to a global drift 6/H of 0.03, where 6 is the roof displacement and H is the structure height. This global drift index corresponds to a roof displacement of 18 in. In all analyses, the P-delta effect is included. 5. GROUND MOTIONS In the portion of the case study discussed in this paper, the ground motions are represented by nine simulated site specific ground motions. The nine records were generated at grid points spaced 1 km apart. The grid points correspond to stations; Station 5 being the center of the grid. The background behind the simulation procedure employed to generate these records has been discussed by Somerville.6 The nine simulated motions are intended to represent the motions during the Northridge earthquake at the site of the 4-story building and to permit an assessment of the variability in expected response. The acceleration spectra of the individual records exhibit large spikes in the short period range, but not around the fundamental period of the 4-story building (between 1.0 and 1.5 s in the N-S direction, depending on the analytical model). Expected values and the dispersion of spectral ordinates are illustrated in Figure 8, which shows the mean and (mean ? a) elastic strength and displacement demand spectra for the N-S component of the 9 motions. The mean elastic strength demand spectrum (elastic strength demand over weight F,/ W is equal to

STEEL FRAME SEISMIC DEMAND EVALUATION


3

13

Genemted Motions for Site #4 (NS). Damping = 2%

. 4
B
1

0 0.0
0.5
1.0

1.5

2.0

2.5

3.0

3.5

4.0

T(=c)
(a)
40

Generated Motions for Site #4(NS).Damping = 2%

C
5 0

C 30
(D

( I )

20

n
I0

0
0.0
0.5
1.0 1.5

2.0

2.5

3.0

3.5

4.0

T(W)

(b)
Figure 8. Mean and (mean + r ~ ) spectra of elastic strength (a) and displacement (b) demands (N-S direction)

spectral acceleration over acceleration of gravity SJg) still shows the peaks in the short period range. The disperion varies with period and is relatively small for periods above 1.0 s, but significant for shorter periods. The mean strength demand spectrum follows the shape expected for typical alluvium ground motions, but the spectral values are high-for two reasons. First, the spectrum is for the 2% damping used in the dynamic analysis and not for the usually assumed 5o/d damping, and secondly, the Northridge ground motions were more severe than is reflected in the present code design spectra. The mean spectra shown in Figure 8 provide preliminary information on expected global seismic demands for the N-S structural system. If the structure were to behave elastically and if higher mode effects are assumed to be negligible, the base shear demand can be estimated as 0 8 x 0 8 W = 064 W (estimate of F,/ W around the period 1 s, which is the fundamental period of the structure without fractured connections-see Table 11-times the estimate of first mode mass fraction times seismically effective weight), and the roof displacement can be estimated as

14

H. KRAWINKLER A N D A . AL-ALI

8.5 x 1.33 = 11 in (estimate of spectral displacement around the period of 1 s times estimate of first mode participation factor). This displacement corresponds to a global drift index of 11/(53 x 12) = 0.017. The dynamic analysis results presented later will show that these rough estimates are close to accurate analytical predictions. The E-W components of the nine motions are significantly smaller in severity than the N-S components. Thus the seismic demands imposed on the E-W structural system are also smaller, which is one reason why almost all fractures occurred in N-S frames.

6. ANALYTICAL PREDICTIONS OF SEISMIC DEMANDS

A comprehensive documentation of the analytical predictions of seismic demands for most of the models is provided in Reference 7. This discussion is concerned only with the N-S structure of the 4-story building, in which most of the fractures occurred. The emphasis is on predictions obtained from dynamic analyses using the 9 N-S components of the simulated site ground motions. Most of the results presented are mean values obtained by averaging results from 9 time history analyses. The focus of the discussion is on an evaluation of sensitivity of the results to various modeling assumptions. All results presented in this section are obtained from two-dimensional analyses, using the computer model shown in Figure 5. Three-dimensional analyses were performed with the elastic benchmark model for the purpose of evaluating torsional effects. The three-dimensional time history analyses indicated that the seismic demands (interstory drifts, D/C ratios) in the MRF on Line D (Figure 1) are about 25% larger than those in the MRF on Line A and are close to those obtained from two-dimensional analyses. Thus, the selected two-dimensional model appears to be a realistic representation of the critical portion of the three-dimensional structure-at least for elastic analysis purposes.

6.1. Static and dynamic characteristics of N-S structure

A perspective of the global lateral load-displacement characteristics of all inelastic models can be obtained from the pushover analysis results presented in Figure 9. The figure shows the relationship between base shear V (normalized by seismically effective weight W) and roof displacement 6 (normaiized by structure height H ) for the six inelastic models. It illustrates how much is gained or lost in strength and stiffness by making different modeling assumptions. Using the Inelastic Benchmark Model as a reference, the following observations can be made on modeling of the undamaged structure. Including the contribution of the floor slab to the beam stiffness (Model 1) increases the elastic stiffness by 16% but has no effect on the structure strength. Including the contribution of the floor slab to the stiffness and strength of the beams in the MRF and simple frame (Model 2) increases the elastic stiffness by 26% and increases the maximum strength (at 6 / H = 003) by 24%. Ignoring yielding in the joint panel zones (Model 1 with Elastic Joints) increases the structure yield strength by 65% and the maximum strength by 35%. In most models of the undamaged structures (Benchmark Model and Models 1 and 2) the differences in strength and stiffness did not lead to very different predictions of seismic demands. The exception is Model 1 with Elastic Joints, which leads to wrong conclusions concerning the deformation demands, as will be discussed later. Using the Inelastic Model 2 as a reference, the following observations can be made on modeling the structure with fractured connections. Modeling fracture at the locations at which fracture was observed in the Northridge earthquake (Model 3-1) decreases the elastic stiffness to

STEEL FRAME SEISMIC DEMAND EVALUATION


0.5 0.45 0.4 0.35
0.3
I

15

5
>

0.25
0.2

Benchmark

0.15 0.1
0.05

Model I(Elas. Jtr.) Model 2

-0

- Model 3-1
Model 3-2

0
0.005 0.01 0.01 5 0.02 0.025 0.03

tiin
Figure 9. Normalised base shear versus roof displacement diagrams from pushover analyses (N-S direction)

65% of the pre-fracture stiffness and has a noticeable but not dramatic effect on structure strength. Even if all connections have fractured and lateral stiffness and resistance is provided only by connection forces in shear tabs and floor slabs (Model 3-2), significant stiffness (40% of pre-fracture) and strength can be mobilized. P-delta effects, which are considered in the analysis, d o not lead to a negative global stiffness at displacements experienced in the dynamic analyses. Thus, P-delta effects are not very important In this structure because of the relatively small values of axial loads. Since stiffness requirements and constructibility issues controlled most of the member sizes, the N-S structure is much stronger than required by the 1988 UBC. The design base shear at the allowable stress level is 0064 W . The elastic base shear strength is in the order of three times as large and the maximum base shear strength is in the order of six times as large. Thus, the elastic D/C ratios and the plastic deformation demands are expected to be relatively small even for the severe seismic inputs used in this study. The effects of the modeling assumptions on natural periods are summarized in Table 11. The effects on the fundamental period reflect the observations made on stiffness effects. For the undamaged structure they are noticeable, but diminish in importance once the uncertainties in evaluating real structure periods are considered. The analytical models do not consider non-structural contributions, and the fundamental period for all models appears to be on the high side. Using the empirical equation T = 0.035(h,)34,a fundamental period 0.69 s is obtained.
Table 1 1 . Computed periods for different analytical models (N-S direction) Elastic Period No. Benchmrak Benchmark Model 1 Inelastic Model 2 Model 3-1 Model 3-2

1 2 3

1.03 0.36 0.19

1.10 0.36 0.17

1.02 0.33 0.17

0.98 0.32 0.16

1.22 0.35 0.17

1.54 0.44 0.19

16
6.2. Story shear force demands

H. KRAWINKLER A N D A. AL-ALI

Story shear force information is plotted in Figure 10. Story shears are separated into the components attracted by the MRF on Line D and the simple frame on Line C . The top two graphs are obtained from elastic time history analyses using the 9 site ground motion records. The simple frame attracts approximately 15% of the shear in the first story even though all beam ends are assumed hinged in the analysis. The shear is attracted because some of the columns in the simple frames are part of an orthogonal MRF and are fixed at the base. These column attract substantial shear in the first story because they have to follow the deflected shape of the MRF and are not free to deform as cantilevers. If the two base shear components for the mean of the 9 records are added up, the result is 0.72 W, which is reasonably close to the base shear demand of 0.64W estimated in Section 5 from the mean strength demand spectrum. The difference comes from higher mode effects. The shear resistance provided by the simple frame gains on importance in the inelastic range of behavior particularly in the post-fracture Models 3-1 and 3-2. This is illustrated in the lower graphs of Figure 10, which present results from the pushover analysis at first yield (Elastic Limit), at 18 in displacement (6/H = 0.03) and from time history analysis using the record at Station 5. The largest contribution from the simple frame is noted in the time history analysis for Model 3-2, where the simple frame contributes 27% of the base shear resistance. The conclusion to be drawn is that in safety evaluations of structures with fractured connections the lateral resistance of simple frames should not be neglected. The large difference in shears between the elastic limit and the 18in displacement of the pushover analyses shows that the two model structures yield early but have large strength reserves beyond yielding. The shear force patterns of the pushover and time history analyses are similar but by no means identical. In the dynamic analysis, the structure attracts higher shear forces because of dynamic redistribution, which is not considered in a static analysis with a constant load pattern. 6.3. Displacement and interstory drgt demands Figures 11 and 12 show displaced shapes and interstory drift demands for the nine site ground motion records and different analytical models. Mean values, maximum values, and the values for Station 5 are plotted. A mean displaced shape is not shown since it has no physical meaning. Of interest are, however, mean values of roof displacement 6, which are listed below for the different models: (G/H)mean = 0.0172 Elastic Model: Inelastic Model 1: ( C ~ / H )= ~ 0.0159 ~~,, Inelastic Model 2: (6/H)mean = 0.0149 Inelastic Model 3-1: ( c ~ / H ) , ,= , ~0.0193 ~~ Inelastic Model 3-2: (d/H)mean = 0.0294 The first value in this list, which is obtained from elastic dynamic analyses, is very close to the value estimated in Section 5 from the mean displacement spectrum. The values for Inelastic Models 1 and 2 show that the roof displacement is not very sensitive to the model assumption, and it follows expected patterns. It is expected that for fundamental periods around 1 s the inelastic displacement will be somewhat smaller than the corresponding elastic one. It is also expected that for stiffer structures (Model 2) the displacement will be smaller than for more flexible structures (Model 1).

STEEL FRAME SEISMIC DEMAND EVALUATION IFkor


NO (MRF)

17
N-S (Slmpla)

Floor

Mean of 9 G.M. -Max. of 9 G.M.


2

0.5

0.1

0.2

Story Shear I W

Story Shear I W

Floor

N-S (Slmplo)

1
0.2

1 I

: !

-d = 18 inches Station 5
0 0.05
0.1

0.4

0.15

Story Shear MT

Story Shear I W

Floor

N-S (MRF)

Floor

N-8 (Slmplo)

F
Elastic Limit TI = 18 inches -----Station 5
0

0.2

0.4

0.05

0.1

0.15

Story Shear I W

Story Shear I W

Figure 10. Story shear force distributions (N-S direction, 4-story building): (a) Elastic Model; (b) Inelastic Model 3-1; (c) Inelastic Model 3-2

18

H. KRAWINKLER A N D A. AL-ALI
Ks

I
Mean of G.M

Floor

KS

- MaxofG. M.
Station 5

0.01

0.02

0.03

0.02

0.04

Displacement I H

Intentory Drift

floor

Y-S

Floor
5 -

Y-S

I '

Mean o f G. M - M a x d G . M. Station 5

a ,

1
5

2 '

I
1,
I

0.01

0.02

0.03

0.02

0.04

DisplacementI H

Interstory Drill

N-f

MeanofG.M
3

Station 5

0.01

0.02

0.03

0.02

0.04

Dlsplacrmmt I H

Interstory Drin

Figure 11. Displaced shapes and interstory drifts for models of undamaged N-S structure (4-story building): (a) Elastic Model; (b) Inelastic Model 1; (c) Inelastic Model 2

STEEL FRAME SEISMIC DEMAND EVALUATION


Floor
N-S

19
N-0

Floor

p
Mean of G. M. -Max of G.M. Station 5 0
0.01

0.02

0.03

0.04

0.02

0.04

0.06

Dlsplacernenl I H

Interstory Drlfl

Floor

N-S

Floor
5 4

N-S

Mean of G.M. -Max of G.hi.

3
2

4'

Station 5
"

: I
1 4
#

0.01

0.02

0.03
0.04

Displacenmnt1 H

0.02

0.04

0.06

Inters!ory Drift

Figure 12. Displaced shapes and interstory drifts for post-fracture models (N-S direction, 4-story building): (a) Inelastic Model 3-1; (b) Inelastic Model 3-2

Considering the similarity of the graphs shown in Figure 1 1 for three of the models, it can be concluded that neither global drift nor interstory drifts are very sensitive to modeling assumptions. Even the Elastic Model predicts these quantities reasonably well. All three models indicate that the drift demand is higher in the second story than in the first one. Inelastic Model 1 has also high demands in the third and fourth story. It is expected that these patterns will also be reflected in the local member demands discussed later. For the post-fracture models, the roof displacement increases considerably-at least for Model 3-2, which has only so-called simple connections. But in view of the severe ground motions it is notable that this weak structure does not deflect by a huge amount. As Figure 12 shows, much of the roof displacement in Model 3-2 comes from interstory drifts in the upper stories, which is not the case for Model 3-1.

6.5. Element D / C ratios and plastic deformation demands


Figure 13 summarizes mean values of demand/capacity (D/C) ratios and plastic deformation demands for all elements of the MRF on Line D. The figure consists of frame drawings on which numerical values are indicated in boxes. The boxes pertain to beam and column ends and to joint panel zones. Results are presented for five of the analytical models. Mean D/C ratios for the Elastic Model are presented in the top graph. A D/C ratio larger

20

H. KRAWINKLER A N D A. AL-ALI

(d) (e) Figure 13. Mean D/C ratios (a) and plastic deformation demands (b)-(e) from dynamic analyses (N-S direction, nine records): (a) Elastic Model;(b) Inelastic Model 1; (c) Inelastic Model 2; (d) Inelastic Model 3-1; (e) Inelastic Model 3-2

STEEL FRAME SEISMIC DEMAND EVALUATION

21

than unity indicates that the demand predicted from elastic analysis exceeds the member capacity. Values larger than unity are noted at all beam ends, at all joints, and at the column bases. A cursory inspection would lead to the conclusion that all these members are expected to yield. This is not all true, as seen in the graphs in the middle row, because of inelastic redistribution. The D/C ratios are largest for joint panel zone shear forces on the second and third floor, indicating that these joints yield first and are severely overstressed in shear. This renders the listed D/C ratios for beams and columns questionable since joint yielding is expected to occur before beams attain their bending strength, and the joints may be weak enough to prevent plastic hinging in the beams. This speculation is confirmed in the graphs in the middle row, which show plastic element deformations (in radians) for two inelastic models of the undamaged structure. There is no yielding at the column bases and at many of the beam ends. Relatively small plastic hinge rotations occur at the exterior beam ends. Plastic deformations are concentrated primarily in joint panel zones that distort in shear. There are some differences between Models 1 and 2, caused by the increased beam strengths in Model 2, but in both cases joint shear distortions are the predominant mode of inelastic deformation. The pattern changes in the post-fracture Models 3-1 and 3-2. Significant plastic hinge rotations are noted wherever rigid connections are assumed to have fractured and are modeled as simple connections (see Section 4.2). Thus, in Model 3-2 (in which all connections are assumed to have fractured), all beam end springs undergo large plastic hinge rotations. It is to be determined through research whether these fractured connections, whose bending resistance comes from the shear tab and the floor slab, would be able to undergo these rotations without shear tab failure. The importance of the consideration of relative member strength is also documented in Figure 14, which shows D/C ratios and plastic deformations obtained from pushover analyses. The upper left graph shows the D/C ratios at the displacement closest to first yield (maximum D/C ratio = 1.01). The D/C ratio for joint shear is close to 1.0 at several locations, whereas the maximum D/C ratio for beams is 0.68. The D/C ratios for columns are much lower again. This clearly demonstrates that many joints yield in shear long before beams attain their bending strength. For this reason, the behavior pattern at beam-column joints is very different from what would be predicted from an elastic analysis and the corresponding D/C ratios for severe ground motions. Because of the limited panel zone shear strength, shear yielding is the predominant inelastic deformation mode, as demonstrated in the lower left graph in which are presented plastic beam rotations and joint distortions at a global drift of 6 / H = 0.03. The graph shows that only few beams have yielded and that most of the plastic deformations occur in joint panel zones. Only at exterior joints where only one beam frames into the column, and at the roof level where the beams have low strength, can plastic hinging of the beams be observed. This is an important observation since weld fractures have occurred at many of the joints in which the beams have not yielded according to analytical predictions. In fact, in several of the joints with fractured welds the beam stress level is much below yielding, as can be seen from the D/C ratios at the global drift of 0 0 3 presented in the upper right graph. For yielded beam springs, which have no strain hardening, the D/C ratio is 1.0 as expected. For yielded panel zone elements the D/C ratio is larger than 1.0 because of the hardening provided by the second stiffness of the panel zone shear model (see Section 4.2). The graph also shows that all columns are far from yielding (except at the fixed bases). Unfortunately it is common practice to ignore the limited panel zone shear strength in analytical predictions of seismic response, even when inelastic analysis programs are employed. The consequences are evident by comparing the lower two graphs of Figure 14. Both graphs shows plastic deformation demands at a global drift of 0.03 obtained from two pushover analyses.

22

H. KRAWINKLER A N D A . AL-ALI

Figure 14. D/C ratios (a) and (b) and plastic deformation demands (c) and (d) from pushover analyses (N-S direction): (a) Inelastic Benchmark Model, first yield; (b) and (c) Inelastic Benchmark Model, S,/H = 0.03; (d) Inelastic Model 1 with Elastic Joints, 6 / H = 0.03

The results shown in the lower right graph are obtained by giving a very large strength to the joints and preventing shear yielding (Model 1 with Elastic Joints). This conventional model increases the strength of the structure (see the V-6 diagram in Figure 9), but forces all inelastic deformation into the beams, indicating high plastic hinge rotation demands. In a more realistic model, which includes joint shear yielding, most of the yielding occurs in the joints and most of the beams remain elastic (lower left graph). It will be difficult to reconcile these differences in behavior by judgment.
6.5. Dispersion in predictions o f demands

The discussion so far has focused on the effects of analytical modeling on demand predictions, Realistic modeling of strength and stiffness properties should keep this effect within acceptable bounds. Another source of dispersion in predictions is the variability of strength properties. For instance, the mean yield strength of A36 and A572 steels is in the order of 49 ksi and 57 ksi, respectively. The variations about the mean may greatly affect the relative member strength and the mode of inelastic deformation. From the perspective of predicting seismic demands for the Northridge earthquake, the absence of recorded motions at the site adds great uncertainties to the predictions. The nine simulated ground motions summarized in Section 5 are used as representations of seismic input.

STEEL FRAME SEISMIC D E M A N D EVALUATION

23

It is not within the scope of this study to assess the reality of these motions. But the fact that nine representative records could be employed for dynamic analysis permits an assessment of the dispersion in demand predictions caused by the differences in the severity and frequency content of the records. As a simple measure, the COV is used for this purpose. The COV of SDOF elastic strength and displacement demands can be deduced from the spectra presented in Figure 8. In the period range of interest (1.0 to 1.5 s), the COV in spectral values is relatively small (in the order of 0.2). For all response quantities computed in this study, the COV is computed as well. Most relevant is perhaps the COV for roof displacement and interstory drifts. For the undamaged inelastic models, the COV is found to be comparable to the COV of the first mode spectral values. It is usually somewhat smaller than that value in the lower stories (in the order of 0.15), and usually somewhat larger in the upper stories (in the order of 0.25). For the post-fracture models, the dispersion becomes 1argi.r. Thus, it appears that there is some higher model effect, but that the response is governed primarily by first mode contributions and that the dispersion of demand predictions can be estimated from the dispersion of the first mode spectral ordinates. 7. PREDICTED DEMANDS VERSUS OBSERVED FRACTURES

A summary of observed connection fractures is presented in Section 3. Almost all the fractures occurred in the N-S frame on Line D of the 4-story building, at the locations identified in Figure 4. Thus, a correlation between predicted demands and observed fractures can only be done for the frame on Line D. Inelastic Models 1 and 2, which predict similar element deformation demands, are suited for this correlation. The mean values of demands due to the nine site ground motions, presented in the middle row of Figure 13, are utilized for this purpose. The primary observation is that the joint panel zones exhibit large deformation demands, but the beams do not; in fact, most of the interior beams have not reached their strength capacity. Since the relative importance of panel zone versus beam plastic deformations is not much dependent on the input ground motion, it must be concluded that large joint shear deformations are a potential cause of weld fracture. This conclusion is not new, since it has been pointed out and observed in experiments already in 1971 (Reference 8). Large panel zone shear deformations will lead to kinks (regions of large curvature) at the corners of the joint. These are the points where beam flanges are welded to the column flanges. The large curvatures will lead to large strains around the welds, which are possible sources of fracture. Accepting this hypothesis, there appears to be a good correlation between predicted demands and observed fractures. At most locations where either the panel zone plastic shear distortions or the beam plastic hinge rotations are large (in the order of O.Ol), the connections have fractured. At the roof level, the plastic deformations are relatively small and no fractures were observed. For the damaged 4-story building the analysis focused on the MRF on Line D, which had many fractured connections. The MRF on Line A, in which no fractures were observed, was not modeled except for the three-dimensional ETABS analysis performed as part of the elastic benchmark analysis. No specific conclusions can be drawn on the behavior of this frame, but general observations can be deduced from the ETABS analysis. The time history analyses did show that the elastic interstory drifts of this frame are about 25% smaller than those for the MRF on Line D. If one argues that the difference in displacement response is likely to increase as the MRF on Line D yields and fractures, then it is plausible that the inelastic demands imposed by the Northridge earthquake on the MRF on Line A have been significantly smaller than those on the MRF on Line D. This argument helps to explain the lack of connection fractures in the frame on Line A.

24
0.7 0.6

H. KRAWINKLER A N D A . AL-ALI

0.5
0.4

0.3
0.2 0.1 0 0

0.005

0.01

0.015

0.02

0.025

0.03

SIH

,002

,001

Figure 16. Mean plastic deformation demands from dynamic analyses (E-W direction, nine records, Inelastic Model 1)

It is easy to rationalize that in the E-W direction of the 4-story building only one connection fracture was observed. Firstly, the severity of the ground motions was much smaller; secondly, the structure is stronger and stiffer in the E-W direction. This is evident from Figure 15, which shows pushover base shear versus roof displacement diagrams for the four inelastic models of the undamaged structure. Mean values of plastic deformation demands for Model 1 are presented in Figure 16. Inelastic demands are limited to panel zone shear distortions and are much smaller than the corresponding demands in the N-S direction (Figure 13, middle row, left graph). Unfortunately, the good correlation between predicted demands and observed fractures obtained for the 4-story building is offset by poor correlation obtained for the adjacent 2-story building. In this building no fractures were observed, but the predicted plastic deformation demands are of the same magnitude as those in the N-S MRF on Line D. For reasons unknown, the quality of the welded connections in the 2-story building must have been superior to that in the 4-story building. It must be concluded that the correlation between predicted demands and observed fractures obtained from inelastic analyses is adequate but by no means reliable. A reliable prediction is impossible because of the large uncertainties in deformation capacities.

STEEL FRAME SEISMIC DEMAND EVALUATION

25

A much less satisfactory correlation is achieved between elastic D/C ratios and observed fractures (see top graph in Figure 13). The D/C ratios give only a vague hint of the relative plastic deformation demands in beams and joint panel zones, since they disregard a redistribution of internal forces. They even predict plastic rotation demands in columns, whereas an inelastic analysis shows that the stress level in columns is low everywhere except at the column bases.

8. CONCLUSIONS

The following conclusions are drawn from this case study. It must be emphasized that many of these conclusions are preliminary, are based on insufficient evidence and are in need of further study. Many of the conclusions are case specific, i.e., they apply to the case study structures and can be generalized only with great caution. The accuracy of some of the assumptions made in the analysis is debatable. But it is conjectured that great accuracy in demand predictions is not necessary in view of the great uncertainties inherent in the prediction of capacities of welded connections.
(1) Global demand parameters that provide useful information on the likelihood of connection fractures include the roof displacement and interstory drifts together with the laici ai--irrdcl-dispIacmirmi--c w p o n 3 e - o ~ a n -fmiii--a-yfiiro-iw ~d -anaigsis: -fir p aiicdq-drt; --interstory drift is a good measure of the severity of demands.

(2) Element demand parameters that provide useful information on the likelihood of connection fractures include elastic D/C ratios (with limitations), plastic rotation demands in beams and columns, and plastic shear distortions in joint panel zones.
(3) Elastic D/C ratios are meaningful indicators in some cases and misleading ones in others. In the case study structures they provided misleading information on the relative demands imposed on beams, columns and joint panel zones. In particular, they provide misleading information on plastic rotation demands in beams in cases in which the joint panel zones are weak in shear. Moreover, they provide wrong indications of plastic deformation demands in cases in which important structure periods coincide with periods at which the elastic response spectrum has a large peak or trough.

(4) Modeling of the shear strength and stiffness properties of joint panel zones is critical for realistic demand predictions in steel moment resisting frames. In modern code designs, the joints are often the weak elements that yield first, because the requirements for joint shear strength have been greatly relaxed in the 1988 UBC.

( 5 ) Excessive shear yielding of joint panel zones may be a source of connection fractures. It causes large curvatures in the column flanges at the corners of joints, which in turn causes large strains around beam-to-column flange welds. In the 4-story case study building, the observed fractures can only be explained by this phenomenon since the stress level in several of the beams with fractured connections is low.
( 6 ) The models employed to predict the behavior of the 4-story N-S structure with fractured connections do not lead to catastrophic behavior under the simulated site ground motions. There is little comfort to be derived from this observation since in the model structure the P-delta effect is small and the fractured connections are modeled with non-deteriorating bilinear springs. Furthermore, the response of a fractured structure is sensitive to the frequency content of the ground motion because of the loss in stiffness and the resulting period shift.

26

H. KRAWINKLER A N D A. AL-ALI

(7) For the case study buildings the predictions of plastic rotation and distortion demands are not very sensitive to refinements made in modeling the structure. Incorporating the contributions of the floor slab to the strength and stiffness of the moment resisting frames, as well as incorporating the contributions of strength and stiffness of the simple frames in the computer analysis, made some but not much difference in the predictions for the undamaged structures. However, the contributions of simple frames gained much on importance in the response predictions for the structure with fractured connections.

(8) A clear distinction needs to be made between demand estimates obtained from a global structural analysis and demands that control the behavior of a welded connection. The latter demands are affected by many local issues, including weld detailing and quality, residual stresses and also the effect of the floor slab on the state of stress and strain around the weld. (9) The correlation between predicted demands and observed connection fractures is adequate but by no means perfect. The low stress level predicted in the columns (by the inelastic analysis) justifies the absence of fractures across the column, which were observed in many other buildings. The observed fractures are at locations at which the predicted plastic deformation demands in either the beam or the joint panel zone are high. The difference in observed damage between the N-S and E-W frames of the 4-story building (many fractures in the more highly stressed N-S frame and only one fracture in the E-W frames) is in line with large differences predicted in deformation demands. However., significant plastic deformation demands are also predicted for the 2-story building in which no fractures were observed. (10) An accurate prediction of the occurrence and location of fractured connections by means of a global elastic or inelastic analysis is an unrealistic expectation. The variations in the frequency content of input ground motions are large, the determination of the structural periods is far from perfect and, most importantly, the scatter in deformation capacities of welded connections is very large and is generally affected by local detailing. Analytical predictions should serve as indicators that assist in identifying the existence of potential problems and in developing an inspection strategy once a potential problem has been identifed. The last conclusion points towards critical issues for further study. Both the seismic demands and the element capacities (connection strength and deformation capacities and yield strength of material) are associated with large uncertainties. Thus, a realistic safety evaluation needs to be reliability based, with due consideration given to the uncertainties inherent in demands and capacities and to the consequences of connection fractures on structural safety. This necessitates experimental work to assess the uncertainties in capacities and to develop better models for the hysteretic characteristics of fractured connections. It also necessitates analytical work to develop computer programs that permit realistic and efficient representations of post-fracture behavior, and it necessitates the development of a systematic approach to the safety evaluation of buildings with steel frame structural systems. There is also a great need to develop procedures that permit a correlation bet ween predicted global element deformation demands (e.g. plastic hinge rotations and joint shear distortions) and localized behavior that is controlled by detailing, workmanship and slab contributions. ACKNOWLEDGEMENTS Funding for this case study was provided by the Federal Emergency Management Agency through the SAC Joint Venture. SAC is a partnership of the Structural Engineers Association

STEEL FRAME SEISMIC DEMAND EVALUATION

27

of California, the Applied Technology Council and California Universities for Research in Earthquake Engineering. The willingness of the owner to make the two buildings available for this case study is gratefully acknowledged. The case study was performed with the collaboration of C. C. Thiel and J. M. Dunlea. Much appreciation goes to them and to a great group of graduate students at Stanfords John A. Blume Earthquake Engineering Center, who have spent many nights building computer models, performing dynamic analyses, writing interfaces for data reduction and presentation and preparing the tables and graphs contained in this paper. These students are Ali A1-Ali, Oliver Collignon, Chris Delp, Akshay Gupta, Abhijit Kakhandiki, Pasan Seneviratna and Tom Perry.
REFERENCES

1. Uniform Building Code, International Conference of Building Officials, Whittier, California, U.S.A., 1988. 2. B. Kato, H. Aoki and Y. Tagawa, Seismic behavior of steel frames with composite girders, Proc. 8rh World Conf. on Earthquake Engineering, Vol. VI, San Francisco, U.S.A. 1984. 3. G. Deierlein and R. Leon, Notes on modeling the moment-rotation behavior of shear tab connections to composite beams, Communication to SAC, 1995. 4. H. Krawinkler, Shear in beam-column joints in seismic design of steel frames, Engineering J., American Institute of Steel Construction, Third Quarter, 1978. 5. Statistical analysis of tensile data for wide flange structural shape, Steel Shape Producers Council, 1994. 6. P. Somerville, Task 4. Characterize ground motions at the sites of the subject buildings, Report submitted to SAC, March 1995. 7. H. Krawinkler, A. AI-Ah, C. C. Thiel and J. M. Dunlea, Task 3.1 Case Study #4: Analysis of a damaged 4-story building and an undamaged 2-story building, Report submitted to SAC, May 1995. 8. H. Krawinkler, V. V. Bertero and E. P. Popov, Inelastic behavior of steel beam-to-column subassemblages, Report No. EERC 71-7, Earthquake Engineering Research Center, University of California, Berkeley, California, U.S.A., 1971.

Das könnte Ihnen auch gefallen