Taiwan - Corner Column Failure
Taiwan - Corner Column Failure
Taiwan - Corner Column Failure
2 STORY RC BUILDING
4 Biography: ACI member Aishwarya Puranam is a Graduate Student in the School of Civil
5 Engineering at Purdue University, West Lafayette, IN. She received her BS and MS in Civil
7 ACI Fellow Santiago Pujol is a Professor in the School of Civil Engineering at Purdue
10 Reinforcement; and Joint ACI-ASCE Committee 445, Shear and Torsion. He received his
11 BSCE from National University of Colombia, Bogota, Colombia, in 1996, and his MSCE and
13 ABSTRACT
14 This paper discusses plausible causes for the failure of a corner column in a 14-story
16 Earthquake of February 6, 2017. It is concluded that the failure occurred because of the
17 effect of discontinuous partition walls. Nevertheless, the analyses presented suggest that, in
18 strong ground shaking, the column would have been subjected to excessive axial compression
21 INTRODUCTION
23 buildings was surveyed in Tainan City following the 2016 Taiwan (Meinong) Earthquake. A
1
1 corner column failed at the basement level in one of the buildings in the complex. A study
2 was conducted to investigate the probable cause of failure and the extent to which
3 conventional design approaches and past field observations can be used to avoid this type of
4 failure.
5 The studied building was one of four 14-story towers that shared a common basement used
6 for parking. It had a rectangular floor-plan and four reinforced concrete (RC) moment frames
7 in each direction. Its long direction coincided roughly with the NS direction. A detailed
8 photographic record of the state of the structures after the earthquake and photographs of
10 The failure in the corner column appeared to have features common in axial compression
11 failures. Other elements in the basement (columns, beams, partitions) were nearly free of any
12 damage, including beams framing into this column. These observations and the stiffening
13 expected from the peripheral retaining walls described suggest that the failure was unlikely to
14 have been caused by lateral displacement and/or shear force. Three approaches were used to
15 investigate the demands in the column that failed: 1) Conventional Design Approach, 2)
16 Limit Analysis and 3) Unconventional Treatment of the Problem as that of a Stiff Cantilever
17 on a Soft Base. Results obtained using the three approaches are discussed.
18 RESEARCH SIGNIFICANCE
19 Brittle failure in columns during an earthquake could lead to catastrophic outcomes. Field
20 evidence has shown that columns at the base of a building can be vulnerable if they support
21 discontinuous walls. Results of this study indicate that even in the absence of said
24 failure.
2
1 ANALYTICAL INVESTIGATION AND DISCUSSION
3 Within 3.5 km from the subject building PGA (Peak Ground Acceleration) reached
4 approximately 0.2 g and PGV (Peak Ground Velocity) reached 60 cm/s (24 in./s). Figures 1
5 and 2 show response spectra (for a damping ratio of 2%) for the records from which these
6 peaks were extracted. In the nearly constant acceleration region, mean spectral acceleration
7 was 0.4 g with the mean plus the standard deviation of the same quantity being close to 0.5 or
8 0.6 g. The period at the transition between the ranges of nearly constant acceleration and
9 velocity (Figure 1) was approximately 1.1 s, indicating the presence of soft soils. Records
10 were provided by CWB (Central Weather Bureau, Taiwan) and are available here:
11 https://datacenterhub.org/resources/14098.
13 The studied building was one of four 14-story towers that shared and common basement used
14 for parking. The perimeter of this basement extended well outside the perimeters of the
15 towers which were labeled A, B, C, D (Figure 3). Towers A and B had similar configurations.
16 This paper focuses on Tower B. Tower B had four reinforced concrete (RC) moment frames
17 in each direction. Its long direction coincided roughly with the NS direction. A typical floor
19 The Exterior columns in lower stories were 60 cm x 120 cm in cross section (24 in. x 48 in.)
20 and had high longitudinal reinforcement ratios (2.7% - from 24 #10 bars with a nominal
21 diameter of 1.27 in = 32 mm) – Figure 5. Interior columns had 32 in. x 44 in. sections (80 cm
22 x 110 cm) and their reinforcement ratio was 2.8%. The transverse reinforcement ratio was
3
1 in.) in cross section. Longitudinal reinforcement ratios ranged from 1.2 % (for bottom
3 reinforcement ratio was approximately 0.6 %. Slab thickness was ~ 15 cm (6in.). Top slab
4 reinforcement ratio near columns was approximately 0.4%. Specified concrete strength was ~
5 28 MPa (4000 psi). Specified steel yield stress was 415 MPa (60 ksi).
6 Partitions were made of reinforced concrete. Their thickness was 12 cm (4.7 in.). The
7 longitudinal reinforcement ratio of RC partitions was 0.4%. Partitions were present in all
8 stories. Partition wall density (relative to unit floor area and considering walls in both floor-
9 plan directions) varied from 3 to 5% in stories above ground. In the basement, where
10 partitions were present around the elevator only, partition wall density was 0.3% (Figure 6).
11 The perimeter of the basement was lined by 24 cm (9.5 in.) thick retaining walls. These walls
12 were cast on the inside (i.e. towards the center of the floor plan) of a series of concrete and/or
13 steel piles cast and/or driven to allow basement excavation. Near grade level, a 20 to 24-cm
14 (8-10-in.) thick slab acted as a diaphragm connecting towers A, B, C, and D to these retaining
15 walls.
16 Buildings in Tainan tend to be relatively heavy. The estimated unit weight (relative to unit
17 floor area) was ~ 14.5 kN/m2 (300 psf). The slab near grade supported additional weight
18 from landscaping materials (soil, stone, and pavement) and vegetation. This additional
21 Building height (H) was 44.8 m (147 ft.). Initial period (To) was estimated to be 1.5 sec. a)
22 for gross un-cracked sections, b) for a modulus of elasticity of 25GPa (3600 ksi), c)
23 considering the slab segments within a 45-deg. projection of beam webs to act as beam
4
1 For the estimated period To=1.5s , and for a value of PGV of 60 cm/s, the expression
2 proposed by Sozen (2003) to estimate peak roof drift ratio for nonlinear systems, as modified
5 Inter-story drift ratio is likely to be 1.5 times this value (i.e. 3%).
7 A detailed photographic record of the state of the structures after the earthquake and
9 https://datacenterhub.org/resources/14098
10 Towers A and B had hairline inclined cracks in partition walls, especially in the first story.
11 But this damage is not commensurate with the drift estimated ignoring the stiffening effect
12 from partitions. According to Algan (1982), at inter-story drift ratios exceeding 1%, partitions
13 have so much damage that they require total replacement (instead of repair).
14 The axial demand caused by gravity in column D4 in the first story was approximately 5340
15 kN (1200 kip). The increase of tributary area at grade level, and the weight of landscaping
16 materials may have caused the demand in the basement to increase to 7560 kN (1700 kip)
17 (close to the axial load at the balanced point). Yet, the nominal axial capacity (estimated for
18 specified material strengths) for concentric loads was ~ 24,500 kN (5500 kip). Three
19 approaches were followed to try to account for plausible increases in axial demand that may
22 Design approaches seldom serve well as tools to understand failure. Nevertheless, they can
5
1 “R factor.” For the case at hand, and given the spectra in Figure 1, it seems reasonable to
2 assume Sa = 0.4g even though the spectra in Figure 1 were produced for a damping ratio of
3 2% and not 5% as it is routinely done in design. For R=8, the resulting seismic coefficient is
4 0.05 (i.e. the design base shear would be 5% of the building weight).
5 In design, base shear is “distributed” along building height in linear proportion to the product
6 of story weight and floor elevation (relative to grade). In late versions of U.S. codes,
7 elevation has been modified by an exponent k related to period. In ASCE-7 (2010), for a
8 period of 1.5 s, k=1.5, resulting in the “equivalent” lateral force distribution shown in Figure
9 8.
10 An analysis of a linear model of the building structure was carried out with 100% of the
11 forces in Figure 8 applied in one direction and 30% of them applied in the other direction.
12 The increase in axial load caused by earthquake demands estimated for column D4 with this
13 approach was nearly 2670 kN (600 kip), for a total axial demand of 7560 kN + 2670 kN =
14 10,230 kN (1700 kip + 600 kip = 2300 kip). This number is still much smaller than the
15 nominal axial strength of 24,500 kN (5500 kip). It is difficult to reconcile the two numbers,
16 therefore, a different perspective and the effects of bending are discussed next.
18 If beams yield at each end under earthquake demands, the vertical reaction at one end
19 increases in direct proportion to the sum of negative and positive moment capacities, and
20 inverse proportion to beam span. This vertical reaction is shear force (adding to gravity
21 shear) in the beam. In interior columns, additional shear forces from opposite beams tend to
22 counteract one another. But in an exterior column, the described increases in beam shear
23 result in net increases in axial load. At a corner column, increases in axial load from
24 perpendicular frames are additive when the building displaces at the same time in the
6
1 directions of both frames and towards the corner column (Moehle, 2015).
2 From Figure 1, it is clear there was enough demand to mobilize the strength of the building
4 For the beams framing into corner column D4 above ground, mean probable negative
5 moment capacity (estimated for a yield stress of 520 MPa or 75 ksi) is approximately 2710
6 kN-m (2000 kip-ft.) and mean probable positive moment capacity is approximately 1350 kN-
7 m (1000 kip-ft). Spans were 7.7 m (25 ft.) in one direction and 11.6 m (38 ft.) in the other.
8 Assuming beams at all levels yielded (an extreme but plausible scenario), the corresponding
9 increase in axial load in column D4 is nearly 12,000 kN (2700 kip), for a total axial demand
10 of 7560 kN + 12,000 kN = 19560 kN (1700 kip + 2700 kip = 4400 kip). Although still
11 smaller, this value is much closer to the nominal capacity for concentric demand (24,500 kN
13 The stark difference in estimates from Approaches 1 and 2 i.e. 10,230 kN v. 19560 kN (2300
14 kip v. 4400 kip) should call the attention of designers and code writers. We design for
15 reduced lateral forces because we realize these forces are not critical as long as the structure
17 “plastic hinges” in beams (preferably). If we design beams to yield, we do not achieve the
18 desired ductile mechanism if the columns supporting them cannot resist the reactions
19 associated with said yielding, especially if we consider that the failure of a column under
20 heavy axial load is one of the least ductile failures there are.
21 Next, we examine to what extent bending moment may have caused enough additional
22 demand to trigger failure. If beams in frames in perpendicular directions had yielded, the
23 first-story corner column could have also yielded under bending about an axis making a finite
24 angle with the principal cross-sectional axes. As an example, imagine the first story column
7
1 bending about an axis at 45 deg. relative to the principal axes (Figure 9). The probable
2 moment capacity of the first-story column under the estimated axial load of 5340 kN or 1200
3 kip (from gravity) plus 12,000 kN or 2700 kip (from the mechanism described above) is 2300
4 kN-m or 1700 kip-ft. (for a yield stress of 75 ksi), with equal NS and EW components of
5 2300/√2 kN-m ( (1700/√2 kip-ft.) each. If one distributes these moment components to the
6 beams at grade level and the basement column in direct proportion to moment of inertia and
7 inverse proportion to span, one obtains 310 kN-m (230 kip-ft.) and 1250 kN-m (920 kip-ft.)
8 as the EW and NS components of the bending moment demand at the top of the basement
9 column. These moment components are represented by the solid black dot in Figure 10. In
10 this figure, the X axis represents moment causing bending about the (weak) EW axis, and the
11 Y axis represents moment causing bending about the (strong) NS axis. Open circles represent
12 demands obtained assuming different angles for the axis of bending in the first-story column.
13 The bilinear “curve” in the same plot represents the nominal capacity of the basement column
14 to resist biaxial moment obtained joining capacities related to bending about the NS and EW
15 axes with the capacity related to bending at 45 deg. Capacities were calculated for an axial
16 force of 19,600 kN (4400 kip) and a yield stress of 415 MPa (60 ksi).
17 The markers representing biaxial bending and axial demands in Figure 10 are also included in
18 moment-axial load interaction diagrams in Figures 11 and 12. Figures 10-12 demonstrate that
19 although bending would have approximated demand to capacity further, the increase in
20 demand caused by bending was not dramatic. The driving factor was the axial load. One can
21 almost argue that plausible variations in material properties and strength estimates would
22 suffice to justify the failure on the basis of Figures 10-12. Nevertheless, as discussed above,
23 had the structure responded as a space frame, the drift would have been much higher than it
24 was apparent from the condition of the partitions. And the bending demands estimated were
8
1 obtained assuming the first-story column reached its flexural capacity. But the first story
2 column would have been working above the balanced point also and, contrary to the
3 basement column, it is likely to have experienced large shear and lateral deformation. And
4 yet, the first-story column had not clear signs of distress. So perhaps a different issue was at
5 play.
7 Soft Base
9 building base: The Caromay Building (19 stories), The Sucre Building (14 stories), The Mene
10 Grande Building (16 stories), and the Altamira Building (14 stories). Reconnaissance reports
11 (Portland Cement association, 1967, Sozen et al., 1968, Hanson and Degenkolb 1969) point
12 to partition walls as the cause. In Caromay, Sucre, and Altamira, the failures took place in the
13 first story or a basement. In all three cases, these stories (or basement) had a reduced amount
14 of partitions (Figure 13). In all three cases, the reported observations suggest that the
15 buildings acted as stiff cantilevers supported by a softer base. In the case of Caromay,
16 Hanson and Degenkolb (1969) suggested that column axial compression failures (Figure 14)
17 occurred because of overturning forces causing bending of the entire building (acting as a
18 large cantilever) about one of the principal axes of the floor plan.
19 The Imperial County Services building (Kreger and Sozen, 1989) also had a discontinuous
20 wall supported by a series of (offset) columns. Kreger studied the demands on these columns
21 assuming their unit axial strains varied in linear fashion relative to distance to a neutral axis.
22 The ideas by Hanson, Degenkolb, Kreger, and Sozen were combined here to study the subject
23 building. The building was idealized as a 14-story stiff cantilever sitting on a series of
24 vertical springs. Each vertical spring represented a basement column. Lateral movement of
9
1 spring ends was restrained by pins representing the foundation and the diaphragm connecting
2 Tower B to retaining walls at grade level. The axial strain in each spring was assumed
3 proportional to distance to a neutral axis parallel to the bending axis in Figure 15. The
4 bending axis in Figure 15 was chosen to be perpendicular to the SW-NE diagonal of the floor
5 plan because this orientation puts column D4 in a vulnerable condition. Other axis
6 orientations yield similar results. Spring or column force was obtained from the stresses
7 associated with the mentioned strains. These stresses were obtained using the stress-strain
9 concrete were ignored. Bar stresses were assumed proportional to strain up to yield stress.
10 The location of the neutral axis was chosen so that the sum of all column forces was equal
11 and opposite to total building weight. The summation of moments about the centroid of the
12 floor plan provided an estimate of the overturning moment capacity of the building. It was
13 observed that nearly at the same time as column yielding (in tension) was reached at the
14 South-West corner of the building, a compressive strain of 0.003 was reached at the North-
15 East corner of the building. This observation suggests that the building (as idealized) had
17 Dividing the estimated overturning capacity by 2/3 of building height provides an estimate of
18 base shear strength. This value was approximately 50% of building weight. Form the Spectra
19 in Figure 1, and assuming period of the building was well within the region of nearly constant
20 acceleration, it seems plausible that what occurred in Tower B was nothing but a repeat of
21 what occurred in Caracas and El Centro decades before. Discontinuous walls created axial
22 demands in supporting columns well beyond their strengths. Why did this mistake have to be
23 repeated in the 21st century? The profession ought to make an effort to learn and remember
24 more from its own experience. In the age of artificial intelligence, we have the means to prop
10
1 up our memories to avoid forgetting again. The website mentioned above
3 engineers preserve and retrieve more quickly what they learn in laboratories and field.
4 Other Considerations
7 Hairline cracks were observed in a number of columns. To reach the compressive capacity of
10 The foundation consisted of a 0.7 m (27.5 in.) RC mat stiffened by 1.2 m (47 in.) RC deep
11 beams (protruding 0.5 m or 20 in. beyond the mat). A 4 m x 4 m (13 ft. x 13 ft.) grid of steel
12 beams was cast in this mat. These beams were connected to 20-m long steel “W-shape”
14 Uplift on one side of the subject tower would have left signs of distress on the floor of the
15 parking area. No such signs were observed. Nevertheless, we estimated the base shear
16 required to uplift the foundation assuming each tower had its own mat isolated from other
17 mats. We found the associated spectral acceleration to be much more than the value obtained
19 B. We also considered whether vertical acceleration could have been one the main drivers
20 behind the observed column failure. Figure 16, nevertheless, shows that this is unlikely to be
23 -Peak vertical accelerations occurred early in the ground motion, before large overturning
11
1 C. How likely was the building to lean towards the column that failed (D4) as assumed in
2 Figure 15?
3 To address this question, we estimated the initial period of the building idealizing it a stiff
4 cantilever pivoting about the elevator core, with columns acting as parallel linear springs
5 producing resistance against rotation in the basement. The estimate ranged from 0.2 s (NS
6 direction) to 0.3 s (EW direction) for un-cracked sections. Figure 17 shows a trace of
8 estimated for a linear SDOF with a period of 0.25 s. The trace was obtained for the ground
9 acceleration record obtained closest to the building. The trace is superimposed on the
10 building floor plan from Figure 4. Figure 17 shows that the majority of the displacement
11 cycles for the SDOF studied occurred in directions approaching the SW-NE diagonal of the
12 building implying that the condition depicted in Figure 15 is plausible. The amplitude of the
13 displacements estimated would require strains in corner columns approaching 0.003 in the
14 basement.
15 An interesting question remains. Why did Tower A not have a similar failure to the one in
16 Tower B? We do not have a good answer to this question. The partition walls in Tower A
17 were nearly a mirror image of those in Tower B. For this reason, the column in the position of
18 column D4 in Tower A did not have partition walls directly attached to it in the first story.
19 Perhaps this difference was sufficient to lessen the demands in the NE corner column in
20 Tower A. Or perhaps the ground motion intensity was a bit smaller for Tower A and, given
21 the discussed brittleness of the system, this small difference in demand led to a large
23 CONCLUSIONS
24 Exterior columns in moment frames designed for reduced earthquake demands can be
12
1 vulnerable to axial compression failure. Limit analysis can be used to obtain, with little
3 The vulnerability of columns supporting discontinuous walls (made with tiles or concrete)
4 has been known for decades. Nevertheless, it seems that our profession is too distracted by
5 other issues to pay enough attention to simple observations from previous earthquakes.
6 Observations can be as or more useful than the most sophisticated mathematical algorithm or
7 method of analysis.
8 ACKNOWLEDGMENTS
9 Discussions with Dr. Mete Sozen and Dr. Luis Fargier were critical to shape our thinking.
10 The EERI student chapter at Purdue University acted as a reliable sounding board to help us
12 REFERENCES
15 2. ASCE (2010) “ASCE/SEI 7 Minimum Design Loads For Buildings and Other Structures,” American
17 3. Hanson, R., and Degenkolb, H. J., (1969) “The Venezuela Earthquake July 29, 1967,” American Iron
19 4. Hognestad, E. (1951) “A Study of Combined Bending and Axial Load in Reinforced Concrete
21 5. Kreger, M. and Sozen, M. (1989) “Seismic Response of Imperial Country Services Building in 1979”,
24 7. Portland Cement Association (1967) “Preliminary Report The Behavior of Reinforced Concrete
25 Structures in the Caracas, Venezuela Earthquake of July 29, 1967,” Portland Cement Association,
26 Skokie, Illinois.
13
1 8. Sozen, M.A., Jennings, P.C., Matthiesen, R.B., Housner, G.W., Newmark, N.M., (1968) “Engineering
2 Report on The Caracas Earthquake of 29 July 1967,” National Academy of Science, Washington DC
14
1 APPENDIX A: APPROACH 3-STIFF CANTILEVER ON A SOFT
2 BASE
3 The building was idealized as a 14-story stiff cantilever sitting on a series of vertical springs.
4 Each vertical spring represented a basement column. Lateral movement of spring ends was
5 restrained by pins representing the foundation and the diaphragm connecting Tower B to
12
13
15
1 Material Properties
7
8
16
1 Neutral Axis Depth
2 Location of the neutral axis was chosen so that the sum of all column forces was equal and
5 Column Forces
10 Note:
12 Stress in concrete at a given strain was obtained using the expression proposed by Hognestad
13 (1951)
17
1 Equilibrium
3 Overturning Capacity
7
18
1 FIGURES
2 List of Figures:
3 Fig. 1 – Spectral accelerations obtained for a damping ratio of 2% and for records within
5 Fig. 2 – Spectral displacements obtained for a damping ratio of 2% and for records within
9 Fig. 5–Column (D4) cross section. Transverse reinforcement was #4 stirrup and cross-ties
10 spaced at 10 cm on center.
12 Fig. 7–Column D4
13 Fig. 8–Equivalent Lateral Forces Used to Emulate Conventional Analysis for Design
15 Fig. 10–Probable moment demands and nominal capacities, basement column D4.
16 Fig. 11–Interaction Diagram for Basement Column D4, Strong Axis Bending.
17 Fig. 12–Interaction Diagram for Basement Column D4, Weak Axis Bending.
18 Fig. 13–Locations of Columns and Partitions in Floor Plans of Caromay Building (Hanson
24
19
1.0
0.9
0.8 Mean + Standard Deviation
0.7
Spectral Acceleration (g)
0.6
Mean
0.5
0.4
0.3
0.2
0.1
0.0
0 0.5 1 1.5 2 2.5 3
Period (s)
1
2
3 Fig. 1–Spectral accelerations obtained for a damping ratio of 2% and for records within 3.5
4 km from building.
6 Fig. 2–Spectral displacements obtained for a damping ratio of 2% and for records within 3.5
7 km from building
20
N
D
B A
21
N
10
11
22
1
2 Fig. 7–Column D4
4 Fig. 8–Equivalent Lateral Forces Used to Emulate Conventional Analysis for Design
23
TENSION
Bending axis
COMPRESSION N
1
30⁰
Probable Moment
45⁰
Nominal Capacity for P=4400 kip
60⁰
4 Fig. 10–Probable moment demands and nominal capacities, basement column D4. Angles
5 marking symbols representing demands refer to the angle between the axis of bending in the
24
60, 45,30°
2 Fig. 11–Interaction Diagram for Basement Column D4, Strong Axis Bending.
30,45,60°
4 Fig. 12–Interaction Diagram for Basement Column D4, Weak Axis Bending.
25
1
5 Fig. 13–Locations of Columns and Partitions in Floor Plans of Caromay Building (Hanson
26
1
27
TENSION
Bending axis
COMPRESSION
54 deg
1
28
1
-1
29