Engineering Structures 69 (2014) 90–101

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Experimental investigation of web-continuous diagrid nodes

under cyclic load
In Yong Jung a, Young Ju Kim b, Young K. Ju a,⇑, Sang Dae Kim a, Sung Jig Kim c
a
School of Civil, Environmental and Architectural Engineering, Korea University, Seoul, Republic of Korea
b
Seismic isolation and vibration control research center, DRB Holding Co., Ltd., Seoul, Republic of Korea
c
Department of Architectural Engineering, Keimyung University, Daegu, Republic of Korea

a r t i c l e i n f o a b s t r a c t

Article history: Applications of the diagrid structural system have recently increased, and substantial analytical studies
Received 20 December 2012 have been conducted. However, the structural behavior under lateral loadings, such as wind and earth-
Revised 25 February 2014 quake loads, cannot be fully understood through only analytical approaches, due to the complexity of dia-
Accepted 26 February 2014
grid nodes, and challenges of modeling the welding properties. Therefore, in this study, four web-
Available online 5 April 2014
continuous diagrid node specimens were tested under cyclic loads. Welding methods and design details
were selected as testing parameters. The effects of the welding methods and design details on the initial
Keywords:
stiffness and yielding stress were found to be not significant. However, the failure mode and energy dis-
Diagrid
Node
sipation of the nodes were significantly affected by the welding methods, and the design details.
Seismic test Ó 2014 Elsevier Ltd. All rights reserved.
Structural performance
Welding

1. Introduction investigations to apply the diagrid structural system to high-rise

Developing structural technology makes it possible to build
high-rise buildings, such as the Burj Khalifa, or Taipei 101. The
technology of constructing high-rise building has concentrated
on height, but recently, the aesthetic aspect of high-rise buildings
is being considered as well. Researchers such as Beghini et al. [1]
tried to combine aesthetics and functionality. The author gave sev-
eral tall building examples to prove the possibilities of combina-
tion between design and engineering. Also, the main topic of the
CTBUH World Conference held in 2006 was ‘‘Thinking outside: Ta-
pered, Tiled, Twisted Towers’’, which shows some of the current
trends of high-rise buildings.
The diagrid structural system, which consists of triangular mod-
ules, is considered as one of the most appropriate structural sys-
tems for this trend. 30 St Mary Axe (London), Hearst Tower (New
York), Capital Gate (Abu Dhabi), and Tornado Tower (Qatar) are
well-known buildings that the diagrid structural system has been
applied to.
The diagrid structural system resists vertical and horizontal
forces with triangular modules of beams and braces, without col-
umns. This structural system is very effective in resisting horizon-
tal forces, such as wind or earthquake. Therefore, active research
⇑ Corresponding author. Tel.: +82 2 3290 3327; fax: +82 2 928 7656.
E-mail address: tallsite@korea.ac.kr (Y.K. Ju).
buildings are in progress.
Moon [2] proved that the diagrid system has almost the same
performance as the brace tube system, which is well-known for
its high lateral strength. Moon et al. [3] also suggested optimal
angles of the diagrid system, by measuring the lateral displace-
ment of diagrid structures of various heights and angles. Leonard
[4] investigated the shear lag of the diagrid system, and showed
that the diagrid system performed better than a framed tube sys-
tem. Kim and Lee [5] revealed progressive collapse characteristics
of the diagrid system. Ko et al. [6] evaluated the seismic perfor-
mance of irregular diagrid buildings by an analytical approach,
and suggested a higher value of seismic performance factor than
currently in use. Kim et al. [7] estimated response modification fac-
tors of various heights of the diagrid system. Bae et al. [8] con-
ducted experimental study of the diagrid frame, and also
suggested a higher factor than the current value.
Several researchers tried various kind of diagrid system. Chao
et al. [9] tested diagrid nodes that are made of concrete filled steel
tube and suggested bearing capacity equation based on the Chinese
design codes for Concrete Filled Steel Tubular (CFST) columns. Kim
et al. [10] developed beam-column connection of H-shape column
and Ju et al. [11] assessed performance of Buckling Restrained Brace
(BRB). Based on these two studies, Lee et al. [12] carried out experi-
mental study of diagrid system that is composed of BRBs. As a re-
sults, the author showed that the BRB increased ductility of diagrid
system.

http://dx.doi.org/10.1016/j.engstruct.2014.02.026
0141-0296/Ó 2014 Elsevier Ltd. All rights reserved.
 I.Y. Jung et al. / Engineering Structures 69 (2014) 90–101
Fig. 1. Two types of diagrid.
However, structural behaviors or welding properties of real dia-
grid nodes under lateral forces cannot be completely evaluated by
analytical studies only. Therefore, a series of studies were con-
ducted by the authors. The authors [13,14] conducted experimen-
tal studies of a diagrid system subjected to lateral force. Seismic
performances of web-separated H-section and Box-section nodes
were investigated, in terms of design details and welding methods.
The web-separated H-section diagrid node is a modified version of
the web-continuous H-section diagrid node. Fig. 1 illustrates the
two kinds of H-section diagrid nodes.
In this paper, seismic performance of the web-continuous
H-section diagrid node is mainly discussed, with an emphasis on
initial stiffness, strength, welding method and energy dissipation.
Test results of the web-separated H-section diagrid node are
introduced, to compare behaviors between web-continuous and
web-separated diagrid nodes.
2. Experimental program
2.1. Specimens
The prototype of testing specimens was selected to be the struc-
tural system used in the Lotte Super Tower in Seoul, which has
been designed with a height of 555 m, using the diagrid system.
A node in the lower level of this tall building was selected for this
study.
Fig. 2 illustrates the details of a test specimen, which consists of
two diagonal brace members (H-Section) with an angle of 24°. The
test specimens are scaled down by a factor of five, due to the lim-
itation of actuator capacity (3000 kN). At the intersection of the
two braces, flanges are attached to each other, while webs are con-
tinuously connected through vertical stiffeners. The testing param-
Fig. 2. 3D view of specimen.
91
eters are welding methods, and distances between vertical
stiffeners. The details are shown in Fig. 3, and listed in Table 1.
For a conventional node, the full penetration (abbreviated as FP)
welding method is commonly used. The proposed node has many
welding areas, and the amount of the welding material will be
great, in the case of tall diagrid buildings. If the stress distribution
of the node is analyzed well, several parts of the FP welding areas
will be substituted for by the partial penetration (abbreviated as
PP) welding method. The amount of welding material can be
reduced significantly, if the structural performance with the PP
welding method is similar to that with the FP welding method.
The distances between vertical stiffeners are considered as
design criteria. A close distance between vertical stiffeners would
result in difficulty of manufacturing. Therefore, longer vertical
stiffeners, which lead to a wider distance between vertical
stiffeners, are introduced as a test parameter.
Fig. 4 depicts details of the four specimens. The MB-01
specimen is designed as a baseline specimen. The length of the
vertical stiffener is 107 mm, and is denoted as L1. The FP welding
method is used for the entire welding. The length of vertical
stiffeners in the MB-02 specimen is identical to that of the vertical
stiffener in the standard MB-01 specimen, while the PP welding
method is used at flange–flange, and web–flange, welding areas.
Since a high concentration of stress is expected at the center of
the node, the PP welding method is not used there.
The MB-03 specimen is designed to make the fabrication easy.
Because of wider distance (126 mm) between the vertical stiffen-
ers, the intersection point of the web and the vertical stiffener is
modified. This point is located beside the V-point, where the
flanges of the braces meet each other. Consequently, the length
of vertical stiffener become longer (238 mm), and is denoted as
L2. The full penetration welding method is used for the entire
welding. The MB-04 specimen has the same form as the MB-03
specimen having the L2 length of the vertical stiffener. The full
penetration welding method is used, except at the web–flange
welding area.
2.2. Loading protocol and measurement schedule
As shown in Fig. 5, two actuators were used for each brace
member, to apply cyclic loads. One was installed horizontally,
and the other was installed diagonally, with an angle of 24°. To
represent the structural behavior of the diagrid structural system
under lateral load, tensile force was applied to one brace, while
compressive force was applied to the other. Tensile force to upper
brace and compressive force to lower brace is the start of the load-
ing protocol. As shown in Fig. 6, cyclic displacements of ±d/dy, ±2d/
dy, ±4d/dy and ±6d/dy were applied twice per cycle, where yield dis-
placement, dy of the brace is about 2 mm, which is derived from the
section area of the brace. The test was stopped when the load
decreases below 80% of the maximum load, or the specimen fails
Fig. 7 shows the details of the measurement schedule. During
the test, two dial gauges were used to measure axial deformation
of the braces, and four LVDTs were used to measure horizontal
and vertical displacements at the center of the node. Uni-axial
and rosette strain gauges were used to measure stress flow and
concentration.
3. Experimental results and observations
3.1. Material properties
The material properties of the specimens were tested by the
Korean Standard Test Method. SM490 (Fy = 325 MPa) steels of
three different thicknesses (6 mm, 12 mm, and 16 mm thick) were
92 I.Y. Jung et al. / Engineering Structures 69 (2014) 90–101

Fig. 3. Test parameters.

Table 1 same. And loading is stopped at the same time. Target

List of parameters. displacements as the loading protocol were set to each member
Specimens Web–flange Length of vertical Flange–flange but only compressive target displacements are achieved. Therefore
welding stiffener (mm) welding the accumulation of these differences made the compressively
MB-01 FP 107 FP leaned load–displacement hysteresis loop.
MB-02 PP 107 PP Applying tensile force on the upper brace and compressive force
MB-03 FP 238 FP on the lower brace caused upper directional displacement of the
MB-04 PP 238 FP node center, as shown in Fig. 8, and this displacement affected
the failure mode of the specimens. Combining the displacement
at the center of node and the applied compressive force made
tested. As listed in Table 2, yield stresses of the coupon test
the P–D effect, and extra moment came up at the outside of the
exceeded 325 MPa, with elongations and yielding ratios in the
brace (Fig. 9). Due to the simultaneous effect of the moment and
range of 23–29% and 0.68–0.81, respectively.
the axial force, fracture occurred from the outside of the brace. Also
this fracture broke the joint between the web and vertical stiffener,
3.2. Overall behavior as shown in Fig. 10. This type of failure was also observed in the
author’s previous study [13] of the web-separated diagrid node.
Axial force–displacement relationships are shown in Fig. 12, Whereas the MB-01 and MB-02 specimens showed moment
where the left and right columns are for upper and lower braces, failure, as described above, the MB-03 and MB-04 specimens
respectively. The displacements of two braces had different history showed a different failure mode. In the case of the MB-03 and
during the test. The compressive side always had bigger displace- MB-04 specimens, joints between the web and vertical stiffener
ment because of vertical displacement of the node center. Whereas are located at the outer side. At this location, the effect of extra mo-
the tensile side had relatively small displacement, and the next ment is much smaller than the effect of axial force. Therefore, frac-
cycle was started. Compressive displacements of members are ture started from the welding point between the web and the
always bigger than tensile displacements. Two actuators applied vertical stiffener, by the axial force. From the center of the flange,
loads simultaneously but compressive displacements always where it makes contact with the web, the fracture spread to the
reached target values first. However, applied load is approximately outside, as shown in Fig. 11.

Fig. 4. Details of specimens and parameters.

 I.Y. Jung et al. / Engineering Structures 69 (2014) 90–101 93

Fig. 5. Test setup.

Table 2
Material properties.

Coupon Yield stress Tensile stress Yield Elongation

(MPa) (MPa) ratio (%)
Plate-16 357 528 0.68 29
Plate-12 378 542 0.70 26
Plate-6 421 518 0.81 23

LVDTs were installed at the upper and lower beam members,

Fig. 6. Loading protocol.
The strain of the V-point was not significant before the fracture
of the brace, but it increased rapidly after the fracture. In the case
of the MB-02 specimen, the brace of the compressive side was torn
off soon after the fracture, due to the PP welding of the V-point
(Fig. 13).
though the beams are not structural members in this study, to
measure out-of-plane displacement and twist. As shown in
Fig. 14, the maximum out-of-plane displacement was 4 mm. This
displacement is not significant compared to the total length of
the specimens of 1200 mm. This represents that there is no
strength degradation due to vertical displacement of node center.
Fig. 15 shows differences between displacements of the upper
and lower beam. The degree of twist can be obtained from this va-
lue; the maximum twist is about 0.3°, and can be neglected.

Fig. 7. Measurement plan.

94 I.Y. Jung et al. / Engineering Structures 69 (2014) 90–101

Fig. 8. Displacement of node center.

Fig. 10. Bending failure of brace.

Fig. 9. Bending motion of node.

3.3. Failure mode

The MB-01 specimen was designed as a standard specimen, and

was manufactured with the full penetration welding method. The
maximum strength of this specimen was higher than the allowable
limit of the actuator (3000 kN), therefore after 8th cycle repetitive
loading within the limitation of the actuator was applied. The rep-
etition can be found in the load–displacement hysteresis loop in
Fig. 12(a). The failure mode of the MB-01 specimen was moment
failure, as described previously. At 15th cycle, fractures that started
from the outside of the flange spread into the center of the node,
and eventually the joint between the web and the stiffener was
broken.
The MB-02 specimen has the same design as the MB-01, but the
partial penetration welding method was used at web–flange
welding and flange–flange welding. Likewise MB-01, after 8th cy-
cle repetitive loading within the limitation of the actuator was ap-
plied. The initial failure mode was the same as the MB-01. Tensile
and extra moment stress caused fracture at the outside of the
flange. As the fracture spread to the joint of the web and the verti-
cal stiffener, the welding of the joint broke. Stress that had flowed Fig. 11. Tensile failure of brace. (Specimens MB-04)
through the web-stiffener joint was concentrated at the flange–
flange joint, at which the partial penetration welding was used. The MB-03 specimen, in which the full penetration welding
The partial penetration welding could not resist the stress concen- method is used, had a longer vertical stiffener. The distance
tration. Therefore, the flange–flange joint broke, and the brace was between two vertical stiffeners was farther than for MB-01 and
torn off from the node at 18th cycle. Failure of the web-stiffener MB-02. The maximum strength of this specimen also exceeded
and the flange–flange occurred almost simultaneously. The failure the limit of the actuator. But the number of cycles was less than
mode of the MB-02 specimen was brittle failure. that of the MB-01 and MB-02 specimen. Three repetitive cycles
 I.Y. Jung et al. / Engineering Structures 69 (2014) 90–101 95

Fig. 12. Load–displacement hysteretic curve (Left: upper member, Right: lower member).
96 I.Y. Jung et al. / Engineering Structures 69 (2014) 90–101

Initial stiffness and yielding strength of each specimen are tabulated

in Table 3, and illustrated in Fig. 17. In comparison with MB-01, dif-
ferences of each specimens range from 1% to 13%. The specimens
MB-02 and MB-04, with PP welding area, showed a little higher ini-
tial stiffness. Theoretical stiffness of two members are same but
measured stiffness is not, due to asymmetric settings of two
actuators.
In comparison with MB-01, differences in yield strength of each
specimens range from 3% to 7%. Calculated yield strength is
2184 kN (6270 mm2  324 MPa) and matches well with test results.
The specimens MB-02 and MB-04, with PP welding area, showed a
little small value. Therefore, it can be said that the welding method
and the differences of the design have little effect on initial stiffness
and yield strength.

Fig. 13. Brittle failure of brace.

after 8th cycle MB-03 endured. The failure mode of this specimen
was tensile failure. Because the tensile stress was greater than the
extra moment stress, the joint welding of vertical stiffener and web
was broken.
The MB-04 specimen has the same form as the MB-03 speci-
men, but the partial penetration welding method was used at the
web–flange. The failure mode of this specimen was the same as
for the MB-03 specimen. This specimen experienced three less cy-
cles than MB-03.

3.4. Initial stiffness and yield strength

Fig. 16 shows skeleton curves derived from the hysteresis loop.

From these skeleton curves, the initial stiffness and yielding strength
were obtained as average values from upper and lower members. Fig. 15. Differences of each beam’s displacement.

Fig. 14. Lateral displacement of beams.

 I.Y. Jung et al. / Engineering Structures 69 (2014) 90–101 97

Fig. 16. Skeleton curves (Left: upper member, Right: lower member).

Table 3 3.5. Strain behavior

Results of each specimen.

Specimen Initial stiffness (kN/mm) Yield strength (kN) 3.5.1. Braces

Fig. 18 shows the strain profile of MB-01, which was measured
Upper Lower Upper Lower
at the center (gauge 10, gauge 12) and outside (gauge 16) of the
MB-01 1811 1377 2283 2245
flange. The strain of outside is greater than that of the center from
MB-02 1942 1662 2151 2053
MB-03 1909 1326 2158 2267
the first cycle, and reaches yielding strain more rapidly. The effect
MB-04 2047 1338 2115 2208 of the extra moment can be verified again in this graph. The tensile
strain of gauge 10 and gauge 12 are similar whereas tensile strain

Fig. 17. Initial stiffness and Yield strength.

Fig. 18. Strains of center/outside of flange. (MB-01)

98 I.Y. Jung et al. / Engineering Structures 69 (2014) 90–101
Fig. 19. Strains of center/outside of flange. (MB-03)
Fig. 20. Shear strain of node center.
Fig. 21. Shear strain of the V-point (MB-01).
of gauge 16 is bigger. These differences are influence of bending
moment as described in Fig. 9.
In the case of the MB-03 specimen, however, values of gauge
10 and gauge 16 are almost the same, while gauge 12 has differ-
ences, as shown in Fig. 19. Gauge 12 was located at the intersec-
tion point of the web and the vertical stiffener. Also, its value
starts to show differences with other values from 8 mm tensile
displacement, but remains the same with compressive displace-
Fig. 22. Accumulated energy.
ment. Due to the fracture of the welding between the web and
the vertical stiffener, the tensile stress that had been transferred
through the web was transferred through the flange. Therefore
only the tensile value of gauge 12 is higher than other values.
Also, it can be noted that the fracture started at 8 mm displace-
ment of tensile stress, and the stress concentration became
intensified.
3.5.2. Center of node
Rosette gauges were attached at the intersection point of the
two center axial lines of the braces, because large shear stress con-
centration was expected. However, deformation was not visible,
and remains within the elastic range, as shown in Fig. 20.
3.5.3. V-point
A huge shear strain concentration at the V-point was expected.
As shown in Fig. 21, which shows a graph of the relationship be-
tween the strain of the V-point and the applied load, the shear
strain is small before the failure of the flange, but it becomes sig-
nificantly larger after the failure. The Full penetration welding on
the V-point helped it to not break, but the partial penetration
welding method caused fracture of the brace, as shown in Fig. 13.
 I.Y. Jung et al. / Engineering Structures 69 (2014) 90–101 99

3.6. Energy dissipation specimen started at the curved point of the outside of the flange,
and expanded to the inside. As the fracture expanded to the in-
Fig. 22 shows the accumulated energy dissipation derived from side, the connection between the web and the vertical stiffener
the area of the load–displacement hysteresis loop. The MB-01 and was broken, letting stress go through only the flange. The weld-
MB-02 specimens, which have closer vertical stiffeners, dissipated
more energy. MB-02 experienced 3 more cycles than MB-01.
Therefore dissipated energy of MB-02 is larger than MB-01. How-
ever dissipated energy of two specimens at same cycle are not
Table 4
quite different. In the case of the MB-03 and MB-04 specimens, Comparisons of MA type and MB type.
fractures started at the welding between web and stiffener. A
Specimen Initial stiffness (kN/ Yield strength Dissipated energy (kN-
welding point has low ductility, so that the MB-03 and MB-04
mm) (kN) mm)
specimens have lower energy dissipation ability. If welding quality
of the welding between web and stiffener is poor, early fracture at MA Type 1914 2017 304
MB Type 1594 2259 785
the point could be occurred. Therefore, it is important for nodes, (83%) (112%) (258%)
like those of the MB-03 and MB-04 specimens, to be guaranteed
a good quality of welding.

4. Discussion

4.1. Effect of the length of the vertical stiffener

At a node of the diagrid, many members are connected to-

gether. Therefore, the center of the node is very complicated. A
node used in this study comes from the intersection of H-section
braces, and its webs are continuous through vertical stiffeners.
Two vertical stiffeners connect the webs of the braces, for
smooth flow of stress and reduction of loss of section area. If
these two stiffeners are too close, fabrication of the node could
be difficult. So a new model that has a wider gap between the
vertical stiffeners was suggested, and structural behaviors of
the new model were measured. The standard models (MB-01
and MB-02) have the intersection point of the web and the ver-
tical stiffener at the curved point of the flange. These models
were fractured by outside-of-flange fractures, due to extra mo-
ment and axial tensile stress. However, the new models that
have a wider gap between the vertical stiffeners (MB-03 and
MB-04) showed a different failure mode. Because of the wider
gap of the vertical stiffeners, the web-stiffener intersection point
is located more to the outer side, than for the standard model.
This point locates near the V-point where the flanges of two
Fig. 23. Length of moment arm.
braces meet each other. In the case of these models, the welding
between the web and the vertical stiffener was fractured, and
the crack expanded to the flange, causing fracture of the speci-
mens. While the failure mode of the standard model is moment
failure, the newly suggested model has a tensile failure mode. In
the case of this new model, the intersection of the web and the
vertical stiffener was located where the tensile stress is higher
than the moment stress. So the welding point, which has low
ductility, is the critical point. These differences of failure mode
affect energy dissipation. The suggested model dissipated half
the amount of energy of that of the standard model.

4.2. Effect of welding method

Nodes that are used for tall buildings are fabricated with very
thick plate, and consume large amounts of welding material. If
the partial penetration method is used, instead of the full pene-
tration method, which is commonly being used, the total amount
of welding could show significant savings. In this study, there-
fore, partial penetration welding was used to ascertain the
amount of welding that could possibly be saved. Partial penetra-
tion welding was used at the flange–web and flange–flange
welding. This parameter has no effect on yielding stress and ini-
tial stiffness of the specimens, but did affect the failure mode of
MB-02. This specimen showed moment failure. Fracture of the Fig. 24. Location of crack.
100 I.Y. Jung et al. / Engineering Structures 69 (2014) 90–101

Table 5
Comparisons of test results.

Specimen Initial stiffness (kN/mm) Yield strength (kN) Dissipated energy (kN-m) Web–flange welding Length of vertical stiffener Flange–flange welding
MB-01 1594 2259 785 FP 107 mm FP
(100%) (100%) (100%)
MB-02 1802 2109 1101 PP 107 mm PP
(113%) (93%) (140%)
MB-03 1617 2231 509 FP 238 mm FP
(101%) (99%) (65%)
MB-04 1693 2189 302 PP 238 mm FP
(106%) (97%) (38%)

ing of flange–flange of the MB-02 specimen was partial penetra-
tion, and soon after the breaking of the web-stiffener connection,
the flange–flange welding broke too. The brace was torn off, and
this was brittle fracture.
4.3. Comparison with MA Type
In the study of Kim et al. [13], a web-transferred model, which
was suggested for the convenience of fabrication, was used and
named as the MA Type. The test plan of the MA Type was the same
as the MB Type. The standard model of the MA Type, MA-00, was
compared with the standard model of the MB Type, MB-01
(Table 4). The initial stiffness of the MB Type is 17% lower, yield
strength is 12% higher and energy dissipation is 158% higher than
for the MA Type.
Vertical stiffeners of the MA Type are located at the perimeter
of the node, and transfer stress of the web through the flange.
But vertical stiffeners of the MB Type are located on the inner
side of the node, and transfer the stress of the web directly to
the opposite side. As show in Fig. 23, the distance between the
vertical stiffeners of the MB Type is shorter than that of the
MA Type. This causes a lower ability to resist moment. Therefore
initial stiffness of the MB Type is lower than of the MA Type. But
yield strength and energy dissipation of the MB Type is greater
than of the MA Type. As shown in Fig. 24, fracture locations of
the two types are almost the same. The failure mechanism of
the MA Type, which was verified by Kim et al. [13], is the same
as the MB Type, namely the extra moment. But in the case of the
MA Type, high stress was concentrated at the transference zone,
where ductility is low due to welding, causing low yield strength
and energy dissipation.
5. Conclusions
Structural behavior of the web-continuous diagrid node was
investigated. The initial stiffness, yielding strength and energy dis-
sipation, compared with MB-01, are also summarized in Table 5.
The results are as follows:
(1) The test results showed that there were two kinds of failure
modes. One is moment failure, which is caused by extra
moment and tensile force, and the other is tensile failure,
which is caused by tensile force only. The MB-01 and MB-
02 specimens, which had short vertical stiffeners, were
fractured by the extra moment and the tensile force,
whereas the MB-03 and MB-04 specimens were fractured
by the tensile force only.
(2) The welding method and design details have no effect on ini-
tial stiffness and yield strength. But if partial penetration
welding was used at the V-point, this would cause brittle
fracture. In the previous study by Ju et al. [15], partial pene-
tration welding at the V-point had no significant effect on
structural performance, but in the case of lateral loading, it
caused the brittle failure mode. Therefore, the full penetra-
tion welding method should be used at the V-point. The par-
tial penetration welding method can be used for the web–
flange, because it has no negative effect.
(3) Structural performances of the MB Type were compared
with that of the MA Type. Initial stiffness of the MB Type
is 17% lower than of the MA Type, whereas yield strength
and energy dissipation is 12% and 158% higher, respectively.
Vertical stiffeners of the MA Type located on the perimeter
of the node cause higher moment resistance. But the end
point of the vertical stiffener, where high stress is concen-
trated, has lower ability, due to the welding. Therefore, the
yield stress and energy dissipation of the MB Type are
greater than that of the MA Type. The characteristics of the
two types are very different. The properties of applied load
and convenience of fabrication should be considered, when
selecting types.
Acknowledgements
This research was financially supported by the National Re-
search Foundation of Korea (R1305331). The writers are grateful
to the authorities for their support.
References
[1] Beghini Lauren L, Beghini Alessandro, Katz Neil, Baker William F, Paulino
Glaucio H. Connecting architecture and engineering through structural
topology optimization. Eng Struct 2014;59:716–26.
[2] Moon KS. Material-saving design strategies for tall building structures.
Presented at the CTBUH 8th world Congress; 2008.
[3] Moon KS, Connor JJ, Fernandez JE. Diagrid structural systems for tall buildings:
characteristics and methodology for preliminary design. Struc Des Tall Special
Build 2007;16:205–30.
[4] Leonard JM. Investigation of shear lag effect in high-rise buildings with diagrid
system, Massachusetts Institute of Technology. Dept. of Civil and
Environmental Engineering, Massachusetts Institute of Technology,
Massachusetts Institute of Technology; 2007.
[5] Kim JK, Lee YH. Progressive collapse resisting capacity of diagrid structural
system. J Archit Inst Kor Struct Constr 2008;24:35–42.
[6] Ko CK, Ju YK, Kim SD. Seismic performance evaluation of freeform diagrid
system. J Kor Soc Steel Constr 2011;23:189–98.
[7] Kim KH, Ju YK, Kim SD. Evaluation of the response modification factor for
the height of diagrid structure. J Archit Inst Kor Struct Constr 2009;25:
11–8.
[8] Bae JH, Ju YK, Kim YJ, Kim SD. Experimental evaluation of seismic performance
factors for tall diagrid structure. J Kor Soc Steel Constr 2010;22:75–84.
[9] Huanga Chao, Hana Xiao-lei, Jing Jia, Tang Jia-min. Behavior of concrete-filled
steel tubular planar intersecting connections under axial compression, Part 1:
Experimental study. Eng Struct 2010;32:60–8.
[10] Kim SD, Kim SS, Ju YK. Strength evaluation of beam–column connection in the
weak axis of H-shaped column. Eng Struct 2008;30:1699–710.
[11] Ju YK, Kim MH, Kim JK, Kim SD. Component tests of buckling-restrained braces
with unconstrained length. Eng Struct 2009;31:507–16.
[12] Lee JH, Ju YK, Kim YJ, Kim SD. Experimental evaluation for structural
performance of diagrid brb structural system. J Kor Soc Steel Constr 2010;
22:261–9.
 I.Y. Jung et al. / Engineering Structures 69 (2014) 90–101 101

[13] Kim YJ, Jung IY, Ju YK, Park SJ, Kim SD. Cyclic behavior of diagrid nodes with H- [15] Ju YK, Kim KH, Chang IH, Kim SD. Compressive strength of diagrid node using
section braces. J Struct Eng-Asce 2010;136:1111–22. H-shape steel. J Kor Assoc Spatial Struct 2008;8:91–9.
[14] Kim YJ, Kim MH, Jung IY, Ju YK, Kim SD. Experimental investigation of
the cyclic behavior of nodes in diagrid structures. Eng Struct 2011;33:
2134–44.