Engineering Structures 41 (2012) 209–217

Contents lists available at SciVerse ScienceDirect

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

Prediction of performance of exterior beam-column connections with headed

bars subject to load reversal
Thomas H.-K. Kang a,⇑, Nilanjan Mitra b
a
Department of Architecture & Architectural Engineering, Seoul National University, Seoul 151-744, Republic of Korea
b
Department of Civil Engineering, Indian Institute of Technology, Kharagpur 721 302, India

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

Article history: Given that both ACI 318-08 provisions and 352R-02 recommendations have been developed based on quite
Received 25 October 2011 limited experimental data, an extensive database was assembled by Kang et al. [12], which contains most
Revised 17 March 2012 of the available test data of reinforced concrete exterior beam-column connections with headed bars sub-
Accepted 20 March 2012
ject to load reversal. In this study, the database has been further expanded by adding the recent data focus-
Available online 27 April 2012
ing on the investigation of design parameters of clear bar spacing and head size, and re-evaluated using a
variety of statistical and empirical techniques. An effort has been made to find a statistical model linking
Keywords:
quantitative design parameters and qualitative connection response. In this study, binomial logistic regres-
Reinforced concrete
Beam-column connection
sion methodology has been applied. The statistical methodology quantifies the effect of each design param-
Headed bar eter in determining the performance of the connection. A reliable and robust goodness-of-fit test, the log-
Anchorage likelihood ratio test, was performed to evaluate the developed logistic regression model. Finally, the recent
Development connection data were used to validate the predictive capability of the developed statistical model.
Clear bar spacing Crown Copyright Ó 2012 Published by Elsevier Ltd. All rights reserved.
Statistical methodology
Binomial logistic regression model

1. Introduction is still a need for additional data for various design parameters such
as headed bar clear spacing, number of layers of beam reinforce-
The use of headed bars is becoming more popular since it pro- ment and head size of headed bars, particularly from well-docu-
vides a solution to the constructional problem of steel congestion, mented studies published in English. Also, no study has yet been
particularly in reinforced concrete beam-column connections (see done to investigate the coupled effect of these design parameters
Fig. 1) [22,24,6,9,18,12]. Relevant provisions and limitations have which probably influence connection response. Quantification of
been provided in the 2008 edition of ACI 318 (Sections 12.6.1 and these design parameters that probably influence connection re-
12.6.2). The limitations or restrictions include bar strength, bar sponse has also not been dealt with in any previous study. Further-
and head size, clear cover and bar spacing, and concrete weight. more, no previous study has provided a statistical prediction of
Prior to this, design guidelines for headed bars in beam-column con- satisfactory seismic connection performance over unsatisfactory
nections were incorporated into the 2002 edition of the ACI 352 re- performance given a set of design parameters as specified by the
port based on both monotonic and cyclic tests. This ACI-ASCE code recommendations.
Committee 352 report recommends the development length for To bridge this gap, a thorough data analysis was conducted
headed bars along with some other details such as the location of using a variety of statistical and empirical techniques. A brief sum-
heads and the amount of head-restraining reinforcement for pre- mary of the two recent testing programs conducted as part of this
venting the prying action of headed bars placed near a free surface analysis has been provided. A statistical model has been developed
of concrete. to predict satisfactory over unsatisfactory performance of the con-
In the last two decades, significant amounts of experimental nections with headed bars given a set of design parameters that
investigation have been carried out to determine the suitability of might influence connection responses. For this, binomial logistic
each parameter restriction and limitation as imposed by the ACI regression methodology has been applied. The statistical method-
documents (compiled by Kang [10,11] and Kang et al. [12]). ology quantifies the effect of each design parameter in determining
Three-quarters of the data were available only in Japanese. There the performance of the connection. Subsequently, the methodol-
ogy has been evaluated by using a goodness-of-fit test. Such
comprehensive investigation is made as part of Joint ACI-ASCE
⇑ Corresponding author. Committee 352 task group efforts. Currently, Joint ACI-ASCE
E-mail address: [email protected] (T.H.-K. Kang). Committee 352 recommendations on headed bars in beam-column

0141-0296/$ - see front matter Crown Copyright Ó 2012 Published by Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.engstruct.2012.03.036
210 T.H.-K. Kang, N. Mitra / Engineering Structures 41 (2012) 209–217

Fig. 1. Reinforced concrete beam-column connection with headed bars.

connections are under revision by the Task Group within the com- energy by beam flexural hinging (Fig. 3). Furthermore, neither
mittee, and the authors are spearheading the Task Group’s efforts. strength degradation nor anchorage failure was signaled until the
end of testing (up to 5% drift). As a result, the test results generally
met the ACI 374.1-05 [1] acceptance criteria.
2. Prior database of exterior connections with headed bars

An extensive database spanning a wide range of design param- 4. Discussion of new data in connection with trends
eters for reinforced concrete exterior and knee beam-column
connections with headed bars had been assembled by Kang et al. Fig. 4 presents the relationship of the provided embedment
[12] (see Table 1). All the included specimens were beam-column depth and head size for headed bars used in the satisfactory spec-
connection subassemblies subject to load reversal. This dataset imens (‘o’ marks), and unsatisfactory specimens (‘x’ marks) that
includes classifications of satisfactory and unsatisfactory seismic were affected by improperly short bond development. Note that
connection performance based on the following performance indi- these marks are consistently used throughout the paper. Addition-
ces: (1) the ratio of measured peak moment to nominal moment ally, the three new data discussed in the preceding section were
capacity; (2) drift ratio at the point of 20% drop from the peak lat- plotted using ‘v’ marks in Fig. 4. Based on the test results [13,14],
eral load; (3) ratio of strain in the headed bar at the joint-member all three ‘v’ marked specimens are classified as ‘‘satisfactorily
interface to yield strain; and (4) joint shear distortion during about performed.’’ A new data point [13] for headed bars with small
3.0% drift cycles. Connection behavior was assumed unsatisfactory heads was obtained, giving a valuable indication of the relationship
if the ratio of peak to nominal moments was less than 1.0 and no between a small head size and development length. Fig. 5 signals
bar yielding was monitored by strain gauges [8,15]. If the specimen that a minimum head bearing area (Abrg) of three times the bar area
exhibited more than 20% reduction in strength until 3.5% drift and (Ab) may be feasible for headed bars terminating in beam-column
exceeded 1.2% of joint shear distortion until 3.5% drift cycles, the connections, provided that the development length of the bar com-
connection was also considered to have exhibited unsatisfactory plies with ACI 318-08 [2], Chapter 12. Additional test data would
seismic performance. More details are available in the papers by be helpful to confirm this.
Kang [10,11] and Kang et al. [12]. Only the interstory exterior con- After examining the two new data of beam-column connections
nections of the database are considered in this study. with closely-spaced headed bars [14], it was found that the clear bar
spacing smaller than 4db did not adversely affect the connection
performance due to significant confinement provided by joint trans-
3. Recent data of exterior connections with headed bars
verse reinforcement and a longitudinal beam, where db is the bar
diameter. For the satisfactory connections (see ‘o’ and ‘v’ marks in
Reversed cyclic tests of two full-scale exterior beam-column
Fig. 5), the small clear bar spacing hardly affected the drift ratio
connection subassemblies were carried out to evaluate the applica-
measured at a drop to 80% of the peak lateral load, which is consid-
bility of headed bars with small heads in exterior connections [13].
ered a seismic performance indicator. Particularly, the new test re-
One specimen had headed deformed bars with a small head size of
sults [14] showed that both the connection having a single layer of
(Abrg/Ab = 2.6), while the other specimen had 90-degree hooked
beam top (or bottom) longitudinal reinforcement with cb of 2.1db
bars, where Abrg is the net head bearing area and Ab is the bar area.
and the connection having two layers of beam top (or bottom) lon-
The performance of the connection with headed bars was satisfac-
gitudinal reinforcement with heads touching each other (cb = 1.3db)
tory as demonstrated by modest joint shear distortions (60.01),
performed excellently (see Fig. 3) and generally satisfied the ACI
and generally complied with the ACI 374.1-05 [1], Acceptance Cri-
374.1-05 [1] acceptance criteria. The new data strengthen the pro-
teria for Moment Frames Based on Structural Testing and Com-
posal to increase the minimum clear bar spacing limit in a horizon-
mentary, even with the lack of joint confinement (Fig. 2). More
tal layer of beam longitudinal reinforcement from 4db to 2db for the
details on the specimen design and test procedures and results
design of exterior beam-column connections.
are provided elsewhere [13].
Recently, two more exterior beam-column subassemblies were
subjected to cyclic load reversals: one with a horizontal layer of 5. Methodological approach used for statistical model
closely-spaced headed bars (cs = 2.1db) and the other with two lay-
ers of headed bars with heads touching each other (cs = 1.3db) [14]. Experimental observations provide a qualitative measure of the
The connections were designed in adherence to ACI 318-08 [2] and impact of various design parameters on connection response.
ACI 352R-02 [3], except for the minimum bar clear spacing require- There is an obvious need to relate the qualitative measures with
ment. Both specimens exceeded the values of Mn by 15–20%, exhib- quantitative design parameters such that the effect of the indepen-
ited limited joint shear distortions, and dissipated substantial dent variables can be quantified and the designer acquires an
Table 1
Test data of beam-column connections with headed bars under cyclic loads (1 ksi = 6.895 MPa; 1 in. = 25.4 mm; 1 kip = 4.45 kN).
 
Author I.D Type F.M. fy, ksi db, in. lp, db ldt, db Abrg/Ab ccb, db P Ag fc0 Vn, kip Vu_j, kip qh =qACI;2 d0.8peak cj
h
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16]
Wallace et al. [25] BCEJ1 Ext. I 70 1 13.0 11.9 4.0 Trans. 0 350.1 261.7 1.19 0.048c N.A.
KJ16 Knee N.A. 71 0.63 22.2 11.9 11.4 5.8 0 190.3 104.9 1.08 0.035 N.A.
KJ17 Knee N.A. 71 0.63 22.2 11.9 11.4 5.8 0 191.4 118.1 1.08 0.06 N.A.
KJ18 Knee N.A. 77 0.79 17.8 11.9 7.0 4.6 0 192.9 181.7 1.08 0.04 N.A.
Chun et al. [6] JM-1 Ext. I 58 0.87 15.6 8.2 3.0 6.1 0.03 373.9 165.6 0.27 0.068c 0.001
JM-2 Ext. I 58 0.87 15.6 8.2 3.0 5.7 0.03 369.0 300.2 0.27 0.04 0.011
JM-No. 11-1a Ext. I 66 1.42 11.5 11.3 4.0 3.4 0 366.9 264.6 1.04 0.075 0.003
JM-No. 11-1b Ext. I 66 1.42 11.5 11.3 4.0 3.4 0 366.9 258.3 1.04 0.065 0.006
JMT-No. 11-1a Knee N.A. 68 1.42 10.1 9.8 4.0 5.6 0 278.0 212.6 0.96 0.03 0.003
JMT-No. 11-1b Knee N.A. 68 1.42 10.1 9.8 4.0 5.6 0 278.0 206.3 0.96 0.03 0.025
JMT-No. 11-2a Knee N.A. 68 1.42 10.1 9.8 4.0 5.6 0 278.0 240.1 0.96 0.035 0.003
JMT-No. 11-2b Knee N.A. 68 1.42 10.1 9.8 4.0 5.6 0 278.0 261.7 0.96 0.042 0.002
Bashandy [26] Specimen Ext. I 65 0.98 12 11.4 9.0 2.7 0 135.6 114.3 1.40 0.053 0.005
Lee and Yu [18] W0-M1 Ext. I 69 0.87 13.6 11.5 5.1 9.9 0.10 230.5 170.0 1.10 0.08 N.A.

T.H.-K. Kang, N. Mitra / Engineering Structures 41 (2012) 209–217

W150-M1a Ext. I 69 0.87 13.6b 11.5b 5.1 3 0.10 199.2 172.0 1.10 0.08 0.001
Matsushima et al. [27] H Ext. II 80 0.98 11 17.2 N.A. 2.3 0.11 184.0 165.9 0.39 0.035 0.030
Hs Ext. II 80 0.98 7.2 17.2 N.A. 2.3 0.11 184.0 155.9 0.39 0.035 0.030
Ishida et al. [28] P1 Ext. II 76 0.87 13.0 19.3 N.A. Trans. 0.12 219.4 236.8 1.61 0.015 N.A.
P2 Ext. II 76 0.87 13.0 19.3 N.A. Trans. 0.12 219.4 236.8 1.61 0.03 N.A.
P3 Ext. II 76 0.87 13.0 19.3 N.A. Trans. 0.12 219.4 263.8 1.61 0.03 N.A.
P4 Ext. II 76 0.87 13.0 19.3 N.A. Trans. 0.12 219.4 280.7 1.61 0.03 N.A.
Kiyohara et al. [17] No. 1 Ext. II 103 1.14 12.1 11.5 5.7 2.9 0 601.3 430.2 0.39 0.04c 0.018
No. 2 Ext. II 103 1.14 12.1 9.4 5.7 2.9 0 601.3 513.9 0.26 0.04c 0.018
No. 3 Ext. I 103 1.14 12.1 15.4 5.7 2.9 0 413.2 300.0 0.69 0.04c 0.011
No. 4 Ext. I 103 1.14 15.3 11.5 5.7 2.9 0 601.3 460.4 0.39 0.08c 0.008
No. 5 Ext. II 103 1.14 9.0 11.5 5.7 2.9 0 601.3 388.3 0.39 0.033 0.018
Kiyohara et al. [16] No. 6 Ext. III 150 1.14 12.1 16.5 5.7 2.9 0 635.0 639.1 0.39 0.04c 0.012
No. 7 Ext. III 150 1.14 12.1 13.5 5.7 2.9 0 725.6 664.9 0.26 0.04c 0.013
No. 8 Ext. III 150 1.14 12.1 22 5.7 2.9 0 424.1 447.0 0.69 0.04c 0.017
No. 9 Ext. III 150 1.14 15.3 16.5 5.7 2.9 0 633.8 753.6 0.39 0.04c 0.012
No. 10 Ext. III 150 1.14 9.0 16.5 5.7 2.9 0 640.1 594.7 0.39 0.04c 0.013
No. 11 Ext. II 100 1.14 12.1 11.5 5.7 2.9 0 626.3 454.6 0.39 0.04c 0.014
No. 12 Ext. I 100 1.14 15.3 11.5 5.7 2.9 0 631.1 634.1 0.39 0.04c 0.008
Yoshida et al. [29] No. 1 Ext. I 81 0.75 14.4 7.2 5.8 3.2 0 144.4 127.8 0.50 0.04 0.007
No. 2 Ext. I 81 0.75 14.4 7.2 4.1 3.2 0 144.4 111.1 0.50 0.04 0.006
No. 3 Ext. I 81 0.75 14.4 8 3.1 3.2 0 130.9 109.2 0.61 0.04 0.007
Adachi et al. [30] J30-12-0 Ext. II 76 0.98 11.4 13.5 5.4 3.5 0.06 224.1 258.8 0.71 0.032 N.A.
J30-12-P1 Ext. I 76 0.98 11.4 13.5 5.4 3.5 0.06 224.1 260.5 0.71 0.045 N.A.
J30-12-P2 Ext. I 76 0.98 11.4 13.5 5.4 Trans. 0.06 224.1 264.7 0.71 0.062 N.A.
J60-12-0 Ext. II 76 0.98 11.4 9.5 5.4 3.5 0.04 320.2 276.4 0.35 0.033 N.A.
J60-12-P1 Ext. II 76 0.98 11.4 9.5 5.4 3.5 0.04 320.2 293.2 0.35 0.034 N.A.
J60-12-P2 Ext. I 76 0.98 11.4 9.5 5.4 Trans. 0.04 320.2 303.5 0.24 0.067 N.A.
Masuo et al. [31] AH12-2-45 Ext. II 148 0.98 11.3 8.6 5.8 2.5 0.02 373.2 225.4 0.28 0.03 N.A.
AH12-2-40 Ext. II 148 0.98 11.3 8.6 5.8 2.5 0.02 331.7 210.9 0.30 0.028 N.A.
AH12-2-45A Ext. II 148 0.98 9.3 8.6 5.8 2.5 0.02 373.2 223.5 0.28 0.03 N.A.
AH8-2-45 Ext. II 148 0.98 11.3 10.6 5.8 2.5 0.03 311.8 201.4 0.42 0.03 N.A.
AH12-8-45 Ext. I 92 0.98 11.3 8.6 5.8 3.5 0.02 459.3 485.3 0.37 0.04 N.A.
AH12-8-40 Ext. I 92 0.98 11.3 8.6 5.8 3.5 0.02 408.3 474.4 0.37 0.04 N.A.
AH12-8-45B Ext. I 92 0.98 11.3 8.6 5.8 3.5 0.02 459.3 514.5 0.37 0.04 N.A.
AH8-6-45 Ext. I 92 0.87 12.9 10.6 5.8 4.0 0.02 383.7 379.2 0.46 0.04 N.A.
Tasai et al. [32] No. 6 Ext. III 105 0.98 11.4 6.9 8.0 2.4 0 242.0 224.2 0.91 0.06c 0.020
No. 7 Ext. II 105 0.98 11.4 6.9 8.0 2.4 0 242.0 137.1 1.00 0.03 0.010
Takeuchi et al. [33] 0–1 Ext. I 65 0.98 11.3 7.6 5.8 2.2 0.10 223.1 133.3 0.96 0.05 0.008

(continued on next page)

211
 212
Table 1 (continued)
 
Author I.D Type F.M. fy, ksi db, in. lp, db ldt, db Abrg/Ab ccb, db P Ag fc0 Vn, kip Vu_j, kip qh =qACI;2 d0.8peak cj
h
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16]
0–2 Ext. II 85 0.98 11.3 6.4 5.8 2.2 0.10 262.2 156.2 0.67 0.033 N.A.
0–3 Ext. I 55 0.98 11.3 10.7 5.8 2.2 0.10 183.0 103.9 1.92 0.05 N.A.
0–4 Ext. I 65 0.98 12.6 7.6 5.8 2.2 0.10 223.1 139.7 0.96 0.05 0.003
0–6 Ext. III 104 0.98 11.3 7.6 5.8 2.2 0.10 223.9 187.8 0.96 0.03 0.040
0–7 Ext. III 104 0.98 11.3 6.4 5.8 2.2 0.10 264.7 221.4 0.67 0.03 N.A.
Kato [34] No. 1 Ext. I 82 0.87 15.4 12.1 5.3 4.9 0 331.8 376.3 1.62 0.04c N.A.
No. 2 Ext. I 82 0.87 15.4 11.2 5.3 4.9 0 358.1 365.6 1.62 0.08c N.A.
Murakami et al. [35] No. 100 Ext. I 54 0.63 13.2 11.7 1.7 3.7 0.04 126.4 62.9 0.26 0.08 N.A.
No. 101 Ext. I 54 0.63 13.2 11.7 6.3 3.7 0.04 126.4 65.1 0.26 0.083c N.A.
B8-M Ext. I 74 0.75 11.1 13.0 6.0 3.1 0.06 109.5 85.6 0.34 0.06 N.A.
B7-M Ext. I 74 0.75 11.1 13.0 6.0 3.1 0.06 109.5 76.2 0.34 0.07 N.A.
No. 102 Ext. III 137 0.75 11.1 11.7 2.1 3.1 0.04 126.4 120.1 0.26 0.04 N.A.
No. 103 Ext. III 137 0.75 11.1 11.7 5.8 3.1 0.04 126.4 93.5 0.26 0.055 N.A.
No. 104 Ext. III 137 0.75 11.1 11.7 13.4d 3.1 0.04 126.4 110.0 0.26 0.05 N.A.
M8D16 Ext. III 145 0.63 13.2 13.3 6.0 3.7 0.06 107.1 109.4 0.36 0.04 0.026

T.H.-K. Kang, N. Mitra / Engineering Structures 41 (2012) 209–217

M4D19 Ext. III 145 0.75 11.1 13.3 6.0 3.1 0.06 107.1 103.5 0.36 0.04 0.024
M3D19 Ext. III 145 0.75 11.1 13.3 6.0 3.1 0.06 107.1 102.8 0.36 0.04 0.026
M2D22 Ext. III 141 0.87 9.6 13.3 6.0 2.6 0.06 107.1 103.4 0.36 0.02 0.021
Tazaki et al. [36] E1 Ext. I 55 1.04 9.7 12.1 6.9 1.7 0.08 111.1 91.6 0.40 0.06 0.008
E2 Ext. I 55 1.04 5.9 12.1 6.9 1.7 0.08 111.1 91.6 0.40 0.06 0.007
Nakazawa et al. [37] J1 Ext. I 99 0.75 15.0 12.0 5.9 3.7 0 282.6 254.5 0.81 0.05c 0.001
J2 Ext. I 99 0.75 15.6 12.0 5.9 3.7 0 282.6 251.0 0.81 0.058 0.001
Ishibashi et al. [38] T345-30-4S Roof I 56 0.75 17.9 12.1 5.4 2.5 0 180.9 111.9 0.66 0.065 0.009
T345-30-3N Roof I 56 0.75 17.9 12.1 5.4 2.5 0 180.9 111.3 0.44 0.053 0.008
T490-45-4S Roof I 82 0.75 17.9 14.1 5.4 2.5 0 221.1 156.9 0.44 0.053 0.010
T490-45-3N Roof I 82 0.75 17.9 14.1 5.4 2.5 0 221.1 155.9 0.29 0.04 0.011
Ishibashi and Inokuchi [39] 2S-2 Roof II 77 1.14 16.4 14.9 N.A. 2.5 0 401.5 312.4 0.55 0.03 0.004
2S-0 Roof II 77 1.14 16.4 14.9 N.A. 2.5 0 401.5 307.1 0.55 0.03 0.003
WN-ST Roof II 77 1.14 16.4 14.9 N.A. 2.5 0 410.9 312.4 0.55 0.03 0.005
Shimizu et al. [40] SN-U Roof II 57 0.87 21.7 8.6 N.A. 2.5 0 520.5 329.7 0.37 0.03 0.004
Hattori et al. [41] T-1 Roof II 57 0.75 16.4 11.6 4.1 2.0 0 203.2 N.A. 0.46 0.03 0.028
T-2 Roof II 43 0.75 16.4 11.6 4.1 2.0 0 153.6 N.A. 0.61 0.03 0.028
T-3 Roof II 71 0.75 16.4 17.0 4.1 2.0 0 242.9 N.A. 0.57 0.03 0.028
T-4 Roof II 57 0.75 16.4 11.6 4.1 2.0 0 203.2 N.A. 0.46 0.03 0.016
T-5 Roof I 65 0.63 18.3 9.9 6.2 2.5 0 210.1 N.A. 0.76 0.04 0.002
T-6 Roof I 51 0.63 18.3 9.9 6.2 2.5 0 217.9 N.A. 0.76 0.04 N.A.
T-7 Roof I 56 0.51 22.3 11.6 7.5 3.2 0 158.0 N.A. 1.22 0.04 N.A.
T-8 Roof I 56 0.51 22.3 11.6 7.5 3.2 0 162.6 N.A. 0.61 0.04 N.A.
T-9 Roof I 56 0.51 22.3 11.6 7.5 3.2 0 165.6 N.A. 0.61 0.04 0.008

F.M. = Failure Mode; I = Category-I (member flexural hinging followed by modest joint deterioration); II = Category-II (member flexural hinging followed by joint failure); III = Category-III (joint failure prior to member flexural
hinging); Trans. = with Transverse beams; Abrg = net bearing area of head; Ash = area of joint transverse reinforcement in principal direction within hoop spacing (sh); ccb = clear cover to bar; db = bar diameter; fy = specified yield
stress of headed bars; lp and ldt = development length provided (p) and required (dt) per ACI 352R-02 [3] (Type 2); h00 = joint core width; P = applied column axial force; sh = joint hoop spacing; Vn = nominal joint shear capacity
calculated based on ACI 352R-02 [3]; Vpu_j = max joint shear demand applied during testing; qh = Ash/shh00 ; qACI;2
h
= minimum value recommended by ACI 352R-02 [3] for a Type 2 joint; d0.8peak = drift ratio at a drop to 80% from peak
lateral load; cj = maximum joint shear distortion during about 3.5% drift cycles.
a
Eccentric connection with an offset of a half of the beam width.
b
Average values for staggered headed bars.
c
At least (i.e., testing was stopped prior to 20% drop from the peak).
d
One bearing plate was used for a group of heads; Ab = bar area.
 T.H.-K. Kang, N. Mitra / Engineering Structures 41 (2012) 209–217 213

-4 -3 -2 -1 0 1 2 3 4 (Provided embedment depth of headed bar) / ( ldt )

400 400 0.5 1 1.5 2 2.5
JD Specimen 12 12
300 (Kang et al., 2010a) 300 ACI 352
10 10
200 200
Applied moment (kN-m)

8 8

( Abrg / Ab )

( Abrg / Ab )
100 100 6 6

0 0 4 ACI 318 4
Proposed
2 2
-100 -100
M n (based on meas. 0 0
-200 material properties) -200 0.5 1 1.5 2 2.5
(Provided embedment depth of headed bar) / (ldt ) t

-300 -300
Fig. 4. Head size versus provided-to-required development length per ACI 352.
-400 -400
-4 -3 -2 -1 0 1 2 3 4
Drift ratio (%) 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

Drift ratio at 20% drop from

9 9
Fig. 2. Moment versus lateral drift ratio relationship for specimen tested by Kang
et al. [13]. 8 8

the peak (%)

7 7
6 6
understanding of how the parameters are to be varied to obtain the
desired response. Linear and/or nonlinear regression is one possi- 5 5
ble approach for developing such a model; however, this is not 4 4
ideal because it requires assigning a quantitative measure to the 3 3
qualitative connection response parameter. Conversely, logistic
2 2
regression is ideally suited for developing this type of model. This 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
method allows for quantification of the conditional probability of a
(Headed bar clear spacing) / (db )
qualitative measure based on quantitative data. In fields other than
structural engineering, such as in medicine, clinical studies, sociol- Fig. 5. Drift ratio at 20% drop from the peak versus headed bar clear spacing.
ogy, economics and transportation, the logistic regression is an
established technique for developing relationships between quali-
tative response variables and a set of independent quantitative connections with headed bars under load reversal (a qualitative
parameters. In structural engineering, the method of logistic measure) and a set of independent design parameters (quantitative
regression has been applied by Mitra et al. [21], Mitra and Samui measures). In comparison with linear/nonlinear regression, the lo-
[20] and Mitra [19] to determine the failure initiation mechanism gistic regression model has less stringent requirements, since it
of reinforced concrete interior and exterior beam column connec- does not assume linearity (or nonlinearity) of relationship between
tions utilizing different types of independent quantitative parame- the independent variables and the dependent variable nor does it
ters. For the current study, the logistic regression was used to require normally distributed variables. Details regarding the meth-
develop a relationship between the performance of beam-column odological approach of application of logistic regression models

-6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6
300 300 300 300
JH-R1 Specimen M n (based on meas. Mn (based on meas.
JH-R2 Specimen
(Kang et al., 2010b) material properties) (Kang et al., 2010b) material properties)
200 200 200 200
Applied moment (kN-m)

Applied moment (kN-m)

100 100 100 100

0 0 0 0

-100 -100 -100 -100

-200 -200 -200 -200

-300 -300 -300 -300

-6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6
Drift ratio (%) Drift ratio (%)

Fig. 3. Moment versus lateral drift ratio relationship for specimens tested by Kang et al. [14].
214 T.H.-K. Kang, N. Mitra / Engineering Structures 41 (2012) 209–217

can be obtained from Mitra et al. [21], Mitra and Samui [20] and The method of maximum likelihood [5], which provides a
Mitra [19]; however, for sake of thoroughness, the methodology means of choosing an asymptotically efficient estimator for a set
has been described briefly in the following subsection. of parameters, was used to compute logistic regression parameters
(bi) in Eq. (1). For each of the independent variables in the models,
Table 2 shows the computed regression parameters (bi). It should
5.1. Development of binomial logit model
be noted that by obtaining the logistic regression parameters (bi)
one can use Eqs. (2a) and (2b) to evaluate the probability of occur-
The logistic models [42,4] employ a regression relationship be-
rence of Events 1 and 0, respectively. Given the definition of Y in Eq.
tween independent quantitative variables and discrete qualitative
(1), the sign of a regression parameter indicates whether an in-
events. In the present model, the two discrete qualitative events
crease in the associated design parameter increases or decreases
are ‘‘satisfactory performance’’ (referred to as Event 1) and ‘‘unsat-
the likelihood of the satisfactory performance of exterior connec-
isfactory performance’’ (referred to as Event 0). The likelihood of
tions with headed bars. A positive regression parameter indicates
observing a discrete event of satisfactory performance is defined
that increasing the associated design parameter increases the like-
by the log of the odds ratio for that event. The odds ratio for Event
lihood of a satisfactory performance. Similarly, a negative regres-
1 is the ratio of the probability of occurrence of Event 1, P E¼1 , to the
sion parameter indicates that increasing the associated design
probability of occurrence of Event 0, PE¼0 . Thus,
parameter increases the likelihood of unsatisfactory performance.
   
PE¼1 PE¼1 PK Based on the signs of the parameters in Table 2, an increase in A,
Y ¼ log ¼ log ¼ b0 þ bk X k ð1Þ B, C and D would result in increase in the likelihood of satisfactory
1  PE¼1 PE¼0 k¼1
performance of the connection response. An increase in other
where bi’s are the logistic regression parameters, Xi’s are the covar- parameters such as E, F, G or H would result in increased likelihood
iates or connection design parameters and K is the total number of unsatisfactory performance of the connection response.
(=8) of design parameters considered. As described in the book by An increase in the ratio of provided embedment depth to
Ben-Akiva and Lerman [5], Eq. (1) may be manipulated to define required development length (A) typically means an improved bond
the probability of occurrence of Events 1 and 0 such that: condition of the specimen, and improved bond capacity would qual-
PK itatively mean better performance of the connection. Fig. 4 graphi-
eb0 þ b X
k¼1 k k cally supports the idea that the provided embedment depth directly
PE¼1 ¼ PK ð2aÞ affects the connection performance. An increase in ratio of the thick-
b X
1 þ e b0 þ k¼1 k k
ness of the anchored head to the diameter of the reinforcing bar (B)
qualitatively means an increase in resistance of the head against
1 deformation which would also equate to better performance of
PE¼0 ¼ PK ð2bÞ
1þe b0 þ b X
k¼1 k k the connection. No comprehensive research on head thickness has
been carried out, nor are standards on head thickness available in
ACI 318 codes and ASTM specifications. Based on the new experi-
5.2. Discussion of binomial logit analysis mental work by Kang et al. [13], the head thickness of at least 1db
is considered reasonable.
The independent quantitative design parameters that were con- If the head size (C) is increased, a better bearing is achieved
sidered to affect the performance of the connection were taken as which qualitatively also results in better performance of the con-
follows: (1) the ratio of provided embedment depth to required nection region. A comprehensive review of the database [10–12]
development length as per ACI 352 (lp/ldt; labeled as A); (2) the ratio showed that both the development length and head size determine
of head thickness to bar diameter (thead/db; labeled as B); (3) the ratio the anchorage capacity of a headed bar, and that the head size
of net bearing area to bar area (Abrg/Ab; labeled as C); (4) the ratio of should be large enough to ensure no anchorage failure (at least
joint shear at probable beam moment to joint shear capacity (Vu_j/ Abrg P 3Ab according to the new experimental work by Kang
Vn; labeled as D); (5) the specified yield strength of the headed bar et al. [13]). Better performance can also be achieved if the joint
(fy; labeled as E); (6) the joint transverse reinforcing ratio in the shear demand (D) is decreased. The results obtained from the sta-
direction of lateral loading (Av_joint/[shh00 ]; labeled as F); (7) the side tistical methodology are in direct agreement with what has been
cover for the headed bar to bar diameter (ccb/db; labeled as G); and observed experimentally (e.g., [6]). Less joint shear deformation
(8) the ratio of column axial load to compressive strength of the con- was monitored for the connection subject to a smaller joint shear
crete (P=½Ag fc0 ; labeled as H). Here, Av_joint is the cross-sectional area demand, but with the same or comparable other conditions.
of joint transverse reinforcement placed in the direction of lateral On the other hand, an increase in yield strength of the longitu-
loading within the joint in a layer, sh is the joint hoop spacing, h00 dinal bar (E) results in an increase in the elastic stiffness of the con-
is the joint core width, Ag is the gross column section area, and fc0 nection along with exhibition of a rather brittle response and
is the specified concrete compressive strength. reduced ductility. The reduced ductility due to the use of high
strength bars (fy > 60 ksi; 430 MPa) results in unsatisfactory perfor-
mance of the connection. All of the specimens with very high
Table 2
strength steel (fy P 120 ksi or 830 MPa) in the database [12] did
Estimated logistic regression parameter (bi) and influence factor for each design
parameter.
not exhibit satisfactory seismic connection performance. Further-
more, only 3 of 16 specimens with fy P 100 ksi (690 MPa) (No. 3,
Design parameter Estimated bi Influence factor
[17]; Nos. 4 and 12, [16]) showed satisfactory seismic performance.
A 5.22 6.01 Therefore, the feasibility of the applications of very high strength
B 1.89 1.1 steel in beam-column connections is questionable, and further
C 1.03 5.77
D 19.88 16.56
scientific research on this topic is required.
E 0.25 24.5 Contrary to the common belief, it was observed that increasing
F 94.25 2.97 the area of the transverse reinforcement within the joint (F) results
G 0.29 0.95 in unsatisfactory performance. On a related note, it has been
H 85.3 2.68
pointed out by many researchers (e.g., [7,23]) that for exterior con-
Constant 1.46
nections with conventional longitudinal reinforcement subject to
 T.H.-K. Kang, N. Mitra / Engineering Structures 41 (2012) 209–217 215

load reversal, transverse reinforcement within the joint does not 1

probability of occurence of Event 1

have any significant influence on the failure mechanism of beam-
column connections under low axial loads (e.g., P 6 0:12Ag fc0 ). All 0.8
of the specimens in the database were subjected to axial loads less
0
than 0:12A
  g fc [12]. Based on a review of the database, a ratio of
0.6
qh =qACI;2
h
even at the level of about 0.3 appeared not to pose a
serious joint shear distress problem under low axial loads, where
qh = (Ash/shh00 ), Ash is the area of joint transverse reinforcement in 0.4
the principal direction within hoop spacing (sh), sh is the joint hoop
spacing, and h00 is the joint core width. 0.2
It was observed that increasing the side cover (G) to the headed
bar beyond the minimum value results in an increased probability
0
of unsatisfactory performance. Typically, the larger side cover is, Event 0 Event 1
the less vulnerable the connection is to side-face blowout failure. performance choice of exterior connections
As noted earlier, side-face blowout is not a concern for headed bars with headed bars
anchored in interstory exterior connections because of its suffi-
cient side cover. Only two specimens in the database had the side Fig. 6. Predictive efficiency assessment of the developed logit model.

cover to the headed bar less than 2db. Therefore, the adverse effect
of increased side cover, shown in Table 2, appears to have no sig- 6.2. Predictive efficiency assessment
nificant physical meaning. Increasing the column axial ratio (H)
also results in an increase in the probability of unsatisfactory per- In order to assess the predictive efficiency of the statistical
formance; however, as mentioned earlier, the level of the applied model, the likelihood of satisfactory performance (Event 1), com-
axial load was quite low for all specimens P 6 0:12Ag fc0 . Thus, puted using Eq. (2a) with bi from Table 2, was plotted versus the
no information regarding the effect of high axial loads is available observed event in Fig. 6. Connections from the dataset exhibiting
from the database. unsatisfactory performance (Event 0) are plotted as circles and con-
The magnitude of a regression parameter multiplied with the nections exhibiting satisfactory performance (Event 1) are plotted
mean of its corresponding design variables (referred to as the as squares. If the model were perfect, all connections exhibiting
‘‘influence factor’’ column in Table 2) indicates the relative impor- Event 0 would have a computed probability of occurrence of Event
tance of the design variables in determining connection failure ini- 1 of 0.0, while all connections exhibiting Event 1 would have a
tiation response. It should be noted that the sign of the influence computed probability of occurrence of 1.0. The data in Fig. 6 indi-
factor is similar to the sign of the regression parameter. Based on cate that although the model is not perfect for exterior connections
results obtained in Table 2, the yield strength of the headed rein- with headed bars, the model is able to predict satisfactory perfor-
forcing bar is the most influential parameter, a decrease of which mance for 94% of the connections and unsatisfactory performance
would result in higher satisfactory performance, whereas the least for 89% of the connections. Overall, the correct prediction for the
influential parameters are the side cover and head size. Again, this model is 92%.
indicates that there is a great and urgent need to carry out research
regarding high-strength headed bars in beam-column connections 7. Validation of the developed logit model using new data
subject to load reversal.
In the preceding section, the logit model performance was eval-
uated using statistical methods. In this section, the developed
6. Statistical evaluation of the applied statistical model
model is validated with new experimental investigation data. The
recommended procedure of using the developed model for predict-
6.1. Goodness of fit of the developed logit model
ing satisfactory or unsatisfactory performance is demonstrated
with the help of three recently tested connections (see Section 3).
To further evaluate the model, log-likelihood ratio test was per-
formed which estimates the overall explanatory power of the mod-
1. Calculate the design parameters considered in the logit model
el and also determines whether the inclusion of the independent
for the new data.
parameters chosen for the model are statistically significant and
2. Obtain the probabilities of satisfactory and unsatisfactory per-
improves the overall model prediction.
formance using Eqs. (2a) and (2b), along with the estimated bi
The likelihood ratio test statistic is as follows:
values listed in Table 2.
X 2 ¼ 2½LLðbR Þ  LLðbU Þ ð3Þ
The aforementioned procedure was applied when predicting
where LL(bR) is the log likelihood at convergence of the ‘‘restricted’’ the performance of three connections tested by Kang et al.
model (i.e., model in which all the parameters except the constant [13,14], which were not included in the database used for the logit
term are equal to 0), and LL(bU) is the log likelihood at convergence model development. Table 3 lists the design parameters calculated
of the ‘‘full’’ or ‘‘unrestricted’’ model (i.e., model under investigation for the connection with small-headed bars [13], the connection
with all parameters). The X2 statistic is chi-square distributed with with horizontally closely-spaced headed bars [14] and the connec-
the degrees of freedom equal to the difference in the number of tion with vertically closely-spaced headed bars [14]. The computed
parameters in the ‘‘restricted’’ and ‘‘unrestricted’’ model. For the probabilities of satisfactory and unsatisfactory performance for the
proposed logit model for exterior connections, the value of the ob- first connection are 84% and 16%, respectively. The probabilities of
tained log-likelihood for the ‘‘unrestricted’’ or ‘‘full’’ model is satisfactory performance of the second and third connections are
11.43 whereas for the ‘‘restricted’’ model it is 42.08. The chi- approximately 99%. Therefore, it can be concluded that the statis-
squared value obtained is 30.65 with 8° of freedom resulting in a tical model, expressed by Eqs. (2a) and (2b) along with the bi val-
p-value of less than 0.001, suggesting that inclusion of the indepen- ues in Table 2, predicts the observed three connection responses
dent parameters in the predicted model significantly improves the quite accurately, given the material and geometric properties as
goodness-of-fit measure. well as the reinforcing details.
216 T.H.-K. Kang, N. Mitra / Engineering Structures 41 (2012) 209–217

Table 3 [8] Han SW, Kee S-H, Park Y-M, Lee L-H, Kang TH-K. Hysteretic behavior of
Design parameters used for specimens tested by Kang et al. [13,14]. exterior post-tensioned flat plate connections. Eng Struct
2006;28(14):1983–96.
Design parameter JD [13] JH-R1 [14] JH-R2 [14] [9] Hong S-G, Chun S-C, Lee S-H, Oh B. Strut-and-tie model for development of
headed bars in exterior beam-column joint. ACI Struct J 2007;104(5):
A 1 1.04 1.04
590–600.
B 1 1.53 1.53
[10] Kang, TH-K. A review of ACI standards and seismic tests of beam-column joints
C 2.6 5.28 5.28 with headed reinforcement. In: The 2008 international conference on
D 0.46 0.76 0.78 advanced in structural engineering and mechanics (ASEM’08), Jeju, Korea;
E 400 MPa 414 MPa 414 MPa 2008a. p. 2299–314.
F 0.00307 0.00696 0.00696 [11] Kang TH-K. Recommendations for design of RC beam-column connections
G 3.55 4.87 4.87 with headed bars subjected to cyclic loading. The 14th world conference on
H 0 0 0 earthquake engineering (14WCEE), Beijing, China (Paper No. 08–01-0017);
2008b.
[12] Kang TH-K, Shin M, Mitra N, Bonacci JF. Seismic design of reinforced concrete
8. Conclusions beam-column joints with headed bars. ACI Struct J 2009;106(6):868–77.
[13] Kang TH-K, Ha S-S, Choi D-U. Bar pullout tests and seismic tests of small-
headed bars in beam-column joints. ACI Struct J 2010;107(1):32–42.
The database spanning a wide range of design parameters for [14] Kang TH-K, Kim W, Shin M. Cyclic testing for seismic design guide of beam-
reinforced concrete exterior beam-column connections with column joints with closely-spaced headed bars. J ⁄⁄⁄Earthq Eng
2012;16(2):211–30.
headed bars has been statistically assessed to obtain forensic [15] Kim J, LaFave J. Key influence parameters for the joint shear behaviour of
evidence of the observed behavior. As part of this study, binomial reinforced concrete (RC) beam-column connections. Eng Struct
logistic regression methodology has been developed to quantify 2007;29(10):2523–39.
[16] Kiyohara T, Tasai A, Watanabe K, Hasegawa Y, Fujimoto T. Seismic capacity of
the effect of each design parameter in determining the perfor- high strength RC exterior beam column joint with beam main bars anchored
mance of the beam-column connection with headed bars subject mechanically. Architect Inst Jpn 2004;27–34 [in Japanese].
to load reversal. In the end, the statistical methodology was evalu- [17] Kiyohara T, Hasegawa Y, Fujimoto T, Akane J, Amemiya M, Tasai A, et al.
Seismic performance of high strength RC exterior beam column joint with
ated by using two robust goodness-of-fit tests and actual experi-
beam main bars anchored mechanically. Architect Inst Jp 2005;33–42 [in
mental data. Japanese].
The following conclusions were drawn: (1) an increase in devel- [18] Lee H-J, Yu S-Y. Cyclic response of exterior beam-column joints with different
opment length, head thickness and head size and a decrease in anchorage methods. ACI Struct J 2009;106(3):329–39.
[19] Mitra N. Failure initiation of reinforced-concrete beam-column connections –
joint shear demand result in a qualitatively better performance binomial logistic regression based probabilistic model. Adv Struct Eng
of the connection; (2) an increase in bar yield strength, joint 2012;15(4):121–38.
transverse reinforcement, and column axial force correlates to in- [20] Mitra N, Samui P. Prediction of inelastic mechanism leading to seismic failure
of interior reinforced concrete beam-column connections. ASCE J Practice
creased probability of unsatisfactory performance; (3) the joint Period Struct Des Construct 2011. http://dx.doi.org/10.1061(ASCE)SC.1943-
shear demand and bar yield stress are two of the most influential 5576.0000115.
design parameters on the connection performance; and (4) the fea- [21] Mitra N, Mitra S, Lowes LN. Probabilistic model for failure initiation of
reinforced concrete interior beam-column connections subjected to seismic
sibility of the applications of very high strength headed bars in loading. Eng Struct 2011;33:154–62.
beam-column connections is highly questionable. These conclu- [22] Park HK, Yoon YS, Kim YH. The effect of head plate details on the pull-out
sions were verified by applying the log-likelihood ratio test. Finally, behaviour of headed bars. Mag Concrete Res 2003;55(6):485–96.
[23] Paulay T, Scarpas A. The behavior of exterior beam-column joints. Bull NZ Natl
the statistical methodology was validated using the actual data, Soc Earthq Eng 1981;14:3.
revealing that the model is capable to reasonably predict the con- [24] Thompson MK, Jirsa JO, Breen JE. Behavior and capacity of headed
nection performance. reinforcement. ACI Struct J 2006;103(4):522–30.
[25] Wallace JW, McConnell SW, Gupta P, Cote PA. Use of headed reinforcement in
beam-column joints subjected to earthquake loads. ACI Struct J
1998;95(5):590–606.
Acknowledgements [26] Bashandy TR. Application of headed bars in concrete members. Ph.D.
dissertation, University of Texas at Austin, Austin, Tex., December 1996.
303pp.
The work presented in this paper was sponsored by US DOT– [27] Matsushima M, Kuramoto H, Maeda M, Shindo K, Ozone S. Test on corner
RITA (Grant No. DTRT06-G-0016/OTCREOS9.1-27), by National Re- beam-column joint under tri-axial loadings. In: Proc. Architectural Institute of
search Foundation of Korea (Grant No. 2012R1A1A1005905), and Japan, September 2000. p. 861–3 [in Japanese].
[28] Ishida Y, Fujiwara A, Adachi T, Matsui T, Kuramoto H. Structural performance
by the Engineering Research Institute of Seoul National University, of exterior beam-column joint with wide width beam using headed bars. In:
Korea. The views expressed are those of authors, and do not neces- Proc. of Architectural Institute of Japan, August 2007. p. 657–60 [in
sarily represent those of the sponsor, assistants or discussants. Japanese].
[29] Yoshida J, Ishibashi K, Nakamura K. Experimental study on mechanical
anchorage using bolt and nut in exterior beam-column joint. In: Proc.
Architectural Institute of Japan, September 2000. p. 635–8 [in Japanese].
References [30] Adachi M, Masuo K. The effect of orthogonal beams on ultimate strength of R/C
exterior beam-column joint using mechanical anchorages. In: Proc.
Architectural Institute of Japan, August 2007. p. 633–4 [in Japanese].
[1] ACI. Acceptance criteria for moment frames based on structural testing and
[31] Masuo K, Adachi M, Imanishi T. Ultimate strength of R/C exterior beam-
commentary (ACI 374.1-05). American Concrete Institute, Farmington Hills,
column joint using mechanical anchorage for beam reinforcement USD590. In:
Michigan, 2005.
Proc. Architectural Institute of Japan, September 2006. p. 25–8 [in Japanese].
[2] ACI. Building code requirements for structural concrete (ACI 318-08) and
[32] Tasai A, Kawakatsu K, Kiyohara T, Murakami M. Shear performance of exterior
commentary. American Concrete Institute, Farmington Hills, Michigan, 2008.
beam column joint with beam main bars anchored mechanically. In: Proc.
[3] ACI-ASCE. Recommendations for design of beam-column connections in
Architectural Institute of Japan, September 2000. p. 857–60 [in Japanese].
monolithic reinforced concrete structures (ACI 352R-02). American Concrete
[33] Takeuchi H, Kishimoto T, Hattori S, Nakamura K, Hosoya H, Ichikawa M.
Institute, Farmington Hills, Michigan, 2002.
Development of mechanical anchorage used circular anchor plate. In: Proc.
[4] Agresti A. An introduction to categorical data analysis. Hoboken, New
Architectural Institute of Japan, September 2001. p. 111–4 [in Japanese].
Jersey: Wiley; 2007.
[34] Kato T. Mechanical anchorage using anchor plate for beam/column joints of R/
[5] Ben-Akiva M, Lerman SR. Discrete choice analysis: theory and application to
C frames. In: Proc. Architectural Institute of Japan, September 2005. p. 277–8
predict travel demand. Massachusetts, Cambridge: MIT Press; 1985.
[in Japanese].
[6] Chun SC, Lee SH, Kang TH-K, Oh B, Wallace JW. Mechanical anchorage in
[35] Murakami M, Fuji T, Kubota T. Failure behavior of beam-column joints with
exterior beam-column joints subjected to cyclic loading. ACI Struct J
mechanical anchorage in subassemblage frames. Concrete Res Technol
2007;104(1):102–13.
1998;8(1):1–9 [in Japanese].
[7] Fujii S, Morita S. Comparison between interior and exterior RC beam-column
[36] Tazaki W, Kusuhara F, Shiohara H. Tests of R/C beam-column joints with
joint behavior. ACI SP 123-12: design of beam-column joints for seismic
irregular details on anchorage of beam longitudinal bars. In: Proc. of
resistance, American Concrete Institute, Farmington Hills, Michigan, 1991. p.
Architectural Institute of Japan, August 2007. p. 653–6 [in Japanese].
145–65.
 T.H.-K. Kang, N. Mitra / Engineering Structures 41 (2012) 209–217 217

[37] Nakazawa H, Kumagai H, Saito H, Kurose Y, Yabe Y. Development on the ultra- [40] Shimizu Y, Ishibashi K, Inokuchi R. Experimental study on T-shaped joints with
high-strength reinforced concrete structure. In: Proc. Architectural Institute of anchor-heads on column’s Rebars. In: Proc. Architectural Institute of Japan, No.
Japan, September 2000. p. 611–2 [in Japanese]. 25171, September 2005. p. 281–4 [in Japanese].
[38] Ishibashi K, Inokuchi R, Ono H, Masuo K. Experimental study on T-shaped [41] Hattori S, Ishiwata Y, Ichikawa M, Takeuchi H, Nakamura K, Hosoya H.
beam-column joints with anchor-heads on columns’ Rebars – Part 1 and Part Development of mechanical anchorage used circular anchor plate. In: Proc.
2. In: Proc. Architectural Institute of Japan, September 2003. p. 533–6 [in Architectural Institute of Japan, August 2002. p. 565–6 [in Japanese].
Japanese]. [42] Hosmer DWT, Lemeshow S. Applied logistic regression. Hoboken, New
[39] Ishibashi K, Inokuchi R. Experimental study on T-shaped joints with anchor- Jersey: Wiley; 2000.
heads on columns’ Rebars – Part 3 and Part 4. In: Proc. Architectural Institute
of Japan, August 2004. p. 819–22 [in Japanese].