Journal of Constructional Steel Research 112 (2015) 155–166

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

Cyclic response sensitivity of post-tensioned steel connections using

sequential fractional factorial design
Saber Moradi a, M. Shahria Alam b,⁎, Abbas S. Milani c
a
School of Engineering, The University of British Columbia, 1137 Alumni Avenue, EME 3213, Kelowna, BC V1V1V7, Canada
b
School of Engineering, The University of British Columbia, 1137 Alumni Avenue, EME 4225, Kelowna, BC V1V1V7, Canada
c
School of Engineering, The University of British Columbia, 1137 Alumni Avenue, EME 4203, Kelowna, BC V1V1V7, Canada

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

Article history: Through the use of post-tensioned (PT) elements in steel beam-column connections, steel buildings under
Received 17 December 2014 seismic excitations can return to their plumb position, displaying negligible permanent deformation. The cyclic
Accepted 26 April 2015 behavior of a PT connection is affected by several design parameters. This paper aims at identifying the significant
Available online xxxx
factors which affect the cyclic response of steel PT connections with top-and-seat angles. A sequential fractional
factorial design-of-experiment methodology is used to statistically evaluate the effects of different design factors
Keywords:
Post-tensioned connection
as well as their interactions on the cyclic response of PT connections. To this end, 3D finite element models are
Steel beam-column connection first developed to accurately simulate the cyclic behavior of the connections. After validating the finite element
Moment resisting frames results with the past experimental data, a two-stage (sequential) sensitivity analysis is conducted. Eight potential
Sensitivity factors, including the material and geometric properties of steel angles, reinforcing plates, and bolts, are
Cyclic response considered. The cyclic response of connections is examined in terms of stiffness, strength, energy dissipation
Finite element analysis capacity, and residual displacement. From this parametric study, the significance of the design factors is deter-
mined with respect to each response. Additionally, regression models are presented to estimate the response
quantities for other PT connection configurations with the same beam and column sections.
© 2015 Elsevier Ltd. All rights reserved.

1. Introduction Previous research highlights the importance of accounting for

In current seismic design codes, steel buildings are designed to
ensure that the collapse of buildings and loss of lives are prevented.
Buildings are expected to resist seismic loads while being ductile
enough to undergo large deformations [1]. Consequently, steel
buildings may suffer extensive damage in the form of material yielding,
fracture and buckling of structural steel elements. Following an
earthquake, the incidence of permanent damage in a building increases
repair costs. Large permanent deformations can even make the repair of
buildings prohibitive, and ultimately, demolition of severely-damaged
buildings could be required. In the 2010 and 2011 Canterbury
earthquakes, over 150,000 homes were damaged. A rebuild cost of
around $20 billion (New Zealand dollars) was estimated, not including
disruption costs and insured losses [2]. In the central business district
and suburbs, 1600 buildings needed to be demolished [3]. For a large
city with damaged, non-serviceable buildings, the economic losses are
significant. The repair of buildings with residual deformation of over
0.5% may not be economically feasible [4].
⁎ Corresponding author.
E-mail addresses: saber.moradi@ubc.ca (S. Moradi), shahria.alam@ubc.ca (M.S. Alam),
abbas.milani@ubc.ca (A.S. Milani).
residual deformations when designing buildings for seismic loads
[5–7]. Along with these efforts, research continues in mitigating residual
deformations of buildings subject to earthquakes [8,9]. Over the past
20 years, there has been a growing interest in seismic resilient buildings
with self-centering capabilities [10–13]. In such buildings, certain
elements are used so as to provide restoring forces, and thus eliminate
(or control) residual deformations after earthquakes. Several self-
centering systems have been previously investigated for steel buildings,
using different mechanisms for energy dissipation and self-centering
capability. Examples of these self-centering systems are steel buildings
with shape memory alloy (SMA) bracing [14–16] and SMA-based
connections [17–19], self-centering energy dissipative braces [20–22],
rocking systems [23–25], and steel plate shear walls [26,27]. Another
approach to self-centering is to use post-tensioned (PT) elements in
steel moment resisting frames [28].
Over the past two decades, several kinds of energy dissipation
mechanisms have been examined for PT steel beam-column connec-
tions. Fig. 1(a) illustrates a PT beam-column connection with top-and-
seat angles. In a PT steel moment frame, beams are compressed to
columns using horizontal PT elements (strands or bars). Ricles et al.
[29] performed nine experimental tests to investigate the cyclic
performance of PT connections with top-and-seat angles. It was
confirmed that self-centering is achieved using PT strands, while

http://dx.doi.org/10.1016/j.jcsr.2015.04.022
0143-974X/© 2015 Elsevier Ltd. All rights reserved.
156 S. Moradi et al. / Journal of Constructional Steel Research 112 (2015) 155–166

Fig. 1. (a) Schematic of an exterior self-centering beam-column connection with PT steel strands, (b) Geometric factors considered in the sensitivity study (factors A, B, E, and F).

damage is essentially confined to replaceable top-and-seat angles. The
test results demonstrated that PT connections display excellent
stiffness, strength, and ductility. Further tests [30] with larger beam
sections introduced PT connections as a vital alternative to conventional
welded connections. Researchers have also studied PT beam-column
connections with other energy dissipation mechanisms. Examples of
those systems are PT connections with buckling-restrained energy
dissipative steel bars (e.g., [31]), PT connections with friction devices
[32,33], and PT connections with web hourglass shape steel pins [34].
2. Motivation and objective of this study
In this paper, the sensitivity of cyclic response of PT connections to
different design variables is examined. The variability of the cyclic
response to a number of individual factors such as thickness of steel
angles has been addressed in previous research [29]. However, to
investigate response sensitivity of multiple factors at the same time, a
factorial analysis is required. By varying different factors together, the
effect of each factor is assessed. Further, any possible interaction
between factors is determined. An interaction is statistically active
when the effect of one factor on the response changes at different levels
of another factor. For the present study, first, an efficient finite element
model is generated and validated against experimental data in [29].
Then, two-level factorial designs with eight factors are generated to
evaluate the influence of each factor on the cyclic response. The selected
factors include geometric and material properties of steel angles
(thickness, gage length, and yield strength), thickness and length of
reinforcing plates, bolt pretension force and bolt yield strength, and
post-tension force in high strength steel strands. The cyclic force–dis-
placement response of PT connections is examined in terms of different
characteristics, including stiffness, load capacity, hysteretic energy
dissipation, and residual displacement. In addition to evaluating the
significance of different factors associated with each response
characteristic, predictive equations are presented. The accuracy of the
predictive equations is verified by comparing the response quantity
from regression models with that from finite element analysis. Overall,
using a two-stage fractional factorial strategy, 56 finite element models
of PT beam-column connections with top-and-seat angles are analyzed,
including the final verification models. The results to be presented in the
following sections would be useful in practice for performance
optimizations of PT steel beam–column connections.
3. Finite element model
To simulate the cyclic behavior of PT steel beam–column connec-
tions, 3D finite element models were developed and analyzed in FE
software ANSYS [35]. In order to calibrate the finite element models
and verify simulation results, two of the PT connection specimens

Fig. 2. Interior PT connections considered (adapted from [29]).

 S. Moradi et al. / Journal of Constructional Steel Research 112 (2015) 155–166 157

Fig. 3. Meshing details of the finite element model for specimen PC3: the angle, the connection interface, and the bolt.

reported in the experimental study by Ricles et al. [29] were modeled,
namely post-tensioned connections PC3 and PC4. A schematic view of
these interior connections is illustrated in Fig. 2. Each specimen consists
of W24 × 62 wide flange beams and a W14 × 311 wide flange column.
The connection also has L203 × 203 × 15.9 steel angles, 25.4 mm A325
bolts, beam flange reinforcing plates, and shim plates.
The finite element model includes all components, such as column,
beams, angles, reinforcing plates, shim plates, bolts, bearing plates,
steel PT strands, and anchorage plates at the beams end. Except for
the PT strands, the finite element models of all the components were
meshed using SOLID185 elements in ANSYS. To improve the efficiency
of the finite element analysis, the PT strands were modeled using
BEAM188 elements. Additionally, mesh density was controlled for
different parts of the model. In this regard, finer meshes were produced
for the volumes close to the connection interface. The finite element
model for a PT connection typically consists of 41432 nodes and
16874 elements. As shown in Fig. 3, a symmetry condition was assumed
where half of the connection was modeled to further improve the
efficiency of the finite element analysis. The computational running
time for a model was typically 5 h, using a 64-bit operating system in
Windows 7 with an Intel Core i5-3570 CPU at 3.40 GHz and with 8 GB
of RAM.
The material properties reported in [29] were used for the finite
element analysis. For the high-strength bolts, a trilinear stress–strain
relationship was adopted, whereas a bilinear elasto-plastic behavior
was assigned for the other components. The modulus of elasticity and
Poisson's ratio for steel sections and bolts were assumed to be
200 GPa and 0.3, respectively. Further, von Mises yield criterion with
kinematic strain hardening and the associated flow rule were employed
[36].
In addition to material and geometric nonlinearity, the finite
element analysis includes the nonlinearity due to contact and sliding
between components. Contact elements, CONTA173, and target
elements, TARGE170, were defined over the areas between the follow-
ing structural components [refer to Fig. 1(a)]: (a) angle vertical leg
and the shim plate; (b) angle horizontal leg and the beam flange (for
the sensitivity study, the reinforcing plates were used outside of the
beam section. Therefore, for those models contact elements were
defined between the angle leg and the beam flange reinforcing plates);
(c) beam web/flanges and the shim plate. A coefficient of friction equal
to 0.33 was assumed for all the contact elements. Furthermore, to model
the contact between PT strands and holes on column flanges, force-
distributed constraints were employed. By using this multipoint
constraint (MPC) capability, the motion of nodes on the contact surface

Fig. 4. Lateral load-displacement response of specimens PC3 and PC4: numerical results in comparison with experimental results (adapted from [29]).
158 S. Moradi et al. / Journal of Constructional Steel Research 112 (2015) 155–166
Table 1
Comparison of the finite element analysis response with test results in [29].
Specimen Ki,FE Kr,FE Fmax,FE Ki,FE/Ki,test Kr,FE/Kr,test Fmax,FE/Fmax,test
(kN/m) (kN/m) (kN)
PC3 12,337.34 873.65 258.95 1.04 0.91 0.99
PC4 13,156.28 1010.85 294.32 1.08 0.94 0.96
(contact elements, CONTA177) was coupled to a pilot node on the target
surface.
Boundary conditions were modeled by restraining node displace-
ments. The vertical and horizontal displacements for the column pin
support and the vertical displacement for the beam roller support
were set to zero. Similar to the reference experimental study in [29], a
displacement-controlled loading was applied on the column nodes at
a distance of 3658 mm from the bottom. The loading protocol involves
symmetrical cyclic story drifts with amplitudes of 0.1, 0.2, 0.3, 0.4, 0.5,
0.7, 1, 1.5, 2, 2.5, 3 and 3.5%. Here, story drift is defined as the displace-
ment applied on the column flange divided by story height (3658 mm).
Further explanations for the finite element analysis are available in
Moradi and Alam [36].
3.1. Validation of finite element analysis
The simulations included creating the solid models, meshing, and
defining contact surfaces. Then, pretension forces of 230 and 88.74 kN
were applied to the bolts and PT strands, respectively. Afterwards,
displacement loads were applied to the column flange as next load
steps. Nonlinear static analyses with automatic Newton–Raphson
procedure were performed. To take into account the geometric
nonlinearity, large-deformation effects were considered in the analysis.
In order to assess the accuracy of the finite element analysis, the
numerical results for two PT connection models are compared with
the experimental test results in [29]. Fig. 4 shows the numerical lateral
load-displacement response for specimens PC3 and PC4 alongside the
experimental test results. A comparison of the numerical and test
results for the initial stiffness (Ki), residual stiffness (Kr), and maximum
lateral load (Fmax) are also given in Table 1 (see Fig. 5 for the response
parameters). These comparisons indicate that the finite element
analysis can accurately capture the cyclic response of PT connections
in terms of stiffness, strength, residual displacement, and dissipated
energy. The modeling and validation of a very similar finite element
analysis approach (but with solid elements for PT strands) can be
found in Moradi and Alam [36].
Fig. 5. Schematic of the lateral load-displacement response with the response variables
considered in the sensitivity study.
Table 2
Factors and levels considered for the sensitivity analysis.
Factors Symbol Low level (−) High level (+) Units
Angle thickness A 12.7 19.05 mm
Gage length B 57 103 mm
Strand post-tension force C 65 131 kN
Angle yield strength D 236 383 MPa
Reinforcing plate thickness E 10 20 mm
Reinforcing plate length F 254 406 mm
Bolt yield strength G 827 930 MPa
Bolt pretention force H 454 520 kN
4. Sensitivity analysis
In order to assess the sensitivity of the cyclic response of PT
connections to potentially varying design parameters, a sensitivity
analysis was conducted using the validated finite element model.
Eight potential factors were considered in a design-of-experiment
framework. Four of the select factors are geometry-related (including
thickness of angle section, A; gage length, B; thickness and length of
reinforcing plates, denoted as E and F, respectively) and others are due
to variability in material properties (namely, the amount of post-
tension force in the strands, C; yield strength of steel angles, D; yield
strength of high strength bolts, G; and pretension force in the bolts,
H). The geometric factors are illustrated in Fig. 1(b). The gage length
(factor B) is defined as the distance between the fillet of the angle
vertical leg and the edge of the bearing plate. For each factor, two levels
were considered in the design. Table 2 lists these factors along with
their low (−) and high levels (+). These levels were selected in view
of practical ranges for PT connections with top-and-seat angles. In this
sensitivity study, the beam and column sections were kept constant
(the same sections as in the reference test specimens) although the
connection performance is highly dependent on the beam size [30].
For the sensitivity analysis models, a few adjustments were
considered in the specifications of the validated finite element model:
(1) the beam flange reinforcing plates were used outside of the beam
section; (2) A490 high strength bolts of diameter 32 mm were used,
which permitted a greater variation for the bolts strength and
pretension force when choosing the levels (or treatments). The purpose
of making these minor adjustments was to generate more general
details for PT connections.
4.1. Initial sensitivity analysis
The Design-Expert software [37] was used to generate two-level
fractional factorial designs. A full factorial design with two levels for
eight factors involves 28 (=256) models, which is computationally pro-
hibitive. Hence, a sequential design approach, consisting of smaller
computer experiments (number of models) was employed. In this
initial experimental design, a fraction (one-eighth) of the full design
was used, which only requires 28 − 3 = 32 models. The treatment
combinations (models) for this fractional factorial design (of resolution
IV) are provided in Table A.1. Note that 32 (or 25) runs contain a full
factorial or basic design for five factors (creating all possible combina-
tions of + and −). Since we have 8 factors, (8 − 5=) three generators
are required to assign the low or high level to these three factors,
thereby constructing the design in Table A.1. These three generators
are: F = ABC, G = ABD, H = BCDE. For example, the F column in
model 1 is obtained by multiplying columns A, B, and C in the table of
plus and minus signs. That is, F = (− 1)(+ 1)(− 1) = + 1, which
represents the high level for factor F. The selected generators results in
a minimum abbreviation design with the highest possible resolution.
The details of such designs for different numbers of factors have been
tabulated in reference design-of-experiment books, for instance in
 S. Moradi et al. / Journal of Constructional Steel Research 112 (2015) 155–166
Fig. 6. Lateral load-displacement response for models 1, 8, and 29.
[38]. For each model in Table A.1, a finite element model was developed
and analyzed under cyclic loading. As the output of each simulation, five
different response characteristics were recorded, while examining the
lateral load-displacement response. These cyclic response quantities
include: (1) the initial stiffness, Ki; (2) residual stiffness, Kr (defined
as the stiffness after gap opening at the connection interface). Kr was
calculated from the slope of a line on the force-displacement response
between 1.5% and 2.5% drifts; (3) load capacity, Fmax (which is the
maximum lateral force generated at the column base when the drift
reached 3.5%); (4) hysteretic energy dissipation capacity (calculated
as the total area enclosed by the hysteretic curves), denoted by HED;
and (5) residual displacement, taken as the maximum displacement
exhibited at the end of loading when the horizontal load is zero.
Residual displacement is denoted by Rdisp. The response variables are
depicted in Fig. 5. As an example for the procedure, the tenth model
(in Table A.1) is the analysis of a finite element model when all the
considered factors are at their low levels, except factor H (the bolt
pretension force). The results of the initial fractional factorial design
are presented in the next section.
4.1.1. Results of initial sensitivity analysis
For each model in the initial design (with 32 models), load-
displacement response curves were generated to quantify the cyclic
response of PT connections. A different range of cyclic performance
was observed from these models. For example, the cyclic response
curves for three models 1, 8, and 29 are presented in Fig. 6. The model
with the factor combination 1 (i.e. model 1 in Table A.1) displays only
2.94 mm residual displacement, showing better self-centering behavior
than models 8 and 29 which exhibit a permanent displacement of 21.31
and 10.63 mm, respectively. The response quantities for each model are
provided in Table A.2.
In order to identify important factors and interactions, which have
significant effects on the response, a half-normal probability plot was
created. This method is based on the theory that unimportant factor
effects are normally distributed with a near-zero mean value. The nor-
mality of the estimated effects is supported by the Central Limit
Theorem, which states that a distribution of averages tends to follow a
normal distribution. Therefore, in a plot of absolute value of effect
estimates versus their cumulative normal probabilities, unimportant
factors fall on a straight line [38]. The standardized effects, on the
horizontal axis of the probability plot, are essentially the effects adjusted
by their standard error values. In a two-level factorial design, the effect
of a factor (or an interaction) is the average of response when the factor
is at its high level minus the average of response at the low level of that
factor. If the effect estimate of one factor is altered depending on the
level of another factor, it indicates an interaction between the two
factors. Further background on factorial design is available in [38]. The
half-normal probability plots of the effects for two of the response
quantities (as examples) are shown in Fig. 7.
Using diagnostic tools in Design-Expert software, it was observed
that the Rdisp data to some extent has departure from the normality as-
sumption. Therefore, before analyzing the data, the Rdisp response data
159
was transformed using an inverse square root transformation (pffiffiffiffiffiffiffiffi
1
).
Rdisp
This type of transformation was selected by trying different transforma-
tion alternatives and checking the results. The Box–Cox method was
also used as a guide to choose the best transformation [38].
In the half-normal probability plots (Fig. 7), factors that have a
positive effect on the response are shown with an orange color, whereas
factors with negative effects have a blue color. The significant main
effects associated with the initial stiffness response are: the reinforcing
plate length and thickness (F and E, respectively), post-tension force in
the strands (C), angle thickness (A), and gage length (B). Similarly, other
plots were used to distinguish important factors for each response. At
this stage of the study, it can be stated that factors G and H (bolt yield
strength and pretension force, respectively) do not have significant
effects on the response parameters.
The results of the initial design demonstrate that, in addition to
significant main effects, there are active interactions. Among the impor-
tant interaction effects, those involving factor A, such as AB, AC, AD, AF,
and AG appear to be more noticeable. However, some of the two-
factor nteractions are aliased with each other due to the fractional as-
pect of the factorial design. In fact, due to the alias, the interaction AF
(for example) represents AF + BC. Thus, from the initial design, it is
not possible to know which interaction effect is influencing the
response. The important effect might be AF or BC or both. In order to
separate (de-alias) the two-factor interaction effects involving factor
A, a secondary factorial design was therefore generated. A complete
discussion of the results is provided in the following sections, where a
combination of both the initial and secondary designs is considered.
4.2. Secondary sensitivity analysis and results: a partial fold-over
As addressed above, in order to distinguish between two-factor
interactions involving factor A, a secondary factorial design was
constructed. The new results were added to the initial group of models.
With this aim, a “partial (semi) fold-over” technique was used to
generate a new set of experimental designs. The partial fold-over design
was to de-alias effects of interest with fewer runs [38]. Table A.3
provides the secondary design, which has 16 models. The response
quantities obtained from each model are listed in Table A.4. Knowing
that factor H is not significantly affecting the response parameters, its
lower-level value was retained in the secondary design. This is part of
the procedure in a semi-fold over experimental design, which leads to
a fewer number of models.
5. Results and discussions
For each response variable, important main factors and interaction
effects were determined using half-normal probability plot. Percentage
contributions of factors are used to demonstrate their importance in a
relative way. Fig. 8 summarizes the obtained percentage contributions
with respect to each response variable. Percent contributions are
calculated as the ratio of the sum of squares of each given factor to the
total sum of squares [38].
160 S. Moradi et al. / Journal of Constructional Steel Research 112 (2015) 155–166
Fig. 7. Half-normal probability plot for (a) the initial stiffness and (b) residual stiffness.
By performing an analysis of variance (ANOVA) for each response
variable using statistical tools in Design-Expert software, the results
from half-normal probability plots were also confirmed. ANOVA is a
technique to statistically test the null hypothesis of no treatment effects
[38]. The significance level chosen for this ANOVA study is 5%. Hence,
the most significant factors have p-values less than 0.05. Fig. 9 shows
p-values associated with each response for the main factors. As seen in
this figure, main factors A (angle thickness) and C (post-tension force)
with p-values less than 0.05, significantly affect all the characteristics
of the cyclic response of PT connections. Further, the p-values for factors
G (bolt yield strength) and H (bolt pretension force) are greater than
0.05, indicating that these factors do not have any significant effect on
the response variables.
5.1. Stiffness of the PT steel connections
In this section, the sensitivity analysis results are examined with re-
gard to the initial and residual stiffness of PT connections (Ki and Kr in
Fig. 5). As shown in Fig. 8(a), the most important factors affecting the
initial stiffness of PT connections are the length and thickness of rein-
forcing plates (F and E), post-tension force in PT strands (C), and thick-
ness of angles (A), with percent contributions of 43.3%, 31%, 17%, and
5.7%, respectively. The next important effect is the interaction between
length and thickness of reinforcing plates, having a contribution percent
of 2.3%. The remaining factors, shown in Fig. 8(a), contribute less to the
initial stiffness response of PT connections (less than 1% each). In Fig. 8,
positive effects are marked with a “+” sign, suggesting that the re-
sponse is directly proportional to the factor. As an example, by increas-
ing the length of reinforcing plates (F) from the low level (254 mm) to
the high level (406 mm), the initial stiffness response is increased. The
overall stiffness of the PT connection is inversely affected by the gage
length (B). However, the Ki response is not very sensitive to the gage
length. The effect of varying the two most significant factors (F and E)
on the Ki response is illustrated in Fig. 10. The vertical axis in this 3D
plot is the response characteristic (i.e. Ki). It can be observed that the
initial stiffness (Ki) increases as the length and thickness of reinforcing
plates increase. The plot also shows that the PT connection possesses a
maximum initial stiffness of 16348 kN/m when all the important factors
with positive effects (F, E, C, and A) are at their high levels and gage
length (B) has a low value of 57 mm.
The residual stiffness response (Kr) is denoting the stiffness of a
PT connection after the gap opening between the beam and column
initiates. The sensitivity study shows that the residual stiffness of a
PT connection is mostly affected by the post-tension force (C), gage
length (B), reinforcing plate length (F), and angle thickness (A).
The percent contributions for these important factors are 55.2%,
24.3%, 8.1%, and 6%, respectively. As shown in Fig. 8(b), the reinforc-
ing plate length (F) has a positive effect on the Kr response. That is,
the residual stiffness increases for a PT connection with longer rein-
forcing plates.
The 3D plot in Fig. 11 illustrates the effect of interaction BC on the
residual stiffness when factors A and F are at their high levels. This
factor combination shows a peak residual stiffness (Kr) of 1680 kN/
m with low levels of the gage length (B) and post-tension force (C).
As seen in Fig. 11, the residual stiffness decreases with the increase
in the post-tension force. This trend is attributed to the damage in
the beam flanges when the post-tension force is high. After the gap
opening occurs in the beam–column interface, the tensile force in
the PT strands increases, which results in the beam shortening. In
the case of higher post-tension forces, this beam shortening is
more pronounced as the beam undergoes greater compressive
forces. Consequently, a reduction of the residual stiffness is observed
for PT connections with a high post-tension force. The sensitivity
analysis also shows that the residual stiffness of a PT connection de-
creases with the increase in the gage length (B) or decrease in the
angle thickness (A). This can be expected as the flexural deforma-
tions of the angles dominate when the ratio between gage length
and angle thickness is large.
5.2. Load bearing capacity of the PT steel connections
The peak horizontal load on the column was monitored as a re-
sponse characteristic, representing the load bearing capacity of PT con-
nections. As illustrated in Fig. 8(c), two important factors influencing
the load bearing capacity (Fmax response) of PT connections are gage
length (B) and angle thickness (A). The former (factor B) with 39.1%
contribution has a negative impact on the Fmax response, while angle
thickness directly affects Fmax. This trend can be attributed to the fact
that the shear deformations of angles (rather than flexural) dominate
the overall response for a connection with smaller ratio of gage length
to the angle thickness. These observations are also consistent with the
findings from experimental test results in [29]. They reported that the
load capacity and energy dissipation of PT connections increased with
the increase in angle thickness or decrease in the gage length.
To examine the effect of interaction between the main factors A and
B on the variability of Fmax response, the 3D plot in Fig. 12 is presented.
It can be observed that the Fmax response increases by approximately
15–20% as the gage length (B) decreases from its high level (103 mm)
to low level (57 mm). Since interactions are involved, the exact amount
of change in the response as a result of varying gage length is different.
That is, this response variability depends on the level of another
important effect. A peak Fmax response of approximately 452 kN/m is
 S. Moradi et al. / Journal of Constructional Steel Research 112 (2015) 155–166 161

Fig. 8. Percentage contributions of important factors and interactions associated with (a) Ki response, (b) Kr response, (c) Fmax response, (d) HED response, (e) inverse square root of the
residual displacement response.

predicted, based on the experimental design, for a combination when all effects on the Fmax response, are 30.3%, 16.9%, 5.7%, and 3.9%, respec-
the important main factors are at their high levels, except gage length tively. The interaction AB has a 1.1% contribution, while the other effects
with its low value. are small (less than 1%).
The percentage contribution plot [Fig. 8(c)] shows that the Fmax
response is also sensitive to the post-tension force (C), yield strength
of angles (D), and the length of reinforcing plates (F). The percentage
contributions obtained for factors A, C, D, and F, which have positive

Fig. 10. The effect of interaction between reinforcing plate length and thickness on the Ki
Fig. 9. P-values associated with each response for the main factors. response.
162 S. Moradi et al. / Journal of Constructional Steel Research 112 (2015) 155–166
Fig. 11. The effect of interaction between gage length and post-tension force on the Kr
response.
5.3. Energy dissipation of the PT steel connections
The hysteretic energy dissipation (HED) of PT connections is mostly
affected by the properties of steel angles, including gage length (B),
angle thickness (A), and angle yield strength (D). This fact can be
expected because energy in a PT connection is mainly dissipated
through the deformation of steel angles. The calculated percentage
contributions for important factors show that these factors (i.e. B, A,
and D) account for 42.3%, 35.6%, and 9.0% of the total response variabil-
ity, respectively. As shown in Fig. 8(d), the HED response also appears to
be affected by post-tension strand force (C), and interactions AB and BD
with percent contributions of 3.9%, 3.7%, and 1.6%, respectively. The
effects of other factors, such as reinforcing plate thickness and length
(E and F) are small (less than 1%).
Depending on the factors combination, a wide range of HED
response from 39.1 to 226.9 kN·m was observed. As with the Fmax
response, the PT connections with smaller gage length (B) and larger
thickness of angle (A) dissipate greater energy. Fig. 13 illustrates how
the AB interaction influences the HED response. For a PT connection
with a gage length of 57 mm and reinforcing plate length of 254 mm,
a peak HED of 215.7 kN·m is predicted, when the other important
main factors are at their high levels.
5.4. Residual displacement of the PT steel connections
The residual displacement response (Rdisp) of the PT connection
models varies from 2.78 mm to 47.82 mm. The high ratio of the
maximum to minimum Rdisp response was suggesting the need for
Fig. 12. The effect of interaction between gage length and angle thickness on the Fmax
response.
Fig. 13. The effect of interaction between angle thickness and gage length on the HED
response.
transforming the Rdisp data. This was also confirmed when checking
the statistical assumptions underlying the data analysis, including
normality and equal variances. Therefore, the inverse square root of
Rdisp data was used for the sensitivity analysis.
Fig. 8(e) shows contributions of important factors associated with
the transformed Rdisp response. The most important factors influencing
the Rdisp response are gage length (B), angle thickness (A), and post-
tension force (C). In addition, yield strength of angles (D) and length
of reinforcing plates (F) are among the important factors. The percent-
age contributions of main factor effects B, A, C, D, and F are 34.8, 21.7,
18.4, 3.8, and 3.5%, respectively. In addition, interactions BC, AC, and
AB appear to be important with percent contributions of 6.4, 3.5, and
1.6%, respectively. Since interactions are important, the effect of each
factor should be studied while taking into account the level of other
effects. Fig. 14 illustrates the effect of interaction BC on the Rdisp re-
sponse. This 3D plot is obtained for a design combination which displays
a maximum residual displacement of 47.8 mm. That is, a design with
high values of A and D, along with a low value for F. As seen in Fig. 14,
an increase in factor C (post-tension force) results in a lower Rdisp
response (by 4%) when factor B is at its low level. On the other hand,
when B has a high value, the Rdisp response increases (by 110%) with
the increase in post-tension force. This opposite trend is due to the
interaction between B and C.
Overall, with the increase in the gage length (factor B) from 57 mm
to 103 mm, the residual deformation reduces by 24–85%, depending on
the other factors. An opposite trend was observed for angle thickness
(factor A). The residual displacement of the PT connections increased
up to four times when varying the angle thickness from 12.7 mm to
Fig. 14. The effect of interaction between gage length and post-tension force on the Rdisp
response.
 S. Moradi et al. / Journal of Constructional Steel Research 112 (2015) 155–166 163

Fig. 15. Distributions of von Mises plastic strain: (a) model 38 alongside (b) model 46 (plots were magnified by 4).

19.05 mm. Therefore, a larger ratio of gage length to angle thickness
might lead to reduction in the amount of residual deformations.
However, load capacity and energy dissipation of the connection will
decrease, as discussed earlier. This observation can be attributed to the
fact that thinner angles with a large gage length provide a low flexural
stiffness. Consequently, when load is applied to the PT connection,
angles would be the main source of energy dissipation, undergoing
inelastic deformations before other structural elements. Therefore,
beams would remain elastic and damage in the beam would be
prevented. To further clarify this statement, Fig. 15 compares the distri-
butions of von Mises plastic strains in two of the finite element models
at the end of loading. The PT connection in model 38, which has thick
angles with a small gage length, displays a large residual displacement
of 47.82 mm. In contrast, the connection in model 46 with thin angles
and a large gage length exhibits only 4.93 mm residual displacement.
As evident in Fig. 15, damage in the beam is significantly mitigated for
connection 46 with a higher ratio of gage length to angle thickness.
The sensitivity analysis also shows that PT connections with angles
of lower yield strength (= 236 MPa) can lead to a decrease in the
Rdisp response. The reduction in the response upon reducing the angle
yield strength is approximately 15–43%, depending on the other factors.
This reduction is due to the fact that the angle sections with lower yield-
strength would resist the load by experiencing deformations before any
other components of the connection, such as beams. However, it should
be noted that percentage contribution of factor D is not significant
(about 3.8%).
Based on the results, it can be observed that connections with longer
reinforcing plates exhibit 14–42% lower residual displacements. As
expected, the use of long reinforcing plates is beneficial in
preventing damage in the beams, and consequently, reducing residual
deformations.
6. Verifications of results and predictive equations
From the data points in the latter factorial design, regression
equations were developed to quantitatively express a relationship
between important factors and each response characteristic. Eq. (1) is
the general form of this type of equations relating important design
factors and the response variables, including initial stiffness (Ki), resid-
ual stiffness (Kr), maximum horizontal force (Fmax), dissipated energy
(HED), and residual displacement (Rdisp). Table 3 provides the regres-
sion coefficients β0, β1, …, β12 corresponding to each response. The re-
gression models contain actual values of each factor (not the coded
values of −1, +1). Given that statistically significant factors are includ-
ed in the regression models, high values of the coefficient of determina-
tion (R2) imply the goodness of regression models.
Response ¼ β0 þ β1 A þ β2 B þ β3 C þ β4 D þ β5 E þ β6 F þ β7 AB
ð1Þ
þ β8 AC þ β9 AD þ β10 BC þ β11 BD þ β12 E F
These regression equations can now be used to predict the response
parameters at different untried points (i.e. to predict the response for
different factor combinations which were not considered in the factorial
designs). In order to assess the precision of the fitted equations, eight
additional finite element models were created and analyzed in ANSYS.
Table A.5 summarizes these verification models. The response quanti-
ties obtained from finite element analysis are presented in Table A.6.
The ratios between the response quantities from the finite element

Table 3
Coefficients of regression models for each response variable.

Response variable
pffiffiffiffiffiffiffiffiffiffiffiffi
Coefficient Factor Ki (kN/m) R2 = 0.998 Kr (kN/m) R2 = 0.964 Fmax (kN) R2 = 0.997 HED (kN·m) R2 = 0.992 1= Rdisp (mm−0.5) R2 = 0.95

β0 – 12,508.315 1513.668 67.051 −147.785 0.781

β1 A 73.211 45.233 14.797 9.787 −0.053
β2 B −0.746 −10.438 −0.474 2.084 3.536E−03
β3 C 6.578 −6.134 0.727 0.136 −1.682E−03
β4 D 0 0 0.084 0.16 −3.055E−04
β5 E −10.093 0 0.952 0.718 0
β6 F 1.416 1.057 0.132 −0.042 2.822E−04
β7 AB −0.221 0 −0.072 −0.123 1.980E−04
β8 AC −0.172 −0.239 −0.035 0.029 2.048E−04
β9 AD 0 0 0.011 0.019 0
β10 BC 0.033 0.045 5.869E−03 −3.951E−03 −3.821E−05
β11 BD 0 0 −1.089E−03 −3.450E−03 0
β12 EF 0.206 0 0 0 0
164 S. Moradi et al. / Journal of Constructional Steel Research 112 (2015) 155–166
Table 4
Comparisons of the response quantities from finite element analysis and regression models.
Model (Ki)FEM/(Ki)Reg. (Kr)FEM/(Kr)Reg.
49 1.000 0.919
50 1.000 1.126
51 1.002 1.029
52 0.999 0.993
53 0.997 0.970
54 1.008 1.063
55 1.001 0.969
56 1.006 0.994
simulations and those using the regression equations are listed in
Table 4. As seen in this table, the regression equations can predict all
the response characteristics with an acceptable accuracy: The average
values of the ratio between the finite element and predicted response
quantities are 1.00, 1.01, 1.00, 0.98, and 0.99, respectively, for initial
stiffness, residual stiffness, maximum load capacity, dissipated energy,
and residual displacement response. It should be noted that these re-
gression equations can be only useful for post-tensioned connections
with W24 × 62 beam and W14 × 311 column sections (which were
kept constant in the sensitivity study).
7. Summary and conclusions
The effects of eight different design factors on the cyclic response
of PT steel beam–column connections with top-and-seat angles were
assessed in this paper. These factors included geometric and material
properties of steel angles (thickness, gage length, and yield
strength); length and thickness of reinforcing plates; post-tension
force in high strength steel PT strands; bolt pretension force and
bolt yield strength. The sensitivity study was conducted using frac-
tional factorial design-of-experiment principles so as to evaluate
the effects of any potential interactions between the factors as well
as main factor effects, with a minimal number of models. First, 3D fi-
nite element models for PT beam–column connections were devel-
oped in ANSYS. The finite element analysis results were then
verified using past experimental test results. Two-level fractional
factorial designs were generated to sequentially identify important
effects. Fifty-six finite element models were performed to do the
screening and to verify the results. The cyclic force-displacement re-
sponse of PT connections was evaluated in terms of stiffness, maxi-
mum load capacity, hysteretic energy dissipation, and residual
displacement. The following conclusions can be summarized:
• By varying the bolt yield strength (from 827 to 930 MPa) and the bolt
pretension force (from 454 to 520 kN), it was observed that the cyclic
performance of a PT connection is not significantly affected. The
sensitivity study showed that the variability in the response quantities
was generally below 1% for the ranges considered for the bolt yield
strength and its pretension force.
• The initial stiffness of a PT connection was mostly influenced by
reinforcing plate length and thickness, post-tensioning force, and
the thickness of angles. These factors account for about 97% of the
total variability in the initial stiffness response. Further, the initial stiff-
ness of a PT connection increased with the increase in any of these
factors.
• The residual stiffness of a PT connection (stiffness after the gap
opening at the beam–column interface) was mostly sensitive to the
post-tension force in the strands, gage length, reinforcing plate length,
and angle thickness. These factors account for about 94% of the total
variability in the residual stiffness response. Higher residual stiffness
was observed for PT connections with a lower post-tension force or
a smaller ratio of gage length to angle thickness. Further, the residual
(Fmax)FEM/(Fmax)Reg. (HED)FEM/(HED)Reg. (Rdisp)FEM/(Rdisp)Reg.
0.995 0.991 0.993
1.002 1.034 0.970
1.009 0.996 1.043
1.004 0.986 1.013
0.985 1.079 0.958
1.007 0.898 0.957
1.009 0.972 0.980
0.993 0.892 0.999
stiffness increased when longer reinforcing plates were used.
• The load capacity of a PT connection was found to be sensitive to gage
length, angle thickness, post-tensioning force, yield strength of angles,
and reinforcing plate length. These factors account for about 96% of
the total variability in the load capacity response. With the exception
of gage length, all these factors have a positive effect on the load
capacity of PT connections.
• The most important factors influencing the hysteretic energy dissipa-
tion of a PT connection were the gage length, angle thickness, angle
yield strength, and post-tensioning force. These factors account for
nearly 91% of the total variability in the response. Energy dissipation
of a PT connection increases with the reduction in gage length,
whereas an increase in angle thickness, angle yield strength, or post-
tensioning force can lead to improved energy dissipation capacity.
• The three most important factors affecting the residual displacement
of a PT connection were gage length, angle thickness, and post-
tensioning force. Increasing the gage length from 57 mm to 103 mm
resulted in approximately 24–85% decrease in the residual displace-
ment of the PT connection. Further, thin angles could reduce the
residual deformation of a PT connection up to four times. The reason
for these trends is that the deformation of a PT connection is more
likely confined to thin angles when a larger gage length is provided.
Thus, damage in the other elements such as beams is prevented. It
was also shown that the interactions between gage length, angle
thickness, and post-tensioning force play important roles in the self-
centering capability of PT connections. Additionally, PT connections
with angles of lower yield strength and also connections with longer
reinforcing plates experience less residual displacements.
• The predicted responses based on the developed regression equations
were very close to the FE verification models.
The results of this study may be useful to designers and engineers as
they provide more insight into understanding the effects of a number of
design parameters on the cyclic performance of PT connections. The
results would be also valuable for future research aimed at optimizing
the performance of these kinds of connections and ultimately enhanc-
ing the seismic performance of steel PT frames. However, future studies
can be conducted to further generalize the results of the sensitivity
analysis by including the effects of other factors, such as the beam and
angle sections, which were deemed to be constant in the presented
research. It should be also highlighted that the results and the estimated
effects derived in this study can be dependent on the selected ranges for
the considered parameters.
Acknowledgments
The financial contribution of the Natural Sciences and Engineering
Research Council of Canada (NSERC) through Discovery Grant (2010-
2015) was critical to conduct this research and is gratefully acknowl-
edged. We would also like to acknowledge CMC Microsystems for the
provision of products and services that facilitated this research, includ-
ing ANSYS Multiphysics.
 S. Moradi et al. / Journal of Constructional Steel Research 112 (2015) 155–166 165

Appendix A Table A.3

Factor combinations (models) for the secondary design.

Table A.1 Factors

Initial 28 − 3 fractional factorial design.
Model A B C D E F G H
Factors (mm) (mm) (kN) (MPa) (mm) (mm) (MPa) (kN)

Model A B C D E F G H 33 12.7 57 65 383 10 406 827 454

(mm) (mm) (kN) (MPa) (mm) (mm) (MPa) (kN) 34 12.7 103 65 236 10 254 827 454
35 19.05 57 131 236 10 406 827 454
1 12.7 103 65 236 20 406 930 520
36 19.05 103 131 383 10 254 827 454
2 19.05 103 65 236 20 254 827 520
37 19.05 57 65 236 20 254 827 454
3 12.7 103 65 383 20 406 827 454
38 19.05 57 65 383 10 254 930 454
4 19.05 57 131 383 10 254 827 520
39 19.05 103 131 236 20 254 930 454
5 19.05 103 65 383 20 254 930 454
40 12.7 57 131 383 20 254 827 454
6 12.7 103 131 383 20 254 827 520
41 19.05 103 65 236 10 406 930 454
7 19.05 57 65 236 20 406 930 454
42 12.7 57 65 236 20 406 930 454
8 19.05 57 131 236 20 254 930 520
43 12.7 103 131 236 20 406 827 454
9 12.7 57 131 383 10 406 930 520
44 19.05 57 131 383 20 406 930 454
10 12.7 57 65 236 10 254 827 520
45 12.7 103 131 383 10 406 930 454
11 19.05 103 131 236 10 406 827 520
46 12.7 103 65 383 20 254 930 454
12 12.7 57 131 236 10 406 827 454
47 19.05 103 65 383 20 406 827 454
13 19.05 103 131 383 20 406 930 520
48 12.7 57 131 236 10 254 930 454
14 19.05 103 131 236 20 406 827 454
15 19.05 103 65 236 10 254 827 454
16 12.7 103 131 383 10 254 827 454
17 12.7 57 65 383 20 254 930 520
18 12.7 103 65 383 10 406 827 520 Table A.4
19 12.7 57 65 383 10 254 930 454 Response quantities for the secondary design.
20 19.05 57 65 383 10 406 827 454
21 12.7 103 65 236 10 406 930 454 Response Quantities
22 12.7 103 131 236 10 254 930 520
Model Ki (kN/m) Kr (kN/m) Fmax (kN) HED (kN∙m) Rdisp (mm)
23 19.05 103 65 383 10 254 930 520
24 19.05 57 131 236 10 254 930 454 33 14,904.87 1424.55 336.77 95.72 5.71
25 12.7 57 131 236 20 406 827 520 34 14,399.49 1046.97 244.63 40.39 4.49
26 19.05 57 131 383 20 254 827 454 35 15,595.59 1137.59 411.37 145.92 16.76
27 12.7 57 65 236 20 254 827 454 36 15,076.91 731.74 353.56 109.02 11.86
28 19.05 57 65 236 10 406 930 520 37 15,183.71 1550.19 373.15 135.37 27.38
29 19.05 103 131 383 10 406 930 454 38 14,741.75 1478.76 400.99 172.82 47.82
30 12.7 57 131 383 20 406 930 454 39 15,451.43 727.58 337.3 93.11 11.89
31 12.7 103 131 236 20 254 930 454 40 15,282.85 856.37 365.95 123.89 15.12
32 19.05 57 65 383 20 406 827 520 41 15,122.29 1231.08 296.84 70.54 3.68
42 15,668.85 1446.99 322.32 76.88 5.29
Table A.2 43 16,121.92 950.27 327.32 48.83 7.77
Response quantities for the initial design. 44 16,356.48 997.02 450.94 218.81 30.65
45 15,399.49 897.98 330.27 57.49 7.75
Response quantities 46 14,821.4 1098.22 269.84 54.82 4.93
Model Ki (kN/m) Kr (kN/m) Fmax (kN) HED (kN∙m) Rdisp (mm) 47 15,825.22 1302.79 340.18 93.7 6.21
48 14,833.06 815.28 336.25 89.49 11.02
1 15,614.36 1155.6 265.06 41.48 2.94
2 15,017.31 1203.94 293.92 76.59 6.13
3 15,618.55 1169.98 282.51 53.14 3.64
4 15,079.64 983.73 420.5 211.7 38.82 Table A.5
5 15,010.02 1193.74 320.54 96.81 7.78 Verification models.
6 15,294.7 652.95 316.02 69.59 10.18
Factors
7 16,019.68 1706.26 395.92 130.53 15.94
8 15,495.53 1057.41 399.15 166.58 21.31 Model A B C D E F G H
9 15,388.55 1053.23 379.94 109.69 11.83
10 14,388.92 1293.55 293.3 76.22 6.93 49 12.7 57 131 236 20 254 930 454
11 15,624.02 956.29 351.38 80.28 9.4 50 12.7 103 131 383 20 406 930 454
12 15,365.41 1069.85 357.46 81.99 9.43 51 19.05 103 131 236 10 406 930 454
13 16,331.15 928.16 384.19 111.46 11.84 52 12.7 57 65 383 10 254 827 520
14 16,308.91 1007.11 362.14 85.76 10.22 53 19.05 103 65 236 20 406 827 520
15 14,622.93 1166.51 285.54 72.67 5.27 54 15.875 80 98 309.5 15 330 878.5 487
16 14,868.42 647.16 309.41 64.91 9.38 55 19.05 57 65 383 10 254 827 520
17 14,887.92 1368.4 329.19 105.2 11.04 56 14.29 61.6 120.71 309.5 16.5 307.2 873.35 516.7
18 14,905.05 1103.8 269.85 50.56 2.86
19 14,417.9 1324.22 319.79 99.49 9.38 Table A.6
20 15,251.69 1725.29 419.64 165.79 38.83 Response quantities for the verification models (obtained from finite element analysis).
21 14,891.93 1083.73 253.3 39.15 2.78
22 14,868.96 636.33 295.56 53.19 8.55 Response quantities
23 14,638.6 1151.67 311.98 92.69 6.94
Model Ki (kN/m) Kr (kN/m) Fmax (kN) HED (kN∙m) Rdisp (mm)
24 15,056.68 1065.04 391.94 153.62 18.39
25 16,121.56 1117.55 368.9 88.65 11.85 49 15,270.46 821.16 343.54 95.23 12.07
26 15,483.32 974.66 426.81 226.92 44.55 50 16,135.78 950.03 341.06 61.92 8.73
27 14,859.12 1339.89 302.29 80.06 6.95 51 15,610.35 958.53 351.49 79.89 8.83
28 15,273.74 1636.74 382.7 123.83 14.3 52 14,415.16 1321.62 319.76 99.81 9.56
29 15,614 909.38 373.85 102.49 10.63 53 15,831.06 1292.84 309.94 74.17 4.57
30 16,134.68 1083.02 390.8 117.96 13.49 54 15,432.84 1192.73 345.86 91.47 8.92
31 15,279.75 643.06 302.07 56.97 9.21 55 14,761.98 1471.57 400.76 172.92 47.78
32 16,050.12 1767.8 428.87 178.82 42.1 56 15,461.64 1014.46 361.91 105 13.33
166 S. Moradi et al. / Journal of Constructional Steel Research 112 (2015) 155–166
References
[1] AISC. Seismic provisions for structural steel buildings. Chicago, Illinois: American In-
stitute of Steel Construction; 2010.
[2] Parker M, Steenkamp D. The economic impact of the Canterbury earthquakes.
Reserv Bank N Z Bull 2012;75:13–25.
[3] Brownlee G. State of Canterbury earthquake recovery. Off Website N Z Gov 2012.
http://beehive.govt.nz/speech/state-canterbury-earthquake-recovery.
[4] McCormick J, Aburano H, Ikenaga M, Nakashima M. Permissible residual deforma-
tion levels for building structures considering both safety and human elements.
Proc. 14th World Conf. Earthq. Eng., Beijing, China; 2008. p. 12–7.
[5] Uma SR, Pampanin S, Christopoulos C. Development of probabilistic framework for
performance-based seismic assessment of structures considering residual deforma-
tions. J Earthq Eng 2010;14:1092–111. http://dx.doi.org/10.1080/13632460903556509.
[6] Ruiz-García J, Miranda E. Evaluation of residual drift demands in regular multi-
storey frames for performance-based seismic assessment. Earthq Eng Struct Dyn
2006;35:1609–29. http://dx.doi.org/10.1002/eqe.593.
[7] Pampanin S, Christopoulos C, Nigel Priestley MJ. Performance-based seismic re-
sponse of frame structures including residual deformations. Part II: multi-degree
of freedom systems. J Earthq Eng 2003;7:119–47. http://dx.doi.org/10.1080/
13632460309350444.
[8] Pettinga D, Christopoulos C, Pampanin S, Priestley N. Effectiveness of simple ap-
proaches in mitigating residual deformations in buildings. Earthq Eng Struct Dyn
2007;36:1763–83. http://dx.doi.org/10.1002/eqe.717.
[9] Sahoo DR, Chao S-H. Stiffness-based design for mitigation of residual displacements
of buckling-restrained braced frames. J Struct Eng 2014:04014229-1–04014229-13.
http://dx.doi.org/10.1061/(ASCE)ST.1943-541X.0001204.
[10] Herning G, Garlock MM, Ricles J, Sause R, Li J. An overview of self-centering steel mo-
ment frames. Proc. 2009 Struct. Congr. — Don't Mess with Struct. Eng. Expand. Our
Role, American Society of Civil Engineers, 1801 Alexander Graham Bell Drive, Res-
ton, VA 20191-4400, United States, Austin, TX, United States; 2009. p. 1412–20.
[11] Hajjar JF, Sesen AH, Jampole E, Wetherbee A. A synopsis of sustainable structural sys-
tems with rocking, self-centering, and articulated energy-dissipating fuses. Boston,
Massachusetts: Department of Civil and Environmental Engineering Reports. Report
No. NEU-CEE-2013-01. Department of Civil and Environmental Engineering, North-
eastern University; 2013. http://hdl.handle.net/2047/d20003216.
[12] Chancellor N, Eatherton M, Roke D, Akbaş T. Self-centering seismic lateral force
resisting systems: high performance structures for the city of tomorrow. Buildings
2014;4:520–48. http://dx.doi.org/10.3390/buildings4030520.
[13] Alam M, Youssef M, Nehdi M. Seismic performance of concrete frame structures re-
inforced with superelastic shape memory alloys. Smart Struct Syst 2009;5:565–85.
[14] McCormick J, DesRoches R, Fugazza D, Auricchio F. Seismic vibration control using
superelastic shape memory alloys. J Eng Mater Technol 2006;128:294. http://dx.
doi.org/10.1115/1.2203109.
[15] Miller DJ, Fahnestock LA, Eatherton MR. Development and experimental validation
of a nickel–titanium shape memory alloy self-centering buckling-restrained brace.
Eng Struct 2012;40:288–98. http://dx.doi.org/10.1016/j.engstruct.2012.02.037.
[16] Moradi S, Alam MS, Asgarian B. Incremental dynamic analysis of steel frames
equipped with NiTi shape memory alloy braces. Struct Des Tall Spec Build 2014;
23:1406–25. http://dx.doi.org/10.1002/tal.1149.
[17] Yam MCH, Fang C, Lam ACC, Zhang Y. Numerical study and practical design of beam-
to-column connections with shape memory alloys. J Constr Steel Res 2015;104:
177–92. http://dx.doi.org/10.1016/j.jcsr.2014.10.017.
[18] Fang C, Yam MCH, Lam ACC, Xie L. Cyclic performance of extended end-plate con-
nections equipped with shape memory alloy bolts. J Constr Steel Res 2014;94:
122–36. http://dx.doi.org/10.1016/j.jcsr.2013.11.008.
[19] Moradi S, Alam MS. Feasibility study of utilizing superelastic shape memory alloy
plates in steel beam-column connections for improved seismic performance. J Intell
Mater Syst Struct 2014;26:463–75. http://dx.doi.org/10.1177/1045389X14529032.
[20] Erochko J, Christopoulos C. Self-centering energy-dissipative (SCED) Brace: overview of
recent developments and potential applications for tall buildings. Sustain Dev Crit
Infrastruct pp 2014;488–495. http://dx.doi.org/10.1061/9780784413470.053.
[21] Christopoulos C, Tremblay R, Kim H-J, Lacerte M. Self-centering energy dissipative brac-
ing system for the seismic resistance of structures: development and validation. J Struct
Eng 2008;134:96–107. http://dx.doi.org/10.1061/(ASCE)0733-9445(2008)134:1(96).
[22] Erochko J, Christopoulos C, Tremblay R. Design, testing, and detailed component
modeling of a high-capacity self-centering energy-dissipative brace. J Struct Eng
2014:04014193-1–04014193-12. http://dx.doi.org/10.1061/(ASCE)ST.1943-541X.
0001166.
[23] Gregory HK, Deierlein G, Ma Xiang, Eatherton Mathew, Sarah Billington JH. Collabo-
rative research on development of innovative steel braced frame systems with con-
trolled rocking and replaceable fuses. 6th Int. Conf. Urban Earthq. Eng., Tokyo; 2009.
p. 413–6.
[24] Eatherton MR, Ma X, Krawinkler H, Deierlein GG, Hajjar JF. Quasi-static cyclic behav-
ior of controlled rocking steel frames. J Struct Eng 2014;140:04014083. http://dx.doi.
org/10.1061/(ASCE)ST.1943-541X.0001005.
[25] Deierlein G, Krawinkler H, Ma X, Eatherton M, Hajjar J, Takeuchi T, et al. Earthquake
resilient steel braced frames with controlled rocking and energy dissipating fuses.
Steel Constr 2011;4:171–5. http://dx.doi.org/10.1002/stco.201110023.
[26] Dowden DM, Purba R, Bruneau M. Behavior of Self-centering steel plate shear walls
and design considerations. J Struct Eng 2012;138:11–21. http://dx.doi.org/10.1061/
(ASCE)ST.1943-541X.0000424.
[27] Clayton PM, Berman JW, Lowes LN. Subassembly testing and modeling of self-
centering steel plate shear walls. Eng Struct 2013;56:1848–57. http://dx.doi.org/
10.1016/j.engstruct.2013.06.030.
[28] Ricles JM, Sause R, Garlock MM, Zhao C. Posttensioned seismic-resistant connections
for steel frames. J Struct Eng 2001;127:113–21.
[29] Ricles JM, Sause R, Peng SW, Lu LW. Experimental evaluation of earthquake resistant
posttensioned steel connections. J Struct Eng 2002;128:850–9. http://dx.doi.org/10.
1061/(ASCE)0733-9445(2002)128:7(850).
[30] Garlock MM, Ricles JM, Sause R. Experimental studies of full-scale posttensioned
steel connections. J Struct Eng 2005;131:438–48. http://dx.doi.org/10.1061/
(ASCE)0733-9445(2005)131:3(438).
[31] Christopoulos C, Filiatrault A, Uang C-M, Folz B. Posttensioned energy dissipating
connections for moment-resisting steel frames. J Struct Eng 2002;128:1111–20.
http://dx.doi.org/10.1061/(ASCE)0733-9445(2002)128:9(1111).
[32] Rojas P, Ricles JM, Sause R. Seismic performance of post-tensioned steel moment
resisting frames with friction devices. J Struct Eng 2005;131:529–40. http://dx.doi.
org/10.1061/(ASCE)0733-9445(2005)131:4(529).
[33] Wolski M, Ricles JM, Sause R. Experimental study of a self-centering beam–column
connection with bottom flange friction device. J Struct Eng 2009;135:479–88.
http://dx.doi.org/10.1061/(ASCE)ST.1943-541X.0000006.
[34] Vasdravellis G, Karavasilis TL, Uy B. Large-scale experimental validation of steel
posttensioned connections with web hourglass pins. J Struct Eng 2013;139:
1033–42. http://dx.doi.org/10.1061/(ASCE)ST.1943-541X.0000696.
[35] ANSYS. ANSYS mechanical APDL, release 15.0; 2013.
[36] Moradi S, Alam MS. Finite element simulation of posttensioned steel connections
with bolted angles under cyclic loading. J Struct Eng 2015. http://dx.doi.org/10.
1061/(ASCE)ST.1943-541X.0001336 (in press).
[37] DX9. Design-Expert® software version 9; 2013.
[38] Montgomery DC. Design and analysis of experiments. Eighth ed. John Wiley & Sons,
Inc.; 2013