Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/282588494

Retrofit of Built-Up Steel Columns: Modeling and Fundamental Behavior

Article in Journal of Bridge Engineering · September 2015


DOI: 10.1061/(ASCE)BE.1943-5592.0000808

CITATIONS READS
5 1,450

3 authors, including:

Zhichao Lai Amit H Varma


Fuzhou University Purdue University
74 PUBLICATIONS 909 CITATIONS 330 PUBLICATIONS 4,195 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Light Water Reactor Sustainability, US DOE View project

Post-Fire Assessment of Prestressed Concrete Bridges in Indiana View project

All content following this page was uploaded by Zhichao Lai on 24 November 2015.

The user has requested enhancement of the downloaded file.


Retrofit of Built-Up Steel Columns: Modeling and
Fundamental Behavior
Zhichao Lai, Ph.D., A.M.ASCE1; Amit H. Varma, Ph.D., M.ASCE2; and Robert J. Connor, Ph.D.3

Abstract: Cover plates are often added to built-up steel columns as an effective retrofit method for increasing strength. For retrofitted
Downloaded from ascelibrary.org by Purdue University Libraries on 10/07/15. Copyright ASCE. For personal use only; all rights reserved.

columns, the original member carries all loads, whereas the added cover plates resist only the loads applied after their addition. As a
result, the original member may yield before the added cover plates. Although some specifications provide guidance for evaluating the
capacity of retrofitted columns, the approaches are conservative and do not address the behavior of retrofitted columns if portions of
the original cross section yield first because of preload. In this paper, the behavior and strength of retrofitted built-up columns, as well
as the influence of the cover plates, are evaluated using two analysis approaches: (1) three-dimensional finite-element analysis and (2)
nonlinear inelastic column buckling analysis. Both approaches include the effects of geometric imperfections, column slenderness,
cyclic loading, and temperature gradient on the behavior and strength of retrofitted columns. The failure modes of the retrofitted col-
umns depend on the column slenderness ratio (KL=r). For short columns, the failure mode consists of compression yielding of the entire
cross section, where the existing section yields first, followed by the cover plates. For slender columns, the failure mode consists of
inelastic flexural buckling about the weak axis. The existing section yields first, significantly reducing the flexural stiffness of the retro-
fitted section. DOI: 10.1061/(ASCE)BE.1943-5592.0000808. © 2015 American Society of Civil Engineers.
Author keywords: Retrofit; Strengthening; Built-up; Columns; Hybrid; Steel; Bridges; Design.

Introduction member section to undergo yielding before reaching strength. In this


method, cover plates are added to the webs of the existing column.
Existing structures sometimes require retrofitting or strengthening This method provides a more economical strengthening strategy
to satisfy increased strength demands or to repair damage after than replacing the original structural member or providing massive
extreme events, such as earthquakes and explosions. There are sev- new steel components to prevent yielding of the original member.
eral retrofit methods available. The selection of the retrofit method Retrofitted members are composite in the sense that they are
depends on several factors, including the retrofit objectives; type of composed of both existing components and new components (i.e.,
structure and component, defects, or deterioration of existing struc- cover plates). Locked-in stresses and strains are present in the exist-
ture; loads induced by the retrofit; and available access. The selec- ing components, whereas the entire composite member sustains
tion of the retrofit method may also depend on the structure service additional loads. There is limited research on the behavior and
life, financial analysis, and contractor expertise (Lai and Patel design of such composite members [the authors were able to find
2004). The retrofit of steel structures may include the replacement of only one related research project by Shek and Bartlett (2008)]. In
existing components or the addition of new components. For exam- that research, Shek and Bartlett developed a refined numerical anal-
ple, chord members in girders or steel-arch bridges are usually retro- ysis model (RNAM) to investigate the behavior of steel wide-flange
fitted by adding cover plates to the flanges or webs of the existing columns strengthened with cover plates added to the flanges. They
section. evaluated the effects of several parameters on the axial compressive
Retrofitting to prevent yielding may not be pragmatic. For exam- strength, such as residual stresses, locked-in stresses due to dead
ple, massive components are typically required to strengthen the load, yield stress of the original steel, cover-plate area, and cover-
original section to prevent yielding. These massive components plate aspect ratio. However, the effects of several other parameters
introduce significant additional dead load. To improve overall effi- were not addressed, such as the magnitude of geometric imperfec-
ciency, it is worthwhile to consider the acceptability of some yield- tions, cyclic loading, and temperature gradient. Moreover, the
ing of the original section before reaching strength. This paper applicability of the RNAM to columns with other cross-section
presents an alternative retrofit method that permits the existing shapes (other than the W-shapes) was not addressed.
In this paper, the behavior and strength of retrofitted built-up
1
Postdoctoral Research Engineer, Lyles School of Civil Engineering, columns are evaluated using two analysis approaches: (1) three-
Purdue Univ., West Lafayette, IN 47907 (corresponding author). E-mail: dimensional (3D) finite-element analysis using Abaqus and (2) non-
laiz@purdue.edu linear inelastic column buckling (NICB) analysis implemented in
2
Professor, Lyles School of Civil Engineering, Purdue Univ., West MATLAB 2011a. The fundamental behavior of retrofitted columns
Lafayette, IN 47907. E-mail: ahvarma@purdue.edu
3
was first investigated by analyzing a prototype of a heavy built-up
Associate Professor, Lyles School of Civil Engineering, Purdue Univ., double-box column strengthened using cover plates on the webs
West Lafayette, IN 47907. E-mail: rconnor@purdue.edu
(as shown in Fig. 1) using the finite-element analysis approach.
Note. This manuscript was submitted on September 12, 2014;
approved on May 15, 2015; published online on September 25, 2015. The NICB analysis program was then developed and bench-
Discussion period open until February 25, 2016; separate discussions must marked using results from the finite-element analyses. Both analy-
be submitted for individual papers. This paper is part of the Journal of sis approaches included the effects of geometric imperfections, col-
Bridge Engineering, © ASCE, ISSN 1084-0702. umn slenderness ratio (KL=r), cyclic loading, and temperature

© ASCE 04015054-1 J. Bridge Eng.

J. Bridge Eng., 04015054


Downloaded from ascelibrary.org by Purdue University Libraries on 10/07/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 1. Cross-section details of the prototype built-up double-box member

gradient on the behavior and strength of the prototype column. The modulus (Es cpl ) of the cover plates were 448.2 MPa (65 ksi)
benchmarked NICB analysis program can be used to (1) further and 200 GPa (29,000 ksi), respectively. Two member lengths
investigate the effects of several other parameters on the behavior [7,620 mm (300 in.) and 27,940 mm (1,100 in.)] were used to
and strength of the retrofitted prototype member, such as residual represent short and slender columns, respectively. The members
stresses, end moments, locked-in dead load ratio (PLD =Pori , the ratio were assumed to have pinned-end conditions for buckling about
of the locked-in dead load PLD to the axial compressive strength of both the strong and weak axes.
the existing member Pori ), cover-plate properties (yield stress Fy cpl ,
width bcpl , and thickness tcpl ), and column slenderness ratio (KL=r);
and (2) investigate the behavior of retrofitted columns with other FEM Model Details
cross-section shapes, such as box sections, channels, and angles.
Three-dimensional finite-element analysis was performed using
Abaqus. Both the existing member cross section and the cover
Finite-Element Analysis plates were modeled using S4R elements. These S4R elements have
(1) six degrees of freedom per node, (2) three section points to com-
pute stress and strain variations through the thickness, and (3)
Prototype Member Details
reduced integration in the plan of the elements. The elements model
Fig. 1 shows the cross section of the prototype built-up double-box thick-shell behavior, but converge to Kirchhoff’s thin-plate bending
member. This type of member is typically used as the compression theory with reducing thicknesses. For additional details regarding the
chord of long-span steel-arch bridges. As shown, the section con- choice of these finite elements, see Lai et al. (2014) and Lai and
sists of eight different types of plate (Plates A–H). Angles are used Varma (2015). The steel material multiaxial behavior was defined
to connect the plates into the desired cross-section shape using riv- using the von Mises yield surface, associated flow, and kinematic
ets. The height and width of the cross section are 1,994 mm hardening. An idealized bilinear curve, as shown in Fig. 2(a), was
(78.5 in.) and 1,505 mm (59.25 in.), respectively. Dimensions of the used to specify the steel uniaxial stress–strain behavior. The hardening
plates and angles are included in Fig. 1. The original member was modulus Et was assumed to be equal to Es ori =200 for the existing
strengthened using cover plates with a width of 1,270 mm (50 in.) member and Es cpl =200 for the cover plates.
and a thickness of 86 mm (3.375 in.). These cover plates were Geometric imperfections were incorporated into the models to
attached to the webs of the existing double-box member along its initiate buckling of the retrofitted column (Lai et al. 2014). For
entire length, as shown in Fig. 1. The means of attaching the cover each model, the shape of the geometric imperfection was obtained
plates were not investigated in this study, but are assumed to be suf- by conducting eigenvalue buckling analysis of the column, and the
ficient to transfer the forces between the cover plates and the origi- amplitude of the imperfection was assumed to be equal to
nal member. The yield stress (Fy ori ) and elastic modulus (Es ori ) of L=1; 500. This imperfection amplitude was based on the value
the original plates and angles were 344.8 MPa (50 ksi) and 200 GPa assumed in developing the column-strength curve in AASHTO
(29,000 ksi), respectively. The yield stress (Fy cpl ) and elastic (2012). Fig. 2(b) shows the first mode shape from the eigenvalue

© ASCE 04015054-2 J. Bridge Eng.

J. Bridge Eng., 04015054


(a) (b)
Downloaded from ascelibrary.org by Purdue University Libraries on 10/07/15. Copyright ASCE. For personal use only; all rights reserved.

Reference Point B

Coupling constraint

Coupling
constraint

3 1

(c) Reference Point A

Fig. 2. (a) Idealized uniaxial stress–strain curve; (b) first buckling eigenmode shape from eigenvalue analysis; (c) example of boundary conditions,
couplings, constraints, and mesh in the FEM model

buckling analysis, which was used to define the shape of the geo- because it uses the backward Euler operator as the time integration
metric imperfection. method. This analysis method provides reasonable results in spite of
The model change technique in Abaqus was used to add cover severe (nonductile) failure modes, such as local buckling and frac-
plates to the existing member; this technique enables adding new ture. Mesh convergence and sensitivity analyses were performed
parts (e.g., the cover plates) into the existing model (e.g., the proto- to finalize the size and distribution of the finite elements. The
type double-box member) through the shared deformed boundary selected mesh size, shown in Fig. 2(c), produces reasonable
while ensuring the added parts remain both stress-free and strain- results without significant computational effort.
free at that point in time. Tie constraints were applied to couple the
cover plates to the existing member, assuming a full bond (compos- Analysis and Behavior of Retrofitted Columns
ite action) between these two components.
Coupling constraints were used to model the pinned-end condi- As expected, the column slenderness ratio (KL=r) influences the
tions as follows. Two reference points (Reference Point A and axial behavior of steel columns. Short columns usually fail because
Reference Point B) were defined first. Each of these reference points of compression yielding of the member cross section. Slender
was coupled to one of the column ends (i.e., left or right end) using columns are more likely to fail because of global column buckling
kinematic coupling constraints, as shown in Fig. 2(c). For example, (i.e., flexural buckling or flexural torsional buckling). The behavior
the left end was coupled to Reference Point A, and the right end was and strength of the retrofitted prototype member shown in Fig. 1
coupled to Reference Point B. By applying coupling constraints were evaluated as both a short column (length of 7,620 mm) and a
between the column end and the corresponding reference point, the long (slender) column (length of 27,940 mm).
translational (U1 , U2 , and U3 ) and rotational (UR1 , UR2 , and UR3 ) For both columns, the locked-in dead load was assumed to be
degrees of freedom of the column end were coupled to the corre- equal to 88,960 kN (20,000 kips) and was applied to the existing
sponding reference point. The pinned-end conditions were then (original) member cross section first. The magnitude of the locked-
specified by applying displacement constraints (U1 ¼ U2 ¼ 0, in dead load was established by reviewing the design plans for
UR2 ¼ UR3 ¼ 0) at both reference points, and the displacement compression chord members of a half-through steel-arch bridge an-
constraint U3 ¼ 0 at Reference Point B. Axial loading was then alyzed previously by Lai et al. (2013). The compression chord
applied to the column by applying axial force (P) or axial displace- members of this existing bridge were similar (almost identical) to
ment (U3 ) to Reference Point A. the prototype member in terms of the cross section and length. The
The implicit dynamic analysis method in Abaqus was used to sources of the dead load include the self-weight of the chord mem-
perform the analysis. This analysis method is unconditionally stable bers, bracing systems, etc.

© ASCE 04015054-3 J. Bridge Eng.

J. Bridge Eng., 04015054


350000 300000
Yielding of cover plates Axial strength
300000 250000

Axial force, kN
250000
Axial force, kN

200000
Yielding of existing section
200000
150000
Yielding of existing section
150000
100000
100000
Addition of cover plates
Addition of cover plates 50000
50000

0 0
0 0.002 0.004 0.006 0.008 0 0.0005 0.001 0.0015 0.002 0.0025
Downloaded from ascelibrary.org by Purdue University Libraries on 10/07/15. Copyright ASCE. For personal use only; all rights reserved.

(a) Axial strain (b) Axial strain

350000 300000
Retrofitted column
300000 Existing colulmn
250000
Cover plates
250000
Axial force, kN

Axial force, kN
200000
200000
150000
150000
100000
100000

50000 Retrofitted column 50000


Existing column
Cover plates
0 0
0 0.002 0.004 0.006 0.008 0 0.0005 0.001 0.0015 0.002 0.0025
(c) Axial strain (d) Axial strain

500 400
450 350
400
300
Axial stress, MPa

Axial stress, MPa

350
300 250
250 200
200 150
150
100
100
Existing column Existing column
50 50
Cover plates Cover plates
0 0
0 0.002 0.004 0.006 0.008 0 0.0005 0.001 0.0015 0.002 0.0025
(e) Axial strain (f) Axial strain

Fig. 3. Analysis results of (a and b) the axial force–strain curves; (c and d) the contribution to the axial strength; and (e and f) the axial stress–strain
curves: (a, c, and e) short column; (b, d, and f) slender column

The cover plates were then added to the webs of the prototype conditions, L ¼ original column length, and rx ¼ radius of gyration
built-up double-box column using the model change technique in about the weak axis. Fig. 3(a) shows the axial force–strain curve
Abaqus. Additional loads were then applied to the retrofitted column obtained from the finite-element analysis. The axial strain (i.e., the
using the displacement control technique. Analysis was terminated abscissa) was calculated as the ratio of the total axial shortening to
when the target axial displacement was reached. This target displace- the original column length L. Three points are identified along the
ment was equal to 56 mm for the short column and 66 mm for the axial force–strain response, where the column stiffness changes sig-
slender column. These target displacement values were selected nificantly. These points correspond to the occurrence of the follow-
based on preliminary analysis results and target equivalent strains. ing events: (1) the addition of cover plates to the existing member,
(2) the yielding of the existing member cross section, and (3) the
yielding of the cover plates. Fig. 3(a) shows that the column devel-
Behavior of Retrofitted Short Column
oped extensive strain hardening after both Events 2 and 3, i.e.,
The slenderness ratio of the short column in the weak axis (KL=rx ) yielding of the existing section and yielding of the cover plates,
is 15.6, where the effective length factor K ¼ 1.0 for pinned-end respectively. The final failure was due to the yielding of the entire

© ASCE 04015054-4 J. Bridge Eng.

J. Bridge Eng., 04015054


retrofitted cross section. The postyielding behavior was ductile, i.e., first yield load (Py ) and a constant load increment (Pinc ). The
there was no decrease in axial strength after the yielding of the retro- load increment (Pinc ) was 4,448 kN (1,000 kips, which is 5% of
fitted (composite) section. the dead load) for the short column and 889.6 kN (200 kips,
Fig. 3(c) shows the contribution of the existing cross section and which is 1% of the dead load) for the slender column.
the added cover plates to the axial compressive strength of the retro- 4. Upon completion of the seventh cycle, the retrofitted column
fitted (composite) column. Only 43% of the yield strength of the was loaded until the specified axial displacement value was
added cover plates had been used when the existing cross section reached. Figs. 4(a and b) show the load amplitude curves used
yielded. A value of 96% of the total yield strength of the added in the cyclic analysis of the short and slender columns,
cover plates had been used when they began to yield at portions respectively.
close to the compression flange. Fig. 3(e) shows the axial stress– Comparisons of the axial force–strain curves from the cyclic and
strain behavior of the existing section and the added cover plates. monotonic analyses are shown in Fig. 4(c) for the short column and
The axial stress was estimated as the force resisted by the corre- in Fig. 4(d) for the slender column. The axial stress–strain curves of
sponding component (existing section or cover plates) divided by the existing section and the cover plates obtained from the cyclic
Downloaded from ascelibrary.org by Purdue University Libraries on 10/07/15. Copyright ASCE. For personal use only; all rights reserved.

the component’s cross-sectional area. Fig. 3(e) indicates that after analysis are shown in Fig. 4(e) for the short column and in Fig. 4(f)
the yielding of the existing section, the incremental axial load was for the slender column. After the first inelastic cycle, plastic shake-
sustained by the added cover plates. down occured to the elastic state in both the retrofitted short and
slender columns. The effects of cyclic loading seem to be con-
Behavior of Retrofitted Slender Column strained by the plastic shakedown behavior of the retrofitted
columns.
The slenderness ratio of the slender column in the weak axis
(KL=rx ) is 57.3. Fig. 3(b) shows the axial force–strain curve
obtained from the finite-element analysis. Three points along the Effects of Temperature Gradient
axial force–strain response where the column stiffness changes sig- Structural members in bridges are exposed to temperature changes
nificantly are indentified. They correspond to the occurrence of the across the cross section and member length. These temperature
following three events: (1) the addition of cover plates to the exist- changes introduce thermal deformations or forces in the structural
ing member, (2) the yielding of the existing member cross section, members depending on their boundary conditions (Usmani et al.
and (3) the reaching of the axial compressive strength of the retrofit- 2001). Uniform temperature change produces thermal strains and
ted (composite) column. The column failed owing to flexural buck- axial deformation (elongation or shortening) if the member ends are
ling about the weak axis. The postbuckling behavior was unstable, axially unrestrained. Axial forces develop in the member if this ther-
i.e., axial resistance decreased significantly after reaching peak mal expansion is restrained. Temperature gradients through the
strength and failure. cross-section depth produce thermal curvature and bowing if the
Fig. 3(d) shows the contribution of the existing section and the member ends are rotationally free. Uniform moments develop in the
added cover plates to the total axial strength of the retrofitted (com- member if the thermal curvature is restrained. The following discus-
posite) column. Only 40% of the yield strength of the cover plates sion focuses on the effects of positive change (increase) in tempera-
was used when the existing section yielded. The cover plates were ture. The findings are applicable to negative change (decrease) in
still in the elastic range, with only 46% of their yield strength being temperature, which is not discussed in detail for brevity.
used, when the axial strength was reached owing to flexural buck- The effects of temperature changes on the behavior of retrofitted
ling. Fig. 3(f) shows the axial stress–strain behavior of the existing prototype members (both short and slender columns) were investi-
section and cover plates, which was determined as explained earlier. gated using the FEM models developed previously in this section.
Flexural buckling occured almost immediately after the yielding of Sequentially coupled thermal-stress analyses were conducted for
the original section.
both the short and slender columns using the following four steps:
(1) apply the specified thermal gradient to the original member, (2)
Effects of Cyclic Loading apply the locked-in dead load to the original member, (3) add the
Cyclic loading causes inelastic deformations to accumulate over cover plates using the model change technique, and (4) apply addi-
subsequent cycles. The cumulative deformations and inelastic dis- tional loading to the retrofitted (composite) column until the speci-
placements may lead to member failure, unacceptable load redis- fied axial displacement is reached. Fig. 5(a) shows the temperature
tribution, or excessive distortion. However, if no additional inelastic gradient used in the analyses. This gradient was applied through the
deformations occur after the first few inelastic cycles, perhaps cross-section depth of the existing member along its entire length.
because of strain hardening of the material, then the structure is said As shown, the maximum temperature is 50°C at the top flange, and
to have experienced plastic shakedown (Chen and Han 2007). Two the minimum temperature is 0°C at the bottom flange. These tem-
cyclic analyses were performed to investigate the potential effects peratures were selected by the authors as conservative (upper
of the cyclic loading on retrofitted columns (one for the short col- bound) estimates of outdoor temperature variations in the field. A
umn and the other for the slender column). thermal expansion coefficient of 1:4  105 = C [as per AISC 360-
The following four steps were used to conduct both analyses: 10 (AISC 2010), Section 4.2.3.1] was used in the analysis.
1. The existing member was first loaded with 88,960 kN (20,000 kips) The retrofitted (composite) column had pinned-end conditions
dead load. and was axially and rotationally free (unrestrained). Thermal expan-
2. Cover plates were added using the model change technique, sion [as shown in Fig. 5(b)] and bowing [as shown in Fig. 5(c)]
and the retrofitted (composite) column was loaded until yield- occurred in both the short and slender columns. Comparisons of the
ing occurred in the existing section (i.e., Py was reached). axial force–strain curves for the retrofitted columns with and with-
3. The retrofitted column was subjected to seven cycles of unload- out thermal gradients are shown in Fig. 5(d) for the short column
ing and reloading. In each cycle, the retrofitted column was and in Fig. 5(e) for the slender column. For members with thermal
unloaded to the value corresponding to the dead load and then gradient, the axial strain represents the mechanical strain due to
reloaded to the value corresponding to the summation of the axial force only.

© ASCE 04015054-5 J. Bridge Eng.

J. Bridge Eng., 04015054


Continued to load until Continued to load until
prescribed displacement Yielding prescribed displacement
Yielding Seven loading cycles reached occurred Seven loading cycles reached
4.0 3.5
occurred at Py
at Py

Applied load / dead load


Applied load / dead load 3.5 Py = 0.05PLD 3.0 Py = 0.02PLD
3.0 2.5
2.5
2.0
2.0
1.5
1.5
1.0
1.0
0.5 0.5
Dead load, PLD Dead load, PLD
0.0 0.0
0 5 10 15 20 0 5 10 15 20
(a) Time (d) Time
Downloaded from ascelibrary.org by Purdue University Libraries on 10/07/15. Copyright ASCE. For personal use only; all rights reserved.

350000 300000
300000 250000
Axial force, kN

250000

Axial force, kN
200000
200000
150000
150000
100000
100000
Cyclic 50000 Cyclic
50000
Monotonic Monotonic
0 0
0 0.002 0.004 0.006 0.008 0 0.0005 0.001 0.0015 0.002 0.0025
(b) Axial strain (e) Axial strain

600 400
350
500
Axial stress, MPa
Axial stress, MPa

300
400
250
300 200
150
200
100
Existing Existing column
100
CPLS 50
Cover plates
0 0
0 0.002 0.004 0.006 0.008 0.01 0 0.0005 0.001 0.0015 0.002 0.0025
(c) Axial strain (f) Axial strain

Fig. 4. (a and b) Amplitude curves; (c and d) comparisons of axial force–strain curves; (e and f) comparisons of axial stress–strain curves: (a, c, and
e) short column; (b, d, and f) slender column

These comparisons indicate that thermal gradients have a negli- Summary of the Behavior of Retrofitted Columns
gible influence on the axial strength of both short and slender col-
umns. This is reasonable because the lateral displacements and The fundamental behavior of retrofitted (composite) columns is
resulting second-order moments introduced by thermal bowing are summarized based on the analytical results and discussions pre-
relatively small. The maximum lateral displacement was 2.8 mm sented in this section, as follows:
(0.11 in.) for the short column and 33.0 mm (1.3 in.) for the slender 1. Thefailure mode for short columns consistsof compression yield-
column. Similarly, the effects of applying a thermal gradient ing of the entire cross section, where yielding of the existing mem-
about the weak axis will also be negligible because the corre- ber section occurs first, followed by yielding of the cover plates.
sponding lateral displacements and second-order moments will 2. The failure mode for slender columns is flexural buckling about
be small relative to the dimensions of the member. The axial stiff- the weak axis, where yielding of the existing section occurs
ness of columns with thermal gradients was marginally greater first, followed by flexural buckling of the retrofitted column
than the axial stiffness of columns without thermal gradients. before yielding of the added cover plates.
This was probably due to the small tensile strains induced in the 3. The effects of cyclic loading and temperature changes on the
original member by the thermal deformations before additional behavior and strength of retrofitted columns are negligible
axial loads were applied. within the range of parameters investigated.

© ASCE 04015054-6 J. Bridge Eng.

J. Bridge Eng., 04015054


Downloaded from ascelibrary.org by Purdue University Libraries on 10/07/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 5. (a) Linearly varying temperature gradient; (b) thermal expansion; (c) thermal bowing; and comparisons of the axial force–strain curve for
(d) a short column and (e) a slender column

Nonlinear Inelastic Column Buckling Analysis L=1; 500, which was similar to the imperfection used in the 3D
This section presents the development and benchmarking of a finite-element analysis presented in the previous section. This sinu-
NICB analysis approach and a computer program for predicting soidal imperfection is a conservative and reasonable assumption for
the critical buckling load (Pcr ) of retrofitted (composite) col- inelastic column buckling analysis (Ziemian 2010). Cyclic loading
umns. The analysis was based on the numerical integration and temperature gradients were not implemented in this program
approach presented by Chen and Lui (1987), and the program because the effects of these two parameters were found to be negligi-
was implemented in MATLAB 2011a. The computer program ble. The following subsections present the algorithm and the bench-
calculated the critical buckling load (Pcr ) for retrofitted (com- marking of the NICB analysis program.
posite) columns using the geometric and material properties
provided by the user. This benchmarked analysis approach (and Algorithm
program) provides a convenient tool for conducting parametric
studies to further evaluate the effects of geometric imperfec- The NICB analysis was conducted in two steps. In the first step, a
tions, residual stresses, locked-in dead load ratio (PLD =Pori ), locked-in dead load was applied to the existing section. In the sec-
cover-plate properties (yield stress Fy cpl , width bcpl , and thick- ond step, cover plates were added to the existing section, and addi-
ness tcpl ), and column slenderness ratio (KL=r) on the behavior tional axial load was applied to the retrofitted column until failure
of retrofitted columns. Other researchers have used similar (i.e., Pcr was reached). In both steps, the loads (locked-in dead load
approaches to evaluate the behavior and strength of columns and additional load) were applied incrementally to the member. For
subjected to complex loading and boundary conditions; for each load increment (Pi ), two analysis subroutines were called con-
example, Hong and Varma (2010) used this approach to predict secutively: (1) the P  M  f subroutine and (2) the D subroutine.
the fire resistance of steel–concrete composite columns. The P  M  f subroutine was implemented to calculate the axial
For the benchmarking analysis, the geometric imperfection was load-moment-curvature (P  M  f ) relationship for the cross
implemented with sinusoidal shape and magnitude equal to section using fiber analysis. The D subroutine was implemented to

© ASCE 04015054-7 J. Bridge Eng.

J. Bridge Eng., 04015054


calculate converged member deflections (Yk ) using the cross section P 2M2/ Subroutine
P  M  f relationships. External moments (Mk ) including the
secondary moments were calculated using the applied axial load The cross section of the retrofitted column was discretized into
(Pi ) and the calculated deflections (Yk ). Column failure occurred layers of fibers, as shown in Fig. 6. For each fiber, the area (Afib ),
when the calculated external moment (Mk ) was greater than the moment of inertia with respect to the centroid of the cross section
cross-section moment capacity obtained from the P  M  f subrou- (Ifib ), and centroid distance (distance from the center of the fiber to
tine. Details of these two subroutines are presented in the follow- the centroid of the cross section, yfib ) were calculated. The proce-
ing section. dures shown in Fig. 7 were then used to obtain the P  M  f
Downloaded from ascelibrary.org by Purdue University Libraries on 10/07/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 6. Cross-section fiber discretization and summation of strains

Fig. 7. Flowchart for the P  M  f subroutine

© ASCE 04015054-8 J. Bridge Eng.

J. Bridge Eng., 04015054


Downloaded from ascelibrary.org by Purdue University Libraries on 10/07/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 8. (a) Discretization of the segments along the length; (b) calculation of deflections

relationship for each load increment (Pi ). The P  M  f relation- station (i.e., Station n). This complication is illustrated in Fig. 8(b).
ship was obtained when (1) the converged values of curvatures As shown, the calculated displacement (yk i;j ) is not the same as the
were found, and (2) the corresponding moments for each load incre- actual deflection (Yk i;j ); therefore, it is corrected to the actual deflec-
ment (Pi ) were calculated. tion (Yk i;j ) as follows:
 
D Subroutine k
Yk i;j ¼ yn i;j  yk i;j
n
The retrofitted (composite) member was discretized into n  1
(e.g., 16) segments, as shown in Fig. 8(a). This resulted in a total • Compare the corrected displacements at each station
of n (n ¼ 17) stations along the length (L), with the length of each (Yk i;j ) with those (Yk i;j1 ) from the previous iteration. If
segment (Dx) equal to L=16. The following iterations were then used Yk i;j  Yk i;j1  tolerance, then the converged displace-
to calculate the member deflections for each load increment (Pi ):
ments have been found, i.e., Yk i;j . However, if Yki,j 
• Assume the lateral displacements (Yk i;j ) in iteration j to be Yki,1j > tolerance, then use the corrected displacements
the same as the lateral displacements in the previous itera-
(Yk i;j ) as the initial value for the next iteration (j þ 1) and
tion j  1. For the first iteration (j ¼ 1), the lateral displace-
continue. The tolerance was assumed to be h/6, 000.
ments were assumed to be the same as the converged
deflections from the previous load increment (i.e., the
deflections for Pi1 ), as shown in Fig. 8(a). For the first Critical Buckling Load (Pcr )
load increment (i ¼ 1), the lateral displacements were Once the converged deflections (Yk i;j ) were found, the applied
assumed to be the same as the sinusoidal imperfections. moment (Mk ) at each station (Station k) was calculated as the prod-
• Calculate the external moment (Mk i;j ) at each station uct of applied axial force (Pi ) and the deflections (Yk i;j ). If the
(Station k) as moment at any station (typically at midheight) had become greater
than the cross-section moment capacity obtained from the corre-
Mk i;j ¼ Pi yk i;j sponding P  M  f subroutine, then the column had failed
because of inelastic column buckling, and Pi was the critical buck-
• Obtain the curvature at each station ( f k i;j ) using the rela- ling load (Pcr ). Otherwise, the axial load was increased from Pi to
tionship calculated from the P  M  f subroutine. Then, the next load increment Piþ1 , and the two subroutines (P  M  f
calculate the rotation (u k i;j ) and lateral displacement (yk i;j ) and D) presented above were recalled until column failure caused
at each station using by inelastic buckling occurred.
u k i;j ¼ u k1 i;j þ f k i;j Dx
Benchmarking
yk i;j ¼ yk1 i;j þ u k i;j Dx The NICB analysis program was used to predict the behavior and
strength of the two retrofitted columns analyzed in the previous sec-
However, using these two equations results in Station 1 being tion. Fig. 9 shows the comparison of the results from the NICB and
fixed with zero rotation and displacement, and all rotations and dis- finite-element analyses. Figs. 9(a and b) show comparisons of the
placements from Station 1 to Station n  1 to cumulate at the last axial force–strain curves for the retrofitted (composite) short and

© ASCE 04015054-9 J. Bridge Eng.

J. Bridge Eng., 04015054


350000 300000
300000 250000
Axial force, kN 250000
200000

Axial force, kN
200000
150000
150000
100000
100000
50000 50000
NICB NICB
0 FEM FEM
0
0 0.002 0.004 0.006 0.008 0.01 0 0.0005 0.001 0.0015 0.002 0.0025
Axial strain Axial strain
(a) (b)
Downloaded from ascelibrary.org by Purdue University Libraries on 10/07/15. Copyright ASCE. For personal use only; all rights reserved.

400 400
350 350
300 300
Axial stress, MPa

Axial stress, MPa


250 250
200 200
150 150
100 100
50 NICB 50 NICB
FEM FEM
0 0
0 0.002 0.004 0.006 0.008 0.01 0 0.0005 0.001 0.0015 0.002 0.0025
(c) Axial strain (d) Axial strain

500 300
450
250
400
Axial stress, MPa
Axial stress, MPa

350 200
300
250 150
200 100
150
100 50
NICB
50 NICB FEM
FEM 0
0
0 0.0005 0.001 0.0015 0.002 0.0025
0 0.002 0.004 0.006 0.008 0.01
(e) Axial strain (f) Axial strain

Fig. 9. Comparisons of results from NICB analysis and finite-element analysis: (a and b) axial force–strain curves for the retrofitted column; (c
and d) axial stress–strain curves for the existing column; (e and f) axial stress–strain curves for the cover plates: (a, c, and e) short column; (b, d,
and f) slender column

Fig. 10. Stress and strain distributions at the midspan section from the NICB analysis program

© ASCE 04015054-10 J. Bridge Eng.

J. Bridge Eng., 04015054


slender columns, respectively. The axial strains were calculated as indicated that the effects of cyclic loading and temperature gradient
the ratio (D0 =L) of axial shortening (D0 ) to the undeformed column were negligible with the range of parameters evaluated.
length (L). Figs. 9(c and d) show comparisons of the axial stress– A NICB analysis program was also developed and then bench-
strain curves for the cover plates of the retrofitted short and slender marked using results from the detailed finite-element analysis. This
columns, respectively. Figs. 9(e and f) show comparisons of the program provides an efficient tool to further evaluate the behavior
axial stress–strain curves for the existing sections of the retrofitted of retrofitted columns by conducting comprehensive parametric
short and slender columns, respectively. In Figs. 9(c–f), the axial studies investigating the effects of (1) the magntude of geometric
strains were extracted from the analysis results at the mid- imperfections, (2) residual stresses, (3) locked-in dead load ratio
height of the columns. The comparisons in Figs. 9(a–f) indi- (PLD =Pori ), (4) cover-plate properties (yield stress Fy cpl , width bcpl ,
cate that the NICB analysis program predicts the behavior and thickness tcpl ), and (5) column slenderness ratio (KL=r). Results
and strength of retrofitted columns with an acceptable level from these comprehensive parametric studies and the resulting
of accuracy when benchmarked against the more detailed design equations will be presented in a future paper.
FEM solutions. The NICB analysis program also generates
Downloaded from ascelibrary.org by Purdue University Libraries on 10/07/15. Copyright ASCE. For personal use only; all rights reserved.

the stress and strain distributions at any station; for example, Acknowledgments
the stress and strain distributions at the column midheight are
shown in Fig. 10. The research presented in this paper was funded partially by HDR
Inc. and by Purdue University. The research, findings, and
conclusions presented in this paper belong to the authors.
Further Work

A number of aspects were not addressed in this study. These include References
(1) the effects of several parameters (such as the magnitude of geo-
metric imperfections, residual stresses, end moments, locked-in AASHTO. (2012). LRFD bridge design specifications, 6th Ed., Washington,
dead load ratio, cover plates, and column slenderness) on the behav- DC.
ior of the retrofitted columns and (2) the behavior of retrofitted Abaqus 6.12 [Computer software]. SIMULIA, Johnston, RI.
columns with other cross-section shapes (such as box sections, AISC. (2010). “Specification for structural steel buildings.” AISC 360-10,
Chicago.
channels, and angles). To address these aspects, comprehensive an-
Chen, W. F., and Han, D. (2007). Plasticity for structural engineers, J. Ross
alytical investigations will be conducted using the benchmarked Publishing, Plantation, FL.
NICB analysis program. Appropriate design equations for evaluat- Chen, W. F., and Lui, E. M. (1987). Structural stability, Prentice Hall,
ing the axial compressive strength of retrofitted columns with vari- Upper Saddle River, NJ.
ous cross-section shapes will also be proposed, using the results Hong, S., and Varma, A. H. (2010). “Predicting column buckling under fire
from the comprehensive analytical investigations. loading using fundamental section behavior.” J. ASTM Int., 7(1), 156–
178.
Lai, D., and Patel, N., eds. (2004). Structure rehabilitation manual:
Conclusions Revision No. 9, Ontario Ministry of Transportation, Toronto.
Lai, Z., and Varma, A. H. (2015). “Noncompact and slender circular CFT
Built-up steel columns were strengthened in the weak axis by add- members: Experimental database, analysis, and design.” J. Constr. Steel
ing cover plates to the webs. Comprehensive finite-element models Res., 106(3), 220–223.
Lai, Z., Varma, A. H., Connor, R. J., and Liu, J. (2013). “Retrofit analysis of
were developed to investigate the fundamental behavior of these
steel built-up members for bottom chords of Bayonne bridge.” Bowen
retrofitted columns. Results from the finite-element analysis indi- Laboratory Research Rep. (BLRR) No. 2013-02, Purdue Univ., West
cated that failure modes for retrofitted columns depend on the col- Lafayette, IN.
umn slenderness ratio. For short columns, failure occurred because Lai, Z., Varma, A. H., and Zhang, K. (2014). “Noncompact and slender rec-
of yielding of the entire retrofitted cross section, where the existing tangular CFT members: Experimental database, analysis, and design.”
section yielded first, followed by the yielding of the added cover J. Constr. Steel Res., 101(10), 455–468.
plates. Consequently, the strength of short columns was improved MATLAB 2011a [Computer software]. MathWorks, Natick, MA.
efficiently by adding cover plates. The full yield strength of the Shek, K. K.-W., and Bartlett, F. M. (2008). “Analysis and design of rehabili-
added cover plates could be achieved before failure. For slender col- tated built-up hybrid steel compression members.” Can. J. Civil Eng.,
35(12), 1375–1387.
umns, failure occurred due to inelastic column buckling of the retro- Usmani, A. S., Rotter, J. M., Lamont, S., Sanad, A. M., and Gillie, M.
fitted column about the weak axis. The existing section yielded first, (2001). “Fundamental principles of structural behaviour under thermal
but buckling failure occurred while the cover plates were still in the effects.” Fire Safety J., 36(8), 721–744.
elastic range. Only a portion of the yield strength of the added cover Ziemian, R. D., ed. (2010). Guide to stability design criteria for metal struc-
plates could be used. The finite-element analysis results also tures, John Wiley & Sons, Hoboken, NJ.

© ASCE 04015054-11 J. Bridge Eng.

View publication stats J. Bridge Eng., 04015054

You might also like