Professional Documents
Culture Documents
(Asce) Be 1943-5592 0000265
(Asce) Be 1943-5592 0000265
(Asce) Be 1943-5592 0000265
on Bridge Columns
G. Daniel Williams1 and Eric B. Williamson, P.E., M.ASCE2
Downloaded from ascelibrary.org by "Indian Institute of Technology, Gandhinagar" on 08/07/23. Copyright ASCE. For personal use only; all rights reserved.
Abstract: Historical data show that terrorist attacks against transportation assets have increased in recent years and that the vast majority of
these attacks have been bombings. As such, there is growing interest in protecting highway infrastructure from blast loads. To address this
concern, the National Cooperative Highway Research Program (NCHRP) sponsored a project to investigate the performance of highway
bridges subjected to the nearby detonation of an explosive, and this paper presents research that advances the understanding of blast loads
acting on bridge columns. Unlike large wall panels for which much of the existing knowledge about blast effects against structures has been
established, the research presented in this paper focuses on slender structural components in which the effects of cross-sectional geometry,
engulfment of blast pressures, and clearing effects strongly influence loading history. Based on the findings obtained from this study,
a simplified procedure for predicting blast loads acting against bridge columns is proposed. DOI: 10.1061/(ASCE)BE.1943-5592
.0000265. © 2012 American Society of Civil Engineers.
CE Database subject headings: Blast loads; Bridges; Columns; Computer models; Terrorism; Predictions.
Author keywords: Blast loads; Bridges; Computational modeling; Terrorism.
ize blast loads on slender members because they lack the ability to (USACE 2001), a computer program developed and supported by
accurately characterize phenomena associated with the relief wave the U.S. Army Corps of Engineers that can be used to predict blast
that travels from the free edges toward the center of a component loads on flat surfaces, indicates a positive pressure. Rickman and
(Rickman and Murrell 2007). Murrell then developed an equation and associated constants for a
Several researchers have addressed the issue of shock waves relief function to subtract the CONWEP wave form from the mea-
interacting with circular members in recent years. Ofengin and sured wave form beginning at the time of arrival of the
Drikakais (1997) studied the diffraction of shock waves over a cyl- relief wave. They also formulated a mathematical expression for
inder for three planar blast waves with short-term, long-term, and the arrival time of the relief wave. Both were combined into a com-
infinite durations. These loads were applied through a high- puter program to calculate reflected impulse curves, and predicted
pressure layer similar to the sudden rupture of a membrane within results compared favorably to both small-scale and large-scale
a shock tube. The results provided by the authors qualitatively char- experimental tests.
acterized the various shock phenomena over a cylinder, and they For the specific case of blast loads acting against highway
also demonstrated that blast wave duration significantly influences bridges, Agrawal and Yi (2009) presented the results of detailed
the pressure variation acting over a cylinder. The findings presented finite element simulations for three levels of blast loads (low,
by the authors agree well with expected results; however, they are medium, and large). While finite element programs such as
not provided in a format useful for engineering applications, LS-DYNA offer the ability to simulate shock wave propagation
especially the results for blasts of infinite duration. through a medium (e.g., air), such models can be computationally
Shi et al. (2007) used Autodyn 3D to model semispherical sur- intensive and require significant time to run. Therefore, rather than
face detonations in air and the resulting shock interaction with re- directly modeling detonation effects in their simulations, Agrawal
inforced concrete square and circular columns at a 10-m standoff. and Yi (2009) proposed a method for generating loads in their finite
The model was validated using results from a blast test reported in element models by using results computed from CONWEP. The
the literature, and the researchers then conducted a parameter study CONWEP-generated loads were then transmitted through an air
to investigate the influence of five parameters on column-shock medium to structural components, thereby making the load simu-
interaction: (1) Scaled distance (Z)—defined as the distance from lation phase of the analysis more efficient than directly modeling
the center of detonation to the target point of interest (ft. or m) detonation effects. This approach was adopted by Agrawal and Yi
divided by the charge weight (lbs. or kg) raised to the 1∕3 power (2009) for use in very detailed finite element models that were used
(i.e., Z ¼ R∕W 1∕3 ); (2) Ratio of supported mass to column mass; to parametrically evaluate the effects of various design variables
(3) The stiffness of the column; (4) The dimension of the column; (e.g., concrete strength, seismic detailing) on the performance of
and (5) The geometry of the column. The results showed that the blast-loaded bridge components. The work presented in the current
column mass ratio, the column stiffness, and the column deflection paper extends the results reported by Agrawal and Li (2009) in two
had no significant effect on the loading. The scaled distance and the significant ways. First, Agrawal and Li (2009) focused their sim-
column dimension, however, were shown to significantly influence ulations on a bridge that comprised square columns, whereas the
reflected pressure and impulse on the columns, where impulse is current work considers the effects of column cross-sectional shape
defined as the area under the pressure-time loading curve. These on the loads that must be resisted in a blast event. Second, the focus
findings are consistent with the research results reported later in of the current study is on the development of simplified load
this paper. prediction methods that can be used for design purposes rather than
Shi et al. (2007) then developed factors to account for shape detailed finite element simulations. Thus, while the research
effects when calculating reflected pressure and impulse on the front presented in the current manuscript utilizes detailed finite element
and back centerlines of a column using the positive incident pres- simulations, the emphasis is on interpreting the results so that they
sure and the positive incident impulse along with the column shape can be readily implemented in the blast-resistant design of bridge
and dimension. While these equations and factors provide insight columns.
into the interaction of blast loads on columns, they have limited Of most interest to the current work, researchers at the State
application for the design and analysis of bridge columns because University of New York investigated a design alternative and ana-
they are restricted to a standoff distance of 10 m and they only lytical approach for blast-resistant bridge columns (Fujikura et al.
consider the loads acting on the centerline of the column. Research 2008). The design utilized a concrete-filled steel tube anchored into
findings presented later in this manuscript and by others have the foundation with steel plates linked to concrete-embedded
shown a significant reduction in net impulse across the entire C-channel shapes. One-fourth-scale prototypes of the concrete-
cross-section of a column due to shape and slenderness effects filled steel tube columns were tested experimentally. The research-
(Williams 2009; Fujikura et al. 2008). ers also developed an analytical method to calculate the response of
Rickman and Murrell (2007) conducted a series of small-scale the bridge columns using an equation-based single-degree-of-
experiments to characterize the relief wave form, the clearing time, freedom (SDOF) method. The energy imparted to the structural
and the resulting loads on walls. This effort was in response to a system by the blast loads was considered to be an impulsive loading
recognition of the following errors in existing methods used to in this research because the blast load duration was much shorter
calculate reflected pressures and impulses on walls: approximation than the natural period of the columns that were tested, which is
reinforcing steel shared nodes with the concrete elements, and as vertical base constraints (i.e., no horizontal base constraints in ei-
such the model did not explicitly consider bar slip. This approxi- ther direction) on base nodes outside the transverse reinforcement
mation was acceptable because highly distorted concrete elements (i.e., in the region of cover concrete) to allow erosion of cover con-
could be eroded from the model, and as such rebar elements did not crete. The pressure-time histories applied to the numerical models
necessarily remain connected to concrete elements for the entire of the blast-tested columns were computed using the computational
analysis duration, allowing them to move independently from fluid dynamics (CFD) numerical techniques discussed above
the concrete elements under large deformations when rebar slip (Williams 2009). Readers are directed to Williams (2009) for addi-
tional information on the finite element models used to simulate the
is most pronounced. All nodal translational displacements were
experimental blast tests carried out under NCHRP 12-72.
constrained at the base of the model to be consistent with the fixed
boundary condition provided by the column foundation and reac-
tion structure. The models in which spalling occurred had only Validation Efforts
3A 76.2 (3.0) 66.0 (2.6) and 3Blast—generated slightly lower displacements relative to
3-Blast 114 (4.5) 78.7 (3.1) the experimental results. The main contributor to the small pre-
dicted displacement for these columns is the inability of the con-
crete constitutive model to accurately represent degradation of
not practical to provide a detailed discussion of each individual concrete in compression and shear. Stress versus strain curves from
numerical column model used to validate the CFD models. The single element stress analyses of unconfined axial compression
results of all models are summarized in Table 3, and one numerical tests at various loading rates using the Karagozian and Case con-
column model is briefly presented herein. crete model showed that the material model exhibits an unrealisti-
Fig. 3 shows both the experimental and numerical damage and cally long unloading phase after reaching the peak concrete
cracking patterns for Column 3A. The damage is shown using a compressive strength with a strain well beyond 0.008, which is sig-
normalized scale of 0 to 2. A value of 1 indicates that the material nificantly larger than the typical range of 0.003 to 0.004
has just reached its peak strength, and a value between 1 and 2 (MacGregor 1997). Concrete constitutive modeling of blast-loaded
indicates that the material has passed its peak strength and is soft- structures remains a computational challenge (DoD 2002), and this
ening. Accordingly, for tension, values close to 2 indicate that the field continues to be an active area of research. Despite this limi-
concrete is cracking, and for compression, such values indicate that tation, the numerical models used in this study performed quite well
the concrete is crushing (i.e., softening) and nearing failure (or per- with the pressure-time history data derived from the CFD analyses
haps has already reached failure). The appearance of failure on the described above.
compression side of the column must be evaluated carefully, as an
Characterization of Reflected Impulse on Columns
indication of large plastic strains in the numerical models do not
always correlate to similar damage in the experimental models Eqs. (5) and (6) were employed to compute the net resultant im-
due to limitations in the constitutive models used to represent pulse (i.e., the difference in impulse between the front and back
the concrete behavior [as described further below and in Williams faces) at a given cross-section using the pressure-time history data
(2009)]. Overall, the numerical models, which were loaded using and the tributary areas shown in Fig. 4.
Fig. 3. Comparison of response shapes and damage patterns for Column 3A: (a) cracking and damage pattern from analytical model; (b) analytical
model with overlay of real cracking pattern; and (c) real cracking and damage pattern
ðT1 T12Þ þ 2ðT2 T11Þ þ 2ðT3 T10Þ þ 2ðT4 T9Þ þ 2ðT5 T8Þ þ ðT6 T7Þ
I nr:square ¼ ð6Þ
10
The designation T# in Eqs. (5) and (6) represents numerically at a position above the ground than at the base of the column.
Downloaded from ascelibrary.org by "Indian Institute of Technology, Gandhinagar" on 08/07/23. Copyright ASCE. For personal use only; all rights reserved.
generated pressure-time history data from the locations labeled Shi et al. (2007) reported similar findings for small scaled distan-
by the corresponding numbers in Fig. 4. For example, considering ces, albeit at a slightly higher position along the column. This phe-
the first term in the numerator of Eq. (6), Fig. 4 shows that locations nomenon is likely due to pressure magnification from reflections of
“1” and “12” are directly opposite each other, and it is the differ- the shock wave off the ground, the nature of which is a function of
ence in loading between these locations that is used to compute the standoff, charge size, and shape.
net resultant impulse. Impulse values are considered to be the most Several attempts were made to normalize these data into equa-
important load parameter for this study because the net resultant tions that could be used to calculate equivalent uniform impulse
impulse typically governs the global response of blast-loaded mem- values acting against circular and square columns for blast-resistant
bers when the duration of loading is much less than the natural design purposes. In this context, an equivalent loading that has a
period of the structure being analyzed (Biggs 1964; DoD 2008; uniform distribution rather than one that varies along the column
Fujikura et al. 2008), which was the case for all scenarios height can be derived using principles of work and energy conser-
considered during this research program. Curves that indicate vation, and this approach is used in various existing software pack-
how blast intensity varies with position along the height of a ages that compute blast loads, including CONWEP (USACE 2001)
column were developed based on the net resultant impulse values and BEL (USACE 2000). Given the uncertainty associated with
computed at 20 equally spaced cross-sections along the height of potential threat scenarios, an approach that utilizes an equivalent
each column. uniform load is often preferred to simplify the design process
Because the numerical CFD parameter study included a variety (DOD 2008; Conrath et al. 1999). The effort to develop equivalent
of charge sizes and standoff distances, actual pressures and uniform impulse values encountered significant difficulty because
impulses acting on the analyzed columns varied considerably. of the highly nonlinear nature of shock propagation and the
To allow for consistency in interpreting the computed results, nor- complex interaction between a shock wave and a column. The best
malized net resultant impulse curves (termed “load curves” in this solution yielded results that could differ up to 35% from the results
paper) were developed by calculating and plotting the net resultant of the numerical models. Thus, this approach was abandoned in
impulse at various vertical locations along the height of the col- favor of an alternative method for computing equivalent blast
umns and normalizing those values to the computed peak net result- impulse acting on a column.
ant impulse value for each column (Fig. 5). This figure is difficult to Methods already exist to calculate pressure wave forms and the
understand with the numerous curves shown on the plot, but all resulting impulses acting against flat panel structures (DoD 2008;
curves are shown to allow a detailed independent investigation USACE 2001; SAIC 2001; USACE 2000), and recent research
of the results if desired. The peak reflected impulse on each column confirms their applicability and accuracy when clearing from free
was chosen as the basis for normalization because peak reflected edges is not a concern (Rickman and Murrell 2007). Therefore, a
impulse can be readily calculated with high confidence using revised method was developed for computing equivalent uniform
several widely accepted methods (e.g., USACE 2000; DoD impulse values for use in analyzing blast-loaded bridge columns.
2008). The curves in Fig. 5 show a higher net resultant impulse The proposed method utilizes existing procedures for computing
equivalent uniform impulse values against flat panels (USACE
2001; SAIC 2001; USACE 2000), coupled with column shape
factors that account for the effects of cross-sectional shape and
slenderness on load history. The procedure used to develop the
column shape factors for slender circular and square members
subjected to blast loads is described in the following section. After-
ward, a simplified design procedure based on these factors is
presented.
Fig. 5. Normalized impulse curves for cylindrical bursts against bridge columns
in Fig. 4. Moreover, the peak incident overpressure is less at a given other factors contribute to the nonlinearity of the data, including
location around the circumference of a circular column than at a varying charge weights, charge sizes, standoff distances, column
similar position along the face of a square column due to the sizes, clearing times, angles of incidence, heights-of-burst, and
increased standoff distance to the position on the circular column ground reflections. Therefore, rather than normalizing the data,
relative to that on a square column. Both an increase in angle of it is more appropriate for design purposes to determine suitable
incidence and a decrease in peak incident overpressure individually upper bounds as shown in Fig. 6. The column shape factors are
produce reduced reflected pressures and impulses, and thus with found using Eq. (7) for a square column and Eq. (8) for a circular
the combination of the two a circular column will experience less column:
net load than a square column. Furthermore, it is important to note
that load clearing around circular columns is different than around R
Ss ¼ 0:013 þ 0:49 ð7Þ
square columns. Shock waves engulf circular columns more easily D
than square columns, making the resultant impulse on the back face
of circular columns greater than that on the back face of similarly
sized square columns, further reducing the net resultant load a
circular column must resist relative to a square column. This find- 0.60
ing is consistent with those of Shi et al. (2007) as described above.
Fig. 6 shows the column shape factors derived from the airblast 0.55
models analyzed for this research. These values were determined
0.50
Column Shape Factor, S
This paper presents research results obtained from computational Holland, C. E. (2008). “Blast-resistant design of highway bridge columns.”
Master’s thesis, Dept. of Civil, Architectural, and Environmental Engi-
fluid dynamics models of blast loads acting against circular and
neering at the Univ. of Texas at Austin, Austin, TX.
square bridge columns, although the findings apply in general to
Jenkins, B. M., and Gersten, L. N. (2001). “Protecting public surface
slender structural members directly exposed to airblast. The com- transportation against terrorism and serious crime.” MTI Report
putational models developed for this research were validated using 01-14, Mineta Transportation Institute, San Jose, CA.
data from experimental blast tests and results reported in the liter- Livermore Software Technology Corporation (LSTC). (2007). LS-DYNA
ature. The numerical results gathered from the computational keyword user’s manual, version 971, LSTC, Livermore, CA.
models showed the influence of column cross-sectional shape Luccioni, B., Ambrosini, D., and Danesi, R. (2006). “Blast load assessment
on net resultant impulse, and this information was used to develop using hydrocodes.” Eng. Struct., 28(12), 1736–1744.
equations for column shape factors that accurately account for the MacGregor, J. G. (1997). Reinforced concrete: Mechanics and design, 3rd
reduced impulses experienced by slender square and circular Ed., Prentice Hall, Upper Saddle River, NJ.
members relative to those acting on a wall subjected to the same Malvar, L. J., Crawford, J. E., and Morrill, K. B. (2000). “K&C concrete
threat scenario. These equations can be combined with an SDOF material model release III—automated generation of material model
analysis model, which provides a simple method for predicting input.” Karagozian and Case Report TR-99-24.3, Karagozian & Case
response, to ensure that bridge columns have adequate capacity Structural Engineers, Burbank, CA.
to resist blast loads. Ofengein, D. K., and Drikakais, D. (1997). “Simulation of blast wave
propagation over a cylinder.” Shock Waves, 7(5), 305–317.
Rickman, D. D., and Murrell, D. W. (2007). “Development of an improved
Acknowledgments methodology for predicting airblast pressure relief on a directly loaded
wall.” J. Pressure Vessel Technol., 129(1), 195–204.
The authors gratefully acknowledge the support of the National Schwer, L. E., and Malvar, L. J. (2005). “Simplified concrete modeling
Cooperative Highway Research Program for providing the finan- with *MAT_concrete_damage_REL3.” JRI LS−DYNA USER WEEK
cial resources that made this project possible. The authors also wish 2005, Karagozian & Case Structural Engineers, Burbank, CA.
Science Applications International Corporation (SAIC). (2001). BlastX
to thank the Southwest Research Institute and Protection Engineer-
version 4.2.3.0, Science Applications International Corporation. San
ing Consultants for their assistance in completing the experimental
Diego, CA.
tests. The opinions expressed in this paper are those of the authors Shi, Y., Hao, H., and Li, Z. (2007). “Numerical simulation of blast wave
and do not necessarily reflect those of the sponsor. interaction with structure columns.” Shock Waves, 17(1–2), 113–133.
Sućeska, M. (1999). “Evaluation of detonation energy from EXPLO5 com-
puter code results.” Propellants Explos. Pyrotech., 24(5), 280–285.
References U.S. Army Corps of Engineers (USACE). (2000). Bridge Explosive
Agrawal, A. K., and Yi, Z. (2009). “Blast load effects on highway bridges.” Loading (BEL), version 1.1.0.3, Engineer Research and Development
Univ. Transportation Research Center, City College of New York, NY. Center. Vicksburg, MS.
〈http://www.utrc2.org/research/assets/123/FR_Blast_bridge1.pdf〉 U.S. Army Corps of Engineers (USACE). (2001). CONWEP, Version
(Apr. 3, 2012). 2.0.6.0, U.S. Army Engineer Research and Development Center,
Alia, A., and Souli, M. (2006). “High explosive simulation using Vicksburg, MS.
multi-material formulations.” Appl. Therm. Eng., 26(10), 1032–1042. Williams, G. D. (2009). “Analysis and response mechanisms of
Biggs, J. M. (1964). Introduction to structural dynamics, McGraw-Hill, blast-resistant bridge columns.” Ph.D. dissertation, Dept. of Civil, Arch.
Inc, New York. and Env. Eng., Univ. of Texas at Austin, Austin, TX.
The Blue Ribbon Panel on Bridge and Tunnel Security (BRPBTS). (2003). Williamson, E. B. et al. (2010). “Blast-resistant highway bridges: Design
“Recommendations for bridge and tunnel security.” Special report and detailing guidelines.” National Cooperative Highway Research
prepared for FHWA and AASHTO, Washington, DC. Program (NCHRP) Report 645, Transportation Research Board of
Boyd, A., and Sullivan, J. P. (2000). “Emergency preparedness for transit the National Academies, Washington, DC.
terrorism.” TR News. No. 208. 12–17. 〈 http://onlinepubs.trb.org/ Winget, D. G., Marchand, K. A., and Williamson, E. B. (2005). “Analysis
onlinepubs/trnews/trnews208_transit_security.pdf〉 (Apr. 6, 2012). and design of critical bridges subjected to blast loads.” J. Struct. Eng.,
Conrath, E. J., Krauthammer, T., Marchand, K. A., and Mlakar, P. F. (1999). 131(8), 1243–1255.