Você está na página 1de 8

Engineering Structures 31 (2009) 28252832

Contents lists available at ScienceDirect

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

Finite difference analysis of simply supported RC slabs for blast loadings


J. Jones a , C. Wu a, , D.J. Oehlers a , A.S. Whittaker b , W. Sun c , S. Marks a , R. Coppola a
a

School of Civil and Environmental Engineering, The University of Adelaide, SA, Australia

Department of Civil, Structural and Environmental Engineering, State University of New York at Buffalo, USA

Department of Civil Engineering, Huaiyin Institute of Technology, China

article

info

Article history:
Received 14 August 2008
Received in revised form
1 July 2009
Accepted 8 July 2009
Available online 24 July 2009
Keywords:
Blast effects
Finite difference analysis
Strain rate effect
RC slabs

abstract
A finite difference procedure that can account for strain-rate effects, both shear and flexural deformations,
permits variations in cross-section geometry and strength and loading over the length of a component is
proposed to accurately and efficiently analyze the dynamic response of a simply supported structural
member under blast loads. A section-based layered analysis model that accommodates varying strain
rates across a members cross-section is used to derive sectional momentcurvature relationships. A
formula is derived to estimate the distribution of strain rate over the depth of a cross-section along
the length of the member, and the corresponding strain rate effects are incorporated into the sectionbased layered analysis model. The Timoshenko beam equations that include both shear deformations
and rotational inertia are solved numerically using an explicit finite difference scheme. The accuracy of
the proposed finite difference analysis model is part validated using results of blast testing of reinforced
concrete slabs with combinations of explosive weights and standoff distances. The results are also
compared with those obtained by conventional single-degree-of-freedom (SDOF) analysis and finite
element (FE) analysis using solid elements. The finite difference analysis procedure is both fast-running
and accurate and most suitable for design office application, combining the speed of SDOF analysis and
the detail and accuracy of FE analysis.
2009 Elsevier Ltd. All rights reserved.

1. Introduction
Structural components such as beams and columns are typically analyzed for blast loadings using one of two methods of
very different complexity: (1) equivalent single degree of freedom
(SDOF) analysis [1,2]; and (2) finite element analysis methods (e.g.,
[39,29]). The SDOF method is easy to implement and numerically
efficient and it is widely referenced in current design guidelines
[1012] for the blast analysis and design of building components.
However, SDOF analysis is incapable of capturing a spatially and
temporally varying distribution of blast loading, cannot allow for
variations of mechanical properties of the cross-section along the
member, cannot simultaneously accommodate shear and flexural
deformations, can only address strain rate effects indirectly, and
can produce very conservative answers. A finite element analysis
using codes such as LS-DYNA [13] and AUTODYN [14] can be applied to analyze the structural response to blast loads [4] but such
an analysis is rarely used for design-office applications because of
its perceived complexity. An accurate, fast-running finite difference analysis method, suitable for design-office applications is described herein. The proposed method captures the key attributes of

Corresponding author.
E-mail address: cwu@civeng.adelaide.edu.au (C. Wu).

0141-0296/$ see front matter 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2009.07.011

rigorous finite element analysis but retains much of the simplicity


associated with SDOF analysis.
Finite difference (FD) analysis divides a member into discrete
segments of length dx as shown in Fig. 1 to find an approximate
solution to the differential equations of member motion. The
segments in partial difference analysis are divided by nodes, with
the displacement at each node expressed in terms of the difference
in displacement of adjacent nodes. An FD technique is then used
to solve the partial differential beam equations. An FD model can
accommodate any distribution of load along the member, can
capture a variation of mechanical properties of cross-section along
the member, can incorporate both shear and flexural deformations,
and can analyze the deflections along the entire length of the
member: all in contrast to a SDOF analysis where none of these
accommodations is possible. However, only a few studies have
been conducted to analyze the response of RC members subjected
to blast loads using FD techniques. Krauthammer et al. used an
explicit FD technique to the Timoshenko beam theory to analyze
the response of RC elements under uniformly distributed [15,16]
and non-uniformly distributed [17] loads. In both studies, the
dynamic increase factor, which is used to account for strain-rate
effects, is estimated by using an average value of the strain rate
over the depth of a cross-section; although the strain rate is far
from constant over the cross-section depth. Recently, Gong and
Lu [18] used Timoshenko beam theory to analyze the response of

2826

J. Jones et al. / Engineering Structures 31 (2009) 28252832

=0

=0

M0 = 0

Mn = 0

0 = 1

i =0

V0 = 2V1 V2

i =2

i =1

n -2

n-1

n = n-1
Vn = 2Vn-1 Vn-2

dx
Fig. 1. Beam discretization for simple supports in finite difference model.

Fig. 3. Layered analysis of a cross-section.


Fig. 2. Equilibrium of a discrete segment [20].

a beamcolumn frame to blast-induced ground shock, but did not


include the effects of the strain rate. Herein we extend and validate
the FD model proposed by Krauthammer et al. [1517] and Gong
and Lu [18] to incorporate strain rate effects to solve the partial
differential equations of motion of a structural member subjected
to blast loads. We acknowledge that similar results would likely
be obtained using FE analysis and Timoshenko beam elements but
emphasize that FE analysis of blast effects has typically involved
solid finite elements to date.
In this paper, the Duhamel integral [19] is used to derive the
formula to calculate the strain rate profile along the depth of a
cross-section of a simple supported structural member. The derived strain rate profile is incorporated into a section-based layered
analysis model for calculating the momentcurvature function of
the member under blast loads. The derived momentcurvature
function of the structural member is then incorporated into the
Timoshenko beam theory [20] and the FD approach is then used
to solve the Timoshenko beam equations. The FD procedure proposed herein can accommodate changes in the mechanical properties of a members cross section along its length and through its
depth in each time step, making it possible to directly incorporate
both strain rate effects (which will vary along the length and depth
of a member) and non-uniform member loading to solve the partial differential equation of motion of the member. The accuracy
of the proposed FD method is validated using data from field blast
testing [21]. The results of the FD analysis are then compared with
those from both a SDOF analysis and a detailed finite element analysis. The FD procedure is implemented easily and enables accurate
predictions of member response.
2. Finite difference analysis of blast effects
The FD procedure involves discretization of a member into a
series of segments joined by nodes to solve the partial differential
equations of motion. It is different from a finite element analysis
that divides a structural member into three dimensional solid
blocks to carry out the dynamic response analysis.

equilibrium of a beam segment of length dx as shown in Fig. 2


were used in this study to model the member responses under
blast loads as they take into account both shear deformations
and rotational inertia. These effects can be significant in blast
loading due to the fast propagation of stresses in the member.
The Timoshenko beam equations [20,22] to describe the member
motion are as follows

M
2
V = m I 2
x
t
V
2v
+ p = m A 2
x
t

(1)
(2)

where M and V are bending moment and shear force, respectively;


I is moment of inertia; m is material density; A is the crosssectional area; p is the any distributed dynamic load transverse to
beam length; is the rotation of the cross section due to bending;
and v is the transverse displacement of the mid-plane of the beam.
2.2. Moment curvature relationship
In Eqs. (1) and (2), nonlinear flexure can occur anywhere along
the length of the member. The resistance function for flexural
nonlinearity is represented by the moment versus curvature
relationship of the member that can be determined using the
section-based layered analysis model of Fig. 3. In the layered
analysis model, the cross-section is sliced into a large number of
layers; stress and strain are assumed to be constant in a given layer.
The resistance of the ith slice is equal to Fi = i (i )DIFi (i )wb s,
where i (i ) is the stress, DIFi (i ) is the dynamic increase factor in
the slice that changes in every step, and wbi and si are the width
and thickness respectively. The momentcurvature relationship
is computed as per [23] but considering the effect of strain
rate as noted above. For the calculation of the ultimate moment
and curvature, a simplified bilinear stress strain relationship for
concrete material is adopted and the maximum compressive
concrete strain is assumed to be 0.0035, the neutral axis depth
is computed by horizontal force equilibrium on the cross section,
accounting for the effect of strain rate, and the strain profile is
assumed to be linear. The flexural resistance of the cross-section
is calculated by moment equilibrium about the neutral axis.

2.1. Timoshenko beam theory


The Timoshenko beam equations that were derived from
considerations of both vertical force equilibrium and moment

2.2.1. Strain rate profile and dynamic increase factors (DIF)


For the SDOF model of Fig. 4 subjected to a triangular force
function with a peak force of P and duration td , the following

J. Jones et al. / Engineering Structures 31 (2009) 28252832

2827

P(t)
P

P(t)

y(t)

td
(b) Simplified blast loads.

y(t ) = yst

y (t td ) =

1 cos t +
y (td )

td

sin t

t
td

0 t td

(3)
(4)

dx2

(6)

Substituting Eq. (5) into Eq. (6) yields

x
sin .
(7)
L2
L
From Fig. 3, the strain at a height of h above the neutral axis of
a cross section is (x, t ) = (x, t )h. The strain distribution at a
distance x for the tensile rebar, s , and for the concrete, c , in the
compressive zone are

(x, t ) = y(t )L/2

s (x, t ) = y(t )L/2


c (x, t ) = y(t )L/2

2
L2

(h0 Ku d) sin
h sin

x
L

(8)

(9)
L
and the corresponding strain rates for the tensile rebar s and
concrete c are

s (x, t ) = y (t )L/2
c (x, t ) = y (t )L/2

L2

2
L2

(h0 Ku d) sin
h sin

x
L

(10)

(11)
L
where Ku d is the depth to the neutral axis from the extreme compression fiber; h0 is the distance from the extreme compression
fiber to the centroid of tensile reinforcement; and y (t )L/2 is the
velocity response of the member at the mid-span of the member.
Substituting Eq. (3) into Eqs. (10) and (11), it yields
L2

L2

hyst

sin t +

1
td

cos t

1
td

sin

x
L
(13)

The strain rates for both rebar and concrete after td < t can also
be derived by substituting Eq. (4) into Eqs. (10) and (11). Using
the above strain rate profile, DIFi (i ) for concrete can be calculated
using CEB code [24],
DIF =

fcd
fcs

DIF =

(5)
L
where x is a distance from the left support of the member; y(t )L/2 is
the displacement response at mid-span of the member, which can
be determined using SDOF using Eqs. (3) and (4). If plane sections
remain plane after bending, the curvature of a cross-section as
shown in Fig. 3 at a distance of x is the second derivative of the
deflection along the member,
d2 y(x, t )

td

(12)

0 t td .

where is the natural circular frequency of the SDOF system,


yst = P /K and K is the stiffness of the SDOF system.
When a simple supported member under blast loading is in its
elastic range of response, its shape function can be assumed to be
sinusoidal,

(x, t ) =

td

c (x, t ) =

sin (t td ) + y(td ) cos (t td ) td < t

y(x, t ) = y(t )L/2 sin

L2



1
x
1
sin
(h0 Ku d)yst sin t + cos t

0 t td

displacement response is given by Clough and Penzien [19]


1

s (x, t ) =

Fig. 4. Single degree of freedom analysis for blast loads.

Fig. 5. Layered analysis of cross section at tensile yielding.

(a) Single degree of freedom


system.


=

1.026s

1/3

for 30 s1

for > 30 s1

(14)

(15)

where fcs and fcd are the unconfined uniaxial compressive


strength in quasi-static and dynamic loading, respectively. s =
10(6.156s 2.0) , s = 1/(5 + 9fcs /10) and is the static strain rate
(30 106 s1 ). The DIF formula for the rebar is [25]


DIF =

104

(16)

where = 0.074 0.040(fy /414) for yield stress and =


0.019 0.009(fy /414) for ultimate stress, where fy is the steel yield
strength (MPa). After DIFi (i ) for the concrete and the rebar have
been determined, the corresponding stress profile and flexural
resistance can be established by cross-section analysis as shown
in Fig. 3.
We note that the assumed displacement shape function is
strictly applicable only to simply supported beams responding
in the elastic range. For incremental inelastic response involving
hinge formation in the middle of the span, the shape function is
bilinear, with incremental deformation associated only with hinge
rotation in the middle of the span. However, as seen in the tables
in Biggs [1], the differences in the shape functions are modest and
so the sinusoidal (elastic) shape function is adopted herein.
2.2.2. Yield and ultimate moment capacity
To develop a simplified momentcurvature relationship of a
reinforced concrete member, the yield and ultimate points are
determined and the relationship between the origin, yield and
ultimate points is assumed to be linear. For analysis at the yield
point, the strain in the tensile steel is assumed to be equal to the
yield strain and an estimate of the neutral axis depth Ku d is made as
shown in Fig. 3. From the strain diagram with DIF values calculated
from the strain-rate formula in Eqs. (12) and (13), a stress profile
can be derived as in Fig. 5(c) from which the longitudinal forces
acting are calculated as per Fig. 5(d). The curvature is then varied
until equilibrium of the cross-section in axial direction is reached
as indicated in Fig. 5.

2828

J. Jones et al. / Engineering Structures 31 (2009) 28252832

Fig. 6. Layered analysis of cross-section at concrete crushing.

The point of failure by concrete crushing is analyzed by a similar


procedure. For this point, the strain in the outermost compression
fibre of the concrete is held constant at the assumed failure strain
of 0.0035 and the curvature is varied as indicated in Fig. 6 until
horizontal equilibrium on the cross-section is achieved, the neutral
axis depth is identified and the flexural resistance corresponding to
concrete crushing is computed.
The yield and ultimate moment capacities of the cross section
and their corresponding curvatures as well as the moment
curvature diagram in the entire domain are then calculated from
the section-based layered analysis model.

Fig. 7. The approximate slope of a continuous function f (x) at point i.

2.4.3. Moment curvature model


Eq. (18) requires the moment at each node to be known.
The moment at a given node is calculated as a function of the
curvature. Since the momentcurvature relationship is nonlinear
and the internal stresses in the member are cyclic, it is computed
numerically. In the finite difference code developed in this study,
the curvature at each node is input to a subroutine that obtains
the moment at each node. The curvature is given by Krauthammer
et al. [15]

2.3. Shear behavior


In the Timoshenko beam equations, the shear force V at a crosssection is calculated as
V = KAxz = KAGxz

(17)

where G is shear stiffness; K is the correction factor which is used to


take into account the assumption that the shear stress is constant
across the cross-section. For rectangular cross-sections, the value
was given by Krauthammer et al. [15] to be K = 2 /12.
2.4. Numerical technique
A first-order Taylor series expansion for Mx and Qx and a
2
2
second-order Taylor series expansion t 2 and t 2 are used to
approximately solve the Timoshenko beam equations (Eqs. (1)
and (2)) in this paper. The magnitude of the error resulting from
truncating the Taylor expansion to the above terms in the finite
difference scheme is shown in Fig. 7. Small increments in time
and space must be used to limit the error as discussed later. A
complete description of the numerical procedure using the finite
difference method for the analysis of a beam member under blast
loads is provided by Krauthammer et al. [15]. A similar explicit
finite difference scheme was implemented in this study as follows.
2.4.1. Calculating displacement and rotation
Eq. (18) describes how the rotation of node i at time t + 1 is
calculated from values at adjacent nodes at previous time steps.
Eq. (19) describes how the displacement of node i at time t + 1 is
calculated.

it +1 = 2it it 1

t +1
i

= 2
t
i

t 1
i

dt 2

m Ii
dt 2

m Ai

Mit+1 Mit1
2dx

Vit+1 Vit1
2dx


V (i)

+ p(i)

(18)


(19)

where all terms have been defined previously.


2.4.2. Calculation of shear force from shear strain
The calculations of rotations and displacements in Eqs. (18) and
(19) require the shear force at each node at the previous time steps
to be known. The shear at a given node is calculated as a function of
the shear strain. The shear strain is calculated from Eq. (20), with
the partial derivative converted into a finite difference term [15].
t
xzi
=

t it1

= i +1
it .
x
2dx

(20)

it =

t it1

= i +1
.
x
2dx

(21)

2.4.4. Member boundary conditions


The member in the program is divided into n segments as
shown in Fig. 1, with the left support corresponding to i = 0
and the right support corresponding to i = n. If one considers
calculating Eqs. (18) and (19) at i = 0 or i = n, it can be seen that
there are no nodes at i 1 or i + 1 to complete the calculation. This
is dealt with by introducing boundary conditions at the supports,
and performing the analysis from the node i = 1 to i = n 1. The
boundary conditions at the simple supports can be seen in Fig. 1
and the moment and transverse displacement at the two supports
are zero. The shear and rotation at the supports are approximated
by linear interpolation. This is a reasonable assumption because
the moment, and therefore the curvature, is zero at the support. As
the rotation is the integral of the curvature, a very small curvature
will result in only a minor change in rotation at the support.
2.4.5. Stability and convergence
An explicit time integration method is used to solve Eqs. (18)
and (19). In an explicit scheme, the values of rotations and
displacements at time step i + 1 are calculated from values at
time steps i and i 1 and the time step must be small to
guarantee stability of the solution process as described by the
CourantFriedrichsLewy condition.
The number of segments of equal length dx along the length
of the member, or the number of equally spaced nodes along the
member length must be chosen carefully because it influences
significantly the runtime and accuracy of the FD analysis. For the
momentcurvature relationship of Fig. 8 and simply supported end
conditions, once the moment at the central node d has reached
the ultimate moment capacity of the cross-section, an increase in
curvature and decrease in bending moment at the center of the
span will accompany an increase in displacement. To maintain
equilibrium, the neighboring non-yielded nodes c and e will be
unloaded with decreasing curvature. A plastic hinge of length 2
dx forms with much of the displacement being associated with
concentrated rotation in the hinge region. Convergence analysis
was conducted by decreasing the size of the segment by half until
the difference in the displacement response for two consecutive
segment sizes was less than 5% and it was found that a length
of approximately 0.75 times the depth of the slab (i.e. a plastic
hinge length suggested by Warner et al. [26]) satisfied the required
accuracy.

J. Jones et al. / Engineering Structures 31 (2009) 28252832

(a) Curvature distribution before yielding.

2829

(b) Curvature distribution after yielding.

(c) Change in curvature and moment at central and adjacent


nodes after maximum moment attained.
Fig. 8. Momentcurvature relationships along the span.

1000

SectionA-A
10

100

2000
1000
Major
Bending
Plane
Minor Bending Plane

Fig. 9. RC test specimens.

2.4.6. Blast load histories


The blast loading in the finite difference program is applied on
the member as a function of space and time. This allows an arbitrary load distribution to be applied to the member. The shape of
the distributed load across the member is dependant on the charge
weight, shape and standoff distance and the pressure history can
be predicted by current code such as TM5 [10] or using the measured pressure history directly.
3. Validation of models using experimental data
To test the utility of the FD analysis model, the predictions were
compared with maximum displacement data from blast testing
results on 2 m long, simply supported RC slabs that can be modeled
as simply supported beams [21]. The RC specimens were designed
with both tension and compression reinforcement using a 12 mm

diameter mesh, with a 10 mm concrete cover (see Fig. 9 for details).


The mesh bars were spaced at 100 mm centres ( = 1.34%) in the
major bending plane and 200 mm in the minor plane ( = 0.74%).
The concrete had a cylinder compressive strength of 39.5 MPa, a
tensile capacity of 8.2 MPa and a Youngs modulus of 28.3 GPa
at the time of testing. The reinforcing bar had a yield strength of
600 MPa and a Youngs modulus of 200 GPa.
A linear variable displacement transducer (LVDT), accelerometers, and pressure transducers were used to record the response of
the specimen under blast loads. Wu et al. [21] presents complete
details. Fig. 10 identifies the locations of the instruments. The pressure transducers were installed at the center of the specimen (PT1)
and near the supports (PT2) to measure the distribution of pressure
over the face of the specimen.
Two specimens NRC (normal reinforced concrete) A and NRC
B were subjected to different explosive loads at different standoff

2830

J. Jones et al. / Engineering Structures 31 (2009) 28252832

Table 1
Experimental air blast program.
Blast

Slab name

Dimension (mm)

Reinforcement ratio (%)

Standoff distance (m)

Scaled distance (m/kg1/3 )

Explosive used (g)

NRC-1
NRC-2
NRC-3

1A
1A
1B

2000 1000 100


2000 1000 100
2000 1000 100

1.34
1.34
1.34

3
3
1.4

3.0
1.5
0.93

1007
8139
3440

3
Support
PT1
2
Slab
PT2

LVDT

Accelerometer
LVDT
Pressure Transducer (PT)

50 mm
0
Support

20 mm

2ir

(22)

where ir is the measured positive impulse at the center and the


edge. For analysis, the reported values of the peak load were
assumed to apply over the full width of the slab. Since the
negative pressure phase does not affect significantly the maximum
transient displacement of the panel and should have only a
small effect on the residual displacement, we chose to ignore the
negative phase for our computations.
The test specimens were analyzed using the finite difference
analysis model with 23 nodes along the length. Fig. 12 enables

(a) Pentagonal loading at time zero.

0.012

Fig. 12. Predicted and measured displacement histories for NRC 1A.

distances. Specimen NRC A was subjected to two shots, one small


and one large. The experimental test program is summarized in
Table 1. The explosive charge was suspended above the center of
the slab as described in [21].
The experimental pressure readings demonstrate that the
pentagonal distributed load as shown in Fig. 11(a) is a better
approximation of the actual blast load; an expected result given
that the standoff distance and angle of incidence change as a
function of location on the panel. Variables Pr max C and Pr max
are the peak pressures at the center (PT1) and the edge (PT2),
respectively. The pressure time histories at the center and edge
are simplified as triangular blast loads as shown in Fig. 11(b). The
duration of the positive pressure wave td was back-calculated as

prmax

0.008
Time (s)

Fig. 10. Specimen Instrumentation.

td =

0.004

Table 2
Comparison of maximum deflections from predictions and tests.
Test

Max deflection (mm)

NRC-1
NRC-2
NRC-3

Experimental

Predicted

1.8
7.9
14.0

1.9
9.6
13.3

a comparison of the predicted and experimental responses for


the NRC 1A. The finite difference model considering strain rate
effect predicted the measured displacement response very well.
Ignoring strain rate effects in this instance had little effect on
the maximum displacement but overestimated the displacements
during the rebound phase of the response history. Table 2
compares the predicted responses from the finite difference model
with the experimental responses [21]. The finite difference model
accurately predicted maximum displacement responses in all
three tests.
4. Comparison of different analysis methods
To illustrate issues associated with analysis of structural components subjected to blast loads, the displacement response of
specimens 1A and 1B of Table 1 were estimated by SDOF, finite
element and the finite difference analyses. The SDOF analysis was
performed using the industry-standard approach first proposed by

(b) Variation of pressure with time.


Fig. 11. Simplified pressure distributions.

J. Jones et al. / Engineering Structures 31 (2009) 28252832

2831

Table 3
SDOF transformation factors.

Mass factor, KM
Load factor, KL
Loadmass factor, KLM

Elastic region

Plastic region

0.50
0.64
0.78

0.33
0.50
0.66

Fig. 14. Two-curve concrete model with damage and failure.

Fig. 13. Bilinear resistance-deflection curve.

Biggs [1] and later adapted in US Army Technical Manuals such


as TM5-1300 [10] and ASCE standards and guidelines [11,12]. The
SDOF analysis is based on the transformation of distributed component mass, stiffness, resistance and loads to equivalent SDOF
values. The derivation of the transformation factors is reported
elsewhere [1] and not repeated here. Table 3 lists the transformation factors for a simply supported structure under a uniformly distributed load. In the SDOF model, a bilinear resistance-deflection
curve as shown in Fig. 13 was used in the analysis. For a SDOF system undergoing elastic deformation, the yield resistance capacity
Ryield of the cross-section is given by ASCE [11] as
Ryield = 8Myield /L

(23)

where Myield is the yield moment and L is the span. The deflection
at which the yield resistance is reached can be calculated from
yyield = Ryield /K

(24)

where
K =

384EI
5L3

(25)

and E is Youngs modulus and I is the average moment of inertia


of the cross-section and equal to (Ig + Ic )/2, where Ig is the moment of inertia of the gross concrete cross-section and Ic is the moment of inertia of the cracked concrete cross-section that can be
determined from layered analysis. The ultimate resistance (Rult ) of
the section is back-calculated from the ultimate moment capacity
(Mult ) that is computed from the layered analysis:
Rult = 8Mult /L.

(26)

The ultimate deflection (yult ) in Fig. 12 is given by


yult =

L
2

(27)

where is the rotation which is derived by assuming that all of the


rotation in the member takes place over the plastic hinge length
[27,28].
The computer code LSDYNA [13] with the Mat_PSEUDO_
TENSOR solid element model was used to perform the finite
element analysis. In the Mat_PSEUDO_TENSOR model, two yield
versus pressure curves, as illustrated in Fig. 14, with the means of

Fig. 15. Finite element model of the specimen.

migrating from one curve to the other, are adopted for reinforced
concrete. The two yield versus pressure curves take the form

= ao + p/(a1 + a2 p)

(28)

where ao and aof are the cohesions for both undamaged and
failed materials, respectively; a1 and a2 are pressure hardening
coefficients, a1f is the pressure hardening coefficient for the
failed material and p is the hydrostatic pressure. In Fig. 14, max
is the maximum yield strength curve and failed is the failed
material curve. The strain-rate-effect relationships for the rebar
and concrete from the CEB code [24] are also input in the
Mat_PSEUDO_TENSOR model.
Simple tensile failure in Mat_PSEUDO_TENSOR model, namely,
the yield strength, is taken from the maximum yield curve until
the maximum principal stress (1 ) in the element exceeds the
tensile-cut off cut . For every time step 1 > cut , the yield
strength is scaled back by a fraction of the distance between
the two curves. The set of parameters required in this function
can be computed through the following formulas where fc0 is the
unconfined concrete compressive strength [13]: ao = fc0 /4; a1 =
1/3; a2 = 1/3fc0 ; aof = 0; a1f = 0.385; and b1 = 0. The
specimen was modeled using solid brick elements as shown in
Fig. 15. Convergence tests were conducted to investigate how
many elements were needed to achieve a reliable estimation.
This was realized by decreasing the size of the element by half
while keeping loads on the specimen until the difference in the
results between two consecutive element sizes was less than 5%.
The convergence tests resulted in the selection of 400,000 solid
elements employed in the simulation.
The predicted maximum deflections of the tests using the SDOF,
finite element and FD model are summarized in Table 4. The
finite element and FD predictions of the maximum deflection are
much closer to the measured maximum deflections than the SDOF
predictions. The use of the SDOF model gave a very conservative
prediction (the smallest error is 38%), in part because Pr max C was
used in the calculation of the equivalent SDOF load. Although the
LSDYNA model also predicted the maximum deflection well, much
time was required to prepare the FE model and reduce the data.

2832

J. Jones et al. / Engineering Structures 31 (2009) 28252832

Table 4
Predictions of maximum deflections.
Test

NRC-1
NRC-2
NRC-3

Max deflection (mm)

Prediction error (%)

Experimental

FDA

SDOF

LSDYNA

FDA

SDOF

LSDYNA

1.8
7.9
14.0

1.9
9.6
13.3

2.9
12.8
19.3

1.9
8.2
13.1

+6
+21
5

+64
+63
+38

+11
+5
6

5. Conclusion
A finite difference analysis model is proposed in this paper for
the analysis and design of simply supported structural members
subjected to blast loads. A theoretical formula for the strain rate
profile along the depth of a cross-section varying with time and
span is derived for concrete and steel. The stain rate effects are incorporated into the section-based layered capacity model for calculation of the momentcurvature relationship of the member.
Using the Timoshenko theory, variation of blast loads, distributed
stiffness and mass, as well as mechanical properties are coded
into the finite difference analysis model to solve partial differential equations of motion of the member. A comparison between the
measured and analytical responses was made and the largest difference was only 21%, indicating that the finite difference analysis
model can accurately predict the response of a simply supported
structural member to blast loads. However, unlike the finite element analysis that divides a member into three dimensional solid
elements, the finite difference method discretizes a member into a
number of segments joined at nodes. Far fewer segments are used
in the finite difference model than elements in the finite element
model, leading to a substantial reduction in the computational
effort. SDOF analysis is straightforward and suitable for use in a design office but the results can be substantially conservative. The finite difference analysis can capture many of the important features
of a finite element analysis, provides accurate results, is computationally efficient and is ideally suited for routine use in a design
office.
References
[1] Biggs JM. Introduction to structural dynamics. New York, NY: McGraw-Hill
Book Company; 1964.
[2] Mays GC, Smith PD. Blast effect on buildingsdesign of buildings to optimise
resistance to blast loading. London, UK: Thos Telford; 1995.
[3] Mosalam K, Mosallam AS. Nonlinear analysis transient analysis of reinforced
concrte slabs subjected to blast loading and retrofitted with CFRP composites.
Composites 2001;32(8):62336.
[4] Luccioni BM, Ambrosini RD, Danesi RF. Analysis of building collapse under
blast loads. Eng Struct 2004;26(1):6371.
[5] Wang ZQ, Lu Y, Hao H, Chong K. Full coupled numerical analysis approach for
buried structures subjected to subsurface blast. Comput & Structures 2005;
83(45):33956.
[6] Wu C, Hao H, Lu Y. Dynamic response and damage analysis of masonry
structures and masonry infilled RC frames to blast ground motion. Eng Struct
2005;27:32333.

[7] Wu C, Hao H. Safe scaled-distance of masonry infilled RC structures to airblast


loads. Engineering structures. J Perform Constr Facil 2007;21(6):42231.
[8] Scherbatiuk K, Rattanawangcharoen N. Experimental testing and numerical
modeling of soil-filled concertainer walls. Eng Struct 2008;30(12):354554.
[9] Louca LA, Ali RMM. Improving the ductile behaviour of offshore topside
structures under extreme loads. Eng Struct 2008;30(2):50621.
[10] Department of Defence (DoD). Structures to resist the effect of accidental
explosions. US Department of the Army, Navy and Air Force Technical Manual,
TM-5-1300, Washington (DC); 1990.
[11] American Society of Civil Engineers (ASCE). Design of blast resistant buildings
in petrochemical facilities. Reston (VA). 1997.
[12] American Society of Civil Engineers (ASCE) Blast protection of buildings. Ballot
version 2, Reston (VA). 2008.
[13] Livermore Software Technology Corporation (LSTC). LS-DYNA. Version 970
Manual. Livermore (CA, USA). 2003.
[14] Century Dynamics. AUTODYN. Theory Manual. San, Ramon (CA). 2003.
[15] Krauthammer T, Assadi-Lamouki A, Shanaa HM. Analysis of impulse loaded
reinforced concrete structural elements I: Theory. Comput & Structures
1993;48(5):85160.
[16] Krauthammer T, Assadi-Lamouki A, Shanaa HM. Analysis of impulse loaded
reinforced concrete structural elements II: Implementation. Comput &
Structures 1993;48(5):86171.
[17] Krauthammer K, Shanaa H, Assadi A. Response of structural concrete elements
to severe impulsive loads. Comput & Structures 1994;53(1):11930.
[18] Gong S, Lu Y. Combined continua and lumped parameter modeling for
nonlinear response of structural frames to impulse ground shock. J Struct Eng
2007;133(11):122940.
[19] Clough RW, Penzien J. Dynamics of structures. New York (USA): McGraw-Hill;
1993.
[20] Weaver W, Timoshenko SP. Vibration problems in engineering. 5th ed. New
York: Wiley; 1990.
[21] Wu C, Oehlers DJ, Rebentrost M, Burman N, Whittaker AS. Blast testing of ultrahigh performance fibre concrete slabs and FRP retrofitted RC slabs. Eng Struct
2009, in press [doi:10.1016/j.engstruct.2009.03.020].
[22] Gregori JN, Prada MAF, Prada MAF, Filippou FC. A 3D numerical model for
reinforced and prestressed concrete elements subjected to combined axial,
bending, shear and torsion loading. Eng Struct 2007;29(12):340419.
[23] Wight JK, MacGregor JG. Reinforced concrete: Mechanics and design. 5th ed.
Prentice Hall; 2008.
[24] Comit Euro-International du Bton (CEB). CEB-FIP model code 1990.
Redwood Books, Wiltshire (UK). 1993.
[25] Malvar LJ, Crawford JE. Dynamic increase factors for steel reinforcing bars.
Twenty-eighth DDESB (DoD explosives safety board) seminar. Orlando (FL).
1998.
[26] Warner RF, Rangan BV, Hall AS, Faulkes KA. Concrete structures. Addison
Wesley Longman Australia Pty Ltd; 1998.
[27] Wu C, Oehlers DJ, Day I. Layered blast capacity analysis of FRP retrofitted RC
members. Adv Struct Eng 2009;12(3):43549.
[28] Haskett M, Oehlers DJ, Mohamed Ali MS, Wu C. Rigid body rotation mechanism
for reinforced concrete beam hinges. Eng Struct 2008;31(5):103241.
[29] Azevedo RL, Alves M. A numerical investigation on the visco-plastic response
of structures to different pulse loading shapes. Eng Struct 2008;30(1):25867.

Você também pode gostar