Sunteți pe pagina 1din 11

Computers and Structures 81 (2003) 547–557

www.elsevier.com/locate/compstruc

Nonlinear analysis of barge crush behavior


and its relationship to impact resistant bridge design
Gary R. Consolazio *, David R. Cowan
Department of Civil and Coastal Engineering, University of Florida, P.O. Box 116580, Gainesville, FL 32611, USA

Abstract
Bridge structures crossing navigable waterways must not only be designed to resist gravity, wind, and earthquake
loads, but must also be capable of resisting ship and barge collision loads. Design specifications used both in the US and
internationally employ empirical models of vessel crush behavior to produce codified procedures for computing
equivalent static design loads due to vessel impact. In this paper, the ADINA finite element code is used to compute
force-deformation relationships for several hopper barge crushing scenarios. Results obtained from the nonlinear finite
element crush analyses are then compared to empirical crush models found in bridge design specifications.
Ó 2003 Elsevier Science Ltd. All rights reserved.

Keywords: Bridge design; Vessel impact; Barge impact; Collision mechanics; Crush behavior; Contact

1. Introduction type are fundamentally dynamic in nature. As such, the


magnitude of the load imparted to the bridge pier will
A core component of designing bridge structures that vary with respect to time. The peak load generated and
cross navigable waterways consists of computing lateral the rate of load oscillation during the impact are func-
design loads due to vessel impact. Similarly, assessing tions of the type, structural configuration, mass, and
the vulnerability of an existing bridge structure to acci- initial velocity of the impacting vessel; the structural
dental impact by vessels––ships and barges––requires configuration and mass of the pier; and the soil condi-
that lateral impact loads be quantified. In most vessel– tions. Given the complex nature of such dynamic events
bridge collision scenarios, a pier supporting the bridge and the degree of scatter found in the loads generated
superstructure is contacted by a portion (most often the [1], bridge design specifications generally provide sim-
bow) of an errant vessel. The resulting lateral impact plified procedures for computing equivalent static loads
force on the pier is transmitted both downward into the as an alternative to conducting fully dynamic impact
pier-footing and upward to the bridge deck. While ele- analyses. These loads are applied to the bridge in a static
ments of a bridge other than the piers may also be manner and are intended to produce approximately the
subjected to impact loads, most bridge collapses attri- same structural response as that due to the application
butable to vessel collision occur as the result of impact of the corresponding time-varying dynamic impact load.
load imparted to a pier. This is especially true for barge In the US, the American Association of State High-
impact loading––the focus herein––due to the low ver- way and Transportation Officials (AASHTO) bridge
tical profile of barge flotillas. design specifications [2,3] contain procedures for com-
Given that ship and barge impact loads are generated puting equivalent static loads for both ship collision and
by moving vessels having substantial mass, loads of this barge collision events. Included in the AASHTO pro-
visions are equations relating vessel kinetic energy, ves-
sel deformation level (crush depth), and the equivalent
*
Corresponding author. Tel.: +1-352-846-2220; fax: +1-352- static impact force to be applied to the bridge [1–5].
392-3394. Separate relationships are given by AASHTO for
E-mail address: grc@ce.ufl.edu (G.R. Consolazio). ship impact loads and barge impact loads because the

0045-7949/03/$ - see front matter Ó 2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0045-7949(02)00474-1
548 G.R. Consolazio, D.R. Cowan / Computers and Structures 81 (2003) 547–557

experimental and analytical bases for each type of vessel only a subset of the options available in Ref. [2]).
differ [1–4]. The ship impact provisions derive primarily However, the load prediction model implemented, which
from experimental studies conducted by Woisin [6] that is the item of interest here, is the same in both of these
focused on protecting reactors in nuclear powered ships documents.
against accidental impact by other ships. The basis for AASHTO-based load calculations for barge impact
the barge impact provisions, however, comes from ex- design begin with selection of the ‘‘design’’ impact
perimental tests conducted by Meir-Dornberg [7] on condition (barge type and impact speed). Factors such
European hopper barges (discussed in further detail as the characteristics of the waterway, expected types of
later). barge traffic, and the importance of the bridge (critical
Comparing ship and barge impact incidents, the lat- or regular) enter in to this selection process. Once
ter type occurs at a higher frequency because a greater the impact conditions have been identified, the kinetic
number of waterways (having bridge crossings) are energy of the barge is computed as [2]:
navigable by relatively shallow draft vessels (barges)
CH WV 2
than by deeper draft vessels (ships). However, despite KE ¼ ð1Þ
this fact, vessel impact studies published in the literature 29:2
focus more frequently on ship collision mechanics than where KE is the barge kinetic energy (kip ft), CH is the
on barge impact behavior. Even fewer studies focus hydrodynamic mass coefficient (a factor that approxi-
specifically on barge crush behavior during collisions mates the influence of water surrounding the moving
with bridge piers. The logistical problems associated vessel), W is the vessel weight (in tonnes where 1
with conducting full-scale barge impact tests on in- tonne ¼ 2205 lbs), and V is the impact speed (ft/s). It is
service bridge structures––i.e. bridges actively carrying noted that Eq. (1) is simply an empirical version (derived
either vehicular or railway traffic––also hamper efforts for a specific set of units) of the more common rela-
to collect data for purposes of studying barge crush tionship:
characteristics.  
Given that new bridge designs are often controlled by KE ¼ CH 12MV 2 ð2Þ
code-stipulated barge impact loading conditions, devel-
oping an improved understanding of barge collision where KE, M (the vessel mass), and V are all dimen-
mechanics is highly desirable. In this paper, nonlinear sionally consistent.
finite element crush simulations have been conducted to Once the kinetic energy of the barge has been deter-
gain a better understanding of barge crush behavior mined, a two-part empirical load prediction model is
during collisions with bridge piers. All crush analyses then used to determine the static-equivalent impact load.
presented herein were conducted using a high-resolution The first component of the model consists of an empir-
finite element barge model developed as part of this ical relationship that predicts crush deformation as a
study. Importantly, the model was developed strictly function of kinetic energy:
based on structural vessel plans obtained from a US rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ! 
KE 10:2
barge manufacturer. The finite element model is in no aB ¼ 1þ 1 ð3Þ
5672 RB
way dependent on the empirical relationships used in the
AASHTO provisions. As such, the finite element crush
results presented can be considered to be an independent In this expression, aB is the depth (ft) of barge crush
source of barge crush data for purposes of making deformation (depth of penetration of the bridge pier into
comparisons to the current AASHTO code provisions. the bow of the barge), KE is the barge kinetic energy
Before describing the finite element barge model in de- (kip ft) and RB ¼ ðBB =35Þ where BB is the width of the
tail, aspects of the AASHTO barge impact provisions barge (ft).
that are pertinent to this study will be presented. The second component of the load prediction model
consists of an empirical barge-crush model that predicts
impact loads as a function of crush depth:

2. AASHTO methodology for prediction of barge impact 4112aB RB aB < 0:34 ft
loads PB ¼ ð4Þ
ð1349 þ 110aB ÞRB aB P 0:34 ft

Barge impact load calculations conducted according where PB is the equivalent static barge impact load (kip)
to the AASHTO provisions involve the use of both an and aB is the barge crush depth (ft). The crush model
empirical load prediction model and a risk assessment represented by Eq. (4) is illustrated graphically in Fig. 1.
procedure. The AASHTO guide specification for vessel Unfortunately, as a consequence of the fact that very
collision design [2] and the AASHTO LRFD bridge little barge collision data has ever been published in the
design specifications [3] differ in the risk assessment literature, Eqs. (3) and (4) are based on a single exper-
methods available in each document (Ref. [3] provides imental study. During the early 1980s, a study was
G.R. Consolazio, D.R. Cowan / Computers and Structures 81 (2003) 547–557 549

Fig. 1. Barge crush model used in AASHTO bridge design specifications.

conducted in Germany by Meir-Dornberg [7] that in- 3. Finite element analysis procedures and models
volved physical testing of reduced-scale standard Euro-
pean (type IIa) barges. Experimental barge crush data 3.1. Finite element code and solution algorithms
was collected by Meir-Dornberg and then used to de-
velop empirical relationships relating kinetic energy, Analysis of structural crushing is very often handled
depth of barge crush deformation, and impact load. using nonlinear contact finite element simulation. In the
AASHTOÕs relationships, Eqs. (3) and (4), are virtually case of vessel crush simulation, the finite element code
identical to those developed by Meir-Dornberg except used must be capable of robustly representing nonlinear
that a width-modification factor––the RB term––has inelastic material behavior (with failure), part-to-part
been added to approximately account for deviations in contact, self-contact, and large displacements (due to the
barge width from the baseline width of 35 ft (the width significant geometric changes that often occur). The
of barges most often found operating in US inland ADINA finite element code [8] meets all of these re-
waterways). quirements and has been shown in previous studies [9,10]
Interestingly, while Eq. (3) utilizes the RB term to to be capable of accurately simulating complex structural
reflect the influence of barge width, no such factor has crushing. Thus, ADINA was employed throughout the
been included to account for variations in either the size present investigation to study crushing of barges.
(width) or geometric shape of the bridge pier being im- In regard to the selection of static, implicit dynamic,
pacted. Furthermore, since Eq. (4) indicates that the or explicit dynamic solution methods for the purpose of
impact load (PB ) increases monotonically with respect to predicting barge crush behavior, one must examine the
crush depth (aB ), the AASHTO provisions implicitly impact conditions at both the global and local levels.
assume that maximum impact force occurs at maximum From a global viewpoint, when a barge impacts a bridge
crush depth and, therefore, peak impact load can be pier, the inertial forces in the barge––as a whole unit,
uniquely correlated to peak crush depth. In this paper, it including the payload mass––and in the bridge pier are
will be demonstrated through finite element crush sim- quite substantial, and therefore, the event is dynamic in
ulation that this assumption does not hold true for all nature. However, at the local level inside the barge bow,
barge crush scenarios. inertial forces arising in individual structural members
A primary goal of this study was to generate––using are typically small relative to stiffness and contact re-
finite element analysis––independent force-deformation lated forces. Thus, while the forces driving the impact
data that would serve to supplement the experimentally are dynamic, the barge crush behavior itself can be
derived crush-relationships reported by Meir-Dornberg. considered pseudo-static in nature. The fact that Meir-
By comparing simulation results to the static crush- Dornberg found little difference between static and dy-
model assumed by AASHTO (i.e. Eq. (4)), areas of namic barge crushing forces [7] supports this assessment.
potential future improvement in the bridge design From a numerical analysis point of view, then, as
specifications can be identified. long as the time intervals between progressive contact
550 G.R. Consolazio, D.R. Cowan / Computers and Structures 81 (2003) 547–557

events are not very small relative to the time scale of the in this paper, special attention was given to all of these
overall crushing problem, then static or implicit dynamic issues. Accurate representation of geometry and struc-
analysis methods will often be more efficient than ex- tural configuration was achieved by basing the finite
plicit dynamic methods for crush simulation. In this element model on structural barge plans obtained from
study, all analyses were conducted as pseudo-static so- a leading US barge manufacturer.
lutions using the automatic time step and response All of the simulations presented herein involved
smoothing algorithms available in ADINA. crushing a barge model with rigid pier-shaped impac-
As implemented in ADINA, response smoothing tors. In such cases, large deformations occur at the bow
involves adding dynamic effects (mass and damping of the barge. In order to accurately model the crushing
terms) to the system model to smooth the system re- deformation expected in this area of the vessel, a high-
sponse and improve solution convergence. Rather resolution mesh was used. However, the rear portion of
than solving the standard static equilibrium equation, the barge does not experience the same level of defor-
ADINA instead solves the following incremental mation during bow crushing, and therefore a coarser
pseudo-dynamic equilibrium and displacement update mesh sufficed in this region. For simulation efficiency,
equations iteratively at time t þ Dt [8]: the barge model developed was divided into three sep-
arate zones (Fig. 2), each represented with a different
€ ðiÞ þ C U_ ðiÞ þ K ði1Þ DU ðiÞ ¼ R  F ði1Þ
MU
ð5Þ degree of mesh resolution.
U ðiÞ ¼ U ði1Þ þ DU ðiÞ Zone-1, the most detailed region of the barge model,
represents the leading 8.75 ft of the barge bow. In this
where i and i  1 indicate the current and previous zone, steel channel and angle members are joined to
equilibrium iterations respectively, M is the mass matrix, form internal stiffening trusses, while steel plates of
C is a damping matrix computed as C ¼ bK (where b is various thicknesses make up the outer hull shell. In order
a user selected smoothing factor), K ði1Þ is the tangent to capture buckling and contact between the internal
stiffness matrix computed at the end of iteration i  1, components of the barge bow, both the internal struc-
U€ ðiÞ and U_ ðiÞ are the current approximations of acceler- tural members and the outer hull plates of zone-1 were
ation and velocity, DU ðiÞ is the set of incremental dis- discretized into meshes of shell elements (Fig. 3). Four-
placements for the current iteration, R is the external node shell elements based on the mixed interpolation of
load vector, and F ði1Þ is the internal force vector cor- tensorial components (MITC) formulation [12,13] were
responding to displacements U ði1Þ established at the end used throughout zone-1 and zone-2 of the model. The
of the previous iteration. While Eq. (5) includes dynamic MITC4 shell element formulation has been shown by
velocity and acceleration terms, static system response Bathe et al. [10] to perform well in large deformation
can be obtained by making use of very slow loading crush analysis and thus was appropriate for the crush
rates. In this study, response smoothing was found to be simulations conducted herein.
beneficial in stabilizing and improving the convergence In addition to modeling the structural members and
of very slow rate (static) barge crushing simulations and the outer hull plates of the barge bow, it was also nec-
was therefore used throughout. Furthermore, a large essary to model the connections between them. Typi-
displacement formulation was used for all cases simu- cally, structural members such as channels and angles
lated, and full Newton iteration was used to iteratively are welded together to create internal stiffening trusses
solve Eq. (5). onto which formed (bent) steel hull plates are welded to
complete the barge bow. In regions where the structural
3.2. Barge finite element model members and outer hull plates were joined together
by welds, ADINA nonlinear (large-displacement) rigid
In the AASHTO barge impact provisions, the jumbo links [8] were used to tie corresponding shell elements
class hopper barge––measuring 195 ft in length, 35 ft in together.
width, and 12 ft in depth, as defined in Refs. [2,3]––is the Contact-pair definitions were also established in the
baseline vessel upon which Eq. (4) was established model to detect contact between internal components of
(RB ¼ 1 for this barge). To permit comparisons to be the barge bow during crushing. Using such algorithms,
made between finite element predicted crush behavior however, requires special care in the spatial positioning
and the empirical AASHTO crush model, the jumbo of elements. Like other contact finite element codes,
hopper barge was chosen as the vessel to be modeled in ADINA performs initial penetration checks at the be-
this study. Simulating the crush behavior of vessels [11] ginning of any analysis involving contact. While most
places special demands on the finite element model de- contact codes, including ADINA, provide features that
velopment process. Crush models must be constructed either ‘‘correct’’ or ignore initial penetration problems,
with sufficient mesh resolution that local buckling, allowing such procedures to take place is usually not
yielding, and internal contact can be adequately repre- desirable as unintended, and oftentimes nonphysical,
sented. In developing the hopper barge model presented model behavior can result. In the barge model developed
G.R. Consolazio, D.R. Cowan / Computers and Structures 81 (2003) 547–557 551

Fig. 2. Overview of mesh zones used in finite element barge model.

Fig. 3. Cutaway view of barge bow revealing mesh resolution used to model structural components in zone-1 and zone-2.

herein, care was taken in positioning the structural contact during bow crushing. Therefore, the trusses were
members and hull plates such that the projected thick- modeled using line-meshes of elastic three-dimensional
nesses from the mid-surfaces of the shell elements did beam elements based on a large-displacement (small-
not overlap and, therefore, initial penetrations did not strain) formulation. As the nodal symbols in Fig. 4 in-
occur. dicate, each physical truss member was broken into
Zone-2 of the barge represents the 19.75 ft of the several interconnected beam elements so that, through
barge bow that lies behind zone-1. The internal stiffening large-displacement analysis, overall member-level buck-
trusses in this zone were not expected to sustain local- ling (but not flange local buckling) could be represented.
ized flange buckling, plastic deformation, or internal Rigid links were used to connect the beam elements
552 G.R. Consolazio, D.R. Cowan / Computers and Structures 81 (2003) 547–557

representing the truss members in zone-2 to the shell appropriate beam elements to represent welded con-
elements representing the truss members in zone-1. nections.
Rigid links were also used to connect the beam elements Zone-3 of the model represents the rear 166.5 ft
representing the truss members in zone-2 to the shell hopper portion of the barge. Significant deformations
elements representing the gusset plates in this same zone. were not expected in this zone during bow crushing, thus
The outer hull plates in zone-2 were modeled using it was modeled using a coarse mesh of eight-node solid
MITC4 shell elements but with a coarser mesh resolu- elements. Nodes along the bottom surface of zone-3
tion than that used in zone-1. Tied contact definitions [8] were fixed against motion to provide restraint against
were used to join the top and bottom hull-plates to the the crushing forces generated at zone-1.

Fig. 4. Finite element modeling of internal stiffening trusses.

100

600
80

500
True stress (MPa)
True stress (ksi)

60
400

300
40

200

20 Large strain steel material model


100

0 0
0 0.05 0.1 0.15 0.2 0.25 0.3
True strain

Fig. 5. Stress–strain relationship used to model barge steel.


G.R. Consolazio, D.R. Cowan / Computers and Structures 81 (2003) 547–557 553

With regard to material modeling, significant yielding rithmic) strain data shown in Fig. 5 was used. In zone-2
was expected to occur during crushing of the hull plates and zone-3, all beam and solid elements were assigned a
and structural members contained in zone-1. As such, all simple linear elastic material model representing steel.
shell elements in the barge model (both in zone-1 and in In total, the barge model contained approximately
zone-2) were assigned a plastic-multilinear material 25,000 shell elements, 2000 beam elements, 250 solid
model matching the stress–strain data shown in Fig. 5. In elements, and required the solution of approximately
all crush analyses presented in this paper, a large-dis- 200,000 finite element equations (including contact
placement, large-strain formulation was used for the shell equations) when merged with an impactor model to
elements. Therefore, the true (Cauchy) stress, true (loga- conduct crush simulations.

Fig. 6. Crush simulation models. (a) Crush model consisting of barge and circular pier impactor, (b) crush model consisting of barge
and square pier impactor.
554 G.R. Consolazio, D.R. Cowan / Computers and Structures 81 (2003) 547–557

3.3. Pier impactor models and crush conditions finite element simulation data. Results obtained from
the circular crush simulations are presented in Fig. 7.
Of particular interest in this study was the crush Forces acting on the pier (i.e. the impactor) are shown to
behavior that occurs when barges collide with concrete gradually and monotonically increase with correspond-
bridge piers. As such, the geometric shapes of the im- ing increases in crush depth. In general, the crush
pactors developed herein matched the two most common characteristics are also shown to be slightly sensitive to
bridge pier shapes––circular and square. Each impactor variations in pier width, but the effect is not strongly
was modeled using eight-node solid elements, was as- pronounced.
signed a nearly rigid linear elastic material model, and In Fig. 8, crush results are presented for the square
was positioned along the longitudinal axis of the barge impactor simulations. Several key differences between
(Fig. 6) at a location that initially produced a small gap the circular and square crush cases are immediately evi-
between the surfaces of the barge and the pier impactor. dent. In the square crush cases, the contact forces are
A contact definition was then defined between these observed to rise very rapidly and––for small deforma-
surfaces with a Coulomb friction value of l ¼ 0:3 (ap- tion levels––the overall crush behavior is seen to be
proximate frictional coefficient for steel sliding on con- much stiffer than in the circular cases. However, after
crete). Nodes on the back face of the impactor (opposite the contact force has maximized, the stiffness of the
the contact surface) were assigned a displacement time barge diminishes rapidly in the square cases. In fact,
history that translated the pier toward the barge at a whereas all of the circular crush analyses predicted a
constant rate of 0.4 in./s to generate crushing at the monotonically increasing relationship between force and
barge bow. Use of the very slow 0.4 in./s displacement crush depth, none of the square analyses exhibited this
rate ensured that the solution of Eq. (5) predicted the characteristic. In addition, whereas diameter had very
static, rather than dynamic, crushing response of the little effect on the crush behavior observed for circular
barge. piers, Fig. 8 indicates that in square crush conditions,
there is a definite relationship between pier width and
force generated.
4. Discussion of crush simulation results That pier geometry can influence the crush behavior
of a barge is not surprising when consideration is given
Static barge crush analyses were conducted for three to the internal structure of such vessels. In the bow of a
circular impactors (diameters: 4, 6, and 8 ft) and for barge, numerous internal trusses run parallel to one
three square impactors (widths: 4, 6, and 8 ft). For each another (recall Figs. 2 and 3) resulting in significant
level of imposed impactor penetration, i.e. barge crush stiffness in the longitudinal direction. During impact
depth, the total force acting at the contact interface with a pier, both the shape and size of the pier determine
between the pier and the barge was extracted from the the number of internal trusses that actively participate in

Crush depth (mm)


0 50 100 150 200 250 300
1000

800 3.5
Crush force (kips)

3
Crush force (MN)

600
2.5

2
400
1.5

4 ft. Circular impactor


1
200 6 ft. Circular impactor
8 ft. Circular impactor
0.5

0 0
0 2 4 6 8 10 12 14
Crush depth (in)

Fig. 7. Predicted barge crush behavior for circular impactors.


G.R. Consolazio, D.R. Cowan / Computers and Structures 81 (2003) 547–557 555

Crush depth (mm)


0 50 100 150 200 250 300
1000

800 3.5
Crush force (kips)

Crush force (MN)


3
600
2.5

2
400
1.5

4 ft. Square impactor


1
200 6 ft. Square impactor
8 ft. Square impactor
0.5

0 0
0 2 4 6 8 10 12 14
Crush depth (in)

Fig. 8. Predicted barge crush behavior for square impactors.

resisting bow crushing. Fig. 9 illustrates the internal simultaneously resist crushing, thus producing a very
deformation produced by 12 in. of crush depth imposed stiff response. However, when the trusses buckle, and
by a 6 ft diameter circular impactor. Fig. 10 illustrates therefore soften (Fig. 10), they all do so at approxi-
the same scenario, but for a 6 ft wide square impactor. mately the same deformation level. As a result, there is
In the circular case, bow deformation is concentrated in an abrupt decrease in the overall stiffness of the barge (as
a narrow zone near the impact point and only the trusses was illustrated in Fig. 8 where the crush forces plateau
immediately adjacent to this location generate signifi- or even decrease after maximizing). In addition, in-
cant resistance. As increasing crush deformation occurs, creasing the width of a square pier increases the number
additional trusses participate and the force increases in of trusses that simultaneously participate in crushing,
an approximately monotonic manner. thus explaining the sensitivity of force magnitude to
In contrast, when a flat-surface square impactor impactor width that was evident in the square crush
bears against the barge, several trusses immediately and simulations. As was noted earlier, the AASHTO crush

Fig. 9. Barge deformation generated by circular pier impactor.


556 G.R. Consolazio, D.R. Cowan / Computers and Structures 81 (2003) 547–557

Fig. 10. Barge deformation generated by square pier impactor.

model given in Eq. (4) includes a correction factor (RB ) While a more extensive review of structural designs
for barges that deviate from the 35 ft width of the used throughout the barge manufacturing industry is
standard hopper barge. However, the finite element re- needed before it can be stated with confidence that the
sults presented here suggest that it may also be appro- AASHTO provisions are overly conservative, the pre-
priate to include parameters reflecting the effects of pier liminary finite element results presented here suggest
shape and pier size in the crush model as well. that such a review is warranted.
A comparison of the empirical AASHTO crush Finally, the square impactor data shown in Figs. 8
model, Eq. (4), and finite element predicted crush be- and 11 demonstrate that, for some pier configurations,
havior (for the 8 ft circular and 8 ft square crush cases) is crush force does not necessarily increase monotonically
presented in Fig. 11. For cases involving significant with respect to crush depth. In such cases the maximum
barge crushing (e.g. more than 6 in.), the forces pre- force generated cannot necessarily be correlated to the
dicted by finite element simulation are significantly less maximum crush depth sustained (e.g. in an impact
than those predicted by the AASHTO crush model. condition). This fact is important because the AASHTO

Crush depth (mm)


0 50 100 150 200 250 300
1600 7

1400
6

1200
5
Force (kips)

1000
Force (MN)

4
800

3
600

2
400
ASSHTO barge crush model
8 ft Circular impactor
200 1
8 ft Square impactor

0 0
0 2 4 6 8 10 12 14
Crush depth (in)

Fig. 11. Comparison of crush simulation data and AASHTO crush model.
G.R. Consolazio, D.R. Cowan / Computers and Structures 81 (2003) 547–557 557

load prediction procedure described earlier in this paper Acknowledgements


(Eqs. (1), (3) and (4)) assumes that the equivalent static
impact force can be uniquely correlated to (or predicted The authors wish to thank the Florida Department
from) peak crush deformation sustained by a barge of Transportation (FDOT) for providing the financial
during an impact event. The square impactor crush re- support that made this study possible. The material
sults presented here indicate that this procedure should presented in this paper is based upon research that was
be reexamined since impact force does not appear to be supported by the FDOT under Contracts BC354
uniquely correlated to maximum deformation. RPWO#23 and BC354 RPWO#56.

References
5. Concluding remarks
[1] Knott M, Prucz Z. Vessel collision design of bridges:
The primary objectives of this study have been to bridge engineering handbook. CRC Press LLC; 2000.
predict, using nonlinear finite element analysis, the static [2] AASHTO. Guide specification and commentary for vessel
crush behavior of hopper barges and to compare the collision design of highway bridges. American Association
finite element predictions to empirical crush models of State Highway and Transportation Officials, 1991.
currently used in bridge design specifications. In order to [3] AASHTO. LRFD bridge design specifications and com-
mentary. American Association of State Highway and
accomplish these objectives, high resolution finite ele-
Transportation Officials, 1994.
ment models of a jumbo hopper barge and multiple pier [4] Knott MA. Vessel collision design codes and experience in
shaped impactors have been developed to conduct static the United States. In: Gluver H, Olsen D, editors. Ship
crush simulations using the ADINA finite element code. collision analysis. Rotterdam: A.A. Balkema; 1998.
The modeling techniques used to represent the structural [5] Whitney MW, Harik IE, Griffin JJ, Allen DL. Barge
components of the barge and the crush analysis proce- collision design of highway bridges. ASCE J Bridge Eng
dures employed have been presented. 1996;1(2):47–58.
Barge crush behavior has been predicted for both [6] Woisin G. The collision tests of the GKSS. Jahrbuch
circular and square shaped pier impactors, and both the Schiffbautech Gesellsch 1976;70:465–87.
shape and size of the pier have been found to influence [7] Meir-Dornberg KE. Ship collisions, safety zones, and
loading assumptions for structures in inland waterways.
the relationship between force and deformation during
VDI-Berichte 1983;496:1–9.
crushing. In cases involving crushing by a square im- [8] ADINA Research and Development Inc. ADINA theory
pactor (or any flat-faced impactor), the width of the and modeling guide. Report 01–7, Watertown, MA, 2001.
impactor has been found to influence the magnitude of [9] Bathe KJ, Guillermin O, Walczak J, Chen H. Advances in
forces generated. Simulation data have also indicated nonlinear finite element analysis of automobiles. Comp
that forces generated during crushing do not necessarily Struct 1997;64(5/6):881–91.
increase monotonically, and therefore impact force [10] Bathe KJ, Walczak J, Guillermin O, Bouzinov PA, Chen
cannot necessarily be uniquely correlated to maximum H. Advances in crush analysis. Comp Struct 1999;72:31–
sustained deformation. Based on these observations, 47.
areas of focus for future improvement to design speci- [11] Lehmann E, Yu X. Inner dynamics of bow collision to
bridge piers. In: Gluver H, Olsen D, editors. Ship collision
fication based load prediction are as follows: the effects
analysis. Rotterdam: A.A. Balkema; 1998.
of pier shape and size parameters in force determination; [12] Bathe KJ. Finite element procedures. Englewood Cliffs,
and relationships that directly predict impact load from NJ: Prentice-Hall; 1996.
impact energy rather than using crushing deformation–– [13] Bathe KJ, Dvorkin EN. A formulation of general shell
which cannot be uniquely correlated to force––as an elements––the use of mixed interpolation of tensorial
intermediate step. components. Int J Numer Meth Eng 1986;22:697–722.

S-ar putea să vă placă și