Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Ocean Engineering
journal homepage: www.elsevier.com/locate/oceaneng
art ic l e i nf o a b s t r a c t
Article history: The Common Structural Rules for Oil Tankers (OT) and Bulk carriers (BC) of IACS mandate the compu-
Received 30 January 2015 tation of shear flow in hull girders. Such computation involves determinate as well as indeterminate
Accepted 25 August 2015 shear flow evaluation due to the multi-celled nature of the OT and BC transverse sections. This neces-
sitates formulation of an algorithm for automatic detection of closed cells within a ship section as well as
Keywords: for automatic determination of positions of virtual slits for computation of indeterminate shear flows. It
Shear flow is also common practice to neglect the effect of longitudinal stiffeners when evaluating the shear flow in
Shear stress distribution cross sections or to consider such effect using a gross approach. In the current work, the computation of
Statical moments shear flow in practical ship transverse sections has been performed. For this purpose an algorithm and
Ship structures
subsequent program was developed to automate shear flow calculations which is applicable to any ship
Thin-walled beams
type. The results of the program were validated against benchmark examples as well as finite element
results. Case studies were performed on VLCC and BC midship sections. The results also prove that
consideration of longitudinal stiffeners does not significantly affect the magnitude of the shear stresses
developed in the ship plating.
& 2015 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.oceaneng.2015.08.048
0029-8018/& 2015 Elsevier Ltd. All rights reserved.
G.K. Choudhary, K.M. Doshi / Ocean Engineering 108 (2015) 678–691 679
Srinath, 2003, Vlasov, 1961). Ship hull girders being multi-celled 2.2. Implementation of automatic calculation of shear stresses in
open-closed cross sections, however, present difficulties for eva- multi-celled sections
luation of shear flow and shear stresses as the automation algo-
rithm for such sections is not adequately clarified from the A number of computer program implementations to automate
aforementioned references. shear flow evaluation have been proposed. Shama (1971) wrote a
Conventional practices for calculating shear stresses have program to perform shear flow analysis specifically for tankers and
sought to consider the cross section areas of the longitudinal used it for optimization of shear carrying members. Marshall (1974)
vertical members, such as longitudinal bulkheads, side shell, gir- developed a FORTRAN program based on Batho-Bredt equations for
ders, and projected areas of inclined members on a vertical plane, the calculation of shear flow in multi-celled thin-walled beams. Its
for e.g., hopper tank, to get the average shear force generated in application to a ship hull girder was demonstrated. However, the
the vertical members. Thus conventional practices are based on program had limitations as it assumed a symmetric structure, and
engineering approximations. However, the maximum stresses are the positions of virtual slits and closed cells had to be manually fed
higher than the average shear stresses, and it is not straightfor- as user inputs to the computer program. Alfano et al. (1996) used
ward to calculate these peak stresses from their average and the Depth First Search algorithm developed by Tarjan (1972) and an
hence, the use of this approach is not advisable. Hence, attempts Open Section Cut algorithm they proposed to automate shear flow
were made to develop empirical formulae and parametric rela- calculations with bare minimum inputs of section geometry, and
tions between ship scantlings and shear flow. demonstrated the application to a ship hull girder. However, the
Shama (1969) used five independent parameters of main formulation of the program was unclear due to the academic nature
plating scantlings, breadth-to-depth ratio and distance of trans- of its description in terms of tensors, vectors and integrals (Edlund,
verse bulkheads from the centerline of the section to study their 1997). Accordingly, Edlund (1997) expanded the work of Alfano
effect on shear stress distribution in ship hull girders. In this et al. (1996) and gave formulae for the integrals and matrices
parametric study, the effect of stiffeners and girders was assumed appearing in the computer program. Utility of the procedure was
to be taken into account by smearing their total area over the main demonstrated by computation of shear stresses for a passenger ship
plating of the section. Shama (1971) concluded from a parametric section, however, without including longitudinal stiffeners. Prokić
study on oil tankers using effective thickness approach that the (2000) also developed a program on the lines of Alfano et al. (1996).
The coordinate system had to be rotated to the principal coordinate
transverse position of longitudinal bulkheads and the relative
system of the beam which decreased the efficiency of the program
effective thicknesses of the side shell and the longitudinal bulk-
by increasing the number of calculations.
heads influence the maximum shear stress values significantly.
Thus computer programs have been proposed for shear flow
Shama (1975) then performed a parametric study of shear stresses
evaluation. However, all the programs except for the one written
in bulk carriers considering variables such as the scantlings of the
by Edlund (1997) have limitations.
hopper plate, top wing tank, internal partitions of the top wing
tanks, width of hatch openings, and depth of the double bottom
2.3. Classification society approaches
structure. Further study (Shama, 1976) concluded that double
bottom plates do not contribute to resistance to vertical shear
Rules of various classification societies (BV, 2013; DNV, 2014; GL,
forces and the side shell between the hopper and top wing tanks
2013; IRS, 2014; LR, 2013; NK, 2013) and CSR (IACS, 2014) prescribe
carries about 35% of the total shear force making it a critical por-
simplistic evaluation of the net shear force carried by the vertical
tion insofar as shear stresses are concerned. Effects of longitudinal
hull girder members for an applied vertical shear force for various
stiffeners were neglected. Shama (2010) again proposed thickness
candidate ship configurations. The effect of stiffeners appears to be
smearing for bulk carriers and oil tankers.
neglected/implicitly considered in all these configurations, and the
In aerospace applications, particularly in aircraft wings, the
empirical formulae given are only valid for a vertically applied shear
general practice is to lump the stiffeners as point masses or areas force. These simplistic evaluations may need to be augmented for
called ‘booms’ assuming the size of the stiffeners to be small consideration of horizontal shear forces as well. However it is
compared to the dimensions of the cross section (Megson, 2012). acknowledged that some of the rules also recommend the use of
References of this methodology being applied to ship structures, explicit shear flow calculations (ABS, 2014; CSR IACS, 2014, etc.).
however, were not available within the references studied in the
present work. Stiffener lumping is a better approach as compared to 2.4. Summary from literature survey
thickness smearing since moments of inertia, position of center of
gravity, etc. are better approximated by lumping rather than by From the literature survey, the following salient findings were
smearing which only approximates the area correctly and not other noted:
section properties. However, if an automatic shear flow calculation
algorithm and a convenient 2-D modeling methodology are avail- 1. Analysis of shear stresses in ship hull girders can be performed
able, then stiffener lumping as well as the as-is consideration of by considering them as thin-walled multi-celled open-closed
stiffeners require roughly the same amount of effort from the user cross sections.
since in the former case, stiffener properties (positions, areas, and 2. While theoretical background for shear analysis of open-closed
moments of inertia) have to be manually fed to the program, thin walled cross sections is available, illustrative examples
whereas in the latter case, stiffeners have to be directly modeled. providing insight into step by step calculation of shear flow in
Also, stiffener lumping is more suited for manual shear flow cal- complex multi-celled open-closed sections is scarce. Also rele-
culations since automating these calculations requires exactly the vant literature with application to ship structures is also rare.
same algorithm as that required for evaluation without lumping. 3. Computer implementation of automatic shear analysis of thin-
Thus has been the gradual shift from average shear stress walled cross sections has been performed previously. However
determination to equivalent thickness by stiffener smearing per- the practical applications attempted were relatively simple
taining to ships. To the best of the authors' knowledge, stiffeners when compared to ship hull girders. Hull girders have been
have not been considered in ships either by lumping or as-is analyzed with such programs (Alfano et al., 1996; Edlund, 1997),
within the literature surveyed. but stiffeners were not considered.
680 G.K. Choudhary, K.M. Doshi / Ocean Engineering 108 (2015) 678–691
4. With regards to analysis of ship sections, longitudinal stiffeners Consecutive nodes: two nodes at the two ends of a member
are either not considered, or are empirically considered by along which a path is being traversed.
smearing the stiffeners as an enhanced thickness as given in
Shama (2010). To the best of the authors' knowledge insofar as
3.3. Symbols
literature is concerned, stiffener lumping used in aircraft indus-
try has not been applied to ship structures.
5. Shear stress evaluation in ship hull girders is based on a sim-
A Area of the beam cross section
plistic approach from the perspective of rules of classification
E Young's modulus of the beam material
societies.
G Shear modulus of the beam material
Iyy, Izz, Moments of inertia about the neutral axes parallel to
The above findings elucidate that shear flow analysis of ship
Iyz the Y and Z axes
hull girder cross sections requires further investigation. Con- N Number of independent loops or circuits in a given
sideration of stiffeners with smearing/lumping approximations beam cross section
can be eliminated by automating shear flow calculations based on Known determinate shear flow at a given point A
qAd
first principles of structural mechanics. It is possible to then obtain
qBd Unknown determinate shear flow to be calculated at
the shear flow at any given location within the ship hull girder
point B
considering the longitudinal stiffeners as well.
qid * Known determinate shear flow in the members of the
The objective of the present paper is to demonstrate an algo-
ith loop
rithm for the automated shear flow analysis of a complete hull
qid Unknown indeterminate shear flow in the jth loop
girder section including stiffeners. The algorithm requires the user j
to supply the hull girder cross section geometry and the applied s Distance traversed along the section from point A to B
shear force as inputs. Complex steps such as detection of closed Sy, Sz Statical moments or first moments of area about the
loops, location of virtual slits, and calculation of determinate and neutral axes
indeterminate shear flow are automated to provide the final shear t(s) Thickness as a function of s
flow output at any given location. Vy, Vz Y and Z components of the applied shear force on the
The remainder of this paper is arranged as follows. Section 3 section
explains the notations and terminology used in this paper, Section yc, zc Y and Z coordinates of the center of gravity of the
4 summarizes the theory and explains the algorithm of the beam cross section
MATLAB program developed in this research, Section 5 demon-
strates the application of the program on two hull girder sections
with stiffeners, and lastly, Section 6 gives the conclusions of this
research. 4. Theory and algorithm
3.2. Definitions The aforementioned steps have been explained in detail below.
4.2. Automatic loop detection and creation of virtual slits this may or may not result in the dynamic creation of additional
free nodes and removal of bridge nodes. This step is repeated till
For a cross section with N independent cells or loops, N virtual no unvisited free node exists in the structure, including the ones
slits are created in the loops such that no loop exists after the created after deletion of members and paths in this step.
creation of the slits, and no portion of the structure gets com- 2. Check if any members are still left after step (i). If none exist,
pletely disconnected from the original structure, i.e., a single loop detection ends at this stage as this means no more loops
structure exists before and after the creation of virtual slits. So exist. If any members still exist, at least one more loop exists,
long as the aforementioned conditions are adhered to, the position and the next step is proceeded to.
of virtual slits can be chosen arbitrarily. 3. At this stage, label all the left over nodes and members as
The loop detection and slit creation algorithm proposed in this unvisited. Start a path from any node and traverse randomly until
research is understood easily due to its intuitive nature. Fig. 1 gives a previously visited node is encountered, which means a loop has
the flowchart of the algorithm. The general steps of the algorithm been detected. While traversing a path, when bridge nodes are
given below are to be performed on a copied set of variables as encountered, any member may be randomly chosen to continue
these have to be modified as per the algorithm: with the path. This further illustrates the generality of the algo-
rithm, knowing that the final solution is path-independent.
1. If no free nodes exist, proceed to the next step. If any free nodes 4. Record the nodes and members of the detected loop in the
exist, one by one, traverse paths starting from free nodes and proper order. Create a temporary slit in that loop randomly at
ending at the first bridge node encountered at the respective any one of its nodes and store its node number in a separate
paths. Along the path, delete all the nodes and members that variable. This node number is later used in the creation of vir-
are encountered except for the last (bridge) node where the tual slits. Creation of temporary slit results in the creation of at
path was stopped. Update the set of bridge and free nodes as least one free node. Go to step (i).
START
YES YES
Stop path and start
a new path Start a new path at There is at least one
any free node: loop. Start anywhere:
i = Current node i = Current node
NO
NO
NO Is ‘j’ a previously
Is node ‘j’ a visited node?
bridge node?
YES
Delete member ‘m’ YES
and node ‘i’
New Loop Found.
Cut the newly found Store the members &
loop at any node. node numbers.
At the end of the aforementioned steps, no members will exist. The temporary slits are created in the same member, i.e., the single
original geometry of the section stored in memory is used to create structure does not get split into two new structures. Thus, at this stage,
virtual slits (if loops exist) at positions same as those of the temporary loops have been automatically detected and virtual slits have auto-
slits stored in memory from the aforementioned steps. To create a matically been created. Note that at this stage at least two free nodes
virtual slit at a given node in a given member, a new node is created at exist in the structure.
the position of the given node and the node number of the given node
appearing in the row corresponding to the given member in the 4.3. Choosing shear flow paths and calculation of determinate shear
member connectivity variable is simply replaced with the new node flow
number. These new node numbers are stored in memory and are later
replaced with original node numbers after paths are chosen. The The steps outlined in Section 4.2 convert originally closed or
coordinates of the new nodes are appended at the end of the existing open-closed sections into (temporarily) determinate open sections,
node coordinate variable. The above algorithm ensures that no two whereas originally open sections remain unaffected. A set of paths
START
NO Is there an unvisited
Stop path and start Is there any member with a
a new path unvisited free node node ‘i’ such that
remaining? vn(i) = nm(i) – 1?
YES NO
NO
Is node ‘i’ a END
bridge node? YES YES
Determinate shear
flow at this node is 0:
Assign:
q d (i) = 0 Determinate shear
i =j
flow going OUT of this
node is superposition
Check which member
Calculate determinate of all shear flows
node ‘i’ belongs to:
shear flow qd (j) at flowing INTO the
m = Current member
node ‘j’ as per (4) node ‘i’ from the
(nm(i) – 1) previously
visited members
Go from node‘i’ to
Assign:
remaining node of
vm(m) = vm(m) + 1
member ‘m’:
vn (j) = vn(j) + 1
j = Next node
END
Fig. 2. Algorithm for choosing paths and finding determinate shear flow.
G.K. Choudhary, K.M. Doshi / Ocean Engineering 108 (2015) 678–691 683
Table 1 Table 2
Principal Particulars of VLCC and S1 BC hull girder sections. Section properties of the VLCC and S1 BC hull girders.
Principal Particulars Unit VLCC S1 BC Item Comp. Units VLCC properties S1 BC properties
⎛V I − V I ⎞ ⎛V I − V I ⎞ This step is only required if loops exist in the given cross sec-
∆q AB = qB − q A = − ⎜⎜ ⎟ ∆Sz − ⎜ z zz ⎟ ∆Sy
y yy z yz y yz
2 ⎟ ⎜ 2 ⎟ tion. Creation of virtual slits results in strain incompatibility at the
⎝ Iyy Izz − Iyz ⎠ ⎝ Iyy Izz − Iyz ⎠ (1)
position of virtual slits as there is slipping there. Using the
determinate flow calculated as outlined in Section 4.3, strain
compatibility conditions are imposed in the form of a no-slip
sB
condition at the nodes where the virtual cuts have been created.
∆Sy = ∫s ( z (s) − zc ) t (s) ds There are N unknowns as the indeterminate shear flows, viz.,
(2)
A
q1id , q2id , …, qNid , and a set of N simultaneous compatibility equations
684 G.K. Choudhary, K.M. Doshi / Ocean Engineering 108 (2015) 678–691
Table 3
Summary of ANSYS and present results of VLCC: starboard part.
Member Maximum shear flow (N/ Maximum shear stress (N/mm2) Horizontal shear force Vertical shear force (kN) % share in vertical
mm) (kN) shear
4.6. Program validation inflow of shear flow at a given node equals net outflow), com-
paring the position of the shear center (obtained by a ‘no net
The proposed algorithm was implemented in MATLAB. A pro- torsion’ equilibrium condition) with that given by ANSYS, and
gram was developed for modeling thin-walled beam cross sections integrating the shear stresses over the section to obtain the net
and for computing the shear flow in response to transverse shear internal shear force generated in the section to compare it against
applied in any direction. The program was validated against ana- the external applied shear force. The program satisfies all the
lytical shear flow solutions of thin rectangular, channel, ‘I’ and aforementioned checks; however, some of these have not been
single-celled hollow rectangular cross sections available in litera- shown here for the sake of brevity.
ture. The program was also validated against results of finite ele- Comparison was also done with available shear flow results
ment (FE) analysis of a user-defined beam section in the com- (IACS, 2014) for a representative bulk carrier cross section con-
mercial FE software, ANSYS, as illustrated in the illustrative sidering unit vertical shear force. The comparison of shear flow
examples given in Section 5. Other validation methods include results is shown in Fig. 3. It is observed that the present results are
checking adherence to the continuity equation of shear flow (net in good agreement with reference results.
686 G.K. Choudhary, K.M. Doshi / Ocean Engineering 108 (2015) 678–691
Table 4
Summary of ANSYS and present results of BC: starboard part.
Member Maximum shear flow (N/mm) Maximum shear stress (N/mm2) Horizontal shear force Vertical shear force % share in vertical
(kN) (kN) shear
at the shear center of the VLCC whereas a vertical shear force of 2.0✕
107 N is assumed to act at the shear center of the S1 BC.
The shear flow in the stiffeners can also be obtained using the
proposed algorithm as well as ANSYS. Since there are 362 stif-
feners in the VLCC, the shear flow results of some sample stiffeners
have been shown below for illustration. It is notable that the total
Fig. 7. Shear flow distribution in BC due to a vertical shear force using present method. contribution to vertical shear resistance of all the 362 stiffeners
combined together was less than 1% of the applied shear.
5. Application and results The comparisons of shear flow results in the webs of stiffeners
at various locations of the VLCC section are shown in Fig. 8(a)–(f).
The application of the program has been demonstrated with the It can be seen from the figures that the present results agree well
help of two ships modeled along with their stiffeners, viz., a Very Large with those given by ANSYS except towards the ends of the stif-
Crude Carrier (VLCC) and a BC. The results were compared with those fener webs where deviation is observed owing to stress con-
given by the commercial finite element software ANSYS with its centration effects due to the presence of sharp corners.
BEAM188 element assigned with the properties of a user-defined
beam cross-section (modeled and stored separately using MESH200
plane elements) of the ship hull girder section. The principal parti- 6. Comparison with rules of classification societies
culars of the VLCC and the S1 BC are given in Table 1. The cross section
properties of both the midship sections calculated by ANSYS and the Classification societies provide simplistic formulae for the cal-
proposed algorithm are given in Table 2. It is seen from Table 2 that culation of shear forces in the side shell and longitudinal bulk-
the properties found by the algorithm agree well with ANSYS. heads of ships. The present results of shear forces obtained have
been compared with the forces that are obtained by the empirical
5.1. Program input data: ship geometry and loads formulae of some of the classification societies. Some society for-
mulae give the combined forces in the inner and outer side shell
The schematic the VLCC and S1 BC ship cross sections can be found whereas others give separate formulae for the inner and outer side
from Figs. 4 and 5. A vertical shear force of 1.0✕108 N is assumed to act shell. The values in Table 5 have been given accordingly.
G.K. Choudhary, K.M. Doshi / Ocean Engineering 108 (2015) 678–691 687
75 40
ANSYS Result ANSYS Result
30 16
15 8
0 0
0 375 750 1125 1500 0 255 510 765 1020
Distance along the Tee web (mm) Distance along the Tee web (mm)
15 200
ANSYS Result ANSYS Result
12 160
Present Results Present Results
9 120
6 80
3 40
0 0
0 87.5 175 262.5 350 0 150 300 450 600
Distance along the Angle web (mm) Distance along the Tee web (mm)
175 12.5
ANSYS Result ANSYS Result
Shear flow in the flat bar (N/mm)
140 10
Present Results Present Results
105 7.5
70 5
35 2.5
0 0
0 82.5 165 247.5 330 0 145 290 435 580
Distance along the Flat Bar (mm) Distance along the Tee web (mm)
Fig. 8. (a)-(f). Shear flow result comparison in stiffeners at various locations of the VLCC hull girder.
Table 5
The following conclusions are drawn based on the analysis of
Comparison of empirical formulae of various societies with present results.
the obtained results in Table 5:
Rules Longitudinal Outer side Inner side Total side
bulkheada shella shella shella 1. The formulae of LR and NK give results within engineering
accuracy.
DNV 20302 (1.56%) – – 29698
(0.36%)
2. The remaining rules (IRS, DNV, GL and CSR) give results within
GL 17492 (12.50%) – – 32508 acceptable accuracy. However, the side shell double skins are
(9.85%) considered together instead of separately calculating the forces
IRS 20075 (0.43%) – – 29925 (1.12%) carried by each skin.
LR 20551 (2.81%) 14970 14479 –
(0.61%) (0.35%)
NK 20075 (0.43%) 15412 14512 –
(2.33%) (0.12%) 7. Conclusions
CSR 20551 (2.81%) – – 29449
(0.48%) Following summarize the findings of this research:
Present 19990 15062 14530 29592
a
Values given in brackets are % deviation between shear forces of the empirical 1. Shear flow and stress distribution can be found accurately using
formulae from the shear force given by MATLAB. the methodology proposed in this research. The method has
688 G.K. Choudhary, K.M. Doshi / Ocean Engineering 108 (2015) 678–691
been validated against analytical solutions of beam sections, A.1. Loop detection
manual calculations and FE analysis of ship hull girder sections.
2. Stiffeners can be modeled in the proposed method. Current 1. The first stage of the algorithm is to delete all members not
methodology of shear flow calculations in ships uses stiffener belonging to any loop.
smearing with the main plating as an increased equivalent thick- 2. The algorithm starts its first path at free node 10, moves to node
ness. However, this restriction is lifted by the proposed method, 9 and detects it as a bridge node. The current path is stopped
enabling the shear flow calculation even within the stiffeners. and member 13 is deleted. Node 9 is no longer a bridge node.
3. Stiffeners do not directly provide significant amount of shear 3. A new path is started at free node 11. Member 12 is traversed
resistance. The total resisting force of stiffeners even after all of and a bridge node is detected at node 8. The path is stopped and
them are considered together is still negligible. This is evident member 12 is deleted. Node 8 is no longer a bridge node.
from the VLCC example wherein the contribution of shear force 4. Another path is started at free node 12. The algorithm traverses
of stiffeners is of the order of 1% of total applied force. member 11 to reach node 8 (which is currently not a bridge
4. The results of analysis of a VLCC hull including the stiffeners node), and continues along member 10 onto node 7 which is
using the proposed method have been compared with the detected as a bridge node. The path is stopped and members 11
empirical formulae of international classification societies to and 10 are deleted.
find the adequacy and accuracy of the given formulae for cal- 5. No new free nodes are detected at this point, and all the
culating shear forces in the hull. It has been found that these remaining members (if any) are now part(s) of some loop(s).
formulae give results with a good accuracy, at least for the hull The second stage of loop detection begins next.
girder considered in this research. 6. A new path is started at a random node, and is randomly tra-
5. Empirical formulae for quick evaluation of shear flow due to versed until a previously visited node is encountered. In this
horizontal shear forces may not be readily available. Manual cal- case, say nodes 1–7–2–3–9–6–5–2 are traversed, and the path is
culations that are hence required are tedious for a complicated stopped at node 2 since it has already been visited. This means
section such as a hull. The proposed method can be used for the that nodes 2–3–9–6–5–2, i.e., members 3–8–9–5–7 form a loop,
quick determination of shear flow even in these cases. which is stored in memory. A cut is made in the loop at a ran-
6. The proposed methodology also has applications in shear flow dom member, say in member 9 at node 9. In this case two free
calculations in thin-walled structures in other fields of engi- nodes are generated at node 9 (say 9 and 9’).
neering, such as aircraft wings in aerospace engineering and 7. Since free nodes 9 and 9’ are detected, the first stage of loop
bridge box girders in structural engineering. detection (step 1) is restarted on the remaining structure.
Members 3, 8, 9, and 5 are deleted at this stage.
8. Similar steps are carried out till the remaining loop with nodes
1–7–2–5–4–1 and members 1–2–7–4–6 is detected and stored in
Acknowledgments memory. A virtual cut is created in member 7 at node 2, which
again results in two free nodes (2 and 2’) and stage 1 is restarted.
The authors thank Mr. A.R. Kar (Vice President – Research) of 9. At this point, the nodes 2–5–4–1–7–(2’) are traversed. All
Indian Register of Shipping (IRS), Mumbai, India, for his support, members are deleted, no structure remains, and the loop
and IRS for funding this research under project number RRD13018- detection stage ends with 3 things stored in memory: nodes
RR. Any views expressed in this publication are solely the views of contained in the loops, members making up the loops, and
locations (member and node) of the virtual cuts.
the authors and not necessarily those of IRS.
Appendix A. : example illustrating the proposed algorithm In this stage, computation of the determinate part of shear flow
begins.
A beam cross-section with relevant details for calculations is
1. Virtual cuts are recreated in members 9 and 7 at nodes 9 and 2,
shown in Fig. A.1. The cross section is composed of closed cells as respectively. Paths for determinate shear flow calculation are
well as members not within closed cells thus ensuring generality. chosen next.
A vertical shear force of 1 MN is assumed to act on the cross- 2. Each new path is started from a node such that the number of
section. Following sections explain the steps followed by the unvisited members from that node is 1 less than the total
algorithm. number of members being joined at that node. Each path is
stopped when a bridge or a free node is encountered. The fol- 2. Table A.2 is created, for generating equations using (4). The
lowing paths are obtained in that order: relevant values are highlighted in bold in the table. For this
a. Path 1: Nodes 10–9–3–2–7 example, Eq. (A.1) and two unknowns viz. the indeterminate
b. Path 2: Nodes 11–8 shear flows in loops, are obtained and solved for.
c. Path 3: Nodes 12–8
d. Path 4: Nodes 9–6–5 (Node 9 in member 9 is a free node due
to the virtual cut)
e. Path 5: Nodes 2–5 (Node 2 in member 7 is a free node due to ⎧ id ⎫
⎡ 600.00 − 100.00⎤ ⎪ q1 ⎪ ⎧ 71855.63 ⎫
the virtual cut) ⎢⎣ ⎥⎦ ⎨ id ⎬ = ⎨ ⎬
− 100.00 466.67 ⎪ ⎩ 144245.03⎭
f. Path 6: Nodes 5–4–1–7 ⎩q ⎪ ⎭ 2 (A.1)
g. Path 7: Nodes 7–8
3. The determinate shear flows are then calculated by formulae 3. The first equation in (A.1) is obtained as follows, for example.
(1)–(3), and are given in Table A.1, starting with 0 shear flows at The first coefficient turns out to be 600 as per (4) and the
all free nodes, and for bridge nodes we make use of the property second coefficient turns out to be -100 because the directions of
that the net directional inflow at any node equals the net the loops overlapping the common member (i.e. 7) are opposite.
directional outflow. Finding the right hand side from (4) is a matter of integration of
shear flow along the members of the loops and adding them up.
(Note that for members of constant thicknesses, shear flow
A.3. Indeterminate shear flow variation is at most parabolic.) On solving, the indeterminate
shear flows in loops 1 and 2 are 177.62 N/mm and 347.16 N/mm,
In this stage, computation of indeterminate part of shear flow is respectively.
performed.
1. The two loops stored by the loop detection algorithm are used Superposition of determinate and indeterminate (directional)
along with the determinate shear flows calculated above in shear flows is straightforward. The final results of shear flow
Table A.1. analysis are given in Table A.3. The schematic diagram of shear
Table A.1
Determinate shear flows calculated by the algorithm.
Path number Member number Nodes traversed Determinate shear flow (N/mm) Remarks
Table A.2
Indeterminate shear flow calculation.
LOOP 1
LOOP 2
flow distribution in the beam section obtained by the algorithm is stresses at the same location coordinates were also evaluated
shown in Fig. A.2. (A few intermediate nodes have been added in using the present algorithm for ensuring a logical comparison. A
the figure for illustration purposes.) minute difference between the stresses at the actual “nodes” is
In a separate FE analysis, stresses generated in the beam cross- observed when comparing Table A.3 and Table A.4 because of this
section are obtained and given in Table A.4 for comparison. For reason. It is observed that the results of the algorithm and FEM
this purpose, the locations of stress evaluation in FE model were are in good agreement with each other. Accuracy is further
chosen to be slightly away from the nodes (by approx. 3 times the improved as one moves away from the member end nodes where
thickness of the applicable member). This is because FE based
stress concentration effects decay rapidly.
evaluation also captures actual stress concentration effects which
cannot be accounted for with the present analytical method. The
Table A.3
Final shear flows calculated by the present method.
Path number Member number Nodes traversed Shear flow by present method (N/mm) Shear stress by present method (N/mm2)
Fig. A.2. Schematic diagram of final shear flow distribution in the cross-section as per the algorithm.
Table A.4
Comparison of results of proposed algorithm and FE analysis (ANSYS).
Member Number Nodes traversed Shear stress by present method (N/mm2) Shear stress by FEM (ANSYS) (N/mm2)
a
Not the actual nodes but locations in close proximity to the nodes indicated in the Fig. A.1.
G.K. Choudhary, K.M. Doshi / Ocean Engineering 108 (2015) 678–691 691
References Nippon Kaiji Kyokai (NK), 2013. Rules for the Survey and Construction of Steel
Ships: Part C: Hull Construction and Equipment.
Pilkey, W.D., 2008. Formulas for Stress, Strain, and Structural Matrices, second ed.
Alfano, G., Marotti De Sciarra, F., Rosati, L., 1996. Automatic analysis of multi celled John Wiley & Sons, Inc., Hoboken, NJ, USA.
thin walled sections. Comput. Struct. 59, 641–655. Pilkey, W.D., 2007. Analysis and Design of Elastic Beams: Computational Methods,
American Bureau of Shipping (ABS), 2014. Rules for Building and Classing Steel first ed. John Wiley & Sons, Inc., New York.
Vessels: Part 3 Hull construction and equipment. Prokić, A., 2000. Computer program for determination of geometrical properties of
Bureau Veritas (BV), 2013. Rules for the Classification of Steel Ships: Part B: Hull and thin-walled beams with open-closed section. Comput. Struct. 74 (6), 705–715.
Stability. Shama, M.A., 2010. Torsion and Shear Stresses in Ships, first ed. Springer.
Det Norske Veritas (DNV), 2014. Rules for Classification of Ships Part 3: Hull and Shama, M.A., 1976. Analysis of shear stresses in bulk carriers. Comput. Struct. 6,
Equipment – Main Class. 75–79.
Edlund, S., 1997. Arbitrary and Thin Walled Cross Sections: Theory and Computer Shama, M.A., 1975. Shear stresses in bulk carriers due to shear loading. J. Ship Res.
Implementation (Licentiate thesis). KTH, Stockholm, Sweden, ISSN 1103-4270. 19 (3), 155–163.
Germanischer Lloyd (GL), 2013. Rules for Classification and Construction: Part 1: Shama, M.A., 1971. On the optimization of shear carrying material of large tankers.
Seagoing Ships. J. Ship Res. 15 (1), 74–96.
Indian Register of Shipping (IRS), 2014. Rules and Regulations for the Construction Shama, M.A., 1969. Effect of ship section scantlings and transverse position of
and Classification of Steel Ships: Part 3: General Hull Requirements. longitudinal bulkheads on shear stress distribution and shear carrying capacity
International Association of Classification Societies (IACS), 2014. Common Struc- of main hull girder. Int. Shipbuild. Prog. 16 (184), 357–369.
tural Rules for Bulk Carriers and Oil Tankers. Slivker, V., 2007. Mechanics of Structural Elements: Theory and Applications.
Lloyd’s Register (LR), 2013. Rules and Regulations for the Classification of Ships: Springer.
Part 3: Ship structures. Srinath, L.S., 2003. Advanced Mechanics of Solids, second ed. Tata McGraw Hill,
Marshall, R.W., 1974. Shear flow distribution in multi-cell girders. Comput. Struct. 4 New Delhi.
(2), 307–325. Tarjan, R., 1972. Depth first search and linear graph algorithms. SIAM. J. Comput. 1
Megson, T.H.G., 2012. Aircraft Structures for Engineering Students, fifth ed. Elsevier, (2), 146–160.
Boston. Vlasov, V.Z., 1961. Thin-Walled Elastic Beams, first ed. National Science Foundation,
Neiman, A.S., 1948. Shearing stress distribution in box girders with multiple webs. Washington, D.C..
Trans. Am. Soc. Civ. Eng. 114 (1), 162–168.