Ship Structures II Lecture Notes
Uploaded by
Dina ANDRIAMAHEFAHERYShip Structures II Lecture Notes
Uploaded by
Dina ANDRIAMAHEFAHERYLecture Notes for Engineering 6003
Ship Structures II
Claude Daley, Professor, D.Sc., P.Eng.
Faculty of Engineering and Applied Science
Memorial University
St. John’s, CANADA
E6003 – Ship Structures II L1 - ii
© C.G.Daley 2014
T
Taab
blle
eooff C
Coon
ntte
ennttss
Topic 1: Introduction......................................................................................................................... 1
Topic 2: Elastic Plate Behaviour - The basic ship structural element .............................................10
Topic 3: more plate behaviour - add yielding...................................................................................16
Topic 4: Mohr: State of Stress .........................................................................................................22
Topic 5: Plastic Bending Limit States in Laterally Loaded Plates ....................................................27
Topic 6: Elastic Bending in Restrained Laterally Loaded Plates .....................................................34
Topic 7: Membrane Behavior in Restrained Laterally Loaded Plates ..............................................39
Topic 8: Low Aspect Ratio Plates, and Permanent Set ...................................................................46
Topic 9: Plastic Plate Strength Based on Plate Folding ..................................................................51
Topic 10: Introduction to Elastic Buckling ........................................................................................58
Topic 11: Elastic Buckling by Energy Methods................................................................................64
Topic 12: Buckling of Transverse Plate Panels ...............................................................................72
Topic 13: Buckling of Longitudinal Plate Panels..............................................................................77
Topic 14: Post-Buckling Strength of Long Plates ............................................................................86
Topic 15: Local Buckling .................................................................................................................94
Topic 16: Overall Stiffener Buckling ................................................................................................97
Topic 17: Fatigue/Fracture in Ship Structures ...............................................................................102
Topic 18: The Fatigue Process in Ships ........................................................................................112
Topic 19: Discussion of Hot Spot Fatigue Analysis .......................................................................118
Topic 20: Section Modulus – a measure of moment capacity .......................................................126
Topic 21: Strain Rate Effects.........................................................................................................133
Topic 22 : Dealing with Uncertainty in structural design ................................................................139
Topic 23: R-Q calculation of the probability of structural failure ....................................................146
Topic 24 : Engineering Safety Metrics ...........................................................................................154
Topic 25 : Partial Safety Factors ...................................................................................................160
Topic 26 : Risk Analysis - trying to think of everything! .................................................................166
E6003 – Ship Structures II 1
© C.G.Daley 2014
Topic 1: Introduction
The course is intended to develop “advanced” knowledge of structures. The
particular focus is on various types of structural failure in ships. Many of the
topics presented in these lectures are the subject of ongoing research. This
means that our understanding of these issues is still developing.
The photo shown below illustrates an important starting point about ship
structures. Ships are a collection of stiffened panels (plates with frames
attached). To understand the structural behavior of ships, it is necessary to
understand the behavior of stiffened panels.
Kvaerner Masa-Yards Inc.
While there are many causes and symptoms of structural damage, there are
three main types of damage that are essential to understand;
Buckling due to compression (from global bending stress)
Fracture (usually resulting from fatigue from global loads)
Local denting (elasto-plastic bending from local loads)
The sketch below highlights these three damage types. In this course we will
examine the mechanisms for these types of failure.
E6003 – Ship Structures II 2
© C.G.Daley 2014
General Concepts and Trends
Humans have been constructing structures for a long time. A structure is a
tool for carrying (carrying what is in or on the structure).
Ship structures have evolved like all other types of structures (buildings,
aircraft, bridges ...). Design was once purely a craft. Design is evolving as we
understand more about the structure itself and the environment that we
subject it to.
Structural Design
Traditional Design
Built by tradition (prior example)
changes based primarily on experience (some analysis)
essentially a “Craft”
Iron Bridge, Coalbrookdale, UK
First iron bridge
The transition from tradition.
E6003 – Ship Structures II 3
© C.G.Daley 2014
Engineering Design
Incorporates analysis based on math/science
Builds upon prior examples
common designs are codified (building code, class rules..)
new designs should follow the “Engineering Method”
methods are evolving (see below)
Superslender Monohull Ropax
Kvaerner Masa-Yards Inc.
E6003 – Ship Structures II 4
© C.G.Daley 2014
Engineering Design: Working Stress Design
‘Working stress design’ refers to designs which employ an analysis of the
intact structure, meaning the as-built structure in an elastic (unbuckled)
state of stress. The design rules ensure that there is sufficient elastic
capacity under the design load.
example : simple beam under uniform load:
we can solve the beam equation to determine the maximum stress, and
require sufficient section modulus to ensure that the max. stress is less
than the yield stress under the load.
There are several concerns/questions with working stress design:
- what is so important about “yield at the extreme fiber”?
o first yield would be important if it led to unacceptable
behaviour, such as a fracture, or an instability.
o Brittle materials fracture without any ductile stretching.
E6003 – Ship Structures II 5
© C.G.Daley 2014
o Very simple structures (single load path) have no way to re-
distribute the load, and thus exhibit signs of structural failure
at the first stage of material failure.
o Some high-precision structures (aircraft turbines, most
‘machines’) loose their functionality with even small permanent
distortions.
o In most steel structures, and certainly in plated structures such
as are found in ships, first yield remains a very localized
behaviour, with no observable deterioration in the functioning of
the structure.
- can we observe “first yield at the extreme fiber” ?
o This is a very difficult thing to do. Structures are full of residual
stresses even in the “unloaded” condition. A visual inspection
will nether show the initial stresses, nor any local yielding. A
system of installed strain gauges would need to be extremely
extensive and costly to ever show local yielding. The gauge
would have to be right over the point of yield (very difficult, as
peak stresses occur at discontinuities), and functioning when the
peak load occurs. We would need to get the gauge to read the
initial stress, which is theoretically possible, but very difficult.
o In practice, I don’t know of anyone ever observing initial yield.
We know it must happen a lot, but we only observe the
distortions that happen much later.
- will all structures designed using ‘working stress’ be equally
satisfactory? (and what does satisfactory mean?)
o no. some will have no reserve, while some will have large
reserve. If structures are “optimized” under working stress, they
will tend to have the least reserve.
- what about variable loads? flaws in the material?
o Variable loads, and many other uncertainties are handled with a
single, standard, factor of safety.
You should be able to explain the concept and weaknesses of “working
stress design”
E6003 – Ship Structures II 6
© C.G.Daley 2014
Engineering Design: Limit States Design
‘Limit states design’ refers to designs which employ an analysis of the
failure of the structure, meaning the processes and mechanisms that
cause unacceptable behaviour. The design rules ensure that there is
resistance to various types of failure.
example: compare two simple beams under a patch load:
we can solve the non-linear structural behaviour (FE or experiment) to
determine the various failure mechanisms, loads and deflections. The
design rules would require sufficient geometry (shear area, plate
thickness, plastic section modulus) to ensure that the member can carry
the load without excessive deformation. Structures that are optimized for
yield strength tend to be relatively tall and slender, with little reserve.
‘Robust’ structures, while having a lower yield point, can have a larger
range of essentially linear behaviour, with significant reserve.
The two frames sketched below illustrate the different types of behaviour
that frames can exhibit.
E6003 – Ship Structures II 7
© C.G.Daley 2014
At a conceptual level, the figure below sketches a general load-deflection
curve for a ductile steel structure. Note that yielding takes place only part
way up the linear portion of the load deflection curve. Redistribution of
stress occurs as the plastic strain field grows. The <essentially> linear
potion of the curve ends when the strain field can redistribute no more
and a mechanism is formed (e.g. a hinge or a fracture). After this the
structure may continue to support further load increase, now in a new
way (e.g. by membrane behavior). With more deflection, the new behavior
may reach a new stress limit, start a new redistribution process and lead
to another mechanism. One of the mechanisms will define the overall
ultimate capacity. It is reasonable to set the design capacity at the first
plastic mechanism, where the capacity is well above yield, but yet prior to
large permanent deformations.
from ISSC 2003 – San Diego
E6003 – Ship Structures II 8
© C.G.Daley 2014
Subtopic: Structural Yield vs Material Yield
Structural yield is point on the structure’s load vs. deflection plot (see
figure below) where non-linearity first occurs. Structural load-deflection
curves for steel structures are notionally similar to the load-deflection
material property curves for steel that we are used to seeing (see second
figure below), however for a structure the load and deflection
parameters are generally gross parameters; e.g. the deflection at the
centre of a grillage vs. the load applied to the whole grillage.
STRUCTURAL LOAD VS DEFLECTION CURVE
MATERIAL LOAD VS DEFLECTION CURVE
There are several concerns/questions with Limit States design:
- are there equations (solutions) that adequately describe the behaviour?
o In many cases we have no general solution of the limit state that
will give an expression of the influence of various parameters.
o There are potentially very many candidate limit states, which
may interact in unknown ways. We may never be able to find
algebraic solutions for all behaviours. In such cases we either
E6003 – Ship Structures II 9
© C.G.Daley 2014
use exact solutions for similar behaviours, or use simulation
(finite element analysis).
- can we corroborate (observe) the predicted limit states?
o Yes. One of the key benefits of limit state analysis, is that the
limit states are generally observable. We can see when collapse
has occurred.
- In limit states design, a probabilistic assessment of loads and strength
is often included. Can we describe the probabilities adequately?
o There are generally two types of uncertainties, objective and
subjective. Objective uncertainty refers to matters that can be
measured and quantified with probability distributions.
Subjective uncertainty deals with the many quantities for which
little or no empirical data exists (worries and hunches).
E6003 – Ship Structures II 10
© C.G.Daley 2014
Topic 2: Elastic Plate Behaviour - The basic ship structural
element
steel frames inside the icebreaker MV TERRY FOX
Introduction
Ships are cut, almost entirely, from steel plate. Only a few structural
elements are not made from plate (e.g. a few castings and some rolled
sections). An understanding of plate behaviour is therefore crucial. Some
components such as deck plate between supports, behave as ideal plates
under simple boundary conditions and loads. Other components, such as the
webs of large frames are subject to more complex loading. Still others such
as deck/stiffener combinations behave somewhat like simple plates, but with
special aspects.
In this lecture we will discuss elastic plate bending in general, simplify this
for the case of long plates and derive the differential equation of a long plate
(bending in one direction).
For now we will assume the following aspects of what is called “small
deflection theory”:
1. Plane sections remain plane, hence – strain is linear
2. Deflections are small (sin ) , typically wmax< ¾ t
3. Plating remains elastic
E6003 – Ship Structures II 11
© C.G.Daley 2014
We start by sketching a small portion of plate bending in two directions, but
with no in-plane stress. In other words the stress at the mid-plane will be
zero, and (equivalently) the average stress in the x or y direction will be zero.
that z 0 .
There is bending stress and strain in both x and y directions. We assume
The general 2D elasticity relationships amoung stress and strain (Hooke’s
Law for 2D case) are;
x y
x =poisons ratio
y x
E E
y
y x , which after substitution gives z 0
E E
z
E
Aside: Note on units of strain
strain = length/length
Strain is unit-less, but we typically use the units or micro-strain
10-6 in/in or 10-6 m/m
steel in ships is 0 to +- 1200
eg. typical range for elastic strain in
corresponds to y =250 MPa
E6003 – Ship Structures II 12
© C.G.Daley 2014
Long Plates
We will focus on the behaviour of “long” plates. A long plate generally has a
length of at least 3x the width;
Long plates bend in one direction, meaning they form a cylindrical shape
(locally). The cylindrical bending occurs towards the center of the plate (say
at least ‘b’ away from the ends). There are end effects, but the maximum
stresses and deflections occur in the center region, and are essentially
identical to that of an infinitely long plate.
A sketch of long-plate bending :
in long plates, the natural symmetry that each unit strip has with each
neighbouring strip prevents any strain in the y direction. This causes a non-
zero stress in the y-direction:
x y
x
y x
E E
y 0 (2)
E E
E6003 – Ship Structures II 13
© C.G.Daley 2014
y x
(2) is re-arranged to: (3)
E E
x
(3) into (1) gives: x x
E E
which is x
x
E
1 2
(4)
x
1 2 x
E
or
which can be written as x E x where E
1 2
E
This means that plates behave like beams, but with a modified modulus. We
can use beam equations for long plates.
e.g. for a beam, length L, inertia I, load q (N/m)the central deflection is:
wmax
5 pL4
384 EI
for a long plate of width b, length a,
pressure p :
5 pab 4
384 E I
wmax
E6003 – Ship Structures II 14
© C.G.Daley 2014
Partly as a review, lets derive the flexural differential equation for a beam,
which is equally valid for a long plate.
Take a unit strip of plate:
The external moment M balances with the internal moment created by the
stress distribution x(z) (see sketch) :
M ( z ) z dz
t
2
x
t
2
Note that we have sketched a shape for x(z), but we need to confirm this. To
find x(z), we will use Hooke’s Law, and a pattern of deformation. We assume
that the plate deforms from straight into a circular curve of radius ‘r’ (at the
NA).
The circular deformation results in a
change in ‘fibre’ length, depending on z:
L
x (z )
L
from similar triangles
z L
x (z )
z r
L L
or
r L
x ( z)
z
which gives:
r
Hence for long plates: x ( z )
1 2 1 2 r
E E z
x ( z )
E6003 – Ship Structures II 15
© C.G.Daley 2014
We substitute x(z) back into the moment equation to get;
M 1 z dz
t
2
E z
2
t
r
r (1 2 ) t
2
z dz
t
2
z3 t 2
r (1 2 ) 3 t 2
E E
2
]
2
M
Et 3
12(1 ) r
1
2 note: M seems to have units
of force (e.g. N). This is
D flexural rigidity
because M is a moment per
We define: unit length of plate (i.e. N-
D
Et 3 m/m).
12(1 2 ) note:
for a beam: D=EI
For a plate : D=E’ t3/12
hence:
M
D flexural rigidity
=
r radius of curvature
Now we need to relate the radius ‘r’ with the plate deflection ‘w’/
curvature (odd, but curvature = 1/radius of curvature)
1
We define
r
a slope at a
For a curved plate, radius ‘r’:
d b a
r
dx
1
dx dx dx dx r
if w(x) is the plate deflection, then
d 2 w( x) d ( x) 1
( x)
dw( x)
dx dx 2 dx r
consequently:
M
This is the moment-
d 2 w( x) deflection differential
D
dx 2 equation
E6003 – Ship Structures II 16
© C.G.Daley 2014
Topic 3: more plate behaviour - add yielding
structural elements in the FPSO Tera Nova
Introduction
In this lecture we will review the general set of differential equations for
plate bending. Then we will discuss yielding of the plate.
In the previous lecture we derived the moment-deflection relationship as:
M D
d 2 w( x)
dx 2
As for a beam there are a family of differential equations as follows;
w(x) =deflection
w = slope
w
1 M
= moment / rigidity
r D
w
V
= shear / rigidity
D
w = load / rigidity
P d 4 w( x) P
D dx 4 D
E6003 – Ship Structures II 17
© C.G.Daley 2014
1st Aside: dynamics
The equation w’’’’(x)=P(x)/D is a static equation (no time!).
We can include time and dynamics quite easily. We could include time
in a slowly varying situation, (called quasi-static) as follows;
w( x, t )
P( x, t )
D
The above equation does not include inertial effects in the plate,
(i.o.w. the acceleration of the plate is not considered in the force
balance).
For a dynamic equation we need to add a term that gives the inertial
P( x, t ) m( x) w
force;
w( x, t )
( x, t )
D
Note that the primes refer to derivatives in x, while the dots refer to
derivatives in time.
2nd Aside: bending in 2 directions
for a plate bending in 2 directions
we can not assume that y=0
we need to express the deflection
as w(x,y). This results in a general equation for static equilibrium in
terms of partial derivatives;
4 w( x, y) 4 w( x, y) 4 w( x, y) P( x, y)
2
x 4 x 2 y 2 y 4 D
this is abbreviated as; 4 w( x, y)
P( x, y)
D
The analytical solutions for this equation can be complex, even for
simple cases. These problems are more often solved numerically (e.g.
FE method).
E6003 – Ship Structures II 18
© C.G.Daley 2014
The equations given above all represent linear elastic behaviour. The
solutions depend on the pattern of load and boundary conditions.
There are many standard solutions tabulated in, for instance, Roark.
We will move on to the post-yield behaviour.
As an introduction, the behaviour of a plate under increasing uniform
lateral load (e.g bottom plating), is shown;
The plate can resist end moments,
but is free to pull in.
elastic deflections
yielding starts at the plate edges
(location of highest stress)
plastic hinges form at edges
yielding starts at the center
plastic hinges form at the center,
creating a mechanism -> collapse
E6003 – Ship Structures II 19
© C.G.Daley 2014
Yield Criteria
There are many material failure criteria, typically dependent on stress or
strain. For ductile materials, yield is the material behaviour of interest. It is
commonly modeled with the von-Mises yield criteria, as described below;
For the case of a uniaxial stress, the stress reaches yield at y.
Biaxial stress is limited by an envelope of points along a curve.
Consider a small cube of material, subject to biaxial stress.
The three spatial coordinates are labeled 1,2,3. One of the
stresses is zero.
We have:1, 2 , 3=0. There are no shear stresses (these are
principal stresses).
For this case the failure envelope is described by the equation;
12 1 2 22 y2
where y is the yield stress.
We can plot this (on 1, 2 coordinates) as;
if 1=0, 2 = y uniaxial in 2
if 2=0, 1 = y uniaxial in 1
if 1= 2 , 1= 2= y
The surface is an ellipse when viewed on the 3= 0 plane.
The maximum possible stress is:
2 2
2= 1/2 12 1 1 y2 1 y 1.154 y
4
2 4 3
The von-Mises yield criteria is best explained as saying that yield occurs
when the elastic energy stored in shear reaches a limit. Every point on the
E6003 – Ship Structures II 20
© C.G.Daley 2014
von-Mises failure surface represents the same amount of elastic potential
energy in shear.
Strength of plates
We can use the von-Mises criteria to determine the load (pressure) that will
case yielding to occur in a plate.
Assume: - long plate
- uniform pressure p
- plate clamped at edges
The stress in the x direction is 1.
The stress in the y direction is 2.
For long plates we know that 2= 1. If we use this in the von-Mises
equation we get;
12 1 (1 ) 2 12 y2
y
or
1 1.125 y
1 2
We see that the plate is apparently stronger than a beam, by 12.5%, due to
the effects of y-direction restraint.
For a clamped plate:
M max
pb 2
(moment per unit width)
12
at yield;
x= 1=
Mc
I
2 6M
M t
3
t t2
12
collecting terms we get:
y
1
pb 2
1 2 2t 2
E6003 – Ship Structures II 21
© C.G.Daley 2014
We can re-arrange this to find the pressure to cause yield;
2t 2 y t
py 2.25 y
2
b2 1 2 b
This equation can be used to determine the capacity of the plate if y, t, b or
to design the plate (determine t) if p, y, b are known.
E6003 – Ship Structures II 22
© C.G.Daley 2014
Topic 4: Mohr: State of Stress
close-up of ship structural test
Introduction
In this lecture we will
review Mohr’s Circle (stress on any plane)
show von-Mises in terms of engineering stresses
describe elasto-plastic behaviour
~~~~~~
We defined the von-Mises failure criteria, and used it for a simple case. In
that case we had no shear stresses (we ignored them).
We called 1=x. This was a case in which x was the maximum stress on
any plane. Normally we use the notation 1, 2, 3 to refer to principal
stresses.
We will review Mohr’s circle of stress to show what principal stresses are
and how they can sometimes be more useful than x, y, z, xy, xz, yz, which
are called engineering stresses.
Consider a small element of material with normal and shear stresses;
E6003 – Ship Structures II 23
© C.G.Daley 2014
We have drawn the stresses on the y-z plane (x, xy) and on the x-z plane
(y, -xy). Mohr showed that the stresses on all planes, when plotted, will
form a circle in vs coordinates:
The stresses on the y-z plane (the x-direction) are plotted on the Mohr-s
circle (point A). The stresses on the x-z plane (the y-direction) are plotted at
point B. These two planes are physically 90 degrees from each other, but are
180 degrees apart on the Mohr’s circle.
The line joining A, B is a baseline. To find the stresses on a cut plane at
angle from the y-z plane, we must move 2 from the x direction around the
Mohr’s circle. This lands us at point C, where the stresses are c and c.
You can see from the drawing that the largest value of occurs where is
zero. The largest and smallest values of are called 1 and 2. They are
sufficient to define the circle, and are called the principal stresses.
E6003 – Ship Structures II 24
© C.G.Daley 2014
The stress 1, occurs on a plane at an angle from the plane of x.
Exercise:
x = 4, y = -2, = 4. Using force vector equilibrium, find the stresses
on a plane whose normal is 15 deg up from the x axis.
Use Rhino to do the vector addition. Check the answer with a Mohr’s
Circle.
We do not need to solve for 1 and 2 graphically. We can use the following
equations:
x y x y
1 x
2
2
2
1 r
or
2 r
E6003 – Ship Structures II 25
© C.G.Daley 2014
The von-Mises yield criterion is;
12 1 2 22 yield
2
If we have x ,y and we can either convert to principal stresses or we can
use the von-Mises criterion expressed in engineering stress.
For the 2D case (z=0) von-Mises is;
x2 x y y2 3 xy2 yield
2
yield
the most general form of von-Mises is;
( x y ) ( y z ) ( z x ) xy yz xz
2
1 2 2 2 2 2 2
6 3
Now that we have a criterion for yield stress we can examine what happened
after yield occurs.
We will assume that we have ideal plasticity. In other words, we will assume
hat the steel will stretch under constant stress. The stress –strain curve
looks like;
With this kind of elasto-plastic behaviour, the bending stress in a plate will
develop as follows;
E6003 – Ship Structures II 26
© C.G.Daley 2014
E6003 – Ship Structures II 27
© C.G.Daley 2014
Topic 5: Plastic Bending Limit States in Laterally Loaded Plates
Introduction
In this lecture we will
Describe and derive equations for the plastic bending collapse
of a long plate
show the behaviour diagram in terms of a load-deflection plot
Long Plate Plastic Bending Collapse
Assumptions
Long plate (i.e. ≥ )
Elastic-perfectly plastic material behaviour
Boundary Conditions
Clamped boundary conditions – the plate is fixed in bending, but free to pull
in.
This implies that we can only have a flexural response (i.e. bending). There
can be no membrane1 behaviour because axial displacement is not restrained
– and therefore axial forces are not possible.
1 We will cover membrane behaviour in later lectures.
E6003 – Ship Structures II 28
© C.G.Daley 2014
Review of Plate behaviour under increasing uniform lateral load
In Lecture 8, we introduced the following succession of behaviours for plates
under increasing uniform lateral load. We review them here and further
refine their definition:
Initial Condition: The plate can resist end
moments, but is free to pull in.
Elastic Deflection: As load increases from
zero we have elastic deflections.
First Yield: As load increases further
yielding starts at the extreme fibre of the
plate’s edges.
Edge Hinges: As load increases further
plastic hinges form at edges (i.e. full yield
at edges).
Initial Yield at plate centre: As load
increases further, yielding starts at the
plate center.
3 Hinge Collapse: As the load increases
further, a plastic hinge forms at the plate
center, creating a mechanism -> 3 hinge
collapse.
E6003 – Ship Structures II 29
© C.G.Daley 2014
Mechanics of Plate Behaviour under Increasing Uniform Load
Remember that the following equations are for a unit strip of a long plate.
plate; which we shall define here as yp :
Definition: In Lecture 8 we found from von-Mises the effective yield strength for a long
y
yp 1.125 y (for = 0.3)
1 2
First Yield
As the pressure, P , increases to PY , the edges yield at the extreme fibre; all else
remains elastic.
The bending moment at first yield is:
M y yp yp yp
I 1 t3 t2
12
c t 6
2
The pressure to cause first yield is:
t
PY 2.25 y or
2
b
t
PY 2 yp
2
b
At this point the central deflection is:
Y
1 PY b 4
384 D
Edge Hinges
As P increases from PY to PEH , edge hinges form:
The moment to cause full yield (i.e.
the plastic hinge moment) is:
M p 1.5M y yp
t2
4 or
M P PEH
b2
12
The pressure to cause edge hinges is:
t
PEH 1.5PY or PEH 3 yp
2
b
At this point the deflection is:
EH
1 PEH b 4
384 D
E6003 – Ship Structures II 30
© C.G.Daley 2014
Problem change
Edge hinges are a mechanism (i.e. the behaviour of the system changes drastically
when edge hinges form). When edge hinges form the problem changes from an
indeterminate elastic plate with fixed supports, to a determinate elastic plate with
plastic hinges as the supports.
Note: all structures lose their indeterminacy as they become damaged and move
towards failure.
Note: We can think of a plastic hinge as a simple (frictionless) hinge with a
constant applied moment.
To summarize: once the edge hinges (i.e. plastic hinges) have formed, we need to treat
the plate as a new structure; which we can break into two parts.
E6003 – Ship Structures II 31
© C.G.Daley 2014
Superposition of two Problems
We now divide the problem into the sum of two simpler problems; + :
Problem is simply supported and carries only the load, PEH . Problem is simply
supported and has only a constant applied moment, M P . Note that each of these two
cases contains components of the real case, such that the cases can be added to give the
real case (i.e. all load, reactions, stresses and deflections can be added).
Problem stays unchanged (i.e. remains constant) as P increases above PEH. Why?
plastic hinge cannot go above yp with this assumption.
Because of the assumption of a perfectly plastic stress-strain relationship. The stress in a
Problem , however, will continue to deflect elastically until a central plastic hinge
forms.
To find the load to cause the central plastic hinge we examine the bending moments in
both and :
We see that in Problem , the bending moment is 3 M P , but in Problem , it is
M P ; therefore the total bending moment (i.e. the sum of ) is M P at the
2
and
edges and 1 2 M P in the centre.
E6003 – Ship Structures II 32
© C.G.Daley 2014
3 Hinge Collapse
When the P increases beyond PEH , the values for increase until the central moment
is 2M P . stays unchanged, so that the sum of the two cases gives edge and central
moments of M P . At this point we have 3 hinge collapse and we call the load level PC .
We can determine PC by ratio:
PC PEH PEH
2M P 4
1.5M P 3
Or we can determine PC by solving the
statically balanced beam:
The area of the shear force diagram (SFD) to
the left of center will equal �� :
2M P PC
bb1
222
which is rearranged to give:
t
PC 4 yp or
2
16M p
b2 b
t
PC 4.5 y
2
b
For PC the deflection is:
C
4 5 PEH b 4 4 PEH b 4 8 1 PEH b 4 2 PC b 4
3 384 D 384 D 3 384 D 384 D
E6003 – Ship Structures II 33
© C.G.Daley 2014
Summary
Load Deflection
t t
PY 2.25 y 2 yp Y
2 2
1 PY b 4
b b 384 D
t t
3.375 y 3 yp EH
2 2
1 PEH b 4
b b
PEH
384 D
t t
PC 4.5 y 4 yp C
2 2
2 PC b 4
b b 384 D
If we normalize the loads by PC and the deflections by C we get:
Load Deflection
Y 1
PY Y 0.5 Y
C 4
~ P ~
PC
EH 3
PEH EH 0.75 EH
C 8
~ P ~
PC
C
PC C 1 C 1
C
~ P ~
PC
PY is 50% of PC and PEH is 75% of PC ; Y is 25% of C and EH is 37.5% of C
Thus we see that:
E6003 – Ship Structures II 34
© C.G.Daley 2014
Topic 6: Elastic Bending in Restrained Laterally Loaded Plates
Introduction
In this lecture we will
Describe and derive equations for the elastic bending of a
laterally restrained long plate (with membrane stresses)
show the behaviour diagram in terms of a load-deflection
plot
~~~~~~
We start by considering a long plate, pinned at the supports but restrained
from pulling in.
Initial situation:
Under uniform load q :
consider a free body diagram
as part of the plate cut at x;
V: shear
M: moment
w : deflection
P: lateral reaction
Summing moments at the cut (at x);
M ( x) P w( x) q x x0
x qb
2 2
which gives a moment of;
M ( x) x q x P w( x)
qb x
2 2
recall that;
E6003 – Ship Structures II 35
© C.G.Daley 2014
M ( x) D
d 2 w( x)
dx 2
so that we can write the differential equation for an elastic restrained plate ;
P w( x) (bx x 2 )
d 2 w( x) q
D 2
dx 2
The general solution for such a D.E. is;
2Ux qb 4 x x 2 1
w( x) C1 sinh C2 cosh 2 2
2Ux
b b 8U D b b 2U 2
where;
U 2Pb 2
4D
if we take the boundary conditions as;
w(0) 0 , w(b) 0
we get the specific solution;
2 x
cosh U 1
qb b qb 2 (bx x 2 )
w( x) 1
4
16U 4 D 8U 2 D
cosh U
unfortunately, this equation is recursive. The deflection w depends on U,
which depends on the axial force P, which depends on the elongation of the
plate, which depends on w. If not for this interdependence we would be done.
To get the membrane force, we use;
P axial A
EA
1 2
b E t
b 1 2
we find b, the elongation of the plate as follows;
E6003 – Ship Structures II 36
© C.G.Daley 2014
b
dw( x)
dx
b 2
2
0
dx
~~~~~~~~~~~~~~~~~
To derive this formula we consider the following;
We define ‘s’ as the coordinate along the plate, while x is the purely
horizontal coordinate.
x: 0 -> b
s: 0 -> b +b
we can relate dx, ds and dw; ds dx 2 dw2
we can say: b dx and b b ds
all all
so we can write: b ds dx ds dx dx 2 dw2 dx
b b b b
0 0 0 0
which we can re-arrange to give:
b
dw
b 1
2
dx dx
0
dx
and
b
dw
b 1 1 dx
2
dx
0
we can make use of the approximation:
1 a 1
a
for a<<1 (i.e. small deflections!)
2
dw 1 dw
to get: 1 1
2 2
dx 2 dx
E6003 – Ship Structures II 37
© C.G.Daley 2014
This allows us to write:
b
1 dw
dx
b 2
2 0 dx
and by symmetry we can write:
b
dw
dx
b 2
2
0 dx
~~~~~~~~~~~~~~~
Returning to the problem, we have;
2 x
cosh U 1
b qb 2 (bx x 2 )
w( x) 1
4
qb
16U 3 D tanh U 8U 2 D
cosh U
with U
Pb 2
4D
The axial load P is;
b E t dw E t
P dx
b 2
2
b 1 2
0
dx b(1 2 )
We can re-arrange this to give:
Pb (1 2 )
0 dx dx E t
dw
b 2
2
Solving the left hand side and re-arranging will give a expression for U
(which contains P) in terms of q (the applied load):
E2 t8
6
q b (1 )
81 27 27 9
2 8 2 7 6 2 8
16U tanh U 16U sinh U 4U 8U
This equation ( labeled #9 in Ratzlaff and Kennedy) can be solved for U, but
only numerically (trial and error) (HINT: Use the GOALSEEK function in
E6003 – Ship Structures II 38
© C.G.Daley 2014
Microsoft(R) Excel(R)).. The expression for w(x=b/2) will be the cental deflection
and when plotted vs load will normally be very similar to the equation;
wcentral
1 qb 4
, which is the elastic deflection equation for a pinned-pinned
384 D
plate in the absence of axial loads.
Summary:
We have considered a pinned-pinned plate and included one small additional
effect, the axial load that arises as the beam deforms. We have a solution, but
as we see, it is quite complicated. The solution is elastic, so it is only really
valid up to edge hinge formation. By this point the deflections are usually
still so small that the axial force is too small to worry about. We would only
want this solution for certain thin plates, of an intermediate thickness.
Thicker plates have too little axial force, and very thin plates have negligible
bending strength (i.e. all axial).
E6003 – Ship Structures II 39
© C.G.Daley 2014
Topic 7: Membrane Behavior in Restrained Laterally Loaded
Plates
Introduction
In this lecture we will
Describe and derive equations for the membrane distortion
of a laterally loaded long plate (with membrane stresses)
show the behaviour diagram in terms of a load-deflection
plot
~~~~~~
We will now look at a solution for plate with just membrane behavior. This
means that there is a local axial force, but no bending moments. For this case
we have the free-body diagram;
The force in a membrane is always
aligned with the surface (no shear, no
moment).
Taking moments at the cut:
q x2 q b x
P w( x) 0
2 2
P w( x) ( x 2 b x) 0
q
2
2 P w( x)
q
x(b x)
Now for x b / 2 (w(x)=wmax);
2 P wmax 8P wmax
q
b (b b ) b2
q b
2 2
q b2
and we can write: wmax and P
2
8P 8wmax
This gives us expressions relating load, membrane reaction and maximum
deflection. If q and b are known, we still have two unknowns, P and wmax.
E6003 – Ship Structures II 40
© C.G.Daley 2014
From the sum of the reaction force vectors we get;
qb
S P
2
2
2
q b2
using P , inserted into the above we get;
8wmax
q 4 qb
S b
2 2
8wmax 2
We can re-arrange this to get;
q
b 4 4wmax b
8wmax S
(this is equation 16 in RK)
2
This gives us load vs length, deflection and membrane force.
Now we need to find S from wmax and b.
Plastic Case (simplest)
When the plate is fully plastic;
y t
S 1.155 y t (note p=0.5)
1 p 2p
hence;
9.24wmax y t
q
b 4 4wmax b
2
Note that this load – deflection is approximately linear.
Why is the plastic case the simplest?
the maximum stress is y . The fact that the stress cannot be higher than y allows us
Because the of the elastic-perfectly plastic assumption: when the plate is fully plastic,
to say that the membrane force is equal to y for a plate (i.e. y / 1 p p2 ) times
E6003 – Ship Structures II 41
© C.G.Daley 2014
the cross sectional area of the plate (which is simply t for a unit strip of plate). In other
words, S does not depend on wmax, it only depends on the plate being fully plastic.
~~~~~
Elastic Case (more complex)
In this case we need to derive S from w(x) because S depends on the stretch,
which depends on w(x).
we derived the equation ;
2 P w( x)
q
x(b x)
Which we can re-arrange to give;
q( xb x 2 )
w( x)
2P
we will be using the formula for b that we used in the last lecture;
b
dw( x)
dx
b 2
2
0
dx
we find the derivative of w(x);
(b 2 x)
dw( x) q
dx 2P
q b2
substituting P we get;
8wmax
(b 2 x)
dw( x) 4wmax
dx b2
This is the slope of the membrane in terms of b, wmax as a function of x.
we can now write;
E6003 – Ship Structures II 42
© C.G.Daley 2014
b max b
4w
(b 2 x) dx 4bx 4 x 2 dx
b 2 b
2 2 2
16wmax
0
2
2
b b4 0
b
2
8 wmax
3 b
so that the strain is;
b 8 wmax
2
b 3 b2
S t
Now we can
E t
1 2
8 w E t
max
2
3 b 1 2
We can now substitute S back into the membrane equation to get;
64 wmax E t
q
b 4wmax
3
3 b 1 2
1
2 2
This is the elastic membrane equation, and is approximately a cubic.
Behaviour Diagram
We have derived 4 different solutions for plates. One requires a trial and
error solution. The other 3 are elastic-plastic bending, plastic and elastic
membrane. We’ll summarize the equations here and plot then for various
example plates. For consistency, We’ll use as the central deflection (instead
of wmax);
E6003 – Ship Structures II 43
© C.G.Daley 2014
Elasto/Plastic bending:
Load Deflection
t
PY 2.25 y Y
2
1 PY b 4
b 384 D
t
3.375 y EH
2
1 PEH b 4
b
PEH
384 D
t
PC 4.5 y C
2
2 PC b 4
b 384 D
9.24 y t
Plastic Membrane:
q
b 4 4 b
2
Elastic Membrane:
64 E t
q
b 2 4
3
3 b 1 2
1
2
Case 1- a thick plate:
Parameters:
b 400 mm
t 30 mm
E 207,000 MPa
0.3
360 MPa
D 511,813,187
25
20
pessure [MPa]
15
10 Bending:
Plastic Membrane:
5 Elastic Membrane:
0
0 10 20 30 40
Deflection [mm]
E6003 – Ship Structures II 44
© C.G.Daley 2014
Case 2- an intermediate plate:
Parameters:
b 500 mm
t 10 mm
E 207,000 MPa
0.3
400 MPa
D 18,956,044
4
3.5
pessure [MPa]
3
2.5
2
1.5 Bending:
Plastic Membrane:
1
Elastic Membrane:
0.5
0
0 10 20 30 40
Deflection [mm]
Case 3- a thin plate:
Parameters:
b 1000 mm
t 10 mm
E 207,000 MPa
0.3
400 MPa
D 18,956,044
E6003 – Ship Structures II 45
© C.G.Daley 2014
1.4
1.2
pessure [MPa]
1 Bending:
0.8 Plastic Membrane:
Elastic Membrane:
0.6
0.4
0.2
0
0 20 40 60 80
Deflection [mm]
E6003 – Ship Structures II 46
© C.G.Daley 2014
Topic 8: Low Aspect Ratio Plates, and Permanent Set
Introduction
In this lecture we will
Define a normalized capacity for plates, and examine how
to extend the capacity description to low aspect ratio plates
(i.e. not long plates)
examine the Clarkson method for designing plates for a
specified permanent set.
~~~~~~
We define a normalized load Q;
Q
y2
qE
For the yield condition in bending we have;
t E
Qy y
2
qy E
y2
2
1 2 b y
2
t2 E
1 2 b y
2
2
b y
We define , as a kind of slenderness ratio, which lets us write;
t E
Qy
2
2 1
1 2
This is for long plates, a>>b (say a>3b);
E6003 – Ship Structures II 47
© C.G.Daley 2014
For low aspect ratio aspect ratio plates (a<3b), we can use the equation;
1 b
Qy 1 0.6
4
2
2
1 2 a
The extra factor results in an increase of;
a factor increase
b
1 1.6 60%
1.2 1.29 29%
1.5 1.12 12%
2 1.04 4%
3 1.007 0.7%
4 1.002 0.2%
The above equation refers to the yield condition. When just brought to yield,
there will be no permanent deformation (set). For load above Qy there will be
increasing levels of permanent set.
So far we have discussed load-deflection curves. We have approached plastic
bending strength from the point of view of load capacity. Another approach to
plastic design is to allow some level of permanent set, as a % of the plate
thickness, or as a % of plate span.
Refer to pages 351-354 in Hughes. Hughes describes the experimental and
analytical work of Clarkson, and develops a set of ‘design’ plots. Using the
plots, you can determine capacity of scantlings to achieve a given level of
permanent set (i.e. a given level of denting for the design load).
Figure 1 below, shows plate capacity Q, vs. the slenderness ratio , for
various aspects ratios (a/b) of plate. This particular plot gives values for a
level of permanent set of;
1.0
wp
t
This is equivalent to;
1.0
wp
y
E
b
for typical shipbuilding steel (y = 250 -> 400), this is approximately
0.039
wp
b
E6003 – Ship Structures II 48
© C.G.Daley 2014
Fig 1.
Example: For a plate with the following parameters, what is the load
capacity and permanent deflection according to Fig. 1?
a = 1000mm
b = 500mm
y = 300MPa
t = 15mm
Step 1 : find 500 15 300 207000 1.59
Step 2 : lookup Q on the a/b=2 curve. Q=4.1
Step 3 : find q q=4.1*3002/207000= 1.78 [MPa] <= ANS.
Step 4 : find wp. wp=1.59*15=19mm <= ANS.
Note that qy = 2.25y (t/b)2=0.39 MPa. The q at 19mm deflection is not only
above yield, it is above the 3 hinge load qc.
E6003 – Ship Structures II 49
© C.G.Daley 2014
The plots in Figure 2 show four additional cases, for a range of permanent
set. These plots can be used just like Figure 1.
Fig 2. Capacity plots for wp/t= 0.2, 0.4, 0.6 and 0.8
E6003 – Ship Structures II 50
© C.G.Daley 2014
We will now explore a variety of ways to use the above plots.
Case 1: find plate thickness and permanent deflection for wp/t= 0.2, with the
given values;
a = 800 mm
b = 400 mm
y = 250 MPa
q = 0.5 MPa
Step 2 : lookup on the a/b=2 curve (top left plot of Fig.2) . =1.9
Step 1 : find Q. Q=0.5 * 207000/2502=1.656
Step 3 : find t t=400/1.9*(300/207000)= 7.31 [mm] <= ANS.
Step 4 : find wp. wp=0.2*1.9*7.31=2.8mm <= ANS.
Case 2: find permanent deflection, with the given values;
a = 1800 mm
b = 600 mm
y = 250 MPa
q = 0.5 MPa
t = 10 mm
Step 1 : find Q. Q=0.5 * 207000/2502=1.656
Step 2 : find 600 10 250 207000 2.083
Step 3 : find wp. We need to find a figure with a match between Q and .
On the five figures, if we look up b=2.083 we get
wp/t= 0.2, Q=1.15
wp/t= 0.4, Q=1.35
wp/t= 0.6, Q=1.45
wp/t= 0.8, Q=1.65 ***
wp/t= 1.0, Q=1.80
***we can see that our Q and b correspond to a wp/t=0.8
this gives us:
wp= 0.8 600 250 207000 16.7 mm <= ANS.
E6003 – Ship Structures II 51
© C.G.Daley 2014
Topic 9: Plastic Plate Strength Based on Plate Folding
Introduction
In this lecture we will
Re-derive plate capacity at collapse based on energy methods
Extend the concept to low aspect ratio plates
~~~~~~
We derived the plastic collapse load for a long plate in lecture 10;
y
Pc Mp
16M p t2
1 2 4
,
b2
Lets re-visit this with an energy approach.
Energy Solution Concept:
The applied load moves as the structure collapses. Consequently,
the applied load does work (force x distance) to the structure.
(work=energy expended). This is called the external work.
As it collapses, the structure forms plastic hinges, which rotate with
a constant moment. Hence the hinges do work (moment x angle),
and absorb energy. This is called the internal work.
We can equate the external and internal work. We need to find the
angles in terms of the movement of the structure. We can solve the
energy equation for the applied force.
We start by describing the pattern of collapse;
The work done by Pc is called external work (WEXT)
WEXT Pc ( x) dx
E6003 – Ship Structures II 52
© C.G.Daley 2014
WEXT Pc ( x) dx
In our case Pc is a constant so;
which is simply;
WEXT Pc b
2
The internal work occurs at the plastic hinges (at discrete locations), and is
the sum of the work done at each hinge;
WINT M p M P 4 M P
8
b
When we equate external and internal work;
WEXT W INT
8
Pc b MP
2 b
solving for Pc gives;
Pc
16M P
b2
which is exactly the result we derived before, though without all the elastic
analysis (differential equations) and the various steps.
We will now look at the real power of this method. We will find the plastic
collapse load for a patch load. This analysis would be quite challenging if we
were to try this using an elasto-plastic analysis.
E6003 – Ship Structures II 53
© C.G.Daley 2014
The work done by Pc is called external work (WEXT)
WEXT Pc ( x) dx
In this case Pc only acts over b/4,
and the average deflection under the load is 7/8 ;
WEXT Pc
b 7
4 8
The internal work occurs at the plastic hinges (at
discrete locations), and is the sum of the work done
at each hinge, and is exactly as before;
WINT M p M P 4 M P
8
b
When we equate external and internal work;
WEXT W INT
8
Pc M P
b 7
4 8 b
solving for Pc gives;
Pc 36.57 2P
256 M P M
2
7 b b
The pressure is 2.3x larger than the previous case, but the load length is ¼ as
long, so the force is only 57% as much as in the uniform load case. In the limit
case of a point load on the center, the force would be exactly half as much as
in the uniform load case.
What pattern of load allows for the greatest load to be applied to a beam?
The above analysis applies to long plates. Now lets consider finite aspect
ratio plates. Consider a plate a x b with a uniform load, fixed at all edges.
E6003 – Ship Structures II 54
© C.G.Daley 2014
We assume that the plate collapses by folding along the lines shown below;
We assume that the deflections (i.e. ) are very small, so that this folding
pattern is kinematically possible. We again approach the problem as a
balance of internal and external work. The total internal work is found from
the sum of the work going into each hinge;
WINT M p ln n
2
For the long edges;
, l 2a
b
2
for the short edges;
, l 2b
b
4
for the central hinge;
, l a b
b
and for the corner hinges;
2 , l 2 2b
2b
2 2
2
b
E6003 – Ship Structures II 55
© C.G.Daley 2014
This gives us a total internal work of;
2 2 4 2 2
WINT M P 2a 2b ( a b) 2 2b
b b b b
8
WINT M P ( a b)
b
The external work is ;
WEXT P da
which for a uniform pressure is;
WEXT P da P V
where V is the volume of the deflection (i.e. the volume of the cottage roof, or
at least the volume of the cottage roof that is under the load). In this case we
have;
b 2
P b(a b)
3 2
WEXT
or upon simplification;
a 1
WEXT P b 2
2b 6
When we equate internal and external work and solve for PC, we get;
a
1
8 M P b
PC
b2 a 1
2b 6
8 M P 5 21.8 M P
for say a/b=4, we have;
PC
b 2 11
=
b2
6
8 M P 2 48 M P
for a/b=1, we have;
PC
b2 2
= . This is 4 x greater than the long plate solution
b2
6
(compared with 1.6 x for the elastic solution).
E6003 – Ship Structures II 56
© C.G.Daley 2014
8 M P 11 18.2 M P
for a/b=10, we have;
PC
b 2 29
=
b2
6
16 M P
For very large a/b this converges to as before.
b2
The above discussion shows the general idea of energy methods and plate
folding method to find plate capacity. The method is very robust and can be
handle a wide range of problems which would be very difficult with other
analytical methods.
There are a few things to keep in mind when using these energy methods.
1) the method works by assuming a collapse pattern. There are an infinite
number of patterns possible. The ‘correct’ pattern is the one that gives the
lowest load. So the method often involves trying several patterns, and
choosing the smallest. In the case given above, we assumed that the corner
angle was 45º. We could have chosen another angle such as 40º. Or we could
have tried the various patterns shown below.
2) the method is called an ‘upper bound’ method. This means that the method
will theoretically bound the strength from above. If the strength is
overestimated, this is non-conservative. In practice, there are issues that
tend to both raise and lower the estimates, so that in practice the result may
be conservative. The strength of the method is that it tends to be close, and is
a handy way to check a value. Work/energy considerations are extremely
strong principles, so that the results are seldom far from reality.
Entrepreneurial idea: develop a software tool to quickly solve this class of
problems. It will find wide use as structural design moves increasingly to
plastic behavior. (User defines geometry of structure, folds and load.
E6003 – Ship Structures II 57
© C.G.Daley 2014
Software solves for internal and external work, and determines collapse load.
Software solves for optimum parameters in pattern.)
E6003 – Ship Structures II 58
© C.G.Daley 2014
Topic 10: Introduction to Elastic Buckling
Introduction
In this lecture we will
Discuss the concept of buckling
review column buckling
~~~~~~
Buckling is a type of structural behavior.
It can often lead to sudden loss of strength and collapse
Elastic buckling may be acceptable, if it does not lead to plasticity
Buckling tends to be more important as steel strength increases
Buckling often dominates design, because buckling control results
in geometric limits
Column Buckling
Consider a column of length L, area A, modulus E, moment of inertia Iz. Iz is
the weak axis of bending, as Iy>Iz. The column will bend about the z axis,
moving laterally in the y direction.
We assume that the column is initially straight, that the load F acts along
the centroid (neutral axis). In the unbuckled state the column is in pure
compression (no bending).
Under low load levels the column stays straight, and slightly shortens;
E6003 – Ship Structures II 59
© C.G.Daley 2014
The compression stress is , the strain is
F F
A E
.
F L
A
The change in length is L L
A E
.
Observations show that at some critical load level Fcrit, the column will
suddenly deform sideways!
Plot Load vs Deflection.
As the load is steadily increased, the column deflects vertically (compresses).
Prior to buckling there is no lateral deflection of the column. At the point of
buckling the lateral deflection increases (and grows large) under no increase
in load. The vertical deflection will also grow large under no additional load.
This case is plotted with the solid lines in the plots below.
There is another case with slightly different post-buckling behavior. Instead
of having a static load (like static weights), we might apply the load by
imposing a vertical deflection on the top of the column. This would be called
‘displacement control’ (vs. ‘load control’). In this case the load-deflection
curves (plotted with dashed lines) would be the same up to buckling, but the
load would fall after buckling. The drop in load is called the ‘unloading’
portion of the curve.
This type of instability is sometimes called ‘Euler’ buckling or ‘bifurcation’
buckling.
E6003 – Ship Structures II 60
© C.G.Daley 2014
The buckling load can be found in various ways;
1) solve the differential equation:
F y ( x)
d 2 y ( x)
EI z
dx 2
2) Assume a shape y(x), and solve for F (shortcut to solution of the D.E.;
3) Use energy methods.
~~~~~~~~~~
Lets try no. 2).
Assume a deflected shape:
n
y ( x) yo sin( x)
L
n n
y yo sin(
2
L
x)
L
n
y y ( x)
2
L
From this the differential equation becomes;
n 2
EI z y ( x) F y ( x)
L
n EI z
F
2 2
L2
for n=1, we have;
2 EI z
Fcrit
L2
which is called the Euler buckling formula.
E6003 – Ship Structures II 61
© C.G.Daley 2014
Use of High Strength Steel:
Notice that the buckling formula has no mention of strength. The
buckling strength seems not to depend on strength. However,
buckling strength does depend on I, which for a plate, depends on
plate thickness, which does depend on strength.
Example: take a plate designed for a lateral load, but which
experiences some in-plane compression. The plate is originally
intended to be build with regular mild steel (y=225 MPa). The
plate was sized using the standard formula;
t 0.67 b
y
P
The critical buckling force is;
2 EI z a t3 2E a t3
Fcrit , I z or Fcrit
L2 12 12L2
If we double the steel strength to 450 MPa, we can decrease the
thickness by 1/2 (= 0.707) with no loss of bending strength.
Unfortunately the buckling strength will decrease to .7073=.353. In
other words the buckling strength of the high strength steel may
be only 35% of what it would be with mild steel. Of course, it may
be that the buckling strength (at 35%) is still high enough. This
example just shows that increase of yield strength does not
increase buckling strength. If buckling is the critical failure
mechanism, then no increase in material strength will help.
Support Conditions and Critical Length
The buckling pattern, and the critical load, depend on the support conditions
for the column. When the ends are pinned, the column can form a simple
half-sine wave between the supports. If both ends of the column are fixed the
column will buckle into a shape with two inflection points. The center half
will form a half-sine curve. The inflection point will signify that there is no
bending moment at the quarter points. In this case the middle half of the
2 EI z
beam will act like an Euler column. The critical buckling load is;
Fcrit
kL2
where k=0.5. kL is called the critical length.
E6003 – Ship Structures II 62
© C.G.Daley 2014
If one end is fixed and the other end is pinned, the column will buckle into a
shape with one inflection point. The column between the inflection point
(which is like a pin) will buckle according to the Euler formula, with k=0.7;
2 EI z
Fcrit
kL2
These various behaviors are shown below.
E6003 – Ship Structures II 63
© C.G.Daley 2014
T.10 – Problems.
1. A column is constructed with an angle section (100x80x10), 2
m. long. The steel has a yield strength of 250MPa. The ends of
the column are considered free to rotate. What is the largest
load that can be applied to the column (without yielding or
buckling)?
2. You want to use a steel pipe with a 5mm wall
thickness as a column to support a deck edge.
The base of the column is fixed, while the top is
free to rotate. The column is 2.5m long. The steel
has a yield strength of 230MPa. You want to
carry a load of 150kN at failure (yield or
buckling), and keep the pipe diameter as small
as possible. What is the outside diameter of the
pipe?
3. What is the buckling load for a
column made from a wide I beam
(110 tall x 150wide x 10 thick),
that is 4m long with pinned ends?
When braced at the center in the
weak direction, what is the new
buckling load?
E6003 – Ship Structures II 64
© C.G.Daley 2014
Topic 11: Elastic Buckling by Energy Methods
Introduction
In this lecture we will
Discuss buckling as an energy balance
derive column buckling in 2 different ways
~~~~~~
Energy Methods
IDEA: As a force is applied to any elastic body, the force does work (expends
energy) which is stored in the body as elastic potential energy. When the
force is low, and the system is stable, the body stores energy by distorting in
the direction of the force. At this point any further movement of the body uses
more energy than it releases. The body will return from any small
disturbances.
stable state
When the load is sufficiently high, the elastic body is highly strained, and is
storing a large amount of potential energy. The body may find that a lateral
distortion releases more energy than is required to cause the lateral
distortion. In a column, this means that a lateral bend will release more axial
energy than is needed to bend the column. At this point the body is elastically
unstable.
unstable state
We can examine any form of elastic buckling in this way.
E6003 – Ship Structures II 65
© C.G.Daley 2014
Consider the energy in a column;
The External Work done 1 2 F
F 2L
2 AE
Note:
The internal Strain Energy is; InternalWo rk
F x
U A EA 2 dx
L 1
1
A dx
2
20
1
EA dx
for a simple column
2
F
1
E AE
EA 2 dx
2
U A EA
2 AE 0
F
1
dx
L 2 L
1 1 F2
2 0 AE
dx 2
hence;
UA
F 2L
<= same as EW
2 AE
E6003 – Ship Structures II 66
© C.G.Daley 2014
The column will distort in any way possible, so that it is storing the least
amount of energy. One way to distort is to bend sideways. Buckling is
bending sideways.
The elastic strain energy in bending is;
d2y
U B EI 2 dx
L 2
1
2 0 dx
(see derivation at end)
The bent column must store UB
The bent column is longer than the straight
column
The bent column has released axial energy and
gained bending energy.
If more energy is released axially than needed to
bend the additional surplus will become kinetic
energy (velocity)
x
Lets assume the buckled shape is;
y( x) yo sin
L
which means;
2 x
y 2
yo sin
L L
1 y o 2 x
U B EI 2 sin
2L
2 L
2
dx (integral part = L/2)
0
L
4 EI y o2
UB
4L3
~~~~
If the structure bends into a sinusoidal shape with amplitude yo, it will
require an input of energy UB.
Where will it get this energy?
E6003 – Ship Structures II 67
© C.G.Daley 2014
The energy comes from the end load.
The are two ways to view the available energy (available to be converted to
bending energy;
1) As the column buckles the load on the end of the column stays constant,
and the end of the column moves.
At the point of buckling the column has the length L1/ As it buckles the load
stays applied, the column stays compressed, and the column stays L1 long.
The straight-line length is L2, and the load drops by b = L1-L2. The
available energy is: U A Fcr b
2) As the column buckles the length of the column stays constant, and the
compression of the column decreases.
At the point of buckling the column is L1 long. As it buckles the ends don’t
move so that the straight line length remains L1. The bending results in a
longer length, L3, along the column axis. The column stretches by r (axial
rebound) where r = L3-L1. The available energy is U A Fcr r
E6003 – Ship Structures II 68
© C.G.Daley 2014
Note: While it may not seem so at first, the two
cases above are identical. This is because we
assume all incremental buckling deflections are
vanishingly small, so that r = b = Lcurved-
Lstraight.
Either way we need to find out the difference in
length between the straight and curved paths.
In other words we want to know The difference
between the path length S and the straight
length L.
S L ds
ds dx
S L
0 0
ds dx 2 dy 2
dy
dx 1
2
dx
1 1dx
dy
L 2
dx
0
again we use the approximation: 1 a 1
a
for small a. Therefore we can
2
write;
dx y dx
1 dy
L 2 L
1
2 0 dx
2
20
To find D we need to integrate the square of the slope of the deflected shape.
x
We will again assume
y( x) yo sin
L
E6003 – Ship Structures II 69
© C.G.Daley 2014
x
y 2 y o cos 2
2
L L
1 y x y o2 2
o cos dx
2L
2 L
2
0
L 4L
Which give the available energy as;
y 2 2
U A Fcr o
4L
We can now equate the available energy to the energy required for bending;
UA UB
y o2 2 4 EI y o2
Fcr
4L 4 L3
Solving for Fcr we get;
2 EI
Fcr <= the Euler formula
L2
Summary: At the point of buckling, the energy required to deform the
structure laterally is just equal to the available energy. The available energy
comes either from 1) the load doing work as the end of the column moves in
response to the bending, or 2) the release of elastic potential energy as the
column elongates as it bows out, while the ends remain fixed. Both
mechanical concepts are identical in the limit, and produce identical results.
E6003 – Ship Structures II 70
© C.G.Daley 2014
Annex - derivation of work done in elastic bending;
each part of the beam dx long bends under a constant
moment;
d d 2 y
dy
dx dx dx 2
d
d2y
dx
dx 2
The work done by the moment M is dUB
dU B M d M 2 dx
1 1 d2y
2 2 dx
recall that M EI
d2y
dx 2
which gives;
1 d2y
dU B EI 2 dx
2
2 dx
The total bending energy is the integral over the whole
length;
d2y
U B EI 2 dx
L 2
1
2 0 dx
E6003 – Ship Structures II 71
© C.G.Daley 2014
T.11 – Problems.
1. Using the energy method, and assuming that the buckled shape of a
column is y=4 yo ((x/L)^2 –x/L), {this is a parabola} find a formula for the
critical buckling force. Compare the result to the Euler formula for a solid
round steel column with a diameter of 20mm, 2 m. long.
2. Using the energy method, and assuming that the buckled shape of a
column is an arc of a circle, can you find the critical buckling force?
E6003 – Ship Structures II 72
© C.G.Daley 2014
Topic 12: Buckling of Transverse Plate Panels
Introduction
In this lecture we will
Discuss plate buckling, analogous to column buckling
Discuss the slenderness ratio and the transition to yielding
Examine the effect of elastic boundary conditions
Show origins of rule buckling checks
~~~~~~
Long Plates/Transverse Buckling
When we consider a long plate loaded in the in-plane transverse direction.
This will typically occur in the deck of a ship with transverse framing.
When we load a plate in this way, the buckling is simple. Each strip of plate
(of some width w) buckles like each neighboring strip. We can write the Euler
buckling load as;
2 EI
Fcr
L2
we can re-write this in terms of stress in the plate;
Fcr 2 EI 2 E 2 A
cr 2
A LA L2 A
which gives us;
cr E
2
L
2
E6003 – Ship Structures II 73
© C.G.Daley 2014
Note: for a plate of width w, I wt3 2 A
1
12
2 0.29t
t2
12
Normally we try to prevent buckling before the stresses reach yield by
limiting the slenderness of the structural member. We will ensure that yield
cr y or cr k y for k>1
will occur first when we set;
The ratio of buckling to yield stress is;
cr E
2 2
2
y L y
1
y
where is a slenderness ratio;
L
E
This ratio is the Euler non-dimensional buckling stress. We can plot the
Euler curve vs slenderness. The curve is invalid for stresses above yield. Due
to imperfections (stress and geometry) the actual behavior tends to smoothly
join the Euler curve and the yield stress limit.
We can effectively prevent buckling by setting <1;
y
1
L
E
or
y
L E
E6003 – Ship Structures II 74
© C.G.Daley 2014
For plate this becomes;
t y
L 12 E
or
.907
y
L E
t
We would use this formula to limit the buckling of a wide plate in
compression (for pinned edges).
Classification society rules have very similar formulae to prevent buckling in
various plate elements.
For example see ABS rules Part 5 App.5/2 AB for buckling restrictions for
longitudinals;
1.5
y
dw E
tw
tee stiffeners :
.85
y
dw E
bulb stiffeners: tw
.5
y
dw E
flat bar stiffeners: tw
Q? Why do the constants range from 0.5 to 1.5?
Ans: Because of the boundary conditions and the effective length.
E6003 – Ship Structures II 75
© C.G.Daley 2014
We’ve seen that the effective length for fixed ends is half that of pinned ends.
When the ends are elastically restrained from rotating, we get effective
lengths somewhere between 0.5 and 1 x the length.
c l
n
le PCR
EI l PE Remarks
0 1 1 Pinned
1 .86 1.35
2.5 .75 1.78
5
10
.66
.59
2.3
2.85 } typical of
ship structures
∞ .5 4 fixed
Plates supported by frames or bulkheads are closer
to fixed than to pinned.
For example : a deck plate
Springs replace the effect of frames and
neighboring plate
cl
Assume that n 10 , CR 2.85 , and .59
P le
EI PE l
.907
y
.59 L E
This gives us a rule: , which can be written as
t
1.53
y
L E
similar to ABS rule.
t
We can also write this as;
y
L 698
which is similar to the format used in DnV rules.
t
T.12 – Problems.
E6003 – Ship Structures II 76
© C.G.Daley 2014
1. For a bulb stiffener with a height (dw) of 200mm, what is the
minimum web thickness to ensure yield occurs prior to buckling.
2. What would be the value of cl/EI for a 20mm plate supported by deep
frames every 600mm. Assume the frames act as simple supports and
consider only one neighbor plate. (hint: this can be found from the
rotary stiffness term k33 for a beam).
3. Find for a plate 15mm on 650mm span. (350MPa steel).
E6003 – Ship Structures II 77
© C.G.Daley 2014
Topic 13: Buckling of Longitudinal Plate Panels
Introduction
In this lecture we will
Discuss plate buckling
Derive the general response based on strain energy
~~~~~~
Long Plates/Transverse Buckling (edges simply supported)
When we consider a long plate loaded in the in-plane longitudinal direction.
This will typically occur in the deck of a ship with longitudinal framing. In
this configuration, the plate can buckle into a rippled pattern as shown
below. We can’t use the simplifying assumption that we used with transverse
plating. We have to consider the whole plate.
E6003 – Ship Structures II 78
© C.G.Daley 2014
The assumption of pinned edges is reasonable in light of the anti-symmetry
that is likely to develop in the buckling pattern. This beans that the
neighboring plates will not transmit a moment (though the frame may).
Plates buckle in two dimensions. We will us the variables m and n to give the
number of half sine waves in the x and y directions.
1 half wave in x : m=1
1 half wave in y : n=1
Other combinations are:
E6003 – Ship Structures II 79
© C.G.Daley 2014
The values of m and n (the buckling shape) will depend on which shape has
the lowest elastic potential energy. The shape will depend on the aspect ratio
(a/b) of the plate.
We can start the analysis by describing the deformed shape. We will assume
the shape to be of the form;
w A
mx ny
mnsinsin
m n a b
(this represents the superposition of many valid wave
patterns, and could represent almost any deformation
pattern for any load.)
For buckling, we will assume that the plate will take up only one sine-sine
form, or in other words, only one value or m or n. This lets us drop the
mx ny
summation and write;
w wo sin sin (1)
a b
Equation (1) satisfies all boundary conditions ay all four edges, which are;
x 0
2w 2w x a
w0 , 0 , at any of
x 2 y 2
0 ,
y 0
y b
and is symmetrical or anti-symmetrical about x=a/2, y=b/2. With this
deformation pattern we can now determine the elastic potential energy (strain
energy). For a plate (also termed an elastic shell), the strain energy U
(internal work) in bending and twist2 is;
2 w 2 w 2 2 w 2 w 2 w 2
U D 2 2 2(1 ) 2 dxdy
x x 2 xy
a b
1
x y
(2)
2 0 0
D
Et 3
12(1 2 )
where
2 Hughes Ch. 12, pg. 404.
E6003 – Ship Structures II 80
© C.G.Daley 2014
We can integrate the 2nd term over the plate and find that it equals zero;
2 w 2 w 2 w 2
2(1 ) 2 dxdy 0
a b
x x xy
2
o 0
This is illustrated in the extract from Maple™ shown below. Note that m and
n can be any two integers and the result will be zero (Exercise: try this).
Maple™ file:
> restart;
> f(x,y):=(diff(w(x,y),x,y))^2-diff(w(x,y),x$2)*diff(w(x,y),y$2);
> int(int(f(x,y),x=0..a),y=0..b);
> m:=2;n:=3;
> w(x,y):=wo*sin(m*Pi*x/a)*sin(n*Pi*y/b);
> f(x,y):=(diff(w(x,y),x,y))^2-diff(w(x,y),x$2)*diff(w(x,y),y$2);
> int(int(f(x,y),x=0..a),y=0..b);
we can expand the terms;
2w m 2w n
w, w
2 2
x 2
a y 2
b
E6003 – Ship Structures II 81
© C.G.Daley 2014
which allows us to write;
2w 2w m2 n2
2 2 4 2 2 w 2
2 2
x y a b
we can now write;
2 n2 2 a
4 m
U D 2 2
b
1
a b o
w2 dxdy
2 0
The integral of w2 becomes;
w2 dxdy
a b
wo2 ab
o 0
4
So we can write;
4 ab m 2
n2 2
U D 2 2 wo
2
8 a b
This is the stored elastic potential energy.
Recall that for a column the external work done is given by;
2 0 dx
1 dy
W Fcr Fcr dx
L 2
Similarly, for a plate with an end stress, the external work is ;
cr t a b w
2 0 0 x
W cr t dxdy
2
This evaluates as;
2 b cr t 2 2
W wo m
8a
As before, we equate internal and external work (W=U), to get;
2b cr t 4 ab m 2 n2
wm D 2 2 wo2
2
a b
2 2
o
8a 8
solving for the buckling stress, we get;
E6003 – Ship Structures II 82
© C.G.Daley 2014
m2 n2
a D 2 2
2
a b
2 2
cr
t m 2
It is obvious that cr will always be a minimum when n=1. In other words, we
will only have one half wave across the short dimension of the plate. We can
re-arrange cr to give;
2 D mba 2D
cr 2 k 2
2
b t a mb b t
where
mb a
k
2
a mb
When we plot k as a function of a/b, for various values of m, we see that the
minimum value of k is always 4, and always occurring where a/b=m.
This means that for a plate with an aspect ratio of, say, 3 we will get the
lowest buckling stress when we have m=3 (3 half-waves).
The minimum buckling stress depends on the minimum k. It is reasonable to
use k=4 for any aspect ratio.
E6003 – Ship Structures II 83
© C.G.Daley 2014
This lets us write the critical buckling stress as;
2D
cr 4
b 2t
We should note that although the plate is loaded in the longitudinal direction,
we are using ‘b’, the transverse plate dimension in the formula. The ‘length’ of
the plate ‘a’ doesn’t really matter. The plate will deform into a pattern of
buckles, each approximately ‘b’ x ‘b’.
Using
D
Et 3
12(1 2 )
we can write;
4 2 t
cr 2 3.62 E
2
Et 3
b t 12(1 )
2
b
Now we can normalize the buckling stress as we did for columns;
cr 3.62
2
y
where
b y
t E
The longitudinal plate buckling equation is similar to the column buckling
equation:
E6003 – Ship Structures II 84
© C.G.Daley 2014
E6003 – Ship Structures II 85
© C.G.Daley 2014
T.13 – Problems.
4. For a longitudinal plate with an aspect ratio of 1.5, what is the
minimum value of k? For this k, what is the value of m?
mx ny
w 2 dxdy for w wo sin
a b
5. Use Maple to evaluate the integral sin
o 0
a b
6. What is the buckling stress for a longitudinal steel deck plate of 600mm x
1200mm x 12mm?
7. What is the maximum yield strength that you should use for the plate in 3)
above, to be sure that the plate yields prior to buckling?
E6003 – Ship Structures II 86
© C.G.Daley 2014
Topic 14: Post-Buckling Strength of Long Plates
Introduction
In this lecture we will
Discuss von-Kármán’s concept of post buckling behavior
Compare to Faulkner’s and Paik’s equations
~~~~~~
Long Plates/Post Buckling (edges simply supported)
While columns have very little post-buckling strength, and certainly less than
their buckling strength, plates (between frames) can exhibit higher strength
after buckling than before. This is due to two factors;
a) The long edges remain straight and intact due to the frames (i.e. the part
of the plate at the edges does not buckle).
b) There is a stabilizing effect of the membrane stresses across the short
dimension of the plate.
E6003 – Ship Structures II 87
© C.G.Daley 2014
Finite element plot of the
buckled shape of a 500mm
x 2500mm x 9mm plate.
The model was a non-linear
analysis, with an imposed
end displacement, solved
using the arc-length
method. The term
TIME=0.894792 refers to
4.475mm (max imposed
deflection was 5mm), which
is well beyond the onset of
buckling.
After buckling occurs the stresses re-distribute themselves across the plate.
This is shown in a sketch below, and in a finite element plot below that. The
ANSYS plot is actually showing the forces at each node pushing on the
surrounding boundary (which is why they are pushing out for a state of
compression)
before buckling
after buckling
E6003 – Ship Structures II 88
© C.G.Daley 2014
Effective Width
The redistribution of stresses can be treated by the ‘effective width’ concept.
o True stresses are replaced by two uniform zones, where the total zone
width is be (note: the next plate is similar so the load patch is be wide,
centered on the frame.)
The true edge stress e is stress in be.
The total force can be expressed in a variety of ways;
F dy e be a b
b
(1)
0
where a is the average post-buckling stress. We can write;
be a
b e
(2)
Why use this effective width concept? The reason is that it will allow us to
determine the edge stress, which is also the axial stress on the frame, and
thus we will be able to determine the total post-buckling force.
E6003 – Ship Structures II 89
© C.G.Daley 2014
The way that we can find both e and be is with a method proposed by von-
Kármán (1924).
von-Kármán was, at the time, working on aircraft
structures, which had a very similar construction to
the ship structures that we are considering (plate
over longitudinal ribs, in both the fuselage and
wings). von-Kármán proposed a very simple (elegant)
way to find the edge stress along with the effective
width. He suggested that as the plate buckles, the
edge stress (over be) would be the same as the Euler
buckling stress for a plate of be width. This idea
turned out to be close to the truth. In effect, he
proposed to see the plate as two ideal parts: the
middle buckled part carrying no load, and a part (the
effective part) near the frame carrying all the load.
The ‘effective’ part would be progressively narrowed
by the progressive buckling of the middle part.
This is a key point. The idea is that plate buckling is a steady progressive
process. When the stress first causes buckling, the plate only just starts to
deform. If the stress is held at a level just above the buckling stress, the plate
is only slightly deformed (barely noticeable). As the average stress increases,
the middle part progressively sheds load to the sides. The whole process is
actually steady and stable, right up to the point where the edge stress (and
frame stress) reach yield. At least that’s the idea, and it is quite close to
reality.
The von-Kármán relationships are developed as follows;
Recall that
cr 2
3.62
(3)
where
b y
(4)
t E
or
t
cr 3.62 E
2
b
(5)
if we let b be , we get ;
t
e 3.62 E
2
be
(6)
E6003 – Ship Structures II 90
© C.G.Daley 2014
or
be 1.9t
e
E
(7)
These are the equations relating e and be.
The effective stress rises as the effective width decreases. This is plotted
below;
As the overall force on the deck increases and the effective width decreases,
the von-Kármán model will remain valid until the edge stress reaches yield.
At this point the frames will collapse. The minimum effective width can be
found by substituting the yield stress into equation (7).
bem 1.9t
y
E
(8)
for E=207000 MPa, y =235 MPa we get;
bem 56t
We have said that the plate buckles progressively and that the system can
take more and more load as the plate buckles. We can check this by
calculating the force on the plate;
E6003 – Ship Structures II 91
© C.G.Daley 2014
F be t e
3.62 t 2
be t E (9)
be2
3.62t 3 E
be
The limit situation occurs when be bem and e y . At this point the force is;
Quite clearly, the force F will increase as be gets smaller.
Fm bem t y (10)
The average stress on the plate in the limit condition is;
y
Fm bem t y y
E
1.9t
m
bt bt
(11)
b
m
1.9
y b y
t E 1.9
(12)
The buckling (3) and post-buckling (12) capacities are shown below;
Faulkner (Prof. at Glasgow), on the basis of experiments proposed a slightly
m 2 1
modified equation;
y 2
(13)
E6003 – Ship Structures II 92
© C.G.Daley 2014
Paik (Prof at Pusan), proposed an even more conservative equation, taking
imperfections and initial stresses into account, and validated with
experimental and numerical data. Paik’s equation is;
.032 4 .002 2 1 for 1.5
m
1.247 / for1.5 3.0
y
(14)
1.248 / .283 for 3.0
2
E6003 – Ship Structures II 93
© C.G.Daley 2014
T.14 – Problems.
1. Find the buckling and post-buckling capacities for the following
plates;
Plate 1 Plate 2 Plate 3
b [mm] 600 800 1200
t [mm] 20 15 10
y [MPa] 235 300 360
E [MPa] 207000 207000 207000
2. For a deck with 10mm plate, longitudinal frames spaced at 700mm,
transverse web frames spaced at 5000mm. The frames are 150x12
flats bars. All steel has a yield strength of 360MPa. Determine the
buckling and post-buckling strength of the plate. Determine the
buckling strength of the frame (assumed pinned). What is the
average stress that the deck (plate/frame) can withstand befor
failure)? (Hint: how much of the plate do you include in the flange
of the frame?)
E6003 – Ship Structures II 94
© C.G.Daley 2014
Topic 15: Local Buckling
Introduction
In this lecture we will
Discuss a variety of local buckling phenomena
~~~~~~
Local buckling refers to a variety of buckling mechanisms involving any part
(web, flange, bracket) of a frame. In all cases there is a local compression
stress that buckles a section of plate. The compression may be the result of
bending, shear or direct pressure.
Shear Buckling
When a plate (typically a web) experiences a shear stress, there is a
compression field on a 45º diagonal.
This type of shear buckling is actually common and expected in deep web
girders (as might be found in a railway bridge). As long as there are stiffeners
on the surface to take the compression, the buckled web will hold the tension
in a kind of truss-like structure.
Web and flange buckling
E6003 – Ship Structures II 95
© C.G.Daley 2014
Another common type of local buckling is web buckling, which may be caused
by direct compression due to an applied load.
Compression in the flange due to bending can cause the flange to buckle
locally.
Local buckling is prevented by limiting the local aspect ratios of parts of the
construction. A typical rule for plate with flanges on both boundaries would
be;
h 1000
y
[y in MPa]
t
or
2.2
y
h E
[y and E in same units]
t
When the plate is connected on one side only (as for a flange) it is called an
‘outstand’. The typical local buckling rule is for an ‘outstand’ is;
o 250
y
[y in MPa]
t
or
.55
y
o E
[y and E in same units]
t
E6003 – Ship Structures II 96
© C.G.Daley 2014
T.15 – Problems.
3. For a ‘tee’ flange made of 25mm steel with a yield strength of
360MPa, what is the limit width to ensure the flange will remain
unbuckled at yield?
4. For the two brackets pictured below, both 400mm (h x w), estimate
the minimum thickness needed to prevent buckling. One has a
flange on the free edge, while the other does not. Compare your
estimate with rule requirements.
E6003 – Ship Structures II 97
© C.G.Daley 2014
Topic 16: Overall Stiffener Buckling
Introduction
In this lecture we will
Derive the solution for the buckling strength of a frame
attached to a plate, assuming that the plate has already
buckled.
Discuss the modified effective width be
~~~~~~
After the shell plate has buckled, the frames act as columns, and may buckle
prior to reaching yield (when they will collapse plastically).
We need to find the moment of inertia I p of the frame with the part of the
plating that is still effective. The effective width be is not what we will use to
find I p . As the frame buckles, it will compress the plate further and reduce
the effective width to something less than be . We need to find be , which we will
do by considering stiffness.
E6003 – Ship Structures II 98
© C.G.Daley 2014
We will start by considering the load carrying capacity of the plate (which is
one flange of the frame). Recall from Lecture 19 that the plate in compression
carries that load;
P e be t
The flange of any normal beam keeps a fixed width, such that as the beam
bends, the flange pushes (or pulls) increasingly hard to resist the bending. We
could say that the flange contributes to the bending of the beam by being
axially stiff. The axial stiffness of any normal flange would be
A E b t E
k
L L
respect to the shortening. The load (after buckling) is a function of both e
We can also write that the stiffness is the derivative of the axial load with
and be .
dP d ( e be t )
k
d d
d ( e be t )
We can expand this;
k
d
d
t e be e e
d d
db
dbe dbe d e
we can write:
d d e d
which gives;
d
k t e be e e
d d e
db We define this
term as be’
d e '
so:
k t be
d
E6003 – Ship Structures II 99
© C.G.Daley 2014
We know that :
be e b avg which results in:
avg
be b
e
We can use this to get a simple expression for be’:
be' be e e
d e
db
avg d avg
b e b
e d e e
avg d avg
b b e
e d e e
d avg
e avg e
avg d e e
b b e
e e
2
avg b d avg
b e avg
e e d e
so :
d avg
be' b
d e
avg
recall that be b
e
. While be is related to the ratio of stresses, be’ is related
to the ratio of changes in stress. We need to re-visit von-Karman to find the
values. Based on von-Karman:
be 1.9t 1.9t E e0.5
e
E
be' be e
d e
dbe
be 0.5 1.9t E e1.5 e
very simple result!
be
2
As the plate buckles, the minimum be is bem. The minimum be’ is bem
'
E6003 – Ship Structures II 100
© C.G.Daley 2014
bem 1.9t .95t
y y
E E
and bem
'
'
For stiffener buckling we use bem as the effective width of plating to
calculate I.
bem 1.9t = 56 t for y=235MPa
y
E
and
.95t = 28 t for y=235MPa
y
' E
bem
2 EI '
Fstiffener 2
where I’ is found with bem
'
buckling L
stiffener
Fstiffener
where A’ uses bem .
buckling
buckling A'
If the stiffener buckling stress exceeds yield, then the maximum stiffener force
will be F = y A’. The maximum stiffener force (yield or buckling) divided by
the total (true) area will give the limit stress in the stiffened panel.
E6003 – Ship Structures II 101
© C.G.Daley 2014
T.16 – Problems.
1. For the plate/frame structure shown, find the average stress that can
be applied. What mechanisms determine this limit?
E6003 – Ship Structures II 102
© C.G.Daley 2014
Topic 17: Fatigue/Fracture in Ship Structures
Introduction
In this lecture we will
Discuss the basic concepts of fracture and relate this to
fatigue.
~~~~~~
General
Fatigue is a failure process caused by repeated application of load, even at
low stress levels. Hull structure and machinery are both subjected to cyclic
stresses and are very prone to fatigue failure. Fatigue damage is common in
older ships and represents a significant maintenance challenge. Fatigue
damage in machinery will normally result in equipment failure, requiring
replacement of the part. Fatigue damage in the hull structure is usually
considered ‘wear and tear’. Hull structures are very redundant, so that
fatigue failure of one element or joint will seldom results in catastrophic
failure. There are exceptions, as in the case of the FLARE, as shown below.
Fatigue cracks almost certainly were the precursor to the brittle fracture that
cut the ship in half.
Figure 1. Forward section of the FLARE [1].
Fatigue starts as the accumulation of micro-structural damage in the
material, depending on the amplitude of stress and the number of cycles. This
is called the ‘crack initiation’ stage of fatigue. After small cracks are formed,
the state of stress in the material is changed by the presence of the crack, and
E6003 – Ship Structures II 103
© C.G.Daley 2014
the crack tends to grow stably. This is called the ‘crack growth’ stage. Once
the crack is sufficiently long, it becomes unstable, and a rapid fracture occurs,
the ‘fracture’ stage. In simple structures and many machine parts, small
fatigue cracks will grow rapidly and lead to complete failure. In such cases,
fatigue can be regarded as a material property. To analyze these situations it
is common to employ the S-N curve, Miner’s rule and the concept of stress
concentrations.
Fracture Mechanics
Fracture, at the material level, is the creation of new surfaces, which
necessarily mean the breaking of atomic bonds. Figure 2 shows the three
basic ways of creating a crack. While it may seem that tension would create
the Type I crack and shear would create the other two, shear results in a
tension field at 45º and can also allow Type I cracks. Type II and III cracks
can be observed at crystal (grain) boundaries in metals an other
polycrystalline materials, but tend to become Type I cracks as they grow to
any reasonable size. The cracks turn to become normal to the tension (max.
principal stress), and thus tend to align with the direction of the minimum
principal stress (which may be compressive).
Figure 2. Three types of cracks at the primary material level.
E6003 – Ship Structures II 104
© C.G.Daley 2014
Figure 3 shows the stress field around a sharp crack in a plate subject to
constant nominal (far field) stress perpendicular to the crack. The stresses
near the crack tip tend to infinity. There is a pattern of stress in the material
beside the crack that follows the equation;
S local
K
2r
(3)
K S nom a
where r is the distance from the crack, and;
(4)
K is called the “stress intensity” and has units of MPa m . Clearly as r goes
to zero the stress goes to infinity. One might imagine that all sharp cracks
would always grow, because the stress at the crack tip is always infinity.
Observations show that sharp crack doesn’t always grow. There is always
some form of local ductility (plasticity) right around the crack tip. It has been
noticed that K is related to the growth of the crack. If K exceeds a critical
value, called Kc, (or K1c. as we would label the critical value of K for a mode I
crack) the crack will grow very fast (unstably). If K is very low, no crack
growth will occur.
Figure 3. Fracture geometry and stress patterns.
E6003 – Ship Structures II 105
© C.G.Daley 2014
The critical value of K is dependent on plate thickness. This is because plate
thickness affects the stress state at the tip of the crack. In thick plates, most
of the crack is in a condition of plane strain (i.e. zero strain in the through-
plate direction just in front of the crack), while in thinner plates, there is a
plane stress condition (i.e. zero stress in the through-plate direction). The
plane strain condition is more conducive to fracture, and so the critical value
K is lowest (see Figure 4.). To get plain strain, we need a plate thickness of at
least ;
K
2.5 1c
2
y
tmin
For example, for carbon steel of say y 235 N/mm2, K1c 7000 N/mm3/2, we get
for tough shipbuilding steels, the minimum thickness is extremenly thick.
a tmin = 2220mm (over 2 meters thick!). Clearly, cracks in ships are not in a
plane strain condition, which probably is part of the reason that they are
normally quite resistant to fracture, even when they contain numerous cracks.
Figure 4. Influence of plate thickness on fracture toughness.
As fracture toughness increases, the type of fracture behaviour changes.
Figure 5 illustrates. For very brittle situations, (some or all of: brittle
materials, thick materials, large cracks, cold temperatures) failure occurs at
stresses well below yield stress. The theory called Linear Elastic Fracture
Mechanics (LEFM), is used to describe this type of failure. As the situation
becomes more ductile, fractures still occur, but with significant plasticity at
the crack tip. In this region we would use Elastic-Plastic Fracture Mechanics
E6003 – Ship Structures II 106
© C.G.Daley 2014
(EPFM). For very ductile materials, substantial plasticity occurs prior to
material tearing, and we use Limit Load Analysis (LLA). This discussion
refers to the final failure of the structure. In the next part we will discuss
fatigue, which is a mechanism by which cracks are created and grown. When
fatigue cracks grow large enough, they may become unstable and would then
be dealt with by either LEFM or EPFM.
Figure 5. Types of Fracture analysis.
Fatigue as a Material Property
With cyclic stresses, and intermediate values of K, cracks will grow according
to how much K changes in each cycle. Paris [2], suggested the ‘Paris Law’ for
crack growth, of the form;
C K n
da
(5)
dN
where K = Kmax - Kmin is the range of K during a cycle of stress. The da/dN
term is the increase in the crack size (defined by the half-length a ) in each
cycle of stress. Figure 6 shows typical experiment data on crack growth. The
applied for K values in an intermediate range.
Paris Law is an idealization of the measured data. And should only be
E6003 – Ship Structures II 107
© C.G.Daley 2014
Figure 6. Crack growth diagrams (Paris Law)
Figure 7 shows an experimental arrangement that places a test cylinder in a
cyclic stress, which is the same along the test section. There are no notches or
other discontinuities in the test. The test section is polished, so that there are
not even small surface scratches. In this ideal case the test will continue with
no apparent damage until the section snaps. The cyclic stress causes micro-
structural damage, either within or between the grains of the metal.
Dislocations in the atomic lattice of the crystals are created and pile up,
eventually forming micro-cracks. A galaxy of micro-cracks will form, and
some of these will eventually coalesce into small cracks. Fatigue tests of this
type would take millions to trillions of cycles, and so would be left
unattended. Just prior to failure (say a few thousand cycles before) a
detectable small crack would exist. Tests of this form are really tests of the
formation of small cracks, and are thus “crack initiation” tests. If we plot
number of cycles to failure (N) and the cyclic stress (S), we get one point on
an “S-N” curve. Repeating the test at varying stress levels allows us to plots
the S-N relationship, as shown in Figure 8.
E6003 – Ship Structures II 108
© C.G.Daley 2014
Figure 7. Experimental investigation of material fatigue.
After N cycles at stress S, the section fails.
E6003 – Ship Structures II 109
© C.G.Daley 2014
Figure 8. S-N curve for ideal polished cylinder.
Figures 9 and 10 show a typical cross section of a fatigue crack after failure.
The location of the small formation crack is labeled the ‘initiation point’. The
crack then grows steadily and leave beach marks showing the crack front.
The marks represent locations where the fracture was temporarily arrested.
When the fatigue crack is large enough, the stress intensity at the crack front
reaches a critical value and the crack becomes unstable. At this point the bar
snaps.
Figure 9. Fatigue failure cross section.
E6003 – Ship Structures II 110
© C.G.Daley 2014
Figure 10. Photo Showing Beach Marks.
References
[1] Marine Investigation Report, “Break-Up and Sinking The Bulk Carrier
"FLARE" Cabot Strait 16 January 1998”, Transportation Safety Board of
Canada, Report Number M98N0001
[2] Paris, P.C., “Fatigue- An interdisciplinary approach”, ed. Burke, J.J.,
Reed, N.L., and Weiss, V., 10th Sagamore Army Mat. Res. Conf. Syracuse
University Press, New York, 1964.
[3] Dover, W.D., “Fatigue Behaviour of Offshore Structures” in Offshore
Structures, V.2, Ed. by Reddy and Arockiasamy, Kriegar Publishing, Malabar
Florida, 1991.
[4] Paik, J.K., Thayamballi, A.K., Ultimate Limit State Design of Steel
Plated Structures, Wiley, 2003
E6003 – Ship Structures II 111
© C.G.Daley 2014
T.17 – Problems.
1. What might the physical reason be to explain why there is a low
stress fatigue limit in steel?
E6003 – Ship Structures II 112
© C.G.Daley 2014
Topic 18: The Fatigue Process in Ships
Introduction
In this lecture we will
Discuss Miners rule for dealing with fatigue in a complex
loading situation.
Next we discuss the stages of fatigue and applications to ship
structural details.
~~~~~~
In complex structures, fatigue cracks will grow in a slow and stable way, and
do not necessarily pose a large risk. In these cases fatigue should be viewed
as a sequence of fracture processes, dependant on material properties, stress
ranges and geometry. To analyze these situations it is common to employ
crack growth models such as the Paris Law, along with fracture mechanics
concepts such as “stress intensity”, and fatigue life estimation. Alternatively,
it is also common to use “detail-specific” S-N curves, which are empirical
fatigue obtained from structural tests of complete structural details.
Miner’s Rule
In most cases, the stress amplitude is not constant. Most stress time histories
are random, and are described by a frequency distribution. With the aid of a
Fourier transform, a random time history can be described as the sum of a
set of sine waves. Miner [2] proposed a simple method of determining the
fatigue strength of a material subjected to variable amplitude cycles. Figure
1 shows the decomposition of a time history into a series of waves. Each wave
has an amplitude Si and a number of cycles ni. Miner suggested that each
wave is causing damage in the metal. Stress S1 would result in fatigue failure
after N1 cycles. After n1 cycles, S1 has caused a damage ratio D1, where;
D1 = n1/N1 (1)
D = all i ni/Ni
The total damage D is the sum of the contributions from all the waves;
(2)
When the damage ratio D becomes 1, failure will occur. Miner’s rule actually
applies only to crack initiation. For cases where this is most of the fatigue
life, it is a good approximation to the fatigue life.
E6003 – Ship Structures II 113
© C.G.Daley 2014
Figure 1. Miner’s Rule for variable stress.
Fatigue as a Process
The above discussion relates to a perfect geometry, with no initial material
flaws, and no sharp corners or other “stress–risers”. In real structures, the
stress is not uniform. There are sharp corners and flaws that result in
localized stress peaks, called ‘stress risers’. Figure 7 shows a typical detail.
The end of a bracket, although tapered and sniped, causes a local stress peak
called a ‘hot spot’. The stresses are highest at the toe of the weld where a
corner exists. The stress in this region is a multiple of the nominal stresses.
So, as the nominal stresses oscillate, the ‘hot-spot’ stresses will oscillate with
greater magnitude. Obviously, the material at the toe of the weld will fatigue
fastest, and cracks will form here first. The relative complexity of the
geometry causes the stress peaks. Figure 8 shows how fatigue data for
welded joints tends to differ from ideal fatigue tests. The fatigue strength, for
the same material, tends to be lower, and the lower cutoff is much less well
defined.
Figure 9 shows that fatigue is composed of three phases; initiation,
propagation, and final failure. In simple tests the initiation phase is the most
important. Once the crack has formed it grows very rapidly and the part fails.
E6003 – Ship Structures II 114
© C.G.Daley 2014
In more complex structures, the stresses that caused the crack to initiate are
quite localized. As the crack grows it tends to relieve the stresses. The crack
can grow slowly or even stop. In such cases the crack propagation phase
dominates the failure process. In real structures, there are often many cracks
left over from the fabrication process. These can be weld flaws or just
geometric discontinuities. In such cases the cracks are already initiated, and
fatigue life depends on the crack growth process.
Figure 2. Fatigue cracks in structural details. Localized geometric
discontinuities cause local stress peaks.
E6003 – Ship Structures II 115
© C.G.Daley 2014
Figure 3. Typical S-N curves for welded connections compared with ideal
specimens.
Figure 4. The 3 stages of fatigue.
Figure 5 The 3 regions of crack growth rate.
E6003 – Ship Structures II 116
© C.G.Daley 2014
Figure 6 Crack growth during ‘service’ life.
References
[1] Dover, W.D., “Fatigue Behaviour of Offshore Structures” in Offshore
Structures, V.2, Ed. by Reddy and Arockiasamy, Kriegar Publishing, Malabar
Florida, 1991.
[2] Miner, M.A., Jnl. Appl. Mech., ASME, Vol. 12. Sept 1945
E6003 – Ship Structures II 117
© C.G.Daley 2014
T.18 – Problems.
1. For the following stress history and SN curve, determine if the fatigue
failure is likely to have occurred or not.
Stress History: 20,000 cycles of amplitude 300MPa
100,000 cycles of amplitude 200MPa
880,000 cycles of amplitude 100MPa
The SN curve: log(N)=12.164 – 3 x log()
2. Discuss why crack propogation is the most significant part of the
fatigue like of ships, even though initiation is the dominant part of the
fatugue life in lab tests.
E6003 – Ship Structures II 118
© C.G.Daley 2014
Topic 19: Discussion of Hot Spot Fatigue Analysis
Introduction
In this lecture we will
Discuss the DnV Recommended Practice for Fatigue Analysis
of Offshore Structures (RP-C203, 2001), especially the idea of
hot spot analysis
~~~~~~
Intro:
Read RP-C203, especially the parts related to plated structures (not the
tubular joints or pipelines). I’ve included some parts below.
Concept
On the basis of experiments, S-N curves for various structural details were
found. Table 2.3.3 is for structures in a seawater environment with cathodic
instance, S-N curve ‘D’. The table shows that log( a ) 11.764 .
protection. Figure 2.3-2 shows the same S-N data in graphical form. Take for
The equation for the S-N curve is;
t k
log N log( a ) m log
t
ref
where
N : number of cycles of to failure.
m: the ‘negative inverse’ slope, (=3 for stresses above the fatigue limit)
For example, at = 200MPa;
N=10^(11.764-3 log (200)) = 72,595 cycles
The figure shows the ‘D’ curve crossing the 200MPa line just above 70,000
cycles, as we would expect.
The letter ‘D’ and all other letters refer to the detail category. Table 5 below
shows some category ‘C’ and ‘D’ details. The more prone the detail is to
fatigue, the higher the letter and the lower the S-N curve.
Category ‘D’ can be used for a more general purpose. Note that in table 2.3.3,
on the right hand side, the stress-concentration factor as derived by the hot
spot method is 1.0. This means that we can use the ‘D’ line with stresses
calculated by the hot-spot method for any structural detail. The hot spot
method is a way to do fatigue analysis on a detail that is not in the table.
E6003 – Ship Structures II 119
© C.G.Daley 2014
E6003 – Ship Structures II 120
© C.G.Daley 2014
E6003 – Ship Structures II 121
© C.G.Daley 2014
E6003 – Ship Structures II 122
© C.G.Daley 2014
The Hot Spot Method
As you can see in the table above, category ‘D’ details are quite simple, while
higher categories (e.g. ‘F’ ) are more complex and have more geometric
discontinuities which lead to stress concentrations. Note that for category ‘D’, the
only significant aspect is the presence of a weld in an otherwise straight section.
The hot-spot method treats fatigue by separating two types of stress concentration
effects. The stress concentration due to the overall structural geometry is
separated from the stress concentration due to the very local geometry of the weld.
The reason for this is that it is very difficult to properly model the detailed shape
of a weld. On the other hand, it is reasonably practical to model the general stress
concentration in a detail. We use the following definitions;
Hot spot stress : the stress in the region of the toe of the critical weld, accounting
for the general detail geometry, but without the very sharp stress concentration
due to the weld toe geometry.
Notch stress: This is the stress at the weld toe, including the stress concentration
due to the weld toe shape. (very hard to model)
When making a finite element model, we often have very large (though very
localized) stresses at discontinuities. For this reason, the hot spot stress is not
taken directly from the FE model right at the discontinuity (the toe of the weld).
Rather, the hot spot stress is projected to the hot spot, from values nearby. This is
a practical way to avoid the spurious (false) results that an FE model gives at
sharp discontinuities. Figure 2.13-1 shows how the stresses are projected from
nearby. Once we get the (projected) hot-spot stress, we can use category ‘D’ to
estimate the fatigue life of the hot-spot (and the detail).
E6003 – Ship Structures II 123
© C.G.Daley 2014
E6003 – Ship Structures II 124
© C.G.Daley 2014
E6003 – Ship Structures II 125
© C.G.Daley 2014
T.19 – Problems.
1. Explain how you would verify that you are doing the hot spot analysis
correctly for your detail of interest.
E6003 – Ship Structures II 126
© C.G.Daley 2014
Topic 20: Section Modulus – a measure of moment capacity
Introduction
In this lecture we will
Review elastic section modulus
Describe plastic section modulus
~~~~~~
Elastic section modulus
A section modulus is a geometric measure of the ability of a cross-section to
carry a moment. An elastic section modulus measures the ability to carry an
elastci moment, while a plastic section modulus indicates the ability to carry
a plastic moment. We denote section modulus with S (or sometimes Z).
Recall that the moment of inertia of an area is;
Ix y 2 da
where y is measured from some baseline. Ix will be a minimum if the baseline
runs through the centroid. In this case we call the Ix : Ina (through the
neutral axis).
For a rectangle w wide and h high, we have Ina = 1/12 w h3. In general we
can find I by integration or by the parallel axis theorm;
I base ay 2 i
where y, in this case, is measured from a baseline
and
I na I base A hna
2
ay
where
A a and hna
a
For elastic stresses, the stress formula is;
E6003 – Ship Structures II 127
© C.G.Daley 2014
My
I
where in this case y is measured from the neutral axis.
at the elastic limit, the moment is My and the stress is the yield stress (y at
y=c, c being the largest distance from the neutral azis on the section), so we
have;
M c
y y
I
We can combine I and c to give the elastic section modulus;
y y where S e
M I
Se c
Note that we can write;
Se
My
y
,
M y Se y
and more importantly;
We can think of the section modulus as the geometric part of the moment
capacity, while y is the material part.
From this perspective, Se can be thought of as the moment per unit strength
(i.e. the moment for y = 1.0)
We can find Se from I/c.
Alternatively we can find Se from moment considerations (graphically). The
figure below shows how to find Se in this way. If we imagine that the stress at
c is 1.0, for elastic bending the stress is linear and zero at the centroid. The
force in any horizontal strip (b wide, dh high) will be equal to b dh y/c. We can
transform the section by shrinking all widths according to y/c. This is a linear
‘squeezing’ of the area, with the new width at the centroid going to zero, the
width at c remaining unchanged, and all other widths adjusted accordingly.
We now have a modified section, with two parts (above and below the
centroid).
E6003 – Ship Structures II 128
© C.G.Daley 2014
The elastic section modulus can be found as the moment of the modified area
(i.e. the half-area times the distance between the centroids of the upper and
lower area).
Why should this be?
In any pure bending situation the axial force is zero. This means that the
compressive and tensile forces balance. By transforming the section, we create
a pattern of force, reflecting the real force, and equivalent to the force that
would occur if there were a uniform pressure over the transformed area. The
center of action of the uniform pressure would be at the centroid of the area.
For a unit pressure, the forces equal the area, and the moment equals the
product of the half-area times the distance between the centroids. It may
seem obvious that in the case shown that the two areas (upper and lower) are
equal, but this is generally true even for unsymmetrical sections as shown
below.
E6003 – Ship Structures II 129
© C.G.Daley 2014
E6003 – Ship Structures II 130
© C.G.Daley 2014
Plastic section modulus
Again, the plastic section modulus is a geometric measure of the ability of a
cross-section to carry moment. In the plastic case we assume that the stress is
uniformly at yield everywhere. As such there is no need to transform the area.
As with elastic moments, there is no net axial force. This means that the
compressive and tensile forces must balance. As the stresses are all at yield
(in both tension and compression) it means that exactly half the area is in
compression and exactly half is in tension. This means that the plastic neutral
axis is at the half-area. For symmetrical sections this is the same as the
centroid, but different for unsymmetrical sections.
Whether symmetrical or not;
Sp
Mp
y
,
M p S p y
and;
Sp, like Se has units of length3, (e.g. mm3). The plastic section modulus, Sp, can
be found as the moment of the half-area (i.e. the half-area times the distance
between the centroids of the upper and lower half-areas).
This gives a physical (graphical) meaning to Sp. For typical ships frames, with
plating, web and flange, it is convenient to have a generic formula for the
plastic section modulus. Its rather long;
ap aw af
C3
C2
Sp
C1
bp
C1 C 2 C3 ap aw af
tw
C1
1
bp tp 2 tw hw2 bf tf 2
C2 bp tp tw hw bf tf
2
1 2
C3 hw bf tf
4
E6003 – Ship Structures II 131
© C.G.Daley 2014
E6003 – Ship Structures II 132
© C.G.Daley 2014
T.20 – Problems.
1. Find the elastic and plastic modulii for the hollow section;
E6003 – Ship Structures II 133
© C.G.Daley 2014
Topic 21: Strain Rate Effects
Acknowledgement: This chapter prepared by B.W.T. Quinton
Introduction
In this lecture we will:
Introduce the concept of material strain rate effects.
Use the Cowper-Symonds expression to resolve a dynamic
yield stress.
So far during our analysis of ship structures we have implicitly assumed that
all loads are applied quasi-statically; that is, applied sufficiently slowly such
that inertial and material strain rate effects are insignificant, and therefore
ignorable. What if we are dealing with a material whose constitutive
equations are sensitive to strain rate (like steel is)?
In this lecture we will learn how to account for material strain rate effects in
the context of our assumption in previous lectures that steel behaves in an
elastic-perfectly plastic manner.
Dynamic Yield Stress
Steel is a material whose behaviour is different for different strain rates; i.e.
it exhibits material strain rate effects. As the term implies, strain rate is the
speed at which the strain, �, in a material changes with respect to time.
Strain is (as we have seen before) given by the change in length of
a geometry divided by the original length:
−
�=
Strain rate is the derivative of � w.r.t. time:
�� � −
�̇ = =
� � ∙
In the limit of time from → , we have:
�̇� = ��
�̇ − =�−�
If we assume that = and � = than we have:
�
�̇ =
E6003 – Ship Structures II 134
© C.G.Daley 2014
Specifically, material strain rate effects affect the constitutive relationships
of a material. In this course we have assumed an elastic-perfectly plastic
constitutive model for steel.
FIGURE 1: ELASTIC-PERFECTLY PLASTIC CONSTITUTIVE MODEL.
In this context, the strain rate affects only the yield stress, �� ; Young’s
Modulus, �, remains unchanged.
Experiments (particularly by Cowper and Symonds3,4) have shown that the
yield stress, �� , of steel increases as strain rate, �̇, increases. We may call
this new value for yield stress, the dynamic yield stress, ��′ , where:
��′ > ��
This phenomenon implies that structures loaded at high strain rates exhibit
linear elastic behaviour at stress levels above the yield stress, �� . This in
turn implies that plastic collapse will occur at ��′ instead of �� .
FIGURE 2: ELASTIC-PERFECTLY PLASTIC CONSTITUTIVE MODEL ALTERED BY MATERIAL STRAIN RATE EFFECTS.
Therefore, for rapidly loaded structures, we may replace the static yield
stress, �� , with the dynamic yield stress, �′� . So how do we calculate the
dynamic yield stress?...
3 Symonds, P.S., Survey of Methods of Analysis for Plastic Deformation of Structures under
Dynamic Loadings, Brown University, Report BU/NSSRDC/1-67 (1967)
4 Symonds, P.S., Viscoplastic behaviour in response of structures to dynamic loading, in
Behaviour of Materials Under Dynamic Loading, p. 106, ed. N. J. Huffington, ASME (1965)
E6003 – Ship Structures II 135
© C.G.Daley 2014
The Cowper-Symonds Relationship
Based on experimental data, Cowper and Symonds5 used the following
regression equation to predict the dynamic yield stress, ��′ :
⁄�
�̇
��′ = �� [ + ( ) ]
�
where: � and � are constants called the Cowper-Symonds Parameters
Note: � has the same units as strain rate, i.e. [ ⁄ ] (or
equivalently [ ⁄ ⁄ ] as sometimes strain is expressed in
units of [ ⁄ ]) and � is dimensionless.
This relationship has attained almost universal acceptance in the
engineering community because analytical and numerical predictions agree
remarkably well with actual experimental data (Jones, 1983).
For hot-rolled mild steel, Cowper and Symonds determined that:
�= . and � =
Values for other materials are given in the following table6:
Material C [-/s] p
Stainless Steel 304 100 10
Alpha-Titanium (Ti-50A) 120 9
Aluminum 6500 4
If we compare hot-rolled mild steel to other material strain rate sensitive
materials, we find that it is particularly sensitive. The figure below shows
the dynamic scale factor, �, for these various materials plotted verses strain
rate. If we notice that the static yield stress, �� , is simply scaled by the term
in brackets then we may say:
��′
=�
��
where: � is the dynamic scale factor and
�̇ ⁄�
�=[ + �
]
5 See the above Symonds references.
6 Taken from:
Jones, N., Structural crashworthiness, p. 321, ed. N. Jones and T. Wierzbicki, Butterworths
(1983)
E6003 – Ship Structures II 136
© C.G.Daley 2014
450%
Hot-rolled Mild Steel
400% Stainless Steel 304
Alpha Titanium (Ti-50A)
Aluminum
Percentage Increase in Yield Stress
350%
300%
250%
200%
150%
100%
50%
0%
1.0E-08 1.0E-06 1.0E-04 1.0E-02 1.0E+00 1.0E+02
Strain-rate [1/s]
FIGURE 3: COWPER-SYMONDS DYNAMIC SCALE FACTOR FOR THE YIELD STRESS OF VARIOUS MATERIALS.
Figure 3 shows us that hot-rolled mild steel is very sensitive to material
strain rate effects, while aluminum is virtually unchanged.
Limits on the Cowper-Symonds Relationship
It should be noted that the above given values for � and � are valid only up to
a few percent strain (e.g. < � < . for hot-rolled mild steel). This is
because the dynamic stress, � ′ , is a local maximum at the dynamic yield
stress, ��′ , and drops off significantly at higher strains for most normal strain
rates.
Since our elastic-perfectly plastic constitutive model assumes that the stress
is constant once yield is achieved, than for strains higher than say, 5%, we
cannot use the above � and � values because they will overestimate the
resulting constant stress. We require new values of � and � for strains above,
about 5%.
Figure 11.5 below illustrates this concept. Notice for the STATIC curve that
all stresses after yield are larger than the perfectly plastic assumption. The
curve above the STATIC curve is for a strain rate of 0.02 [-/s] and shows that
the stress after the dynamic yield stress, ��′ , is significantly less than the
perfectly plastic assumption.
E6003 – Ship Structures II 137
© C.G.Daley 2014
Estimating Strain Rate
Do to non-linear geometry effects, strain rate is difficult to determine for a
ship’s structure; however we may estimate the average strain rate by
estimating the dynamic loading time.
We get the dynamic loading time from either experiments or by consulting
published data. For example, Giannotti and Associates, Inc.7 estimate that
the collapse of structural members of a container ship occurs within 0.18
seconds. Thus if we assume that the local maximum strain is 0.36, then the
average strain rate in [-/s] is:
� .
�̇ = = =
. 8
Thus for hot-rolled mild steel:
⁄5
�̇ ⁄�
� =[ +( ) ]=[ +( ) ]= .
� .
and:
��′ = . ��
7Giannotti and Associates, Inc., Determination of Strain Rates in Ship Hull Structures. A
Feasibility Study, Report 81-114 for Ship Structure Committee (1982)
E6003 – Ship Structures II 138
© C.G.Daley 2014
Concluding Remarks
We have seen that we may use the Cowper-Symonds relationship to
determine the dynamic yield stress, ��′ for structural analyses that involve
high strain rates. Since the dynamic yield stress is greater than the static
yield stress it is advantageous to use this relationship when applicable, as
significant weight savings in structural steel may be possible during ship
design.
Further to the above lecture, the following two items should be noted:
1. The Cowper-Symonds relationship may be significant for quasi-static loads as the
�̇
may be sufficiently large in cases where inertial effects can still be ignored.
�
2. The Cowper-Symonds relationship is based on uniaxial data, but its application
has been extended to multi-dimensional stress states in practice; with little to no
experimental data to support its applicability.
E6003 – Ship Structures II 139
© C.G.Daley 2014
Topic 22 : Dealing with Uncertainty in structural design
Introduction
We live with uncertainty. We build structures to limit that uncertainty.
The earliest structures were built to cope with the uncertainties of the
weather. Ships must cope with the uncertain sea.
The theory of probability and statistics gives us a way of understanding and
coping as best we can, with uncertainty. There are several theories and
methods developed to specifically deal with engineering design problems
involving uncertainty. These methods have become popular in recent years
and are being implemented in many codes and standards, as well as in
engineering design and analysis. Not all the developments are sensible. Few
professionals can really claim to understand the issues involved, yet many
believe it to be the correct approach. Much further development of these
methods will likely occur. An understanding of the methods and issues is a
crucial element of a professional education in structures.
Somewhere out there
E6003 – Ship Structures II 140
© C.G.Daley 2014
Probability Concepts
ray-traced image by Jeremy Ginsberg and Matt Ginzton June 1998
Uncertainty vs Certainty.
When you roll a single die:
- You are certain to have one of the following: 1,2,3,4,5 or 6
- You are certain not to have 7 or 3.2
- You have an equal chance of getting any of the 6 numbers
- What if the die is flawed? What changes?
Deterministic vs Random.
- Deterministic implies determination by exact natural laws. The
outcome is conditionally pre-determined. e.g. drop an object and it
accelerates at 9.81 m/s2
- Random means that the outcome is determined by a fully or
partially unknown cause. e.g roll a die.
Dynamic vs Static
- Dynamic means varying in time.
- Static means constant in time.
- Probability concepts can be applied to both static (is there oil
6000m below me) and dynamic (what will be the largest wave to
strike the Hibernia platform tomorrow?)
A “Random Variable” means an uncertain dynamic variable, such as the sea
surface.
E6003 – Ship Structures II 141
© C.G.Daley 2014
Histogram: plot of variable occurrences
- May be expressed as observation counts.
- May be expressed as % (observed or predicted chance).
- Cumulative probability for 1 roll of a ‘fair’ die:
E6003 – Ship Structures II 142
© C.G.Daley 2014
While dice and cards have discrete values (integers), waves have
continuous values (real numbers), and can be modeled with smooth
distributions:
The most common distribution is the Normal distribution:
E6003 – Ship Structures II 143
© C.G.Daley 2014
- The normal distribution is denoted X=N()(x),
x
1 / 2
N ( , )( x)
2
2
1
e
where
= the mean = ( xi ) / n
= standard deviation = (( (xi - )2 / (n-1))0.5
(in excel this is STDEV(n1,n2,n3…) )
X is the random variable
x is one specific value (measurement) of X
The “Standard” Normal distribution is N(0,1). All other Normal
distributions are just stretched/shifted versions of N(0,1).
We usually want to find the probability that X is in a certain range.
(e.g. pr(X<10) or pr(3<X<5). For this the more useful function is the
cumulative Normal distribution : (x)
(x) : the probability that the value of X will be less than x.
x
e
1 / 2
( , )( x)
2
1
2
dx
There is no closed form solution for this integral. Instead we use tables
or built-in functions.
(in excel (x) is “=NORMDIST(x,,1)”
See the handout for the Normal Distribution.
e.g. for (0,1) pr(x<1.52) = .9357
for (2,3) pr(x<0) = 1- (0,1)(2/3) = 1-.748 = 0.252
E6003 – Ship Structures II 144
© C.G.Daley 2014
Discussion : Design for safety:
You are asked to design a ferry in which the chances of dying are
“acceptable”.
Define “acceptable” in a rigorous, defensible way.
What are your chances of dying?
- today?
- this year?
- if you go on a ferry?
- What about specific risks – e.g. collision with another ship?
Are current ferries safe enough?
Who should say what is safe enough?
The Herald of Free Enterprise
Extra Reading:
See pdf extract from Target Risk by Gerald J.S. Wilde
See pdf : Personal observations on the reliability of the Shuttle by R. P.
Feynman
L.2 – Problems.
1. What are the chances of drawing a Queen from a deck of cards?
2. What have you assumed in the previous question?
3. What if you thought that someone had lost some of the cards in a
standard deck of playing cards. Then what would you estimate the
chances of drawing a Queen be?
4. A rocket booster is expected to crash in the ocean. The potential crash
zone is 170 x 80 km. Estimate the chance of the rocket hitting a
200,000 tonne ship, moored in the crash zone?
5. Find: (7,4)(10), (7,4)(4), (7,4)(-1)
E6003 – Ship Structures II 145
© C.G.Daley 2014
Standard Normal Probabilities, (mean = 0, SD = 1)
(probability that a value, randomly chosen from X, is less than x)
Pr(X<x)
x 0 1 2 3 4 5 6 7 8 9
0 0.5 0.503989 0.507978 0.511967 0.515953 0.519939 0.523922 0.527903 0.531881 0.535856
0.1 0.539828 0.543795 0.547758 0.551717 0.55567 0.559618 0.563559 0.567495 0.571424 0.575345
0.2 0.57926 0.583166 0.587064 0.590954 0.594835 0.598706 0.602568 0.60642 0.610261 0.614092
0.3 0.617911 0.621719 0.625516 0.6293 0.633072 0.636831 0.640576 0.644309 0.648027 0.651732
0.4 0.655422 0.659097 0.662757 0.666402 0.670031 0.673645 0.677242 0.680822 0.684386 0.687933
0.5 0.691462 0.694974 0.698468 0.701944 0.705402 0.70884 0.71226 0.715661 0.719043 0.722405
0.6 0.725747 0.729069 0.732371 0.735653 0.738914 0.742154 0.745373 0.748571 0.751748 0.754903
0.7 0.758036 0.761148 0.764238 0.767305 0.77035 0.773373 0.776373 0.77935 0.782305 0.785236
0.8 0.788145 0.79103 0.793892 0.796731 0.799546 0.802338 0.805106 0.80785 0.81057 0.813267
0.9 0.81594 0.818589 0.821214 0.823814 0.826391 0.828944 0.831472 0.833977 0.836457 0.838913
1 0.841345 0.843752 0.846136 0.848495 0.85083 0.853141 0.855428 0.85769 0.859929 0.862143
1.1 0.864334 0.8665 0.868643 0.870762 0.872857 0.874928 0.876976 0.878999 0.881 0.882977
1.2 0.88493 0.88686 0.888767 0.890651 0.892512 0.89435 0.896165 0.897958 0.899727 0.901475
1.3 0.903199 0.904902 0.906582 0.908241 0.909877 0.911492 0.913085 0.914656 0.916207 0.917736
1.4 0.919243 0.92073 0.922196 0.923641 0.925066 0.926471 0.927855 0.929219 0.930563 0.931888
1.5 0.933193 0.934478 0.935744 0.936992 0.93822 0.939429 0.94062 0.941792 0.942947 0.944083
1.6 0.945201 0.946301 0.947384 0.948449 0.949497 0.950529 0.951543 0.95254 0.953521 0.954486
1.7 0.955435 0.956367 0.957284 0.958185 0.959071 0.959941 0.960796 0.961636 0.962462 0.963273
1.8 0.96407 0.964852 0.965621 0.966375 0.967116 0.967843 0.968557 0.969258 0.969946 0.970621
1.9 0.971284 0.971933 0.972571 0.973197 0.97381 0.974412 0.975002 0.975581 0.976148 0.976705
2 0.97725 0.977784 0.978308 0.978822 0.979325 0.979818 0.980301 0.980774 0.981237 0.981691
2.1 0.982136 0.982571 0.982997 0.983414 0.983823 0.984222 0.984614 0.984997 0.985371 0.985738
2.2 0.986097 0.986447 0.986791 0.987126 0.987455 0.987776 0.988089 0.988396 0.988696 0.988989
2.3 0.989276 0.989556 0.98983 0.990097 0.990358 0.990613 0.990863 0.991106 0.991344 0.991576
2.4 0.991802 0.992024 0.99224 0.992451 0.992656 0.992857 0.993053 0.993244 0.993431 0.993613
2.5 0.99379 0.993963 0.994132 0.994297 0.994457 0.994614 0.994766 0.994915 0.99506 0.995201
2.6 0.995339 0.995473 0.995603 0.995731 0.995855 0.995975 0.996093 0.996207 0.996319 0.996427
2.7 0.996533 0.996636 0.996736 0.996833 0.996928 0.99702 0.99711 0.997197 0.997282 0.997365
2.8 0.997445 0.997523 0.997599 0.997673 0.997744 0.997814 0.997882 0.997948 0.998012 0.998074
2.9 0.998134 0.998193 0.99825 0.998305 0.998359 0.998411 0.998462 0.998511 0.998559 0.998605
3 0.99865 0.998694 0.998736 0.998777 0.998817 0.998856 0.998893 0.99893 0.998965 0.998999
3.1 0.999032 0.999064 0.999096 0.999126 0.999155 0.999184 0.999211 0.999238 0.999264 0.999289
3.2 0.999313 0.999336 0.999359 0.999381 0.999402 0.999423 0.999443 0.999462 0.999481 0.999499
3.3 0.999517 0.999533 0.99955 0.999566 0.999581 0.999596 0.99961 0.999624 0.999638 0.99965
3.4 0.999663 0.999675 0.999687 0.999698 0.999709 0.99972 0.99973 0.99974 0.999749 0.999758
3.5 0.999767 0.999776 0.999784 0.999792 0.9998 0.999807 0.999815 0.999821 0.999828 0.999835
3.6 0.999841 0.999847 0.999853 0.999858 0.999864 0.999869 0.999874 0.999879 0.999883 0.999888
3.7 0.999892 0.999896 0.9999 0.999904 0.999908 0.999912 0.999915 0.999918 0.999922 0.999925
3.8 0.999928 0.99993 0.999933 0.999936 0.999938 0.999941 0.999943 0.999946 0.999948 0.99995
3.9 0.999952 0.999954 0.999956 0.999958 0.999959 0.999961 0.999963 0.999964 0.999966 0.999967
4 0.999968 0.99997 0.999971 0.999972 0.999973 0.999974 0.999975 0.999976 0.999977 0.999978
E6003 – Ship Structures II 146
© C.G.Daley 2014
Topic 23: R-Q calculation of the probability of structural failure
Introduction
We must design structures for random loads. In such circumstances there can
be no absolute guarantee of safety. The best we can do is to create a structure
with a known and acceptable probability of failure.
We define failure as any circumstance in which load>strength.
This would be true for any internal (load effect) as well as any external load.
Both load and strength are random quantities (random variables).
The probability of failure is thus the probability that load exceeds strength.
We have a system of two random variables, and failure will occur for every
cases in which load>strength.
Typically we use open-tail probability density functions (eg Normal) so that
values theoretically go to ∞.
The total probability of failure must cover all possible values of load and
strength. Failure can occur for any strength level, because it is always
possible that the load is even higher.
Somewhere out there
E6003 – Ship Structures II 147
© C.G.Daley 2014
A Simple approach to safety
We will derive a mathematical expression for the probability of failure,
assuming that we have of a model of the probabilities of both load and
strength.
Define:
Q load (action, demand)
R strength (re-action, capacity)
q a specific load (a particular occurrence or example of..)
r a specific strength
we say that: q Q, q is a member of Q
and : r R
we have failure when q>r (we can’t say Q>R)
the mean load is Q
the mean strength is R
Q < R so ‘on average’ we are safe,
but if we choose examples:
q1 <r1 safe
q2 >r2 failure !
R
note : the factor of safety on the mean, FOSm, is
Q
unfortunately the FOSm has little relation to the probability of failure.
E6003 – Ship Structures II 148
© C.G.Daley 2014
To determine to probability of failure, we need to consider all possible
circumstances. We start by picking either load or strength (lets pick
strength – but it doesn’t matter which). For each possible value (of
strength), we calculate the chance of failure. Once we’ve done this for all
possible values (of strength), we add up the partial chances to get the
total probability of failure. It would give the same answer if we picked
load. The calculation can be visualized with the following plot of Q, R, q,
r:
all possible values of load and strength are values along the x axis. The
chances of either of these values are given by the Q(x) or R(x) probability
density curves. The chances of r falling in the small range ‘dx’ at x is:
Prob(r=x) = R(x) dx (remember R is probability density)
for r=x, the (conditional) probability of failure is the same as the chance that
q>x. We can express this as:
Prob(q>x) = area under Q curve to the right of x
Q( x)dx
= (this is the shaded area)
x
Note: for two independent things, A,B, to happen together, the chance is
Prob(A and B) = Prob(A) x Prob(B)
Therefore the chance of failure, when r=x is:
Q( x)dx
Prob(fail)|r=x = R(x) dx
x
The total probability of failure must include all possible values of r. We need
to sum the above probabilities for all r. This is written as:
Q( x)dx ]
Prob(fail) = [ R(x) dx
x
This integral of an integral is called a convolution integral. However, we can
simplify this to a simple integral.
E6003 – Ship Structures II 149
© C.G.Daley 2014
Qcumulative( x) Q( x)dx
x
therefore
Q( x)dx = 1- Q( x)dx
x
x
= 1- Qcumulative(x)
The total probability of failure is:
Prob(fail) = R( x) (1 Q
all x
cumulative ( x))dx ]
A word about geometry ----------------
It is always nice to see an idea geometrically. In this case there are a couple of
geometric analogs to the probabilities.
In the case of a probability density function, the probability of having a value
between any two points (on the x axis) is equal to the area between those
points.
In the case of the failure probability, the value is determined by combining all
possible probabilities for R and Q. A common misunderstanding is that this is
equal to the area of the overlap between the two curves. This is NOT true.
E6003 – Ship Structures II 150
© C.G.Daley 2014
However, there is a geometric analog to the total probability of failure. We
plot R on one horizontal axis, Q on another, and probability density on the
vertical axis. In this space, the probabilities of all combinations of R and Q
form a surface (like a Mexican hat). The volume under the whole surface is 1.
A line R=Q forms a diagonal line on the horizontal plane. One side of this
line represents Q>R and includes all the combinations that result in failure.
The failure probability is the volume under the surface in this region.
Summary
We have calculated the exact probability of failure. To do so we used the
probability density functions for both load and strength. For simple cases of
one type of load and one failure mechanism, we need good data for both, so
that we could fit good probability distributions. For cases of multiple types of
load or multiple failure mechanisms, we would need to combine these factors
in some way. This gets complicated in practice.
E6003 – Ship Structures II 151
© C.G.Daley 2014
Exercise 1: do on your own.
Recall that when we derived the failure equation, we said that we could start
with either R or Q, but we started with R and we got:
Prob(fail) = R( x) (1 Q
all x
cumulative ( x))dx ]
If we had started with Q, which of the following formulas would we have
derived?
a) Prob(fail) = Q( x) (1 R
all x
cumulative ( x))dx ]
b) Prob(fail) = Q( x) R
all x
cumulative ( x)dx ]
c) Prob(fail) = R
all x
cumulative
( x) (1 Q( x))dx ]
E6003 – Ship Structures II 152
© C.G.Daley 2014
Example: Calculation of Failure Probability
For these load and strength distributions, calculate the exact probability of
failure;
If a failure takes place, it must occur with load and strength in the 34
range. We will use the equation
Prob(fail) = R( x) (1 Q
all x
cumulative ( x))dx
But before we do, lets estimate (guess) what the answer should be.
1/2 of all R values are in the 34 range
1/8 of all Q values are in the 34 range.
Failure happens when both Q and R are in the 34 range. The chance of this
is:
P_fail = P(Q 34 AND R 34) = P (Q 34) x P (R 34)
= 1/2 x 1/8 = 1/16
But even in the 34 range, r could be greater than q. A quick estimate is that
1/2 of the time we have r>q, so
P_fail = 1/2 x 1/16 = 1/32
But most of the Q values are in the lower part of the range, so lets guess that
in the 34 range P(R<Q) = 1/3. This gives us an estimate of:
P_fail = 1/3 x 1/16 = 1/48 our best guess
Now lets calculate the value exactly.
First we need to find the height of the triangle and rectangle. Both have area
of 1, so, by inspection, both have a height of 0.5.
In the range x 34
R(x) = 0.5 (constant)
Q(x)= 1-.25x (check @x=4 Q=0, and @x=2, Q=0.5 OK)
E6003 – Ship Structures II 153
© C.G.Daley 2014
We need to get a mathematical expression for Qcumulative . Note that there are two
branches for this. We must know the area up to x=2, as this will be the
constant in the integral equation for the x 24. It is obvious that the area
up to x=2 is 0.5. Therefore:
Qcumulative 0.5 Q( x)dx 0.5 (1 .25 x)dx 1 x x 2 / 8
x x
(check, Qcumulative @ 4 = -1+4-16/8 = 1 OK)
2 2
R( x) (1 Qcumulative( x))dx = 0.5 (2 x x2 8)dx
4
Prob(fail) =
all x 3
= 0.02083 (=1/48)
So our 2nd guess was correct!
Note that the area of overlap was 1/8, so obviously the area of overlap is not
the failure probability.
L.3 – Problems.
1. For these load and strength distributions, calculate the exact
probability of failure;
E6003 – Ship Structures II 154
© C.G.Daley 2014
Topic 24 : Engineering Safety Metrics
Introduction
To implement the concepts of probability into engineering design, without
making the process too complex, we need to extract simple metrics from the
statistical data. In this lecture we will define the Margin of Safety, and
contrast this with the Factor of Safety. Next we will define what we mean by
Characteristic Values and show how the use of these is related to safety.
Margin of Safety
The margin is the reserve (in units of load) between the load and the
strength:
Margin = Strength - Load
M=R-Q
It is easy to see that failure will occur if the margin is negative.
M<0 = Failure
Contrast this with Factor of Safety FOS, which is the % reserve:
E6003 – Ship Structures II 155
© C.G.Daley 2014
FOS = Strength/Load = R/Q
FOS<1 = Failure
The Margin of Safety is more useful when trying to determine the
probability of failure. The reason has to do with one of the properties of the
Normal distribution. When variables with Normal distributions are added (or
subtracted), the result is also normal. For example:
C = A + B, with A, B Normal
then C is Normal with:
mean(C) = mean(A) + mean(B)
(stdDev(C))2 = (stdDev(A))2+(stdDev(B))2
For Strength R and load Q, Normally distributed, with :
Mean strength = R Mean load = Q
std. dev of strength = R std. dev of load = Q
M=R-Q
M will also be Normally distributed with:
M = R - Q
M R Q
2 2
The probability of failure is the probability that M<0. As M is Normal, it is
easy to find the Prob(M<0).
We want to find the probability that M<0. We know M and M
We can define as:
E6003 – Ship Structures II 156
© C.G.Daley 2014
= M/M
This term gamma , is the number of standard deviations that the mean of
the margin is above zero. This sounds awkward, but it is quite simple and
useful. The standard Normal table lets us find the probability of a value
being below n standard deviations above the mean. We can use this to find
the faiulre probability as follows:
Probability of Failure = 1- NC()
where NC is the standard Normal Cumulative function (see Table)
e.g. if = 1.5, Prob. Failure = 1-0.933193 = 0.066807 = 6.7%
There is always a unique relationship between g and the probability of
failure. This is not true for the Factor of Safety (why?).
We call the Safety Index.
Note : the term ‘coefficient of variation’ or COV for a random
variable A is defined as: COV(A) = A/A
We can see that the safety index = 1/COV(Margin)
E6003 – Ship Structures II 157
© C.G.Daley 2014
Characteristic Values
We often describe certain things with values that are actually not averages,
but special or rare values. Examples include, say, water depth in a lake, or
heights of a mountain range. We might describe a lake as being ‘deep’
because there are spots with over 100ft depth. The actual average depth may
be only 20 feet, with the single deepest spot may be 120 ft. Yet we might say
the lake is 100feet deep. Similarly we might say that the Rocky Mountains
are over 10,000 ft, when the actual average elevation is 3000ft (within the
overall territory called the Rocky Mountains). What we are doing focusing on
a typical rare (high) value, but not the most rare (highest) value.
This approach is very common in engineering. When we say that steel has a
yield strength of 300MPa, we don’t mean that the average strength is
300MPa, we mean that no more than 5% of the tested values fall below
300MPa. This is because if more than 5% are below, the steel would fail to be
rated at 300MPa. There are cases where none of the samples would be below
300, possibly with a minimum test at 320MPa, but yet the steel is rated as
300MPa.
We define values such as the 5% value as Characteristic Values. We might be
more specific and call this the 5% characteristic value. In naval architecture
we use the term ‘significant wave height’ which we might call Hs. Hs, is
approximately equal to the average of the highest one-third of the waves.
This is a ‘characteristic value’, at about the 10% level (est.).
Note: Hs is calculated precisely using:
Hs = 4.0 * sqrt(m0)
where m0 is the variance of the wave displacement time series
acquired during the wave acquisition period.
E6003 – Ship Structures II 158
© C.G.Daley 2014
The “Design Point’
Typically we define load and strength with characteristic values. For strength
we use a rare low value, and for load a rare high value. In the sketch below
we have used the 5% value of each. Further we have defined a design point at
which the two characteristic values are equal.
You might wonder how we can set the 5% strength to equal the 5% load.
Usually the load is defined by nature, though it may be set as a design
decision (e.g. allow 500 lb/ft2 of equipment to be placed on the floor of a room).
Strength is an outcome of the design process, so that once we know the 5%
load (the design point) we can adjust the size (and material) of the structure to
give a strength such that the 5% characteristic value is at the design point.
Example: for the long plate sketched below, the value pc is calculated using
average values of t and L and the 5% value for y . Consequently, Pc is the 5%
characteristic value of strength (i.e. R5%)
E6003 – Ship Structures II 159
© C.G.Daley 2014
When we use characteristic values, the probability of failure can be found. For
example, in the case where Q5% = R5% we express the characteristic values as
the mean plus 1.645 standard deviations (see the N(0,1) table):
Q5% Q 1.645 Q
R5% R 1.645 R
1.645 Q R R Q M
which lets us write:
We need to find M but to do so we need to combine Q and R .
Q 2 R This is used to give;
Typically we have more variation in Q than in R, so lets assume that
M 1.645 3 R
and
M 4 R 2 R 2 5 R
consequently;
M 1.645 3 R
2.21
M 5 R
from which we get;
Probability of failure = 1-N(0,1)(2.21) = 0.0136 = 1.36%
The number would be a bit different if we had chosen a different ratio for
Q/R. In a ‘rule’ development exercise, the rule developers can look at the
typical ratios for Q/R and specify the appropriate characteristic values to
achieve the desired target probability of failure. Use of characteristic values
allows for the consideration of variability in a way that is easy to formulate as
a rule, and will look similar to current rule formats. They are one aspect of a
formal Limit States Design approach.
Exercise: Find Prob(Failure) for R2%, Q1% with Q/R = 3.
E6003 – Ship Structures II 160
© C.G.Daley 2014
Topic 25 : Partial Safety Factors
or – dealing with the unexpected!
Introduction
If all the randomness in the world could be measured and quantified, we
wouldn’t need anything other than a good model of the distribution of Q&R
(load & strength). Unfortunately, a lot of uncertainty can not be readily
measured. Some of it hasn’t even been thought of. As a result of this, we
define two broad categories of uncertainty. One is called statistical
uncertainty. This is the type of uncertainty that we can measure and
quantify, including things like steel strength, wave heights and the size of
welding distortions.
The other type of uncertainty is called approximational uncertainty. This is
the type of uncertainty that we haven’t measured. Quite often we can’t
practically measure this type. One example is called model uncertainty. This
is the uncertainty in our theoretical models (i.e. what if our understanding of
the world is wrong?). Another type of approximation uncertainty covers
unmeasured variables. There is an almost endless list of factors that might
lead to an accident, but haven’t been formally studied and measured (i.e what
% of people are able to execute a successful iceberg collision avoidance
maneuver, on the first actual attempt?). We have a name for all the things we
E6003 – Ship Structures II 161
© C.G.Daley 2014
don’t know, and we do have a standard method of dealing with this
approximational uncertainty. We use partial safety factors.
Why do we call these “partial” safety factors. The reason is that we have
several of them, to cover different concerns, and the partial factors combine to
S1 ,S2 ,Q ,R
give a total factor of safety. Lets call these partial safety factors:
T
and the total factor of safety:
TS1 S2 Q R
where:
Once we have T we use it to increase the strength so that Rc (the
characteristic value of R) is above the design point. Without partial safety
factors, we would set Rc=Qc at the design point. With partial safety factors
Rc = T Qc
we set:
Now for important point #1: These partial safety factors do help our situation,
and do help safety, but we can not say by how much. In part we add them to
account for things that we can’t quantify, so we can only modify our calculated
probability of failure (as we found when we set Rc=Qc), for some of the partial
safety factors.
Typical Partial Safety Factors
E6003 – Ship Structures II 162
© C.G.Daley 2014
Different codes use different specific partial safety factors. The four specified
below are typical of the kinds of issues covered. Each structural component
may have a different set of partial safety factors. The four factors are:
S1 – this factor reflects the degree of seriousness of the
items failure in terms of safety (of people or
environment, e.g. hull rupture, fire.)
S2 – this factor reflects the degree of seriousness of the
items failure in terms of serviceability (operability,
usefulness, e.g. crane damage, loss of processing
equipment)
Q – this factor reflects our confidence in the loads (e.g.
weights are relatively certain, wave loads are less
certain, ice loads are not certain.)
R – this factor reflects our confidence in the strength
(easy to predict plastic hinge, harder to predict
bucking, very had to predict structural aging effects)
Examples for component failures:
S1=1.0
Hull plate plastic set
S2=1.02
safety: no safety concerns
service: minor speed loss
S1=1.4
Oil Valve Rupture
S2=1.02
safety: possible major fire
service: delay to clean mess
E6003 – Ship Structures II 163
© C.G.Daley 2014
In codes that use partial safety factors, various components and structures
would be classed as being more or less important, or more or less uncertain.
The example below is from a Danish paper describing the partial safety
factors in the Danish Structural Codes.
The partial safety factor m is determined by m =0 where 0 takes into account
the consequences of failure (safety class: low: 0.9, normal: 1.0 and high: 1.1), 1 takes into
account the type of failure (ductile with reserve: 0.9, without reserve: 1.0 and brittle: 1.1), 2
the material parameter (uncertainty), 3 takes into account the uncertainty in the
takes into account the possibility of unfavorable differences from the characteristic value of
computational model (good: 0.95, normal: 1.0 and bad: 1.1), 4 takes into account the
basis of the controlled material parameter (large: 0.95, average: 1.0 and small: 1.1) and 5
uncertainty in connection with determination of the material parameter in the structure on the
takes into account the amount of control (extended: 0.95, normal: 1.0 and reduced: 1.1).
from : “Calibration of Partial Safety Factors and Target Reliability Level in Danish Structural
Codes” by J. D. SØRENSEN, S. O. HANSEN, T. A. NIELSEN
Summary
Partial safety factors provide a way to raise the structural strength for those
cases in which a rational person would have increased concerns (i.e. cases
where the designer is more worried about either the design parameters, or the
consequences of failure).
Pros and Cons
There are many good reasons for taking a partial safety factor approach to
structural design. The most obvious benefit comes when we compare two
different designs built to the same code. Partial safety factors let us express
our reasonable concerns about uncertainty and consequences. We will,
hopefully, achieve a consistent level of safety, for both critical structures and
for relatively unimportant components.
Nevertheless, there are some serious flaws with the idea of partial safety
factors. Designers should be sensitive to the limitations of the approach, as
they are always ultimately responsible for the design.
An obvious flaw with partials safety factors is the conceit that we can quantify
what we don’t know. This suggests that knowledge is incremental, such that
our current knowledge is a large % of the truth, with some small % of
uncertainty. Unfortunately, the truth may be a long way from our current
E6003 – Ship Structures II 164
© C.G.Daley 2014
understandings. The pretense that we can use statistics to quantify and, in
effect, sanitize this type of ‘approximational uncertainty’ is quite irrational.
Only by studying the issue and removing our ignorance can we properly deal
this approximational uncertainty. When we don’t know something we should
just state our assumptions and admit the potential for error.
Another problem has to do with the way probability models commonly work.
Most are essentially linear, in that they assume that different effects are
independent and add. The Normal distribution was derived on the basis of
addition of small independent errors. Reality can often be quite different. We
still have little idea of how large systems behave. It is certain that most
systems are non-linear and are full of interacting (non-independent)
components. We use the models we have, because we have nothing better, not
because they are correct.
The most obvious problem with application of a partial safety factor approach
is the assumption that more (specified) strength will result in greater safety.
Most major failures involve some sort of gross error, often some human error.
Gross errors, such as not following the approved welding standard, or leaving
out a crucial structural member, or driving a vessel onto the rocks, can never
be accounted for with small factors. Safety is a multi-dimensional thing. The
people involved at all stages must be trained, alert and cautious. The design
must have layers of independent backup systems. If the designers are using a
code that claims to be capable of producing an extremely safe design (e.g. 10-6
failure per year per structure), there may be a tendency for the designers to
become overly confident that all will be well. Such target levels of safety are
common in probability-based codes, and are overly optimistic. This is
particularly true with offshore structures. The operational experience with
many designs is quite limited, and the actual failure rate for offshore
structures in general is very high.
The following is taken from the abstract of a Norwegian paper, calling for caution in
the area of offshore structures:
The purpose of this paper is to review the worldwide historical structural failure
data in the 1990s on offshore structures, and compare this with the present risk analyses
of Norwegian offshore structures.
The paper describes an overview of registered accidents to offshore structures
based on the databases WOAD and CODAM. The accident data is given for fixed
platforms, jack-ups and for floating platforms. Estimates of risk level in annual
frequencies and PLL values are given for each platform type.
The paper concludes that:
The risk connected to marine operations and structures give a significant
contribution to the total risk.
The historical risk to marine operations and structures is significant higher than
the results from risk analyses.
E6003 – Ship Structures II 165
© C.G.Daley 2014
Neither component nor system based reliability analyses of structures give
adequate descriptions of the real risk connected to structures.
Human errors are probably the dominating cause of accidents connected to
structural failure.
from “On the Risk of Structural failure on Norwegian Offshore Installations” By A.
Kvitrud, G. Ersdal and R.L. Leonhardsen, ISOPE 2001
E7002 – Ship Structures II 166
© C.G.Daley
Topic 26 : Risk Analysis - trying to think of everything!
Introduction
Until now we have talked about the relatively simple question of whether
load exceeds strength. Accidents most often occur as a complex sequence of
events. Further, there are many possible unwanted events.
We can say that risk is the combination of probability and consequences.
More accurately we could say that risk is the combination of probability,
consequences and context. The addition of context will be discussed after we
discuss the probability/consequences.
We would like to use risk analysis to make decisions. We are trying to make
intelligent forecasts, so that we can make the best decisions. It may seem
odd, but we must start with a good definition of the word ‘best’. In a complex
and multi-dimensional world, how do we compare apples and oranges? We
need a way to compare outcomes, so that we can say which one is ‘best’.
E7002 – Ship Structures II 167
© C.G.Daley
Utility and Expected Value
The usual way to compare outcomes is to place a value, either in terms of
‘utility’ or money, on all possible outcomes. Utility is an economist’s term
meaning anything of value, whether actual or perceived, such that at least
some human being will want it. It includes things like ‘satisfaction’ and
‘reputation’, as well as things with a market value.
Once we know the value of each outcome, we can calculate the ‘expected
value’ of the utility, as follows;
Given a choice A, we determine that there is a Pr(B) chance of outcome B,
with a utility of UB . There is also a Pr(C) chance of outcome C, with a utility
of UC.
The expected value of utility of choice A is therefore;
EV(UA) = Pr(B) x UB + Pr(C) x UC
The actual utility if we take choice A will be UB or UC, depending on whether
B and or C actually occur. It is important to understand the idea of expected
value. To further illustrate, let’s take an example from gambling.
Case 1: you and 5 friends each place $1 on the table. Each or you are sitting
in a numbered seat (1 to 6). Someone rolls a single die. The person sitting at
the seat with the # on the die gets the $6. If the odds are fair, the expected
value of the payout for any individual is 1/6x$6 = $1. But each person paid
$1, so the expected net winnings is 1/6x$6 - $100 = 0$.
Universal truth #1: your expected net value when gambling with honest
friends is $0. All the money stays with the group! That does not mean that
after a long night of gambling you will go home with all the money you came
with. See the plots below.
These plots show the cumulative winnings (or loses) that might occur for one
of the friends.
E7002 – Ship Structures II 168
© C.G.Daley
100
80
60
40
20
net $
0
-20
-40
-60
-80
-100
0 1000 2000 3000 4000 5000 6000 7000
# of bets
things might average out, or…
100
-100
net $
-200
-300
-400
-500
0 1000 2000 3000 4000 5000 6000 7000
# of bets
they might not!
Case 2: you place a $1 bet in a typical game in Las Vagas. You may win $6 if
you win, but the odds of winning are 1/7. The expected value of the winnings
is 1/7 x $6 = $0.86. But your expected net winnings is 1/7x$6 - $1 = -$0.14
A 2nd universal truth: your expected net value when gambling in Las Vegas is
negative. If you gamble long enough, you will surely lose your money! Below,
the plot shows the outcome of a series of 1$ bets, where each bet had a 1/7
chance of winning $6. The gambler was ahead for a while, but after about
1000 bets, all was downhill. One would expect that after 6500 bets, our
gambler would be at a net value of 6500 x –$0.14 = -$910, quite close to what
happened!
200
-200
-400
net $
-600
-800
-1000
-1200
0 1000 2000 3000 4000 5000 6000 7000
# of bets
E7002 – Ship Structures II 169
© C.G.Daley
A word about Context
When calculating expected value, we are implicitly assuming that ‘a buck is a
buck’, meaning that winning a dollar would produce a positive effect that
would just balance the negative effect of loosing a dollar. If such were true,
we would be happy to risk loosing one year the same as we might win the
next. This is only true for people betting small amounts (small relative to
their wealth). For most of us, the loss of $10,000 hurts far more than the
positive effect of winning $10,000. The loss of $10,000 threatens our well
being by far more than $10,000 would add to our well being. This winning
and loosing is not symmetrical.
E.g. We will not be happy betting $1Million even if we would win $2.2 Million
on a coin toss. (expected value is: EV=.5 x 2.2 – 1 = + $0.1M). We’d be foolish
not to take such odds if the bet was $1.00, because we could afford to loose.
We could not afford to loose the $1M.
The size of the outcome is part of the context that we need to consider. There
are many other issues where two things may seem to be equal, but are not, or
are asymmetrical. The death of an employee should have the same
consequences regardless of context, but this is not the case. Context is very
important in determining the utility of various outcomes.
Conditional Probability
Before we formulate a risk model we need one more concept. We need a way
to combine various events. We need the idea of ‘conditional probability’. The
condition is the circumstance for which we want the probability. The
probability of winning the season is not the same as the probability of
winning the finals. We need to know the probability of getting through the
semi-finals and the probability of winning the final game. we say:
Prob(A) = Prob(B) x Prob(A|B)
read Prob(A|B) as probability of A given B.
E7002 – Ship Structures II 170
© C.G.Daley
Events can have one or several precursors:
in general:
Prob(A) = Prob(i) x Prob(A|i)
all i
We need this for a variety of probability models. In a “State” based
probability model, can (may) move from any state to any other state.
Now back to risk models. All risk models aim at calculating the expected
value of utility for the choices and random events considered. Lets look at
two examples. In the first example we want to estimate the chance of
structural damage to a ship traveling along a certain route. The event tree is
:
E7002 – Ship Structures II 171
© C.G.Daley
To get a collision we either do not detect the berg in front, or we detect but fail
to avoid. Then we recognize that some of the collisions will cause damage. The
numbers (conditional probabilities) given here are taken from thin air!
Next we will examine a slightly different type of model. This is called a
decision network. In it there are decision nodes (rather than all random
branches). We examine both sides of the decision branch (or branches), and
pick the branch with the lowest cost (greatest utility).
In this case we are trying to compare a GBS with an FPSO for use on the
grand Banks. To do this calculation we need both the conditional probabilities
and the costs of the various outcomes.
Again, these numbers are hypothetical.
In general, risk analyses only focus on a specific set of risks. One will normally
go through a process of hazard identification (HAZID) to find the set of risky
events that you want to model. As mentioned in Lect. 5, this set is often
smaller than the actual set of risks that the design is subject to. The attached
extract from a paper on nuclear waste re-cycling gives a description of the
formulation and use of a risk model.
E7002 – Ship Structures II 172
© C.G.Daley
You might also like
- The Principles of Naval Architecture Series - Strength of Ships and Ocean Structures (PDFDrive)No ratings yetThe Principles of Naval Architecture Series - Strength of Ships and Ocean Structures (PDFDrive)251 pages
- Panel M-16 Propulsio - Practices and Proced - jan.2007.T-RNo ratings yetPanel M-16 Propulsio - Practices and Proced - jan.2007.T-R63 pages
- Plasblak Masterbatches: Product Selection GuideNo ratings yetPlasblak Masterbatches: Product Selection Guide8 pages
- In-Service Performance of Aluminum Structural DetailsNo ratings yetIn-Service Performance of Aluminum Structural Details114 pages
- Manoeuvring Prediction Based On CFD Generated Derivatives: April 2016No ratings yetManoeuvring Prediction Based On CFD Generated Derivatives: April 201610 pages
- Dynamic Analysis of Multiple Body Floating Platforms Coupled With Mooring Lines and ReisersNo ratings yetDynamic Analysis of Multiple Body Floating Platforms Coupled With Mooring Lines and Reisers223 pages
- Comparative Structural Requirements For High Speed CraftNo ratings yetComparative Structural Requirements For High Speed Craft111 pages
- Seam 6: Ship Construction, Trim and Stability in A Cargo VesselNo ratings yetSeam 6: Ship Construction, Trim and Stability in A Cargo Vessel10 pages
- Early Stage Integrated Parametric Ship Design PDFNo ratings yetEarly Stage Integrated Parametric Ship Design PDF14 pages
- Naval Architect & Ship Construction: Non Destructive TestNo ratings yetNaval Architect & Ship Construction: Non Destructive Test109 pages
- Numerical Simulation of Ship Motion Due To Waves and Manoeuvring50% (2)Numerical Simulation of Ship Motion Due To Waves and Manoeuvring48 pages
- Aluminum Welding On Ship Structure ProblemsNo ratings yetAluminum Welding On Ship Structure Problems14 pages
- Design, Manufacturing, and Testing of WT DoorsNo ratings yetDesign, Manufacturing, and Testing of WT Doors11 pages
- Estimation of Wave Making Resistance by Wave Analysis Ship Wave at Sea100% (1)Estimation of Wave Making Resistance by Wave Analysis Ship Wave at Sea11 pages
- Strength of Ships and Ocean Structures 183No ratings yetStrength of Ships and Ocean Structures 18319 pages
- Design of Double Hull Tankers-GL Presentation May 2005100% (1)Design of Double Hull Tankers-GL Presentation May 200542 pages
- Tips For Building Global Ship Structure Finite Element (FE) Model in ANSYS MAPDL - IIINo ratings yetTips For Building Global Ship Structure Finite Element (FE) Model in ANSYS MAPDL - III3 pages
- Sustainable Marine Structures - Volume 05 - Issue 01 - March 2023No ratings yetSustainable Marine Structures - Volume 05 - Issue 01 - March 202352 pages
- Westlawn Lecture 1 - The Design Spiral For Computer-Aided Boat DesignNo ratings yetWestlawn Lecture 1 - The Design Spiral For Computer-Aided Boat Design45 pages
- Hydrodynamic Response of A Floating Dry DockNo ratings yetHydrodynamic Response of A Floating Dry Dock17 pages
- Role of Naval Architects in Ship RecyclingNo ratings yetRole of Naval Architects in Ship Recycling8 pages
- SSC-400 Weld Detail Fatigue Life Improvement Techniques: Ship Structure CommitteeNo ratings yetSSC-400 Weld Detail Fatigue Life Improvement Techniques: Ship Structure Committee143 pages
- SSC-436 The Effect of Fabricaton Tolerances On Fatigue Life of Welded JointsNo ratings yetSSC-436 The Effect of Fabricaton Tolerances On Fatigue Life of Welded Joints100 pages
- Ship Structural Integrity and Performance SystemNo ratings yetShip Structural Integrity and Performance System107 pages
- Hughes, Owen F. Paik, Jeom Kee Ship Structural Analysis and Design 2010100% (2)Hughes, Owen F. Paik, Jeom Kee Ship Structural Analysis and Design 201042 pages
- Naval Innovation for the 21st Century: The Office of Naval Research Since the End of the Cold WarFrom EverandNaval Innovation for the 21st Century: The Office of Naval Research Since the End of the Cold WarNo ratings yet
- Damage Mechanics in Metal Forming: Advanced Modeling and Numerical SimulationFrom EverandDamage Mechanics in Metal Forming: Advanced Modeling and Numerical Simulation4/5 (1)
- Private Ones Or: A Paper For Presentation at The January 21., and Marine EngineersNo ratings yetPrivate Ones Or: A Paper For Presentation at The January 21., and Marine Engineers49 pages
- Tensile Properties of Reinforced Thermosetting Plastics Using Straight-Sided SpecimensNo ratings yetTensile Properties of Reinforced Thermosetting Plastics Using Straight-Sided Specimens8 pages
- Determine The Moisture and Volatile Content in The Green Sand100% (2)Determine The Moisture and Volatile Content in The Green Sand3 pages
- Particle-Size Reduction - Chemical Engineering - Page 1No ratings yetParticle-Size Reduction - Chemical Engineering - Page 17 pages
- SHANSEP Approach For Slope Stability Assessments of River Dikes in The NetherlandsNo ratings yetSHANSEP Approach For Slope Stability Assessments of River Dikes in The Netherlands9 pages
- A Versatile Window Function Fro Linear Ion Drift Model-A New ApproachNo ratings yetA Versatile Window Function Fro Linear Ion Drift Model-A New Approach10 pages
- Introduction To Fundamentals of Modeling For Metals ProcessingNo ratings yetIntroduction To Fundamentals of Modeling For Metals Processing4 pages
- UNIT-5: Dielectrics AND Magnetic MaterialsNo ratings yetUNIT-5: Dielectrics AND Magnetic Materials139 pages
- Datasheet BlueSolar Monocrystalline Panels ENNo ratings yetDatasheet BlueSolar Monocrystalline Panels EN1 page
- A Review On Seismic Assessment of Post-Tensioned Flat SlabNo ratings yetA Review On Seismic Assessment of Post-Tensioned Flat Slab5 pages
- 12th Chemistry Half Yearly Exam 2024 Answer Key Dindigul District English Medium PDF Download 1No ratings yet12th Chemistry Half Yearly Exam 2024 Answer Key Dindigul District English Medium PDF Download 112 pages
- The Principles of Naval Architecture Series - Strength of Ships and Ocean Structures (PDFDrive)The Principles of Naval Architecture Series - Strength of Ships and Ocean Structures (PDFDrive)
- Panel M-16 Propulsio - Practices and Proced - jan.2007.T-RPanel M-16 Propulsio - Practices and Proced - jan.2007.T-R
- In-Service Performance of Aluminum Structural DetailsIn-Service Performance of Aluminum Structural Details
- Manoeuvring Prediction Based On CFD Generated Derivatives: April 2016Manoeuvring Prediction Based On CFD Generated Derivatives: April 2016
- Dynamic Analysis of Multiple Body Floating Platforms Coupled With Mooring Lines and ReisersDynamic Analysis of Multiple Body Floating Platforms Coupled With Mooring Lines and Reisers
- Comparative Structural Requirements For High Speed CraftComparative Structural Requirements For High Speed Craft
- Seam 6: Ship Construction, Trim and Stability in A Cargo VesselSeam 6: Ship Construction, Trim and Stability in A Cargo Vessel
- Naval Architect & Ship Construction: Non Destructive TestNaval Architect & Ship Construction: Non Destructive Test
- Numerical Simulation of Ship Motion Due To Waves and ManoeuvringNumerical Simulation of Ship Motion Due To Waves and Manoeuvring
- Estimation of Wave Making Resistance by Wave Analysis Ship Wave at SeaEstimation of Wave Making Resistance by Wave Analysis Ship Wave at Sea
- Design of Double Hull Tankers-GL Presentation May 2005Design of Double Hull Tankers-GL Presentation May 2005
- Tips For Building Global Ship Structure Finite Element (FE) Model in ANSYS MAPDL - IIITips For Building Global Ship Structure Finite Element (FE) Model in ANSYS MAPDL - III
- Sustainable Marine Structures - Volume 05 - Issue 01 - March 2023Sustainable Marine Structures - Volume 05 - Issue 01 - March 2023
- Westlawn Lecture 1 - The Design Spiral For Computer-Aided Boat DesignWestlawn Lecture 1 - The Design Spiral For Computer-Aided Boat Design
- SSC-400 Weld Detail Fatigue Life Improvement Techniques: Ship Structure CommitteeSSC-400 Weld Detail Fatigue Life Improvement Techniques: Ship Structure Committee
- SSC-436 The Effect of Fabricaton Tolerances On Fatigue Life of Welded JointsSSC-436 The Effect of Fabricaton Tolerances On Fatigue Life of Welded Joints
- Hughes, Owen F. Paik, Jeom Kee Ship Structural Analysis and Design 2010Hughes, Owen F. Paik, Jeom Kee Ship Structural Analysis and Design 2010
- Composite material Complete Self-Assessment GuideFrom EverandComposite material Complete Self-Assessment Guide
- Naval Innovation for the 21st Century: The Office of Naval Research Since the End of the Cold WarFrom EverandNaval Innovation for the 21st Century: The Office of Naval Research Since the End of the Cold War
- Damage Mechanics in Metal Forming: Advanced Modeling and Numerical SimulationFrom EverandDamage Mechanics in Metal Forming: Advanced Modeling and Numerical Simulation
- Private Ones Or: A Paper For Presentation at The January 21., and Marine EngineersPrivate Ones Or: A Paper For Presentation at The January 21., and Marine Engineers
- Tensile Properties of Reinforced Thermosetting Plastics Using Straight-Sided SpecimensTensile Properties of Reinforced Thermosetting Plastics Using Straight-Sided Specimens
- Determine The Moisture and Volatile Content in The Green SandDetermine The Moisture and Volatile Content in The Green Sand
- Particle-Size Reduction - Chemical Engineering - Page 1Particle-Size Reduction - Chemical Engineering - Page 1
- SHANSEP Approach For Slope Stability Assessments of River Dikes in The NetherlandsSHANSEP Approach For Slope Stability Assessments of River Dikes in The Netherlands
- A Versatile Window Function Fro Linear Ion Drift Model-A New ApproachA Versatile Window Function Fro Linear Ion Drift Model-A New Approach
- Introduction To Fundamentals of Modeling For Metals ProcessingIntroduction To Fundamentals of Modeling For Metals Processing
- A Review On Seismic Assessment of Post-Tensioned Flat SlabA Review On Seismic Assessment of Post-Tensioned Flat Slab
- 12th Chemistry Half Yearly Exam 2024 Answer Key Dindigul District English Medium PDF Download 112th Chemistry Half Yearly Exam 2024 Answer Key Dindigul District English Medium PDF Download 1