Professional Documents
Culture Documents
Framework Theory Manual - Wind Fatigue
Framework Theory Manual - Wind Fatigue
Framework
Wind Fatigue Design
Framework
This publication or parts thereof may not be reproduced or transmitted in any form or by any means,
including copying or recording, without the prior written consent of DNV GL AS.
FRAMEWORK SESAM
Program version 2.8 i
TABLE OF CONTENTS
Page
1.0 PURPOSE 1- 1
2.0 SCOPE 2- 1
3.0 REFERENCES 3- 1
4.0 OVERVIEW AND ASSUMPTIONS 4- 1
4.1 Overview 4- 1
4.2 Theoretical Assumptions 4- 5
4.3 Scope of Analysis 4- 5
5.0 THE SPECTRAL APPROACH 5- 1
6.0 THE NATURE OF THE WIND 6- 1
7.0 WIND SPECTRA 7- 1
8.0 A SPECTRAL FORCING FUNCTION 8- 1
8.1 Wind Force on a Member 8- 1
8.2 The Wind Force on a Degree Of Freedom 8- 3
8.3 Spectral Relationships 8- 4
9.0 VORTEX SHEDDING INDUCED VIBRATIONS 9- 1
9.1 Determination of Mode Shape and Fundamental Frequency of a Brace 9- 1
9.2 Brace Oscillation Amplitude Calculation 9- 4
9.3 Member End Damage Calculation 9- 6
9.4 Member Centre Damage Calculation 9- 7
10.0 CONDENSATION OF FORCES 10- 1
11.0 STRESS CONCENTRATION FACTORS 11- 1
11.1 An Overview 11- 1
11.2 Components of the HSS 11- 1
11.3 The SCF Schemes 11- 3
11.4 Non-standard Joints 11- 4
11.5 Efthymiou Scheme 11- 4
12.0 THE HOT SPOT STRESS TRANSFER FUNCTION FROM POINT FORCING 12- 1
12.1 The Dynamic Equation 12- 1
12.2 Results from the Eigenvalue Problem 12- 1
12.3 The Structure Displacement Vector from a Point Force 12- 2
12.4 Hot Spot Stresses from a Point Force 12- 4
13.0 HOT SPOT STRESS POWER SPECTRUM 13- 1
13.1 The Hot Spot Stress Power Spectra 13- 1
13.2 Integrating the Hot Spot Stress Power Spectra 13- 3
14.0 CALCULATION OF FATIGUE LIFE 14- 1
14.1 Assumptions 14- 1
14.2 Damage Evaluation 14- 1
FRAMEWORK SESAM
ii
1.0 PURPOSE
The purpose of this manual is to describe the theory underlying the Framework wind fatigue
application.
Wind fatigue of Framework is according to the wind fatigue program Gusto (Reference 3.9),
which has been implemented into Framework for the purpose of making wind fatigue
application available in SESAM.
This manual is a reprint of the theory manual of Gusto (Reference 3.10), except for matters
and references related to external programs, which are not relevant for Framework
applications.
FRAMEWORK SESAM
Program version 2.8 2-1
2.0 SCOPE
This manual describes the calculation methods used in Framework for the determination of
fatigue damage to frame structures subjected to wind loading. Buffeting loads due to gusting
are treated by the power spectral density approach. The damage is a function of the overall
structural response. The effects of vortex shedding induced fatigue due to steady winds are
treated by separate evaluation of individual member responses. Framework calculates the two
effects on the assumption that they are uncoupled. The resultant damages are summed to give
overall fatigue lives of joints and members.
FRAMEWORK SESAM
Program version 2.8 3-1
3.0 REFERENCES
3.15 A Criterion for Assessing Wind Induced Crossflow Vortex Vibrations In Wind Sensitive
Structures.
Robinson R. W. and Hamilton J. Health and Safety Executive Offshore Technology
Report OTH 92 379, 1992.
3.17 SESAM, Wajac, Wave and Current Loads on Fixed Rigid Frame Structures, User
Manual, May 2001.
3.18 NPD, Regulations and provisions for the petroleum activities, June 1997
FRAMEWORK SESAM
Program version 2.8 4-1
4.1 Overview
Framework is an application that evaluates wind buffeting and vortex shedding induced
fatigue damage to a structure. It receives input information relating to the eigenmodes of a
structure, coupled with statistical data on the annual wind distribution and associated drag
factors.
The annual wind data are characterized by a set of wind states, considered to represent the
climate for the year. For each of these wind states, the response stress power spectra at a local
“hot spot” within a particular joint are evaluated.
For buffeting calculations the hot spot power spectrum response is divided into the quasi-
static response and the dynamic response.
The dynamic response consists of the excited resonant modes. It is partitioned into the
separate resonant modal responses; for each of these an independent damage assessment is
made. This assumes that each response is narrow band and independent of the others, but
sometimes several modes, very close in frequency, are taken as one.
Vortex shedding from brace members may induce oscillations in individual braces. These are
local modes rather than overall structural modes.
The quasi-static part of the power spectra covers the low frequency non-resonant response.
The wind spectrum has a broad peak at low frequencies but is treated as narrow band at its
peak frequency with one third of the stress variance of the low frequency broad band peak
stress spectra. The resultant damage is then multiplied by 10. This approach assumes that the
quasi-static contribution to damage is small, so that a rigorous evaluation is not required.
For each of these dynamic and static partitions the narrow band assumption implies a
Rayleigh distribution for the “hot spot” stress range versus number of cycles. The variance is
given by the integral under the power spectrum. Fatigue damage may then be evaluated by
application of the Palmgren-Miner relationship and reference to a recognised S-N curve (see
Section 14.0).
WIND
HOT SPOT
SPECTRUM
7 STRESS
LOADING
TRANSFER
FUNCTION 10, 11, 12
GEOMETRY
WIND
FORCE CROSS
SPECTRA
8
HOT SPOT
STRESS
SPECTRUM
13
Figure 4.1 Generation of Hot Spot Stress Spectrum with Cross Reference to Principal
Sections of this Manual.
This implies:
• the greatest hot spot stresses within a modal response cycle occur at maximum
modal amplitude
• Guyan reduction may be used to reduce the mass and stiffness matrices without
significant loss of accuracy to the low modes, see Section 8.0. (Note! The
Framework wind fattigue assumes all non-fixed translational degrees of freedoms
(dof) as master dof.)
e) wind gust velocities in the mean wind direction and normal to the mean wind both
horizontally and vertically are statistically independent.
f) member drag coefficients are invariant under the fluctuating wind component and are
appropriate to the mean wind speed.
g) vortex shedding induced member oscillations and fatigue are uncoupled from any
buffeting induced vibrations and damage.
The most appropriate technique for determining wind-induced cyclic stresses is referred to as
the frequency-domain or power-spectral density approach.
A power spectrum describes a time-dependent variable relating the energy distribution over a
range of frequencies. All phase information is averaged out. Analysis methods whereby
output spectra are obtained from input spectra via transfer functions are required for random
processes such as wind or wave loading, where only a statistical description of the
environmental forces can be given. In the spectral analysis method of fatigue due to wind,
the stress spectrum is obtained from the input wind spectrum via the structure stress transfer
function. Because of the nature of the fluctuating wind force, there is, to good accuracy, a
direct linear relationship between the wind speed and force spectra allowing structure stress
spectra to be linearly related to wind speed spectra.
FRAMEWORK SESAM
Program version 2.8 6-1
The energy content of wind as a function of frequency is demonstrated in the form of a power
spectrum averaged over a year, as shown in Figure 6.1, taken from Reference 3.1. Two
dominant peaks occur in this figure. The lower frequency peak is associated with cycles with
a period of the order 2-3 days and is due to the passage of large-scale atmospheric
depressions. The higher frequency peak has significant energy in the range 10 minutes to 1
second. It is this part of the spectrum, which is of interest to structural designers and is
commonly known as the gust spectrum. There is apparently a lack of wind energy with
periods between 10 minutes and two hours. It can clearly be assumed that fatigue, which
depends on changes in wind speed, is not significantly affected by cycles of length greater
than two hours.
For the purpose of fatigue analysis, the wind speed is averaged over a suitable period and the
wind is then represented in that time as having a constant mean value and direction, upon
which fluctuations or gusts are superimposed. Although a period of about ten minutes would
appear to be desirable as an averaging period in order to reflect the influence of the short-
term storms, a period of one hour has traditionally been used. It is for this time that data are
usually available, see, for example, Reference 3.2.
While a speed and direction represent the mean wind in any given hour, the gust components
are statistically described by three parameters: probability distribution, power spectrum and
cross-correlation function.
The probability distribution describes the ratio or percentage of time a certain wind speed is
likely to occur, the power spectra reflect the energy content of the wind as a function of
frequency, and the cross-correlation function indicates the way in which the gusts are
spatially correlated.
The probability distribution can be obtained from measurements; but standard formulae exist
for describing this as a function of mean wind speed, (References 3.3 and 3.11).
FRAMEWORK SESAM
6-2 Program version 2.8
FIGURE 6.1
Framework requires the definition of a set of hourly mean wind speed vs. height profiles,
their bearings and for what fraction of a year they exist. For each of these, three
parameterized gust spectra are calculated, and a resultant damage assessment made. The total
annual damage may be obtained by adding these damage assessments in proportion to the
fraction of a year in which they are generated.
Obtaining a set of such hourly mean profiles may in itself be a major undertaking when the
measured data are not totally adequate. Engineering judgement may be required in assessing
what approximations are valid. To gain the necessary knowledge base from which to do this,
the reader's attention is drawn to References 3.3 and 3.4.
In general, over a period of one year, wind measurements can be taken to show the number of
hours per year the hourly mean wind is blowing for each speed and direction. These
measurements are taken at 10m above ground or sea level. By applying a power law to
represent the variation of mean wind speed with height relationship based on the drag at the
earth's surface, (Reference 3.3), the requirements are met.
Typically the wind data are such that the easiest set of hourly mean wind speeds to find are over
the eight major compass points and the twelve/fourteen divisions of the Beaufort scale. A
hypothetical table for the percentage of the year occupied by each such wind state is shown in
Table 7.1.
DIRECTION
BEAUFORT
N NE E SE S SW W NW TOTAL
SCALE
1 0.24 0.13 0.14 0.14 0.18 0.16 0.19 0.23 1.44
2 0.77 0.64 0.61 0.46 0.77 0.59 0.74 0.68 5.27
3 1.52 1.42 1.27 1.30 2.08 1.82 1.84 1.71 12.97
4 3.14 2.59 1.67 2.22 4.77 3.64 4.01 2.94 24.98
5 2.57 1.95 0.48 1.71 4.26 3.12 2.74 2.06 18.89
6 2.90 1.47 0.39 1.84 4.72 2.85 2.41 1.80 18.37
7 1.58 0.73 0.14 1.42 2.84 1.52 1.36 1.23 10.82
8 0.67 0.20 0.04 0.75 1.57 0.62 0.82 0.58 5.26
9 0.12 0.02 0.00 0.27 0.41 0.16 0.18 0.13 1.28
10 0.02 0.01 0.00 0.13 0.16 0.02 0.06 0.04 0.44
11 0.00 0.00 0.00 0.02 0.05 0.00 0.00 0.02 0.09
12 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
100.0
TABLE 7.1 PROBABILITY DISTRIBUTION OF MEAN WIND SPEED EXPRESSED AS % OF YEAR
FOR WHICH IT OCCURS
FRAMEWORK SESAM
7-2 Program version 2.8
In such a case the bearing of the normal to one of the structure faces does not usually
correspond to one of the compass points. For convenience, the data may be linearly
interpolated to provide wind directions meeting this criterion.
Towers, which exhibit symmetry, may allow opposing wind directions to be treated as one.
Table 7.2 shows data for the 8 directions of Table 7.1 expressed as four composite directions.
The annual mean wind distribution is therefore fully characterized in Table 7.2 by 48 wind
states, that is 12 in each of four directions.
TABLE 7.2
For each wind state, the wind speed at a height of 10m above the earth's surface is used to
help compute a single-sided gust spectrum S s w( f ) , where f is in cycles/sec. There are five
available spectra for wind fatigue:
The first three are spectra representing the gusts in the same direction as the mean wind, the
fourth is for lateral (horizontal) gusts across the mean wind and the fifth is for vertical gusts
across the mean wind. Their formulae are presented in Appendix 7.0.
Figure 7.1 shows a plot of the HARRIS spectra, associated with wind cases 1-11 in Tables
7.1 and 7.2.
• at typical tower modal frequencies (~ 1Hz) gust energy increases greatly with mean wind
speed. Typically damage per unit time is proportional to the mean velocity raised to a
power between 8 and 12
• the eddy spectra is linearly dependent on the drag at the earth's surface
s
From these single-sided spectra S vv ( f ) in cycles/sec, the double-sided spectra in rads/sec are
then
1 s
S vv (w) = S vv ( f ) f = ω / 2π
4π
FRAMEWORK SESAM
7-4 Program version 2.8
FIGURE 7.1
GUST SPECTRA FOR WIND STATES 1-11 IN TABLES 7.1 AND 7.2 (SI UNITS)
FRAMEWORK SESAM
Program version 2.8 8-1
In this section, an approximation to the cross-power spectral density function of the buffeting
wind loads is presented in terms of the power spectra for the fluctuating wind. This may then be
used in the derivation of hot spot stress power spectra. The linear relationship between the forcing
spectra and hot spot stress spectra forms the backbone of the spectral approach to wind fatigue
analysis.
In Sections 8.1 and 8.2 a linear relationship between the fluctuating wind components and the
fluctuating forcing at a given node is established. This involves first evaluating the forcing on
the members, and then defining a relationship between member forces and nodal forces.
Because the linear relationship exists and the wind spectra applied in the three directions may
be regarded as statistically independent the cross-power spectral density function of the
forcing from all three spectra is the linear sum of that from the individual spectra.
In Section 8.3 the cross-power spectrum of forcing is found from the wind cross-power
spectral density function of the wind. The wind-cross power spectral density function is then
expressed in terms of the power spectrum of this wind, multiplied by a simple function.
For detailed descriptions of the mathematical definitions of the various spectral functions
above, see Appendix 2.0.
The general form for the wind force on a member is given by:
F = 12 ρ Cd DL U n U n
Where ρ , Cd , L, D and U n are the air density, member drag coefficient, member length,
member diameter and vector normal velocity respectively. All structural members are
assumed to be tubular with drag coefficients of 1.2 in the sub-critical regime and varying as
per Figure 8.1 in the critical and super-critical regimes. Although Reynolds' number varies
with wind speed, the assumption is made that the drag coefficient is not time dependent.
When the above form is expanded out and the fluctuating terms considered small compared
to the time mean terms, the wind force may be expressed as
U ( z ,t ) 2
2 b c U ( z ,t ) .U 1
F (t ) = 2 ρ Cd DLa A ⋅
1
0 + 1 ρ Cd La A 0 1 0 U ( z ,t ) .V 1
2 0 0 1 U ( z ,t ) .W 1
0
FRAMEWORK SESAM
8-2 Program version 2.8
FIGURE 8.1
where A is a transformation matrix, and a, b and c are constants depending on the mean
wind direction and member orientation. The wind velocity vector over a period of time has
U (z ,t ) U1
been split into a time mean, 0 , plus fluctuations V 1 .
0 1
W
Re ≤ 2.5x105 C d = 1 .2
2.5x105 < Re ≤ 3.85x105 Cd = 10(10.8599 −1.997199 log10 (Re))
3.85x105 < Re Cd = 10( 0.2557 log10 (Re) −1.7109)
By summing these member forces and distributing them to the degree of freedom (dof), the
fluctuating force at dof r from wind fluctuations may then be expressed in the form
g r (t ) = E UU V ( zr ,t )
2
r Equation 8.2.2
where EUUr , Er
VV
and EWWr are constant at each dof. They depend upon ρ and Cd and the
values of D, L, A , a, b, and c for each member connected to that degree of freedom. z r is the
height of the dof r.
FRAMEWORK SESAM
8-4 Program version 2.8
The cross-power spectral density function is the Fourier transform of the cross correlation-
function. The cross-power spectral density function S gg (r ,s : ω ) of the forcing between dof r
and dof s is therefore
1 ∞ − izω
S gg (r ,s : ω ) = ∫ e g r (t ) g s (t + z ) dτ Equation 8.3.1
2π − ∞
g r (t ) g s (t + z ) = g r′ (t ) g ′s (t + z ) + g r (t ) g ′(t ) g s (t + z ) + g r (t ) g s (t + z )
As g ′r (t ) = g ′s (t + z ) = 0 this becomes
g r (t ) g s (t + z ) = g ′r (t ) g ′s (t + z ) + g r (t ) g s (t )
1 ∞ − jzω 1 ∞
S gg (r ,s : ω ) = ∫e g ′ (t ) g ′ (t + τ ) dτ + g (t ) g (t ) ∫e
− jτω
dτ
2π − ∞ r s
2π r s
−∞
1 ∞
S gg (r ,s : ω ) = S g ′g ′ (r ,s : ω ) + g r (t ) g s (t ) ∫ e − jτω dτ
2π −∞
δ (ω ) = ∫ e− jτωdτ
δ (ω ) = 1 ω = 0
=0 ω ≠0
g r (t ) g s (t ) ∞ δ (ω )
∫ exp (− jτω ) dτ = g r (t ) g s (t )
2π −∞ 2π
FRAMEWORK SESAM
Program version 2.8 8-5
δ (ω )
S gg (r ,s : ω ) = S g ′g ′ (r ,s : ω ) + g r (t ) g s (t ) Equation 8.3.2
2π
The first and second terms in the above, may be further expanded in terms of wind data. The
first term S g ′g ′ (r ,s : ω ) is defined as
1 ∞
S g ′g ′ (r ,s : ω ) = ∫ exp (− jτω ) g ′r (t ) g ′s (t + τ ) dτ
2π − ∞
UU UU U 1 (t ), U 1 (t + τ )
Er Es r s
1 ∞ UV UV 1
S g ′g ′ (r ,s : ω ) = 4 V (τ ,r ,t ) V (τ ,r ,t ) ∫ exp (− jτω ) + E r E s V r (t ), V s (t + τ )
1
dτ
2π − ∞ UW UW 1
+ E r E s W r (t ), W s (t + τ )
1
r E s S u ′u ′ (r , s : ω )
E UU UU
= 4 V ( zr , t ) V ( zs, t ) EUVr E s S v ′v ′ (r , s : ω )
UV
EUW EUW S (r , s : ω )
r s w ′w ′
where S u ′u ′(r , s : ω ) etc. are the cross-power spectral densities of the wind in the directions
along, laterally across and vertically across the mean wind respectively
The variation of the mean wind with height, may be parameterized by the following power
law identity (Reference 3.3)
α
z
V (z, t ) = V (10, t ) Equation 8.3.3
10
S UU (r ,s : ω ) = coh(r ,s : ω ) S UU (ω )
FRAMEWORK SESAM
8-6 Program version 2.8
Giving
α α
ErUU ESUU Su 'u ' (ω )
2 z z
S g ′g ′ (r ,s : ω ) = 4 V (10, t ) s r coh (r , s : ω ) + EUV
r E S S v 'v ' (ω )
UV
10 10 + UW UW S (ω )
Er ES w' w'
The forms of the coherence function used are presented in Appendix 1.0. The cross-power
spectral density function of the fluctuating forcing is then given by
α α
2 z z
S g ′g ′ (r , s : ω ) = 4 V (10, t ) s r .
10 10
EUUr E s S u 'u ' (ω )
UU Equation 8.3.4
UV UV
coh (r ,s : ω ) + E r E s S v 'v ' (ω )
UW UW S (ω )
+ E r E s w' w '
Where all the terms on the RHS may be readily evaluated from the structure geometry and
the basic wind data presented in Section 7.0.
S gg (r ,s : ω ) = S g ′g ′ (r , s : ω ) ω ≠0
(t )
= gr g s (t ) ω =0
2π
Now substituting the values of g r (t ) and V ( z ,t ) from the Equations 8.2.2 and 8.3.3.
α
z
g r (t ) = E r V (10, t ) r
10
giving
S gg (r , s : ω ) = S g ′g ′ (r , s : ω ) ω ≠0
α α
EVV EVV 2 z z Equation 8.3.5
= r s
V (10, t ) r s ω =0
4π 2
10 10
FRAMEWORK SESAM
Program version 2.8 9-1
In this section the procedures followed to determine the amplitudes of oscillation excited by
vortex shedding in steady winds are described. The determination of the first natural
frequency and its associated mode shape are described, followed by the estimation of the
amplitude of oscillation in a given steady wind. The stress levels at the member ends and
centre then follow from application of the stress concentration factors to the raw member
behaviour.
It is assumed that the vortex shedding effects are only of any significance if they induce
oscillations in the first mode of the brace. This is a reasonable assumption for tubular
structural steel members that are used in typical flare towers. This assumption would not be
valid if applications were to assess the vibration amplitude and stresses associated with long
slender tie rods or guy ropes, where a higher mode may be excited.
The mode and frequency are highly dependent on the conditions of member end fixity. In
general these are not known to any great degree of accuracy, so Framework allows the user to
investigate ranges of fixity. Low end fixity reduces the natural frequency and the member end
damage that occurs. In the extreme a pin-ended member suffers no end damage because the
pure bending deformation induced by vortex shedding produces no end moments, stresses or
damage. High end fixity produces a higher natural frequency and associated with it the
possibility of higher end moments. The amplitude of excitation may, however, be much
smaller because there may be no resonance between the frequency of shedding of vortices
and the natural frequency of the member. For a pin-ended member the damage at the member
centre will exceed that of the member ends, for fixed end members the damage needs to be
checked at the member centre and at both ends.
The brace may be considered as a beam element with end supports. The ends are assumed to
be restrained against lateral translation. The rotational supports may be different at each end
and are allowed to vary between pin-ended (i.e. no rotational resistance) and fully fixed (i.e.
fully restrained against rotation). The basic theory for the solution of this class of problem is
given in Timoshenko and Young (Reference 3.12).
The fundamental equation for the dynamic bending behaviour of a thin beam (i.e. one in
which shear deformations are negligible) is given by
∂ 4w ∂ 2w
EI = − m
∂ x4 ∂ t2
where E is the Young's modulus of the material; I is the second moment of area of the beam,
w is the transverse deflexion of the beam, m is its mass per unit length, x is the co-ordinate
along the beam's neutral axis and t is time.
FRAMEWORK SESAM
9-2 Program version 2.8
where A, B, C and D are constants that will be determined from the boundary conditions
applied at the beam ends. ω is the natural frequency of the response (in radians per second),
φ is the phase lag and k is defined by the relationship
mω 2
k= 4
EI
For a brace member in a flare tower the displacement boundary conditions applied at both
ends are that the deflexions, w, are zero. For a beam of length L this gives two equations
A+C = 0
A cos kL + B sin kL + C cosh kL + D sinh kL = 0
− A+C = 0
− A cos kL − B sin kL + C cosh kL + D sinh kL = 0
For fixed ends the corresponding boundary conditions, which are that the slopes are zero at
either beam end, leads to
B+D = 0
− A sin kL + B cos kL + C sinh kL + D cosh kL = 0
For more general support conditions, where there are dissimilar rotational springs at either
end of the beam, the following relationships apply
K 0 {B + D} = EIk{− A + C}
K L {− A sin kL + B cos kL + C sinh kL + D cosh kL} =
EIk{− A cos kL − B sin kL + C cosh kL + D sinh kL}
These equations are derived from the ratios of the end moments to end rotations, which are,
by definition equal to the spring stiffnesses for linear elastic spring supports. The rotational
spring stiffnesses at x = 0 and x = L are given by K0 and KL respectively. The relationships for
simply supported or fully fixed ends may be derived as special cases of the last pair of
equations, with zero or infinite spring stiffnesses substituted as appropriate.
From the set of four equations, i.e. two displacements and two rotational boundary conditions
selected as appropriate, it is possible to solve for any three of the unknowns in terms of the
FRAMEWORK SESAM
Program version 2.8 9-3
fourth. This solution gives the mode shape, but not its amplitude, and the corresponding
frequency. Strictly speaking the process will give an infinite number of solutions for the
pairs of mode shapes and frequencies, as there is an infinite number of choices of kL that will
satisfy the relationships. As noted above, the fundamental frequency, corresponding to the
first mode is the only one likely to be of any significance for structural applications.
The solution of this set of four equations is carried out iteratively in Framework. The starting
approximation to the value of kL is made using a simplified assumption, which is close to the
value corresponding to the fundamental mode. This ensures that convergence is to the lowest
frequency and not to a spurious higher mode.
The iterative scheme is based upon the Newton-Raphson approach. The determinant of the
matrix of the four simultaneous equations is evaluated together with the determinant of the
derivative of the matrix with respect to the variable kL. The objective is to produce a zero
value for the matrix. If FN is an approximation to the numerical value of determinant of the
four homogeneous equations then
∂ FN
FN +1 = FN {1 − }
∂ (kL)
is likely to be a better approximation. FN and its derivative are evaluated at the current (trial)
value of kL. This process is repeated until the values of FN and FN+1 differ by less than a
suitably small relative value. Currently this is set to 10-6 in Framework. The iteration limit is
set to 100.
As an example consider the set of four equations for a simply supported beam. In this case
the solution may be derived by inspection. The conditions are that
A=0
B sin kL = 0
C=0
D sinh kL = 0
The only solution for kL that gives deflections that are other than zero at all times is
kL = nπ
D=0
Substitution of integer values of n in the above relation and referring to the equation for k
above gives the relationship for the natural frequency corresponding to mode n. This equation
is valid for all positive integer values of n.
Similar procedures may be followed for other boundary condition sets. The resulting
relationships for kL are more complex and generally involve the use of iterative solution
FRAMEWORK SESAM
9-4 Program version 2.8
techniques. Reference. 3.12 gives details of many more examples, including free ended
beams and cantilevers.
The frequency at which vortices are shed from opposite sides of the brace member is
dependent on the Reynolds' number of the fluid flow. If the vortex shedding frequency is
sufficiently removed from the natural frequency of transverse oscillations of the brace there
will not be any resonance and the amplitude of any oscillations will be negligible. If the ratio
of the two frequencies is close to unity, the amplitude of the oscillations will be significant,
that is high stress levels and hence structural fatigue will be caused. The critical velocity of
the flow is defined as that which will cause resonant vortex shedding. Wind velocities in the
range of 60 to 140 per cent of the critical velocity will excite oscillations that cause damage.
Velocities outside this range do not cause appreciable damage and their effects may be
ignored.
Depending on the member end fixities the damage may be higher at the brace ends or at the
member centre.
For each wind attack direction the wind is resolved into components normal and tangential to
each brace. The velocity used is that computed at the member centroid, it being assumed that
any variation of velocity, with height, along the member length will be relatively
unimportant. The normal component of the velocity is used to calculate the Reynolds' number
in conjunction with the outer diameter of the brace. Note that the value used is the member
total diameter, which will include any non-structural cladding or insulation material.
From the Reynolds' number the vortex shedding frequency may be estimated.
This frequency is then compared with the natural frequency of the member itself, again
taking into account the effects of any non-structural mass due to cladding.
Vcrit = ω 0 Dinc / St
where ω0 is the natural frequency of the brace member, Dinc is the diameter of the member
including any coating material and St is the Strouhal number.
For each brace member the wind velocities that occur throughout the year are resolved into
normal components. This is done by decomposing the statistical data on wind speeds,
directions and the proportion of the year that such winds occur, into discrete ranges at
constant speeds. The effect of each wind range and its associated velocity is then considered
in isolation. The total damage induced by each wind speed range from each direction is then
summed to give the total structural damage. Note that the effect of wind from opposing
directions will be identical so use of the composite wind data, as described in Section 7.0,
will reduce the volume of data required.
FRAMEWORK SESAM
Program version 2.8 9-5
If the normal component of a selected range of wind speeds lies in the range
then the contribution to the member end and centre damage is accumulated. If the normal
velocity is outside this range there is no damage contribution. For each wind speed range the
following process is used. Reference 3.15 contains background information.
K s = 4πε me /( ρ Dinc
2
)
where ε is the structural damping ratio, me is the effective mass per unit length of the brace
and ρ is the density of air.
where Dext is the external diameter of the structural member, i.e. excluding any insulation or
cladding material.
SG = 2π ( St ) 2 K S
The structural lift coefficient, CL , is dependent on the Reynolds' number of the flow.
4
Re 2 / 1012 )
Re ≤ 2.0x104 CL = 0.42 − 0.33e( −2.0 x10
6
/ Rtran )10
2.0 x104 < Re ≤ 4.71x105 / Rtran CL = 0.09 + 0.33e −3200(Re x10
Re > 4.71x105 / Rtran CL = 0.15
L 2
∫0 w dx
CVDP = L 4
∫0 w dx
where w(x) is the mode shape determined from the frequency equation described above.
FRAMEWORK SESAM
9-6 Program version 2.8
The amplitude of the response at the resonant vortex shedding frequency is then given by
At wind velocities away from the critical the non-resonant amplitude for a broad band
response is given by the expression
5
2
/ ( 0.02 + 0.1 ( 0.4 − 0.32 e −25000 turb ) ) )
AB = e ( − (1−VR ) Ar
VR = V / Vcrit
For narrow-band response the amplitude is dependent on the value of the stability factor, KS.
2
K s < 4π AN = e − 30 (1−VR ) Ar
2
( K S / 4π )1.8
4π ≤ K S ≤ 12π or 0.725 < VR < 1.0 AN = e − 30 (1−VR ) Ar
2 1.8
12π < K S AN = e − 30 (1−VR ) 3.0
Ar
Note that in all cases the amplitude of the response is given as a factor of the resonant
amplitude. The factor defines a bell shaped curve between the lower and upper limits of the
velocity ratio, with maximum amplitude at the resonant frequency. A broad band response
gives a flatter bell shaped curve than the pronounced peak of the narrow-band response
curve.
The member end damage calculation closely mirrors that used in the buffeting damage
calculations. From the calculated forcing frequency, as given by the vortex shedding
characteristics of the brace, and the time per year that the wind blows, the total number of
oscillations of the brace may be determined. The amplitude of the vibrations is determined
using the approach outlined in Section 9.2 above. From the displacement amplitude and the
mode shape the beam section properties are used to calculate the member stresses at the two
ends.
The raw member stresses are then factored by the stress concentration factors (SCFs) to give
the local hot spot stresses. The evaluation of the SCFs is described in Section 11.0. Note that
the stress range, which is twice the stress amplitude, is needed for fatigue damage
calculations. The damage is evaluated using the Miner's law approach in an analogous
manner to the buffeting damage.
FRAMEWORK SESAM
Program version 2.8 9-7
The member centre damage is calculated in a similar manner to the member end damage. The
SCF for the member centre is applied as a blanket value to the entire structure. This value is
supplied from the input data and there are no calculations involved to derive the value. This
user specified SCF should represent the typical value that would be associated with a single-
sided girth closure weld. It will depend on the quality control of the welding process, the out
of roundness and the mismatch that are permissible in the fabricated tubular structure.
The approach used is conservative. The damage is evaluated at the section on the brace's
length that has the maximum curvature, and hence bending moment. The member's displaced
shape is examined at 100 equally spaced positions along its length to determine the greatest
curvature. It is unlikely, although possible, that the position of maximum moment would
coincide with a closure weld.
FRAMEWORK SESAM
Program version 2.8 10-1
The transfer function linking the cross-power spectral density function of fluctuating forcing
to the power spectral density function of the response depends upon knowledge of the
structure's modal frequencies and shapes.
The extraction of a full set of eigenvalues and mode shapes for a structure with a large
number of degrees of freedom is computationally expensive. Consequently, the technique
known as static condensation, or Guyan reduction, is used (Reference 3.5). This reduces the
analysis to a more manageable size by retaining a set of master degrees of freedom, which are
chosen to characterize the structure's kinetic energy. The degrees of freedom are reduced out
are known as the slave freedoms. The algebra of the reduction to the master degrees of
freedom of the forcing vector is given in Appendix 3.0
The Framework wind fatigue keeps all free translational degrres of freedom as master
freedoms and the user can not select the master freedoms.
FRAMEWORK SESAM
Program version 2.8 11-1
11.1 An Overview
Stress concentrations occur in the welded tubular joints. For fatigue calculations it is
therefore important to take them into account. To evaluate the stress concentrations or “hot
spot stresses” (HSSs), empirically derived stress concentration factors (SCFs) based on joint
geometry are used.
Using these the distribution of the HSSs around the welds is found, under loading produced
by a mean wind state (static loading). When the tower is subject to buffeting, i.e., dynamic
eddy loading, the maximum HSS at a joint, for each mode of response, is assumed to occur in
the same place as for the static loading. This is a reasonable assumption, in that buffeting
fatigue effects on flare towers are normally dominated by the cantilever modes of response.
These strongly resemble the tower's static response.
At each joint, connected members may be either braces or chords. The chord is taken as the
pair of elements of greatest diameter which are collinear. (If there is more than one pair of
collinear elements of the same maximum diameter, the chord is assumed to be the pair with
the greatest thickness.) All other members are assumed to be braces. HSSs are found for each
brace/chord intersection separately, for both chordside and braceside of the weld.
If collinear elements are not found, the joint classification of Framework is tried. A chord
may be identified in Framework without the presence of collinear elements at the joint.
For each brace the axial, in-plane bending (IPB) and out-of-plane bending (OPB) stresses
( σ AX , σ IPB and σ OPB respectively) are all deemed to contribute to both the chordside and
braceside HSS distribution around the weld. Chordside and braceside IPB and OPB
components are approximated by:
where SCF IPB and SCF OPB are the IPB (crown) and OPB (saddle) SCFs respectively and φ
is the angle around the weld, measured from the saddle point (see Figure 11.1). The axial
components are approximated by:
where SCFAS and SCFAC are the axial saddle and crown SCFs respectively.
FRAMEWORK SESAM
11-2 Program version 2.8
FIGURE 11.1
FRAMEWORK SESAM
Program version 2.8 11-3
The wind fatigue module of Framework has three schemes by which the HSSs may be
evaluated, the “I”, “E” and "O" schemes. They use different SCF formulae, and combine the
component HSSs in slightly different manners. Details of the equations used to derive the
SCFs may be found in Appendix 1.
• T joints
• K joints
• KT joints
• X joints
• or non-standard joints (see Section 11.4)
The "I" and "O" schemes will not recognise an X-joint, which is classified as a non-standard
joint for these schemes. Under the "E" scheme, X-joints may be included.
• treats a KT joint as a K joint plus a T joint. The K joint makes up the outer
braces, while the T joint makes up the middle brace
(φ ) = (φ ) +
[ B HSS OPB (φ )] (φ )
B HSS B HSS IPB + B HSS A
(≡ 0)
(φ ) = (φ ) +
[ C HSS OPB (φ )] + (φ ) + σ *CHORD
C HSS C HSS IPB C HSS A
(≡ 0)
where the new B subscript denotes braceside of the weld and the C subscript denotes
chordside. σ *CHORD is the chord stress, with sign. C HSS (φ ) is the locally enhanced stress.
• evaluates the HSS distribution around the weld, both chordside and braceside as
• the gap between braces is either zero or negative (Figure 11.1 shows a positive
gap). This is called an OVERLAPPING joint.
For both of these joint types each brace will be treated as a T joint, and then take the
maximum HSS over all joints.
The Efthymiou equations which allows X-joints as well as K, KT, and T are made available
in Framework. This is the "E" scheme. Details are given in Appendix 4.0. Note that there
are some differences in the implementation of the Efthymiou equations and treatment of valid
ranges of geometric parameters of the equations in Framework basic and the wind fatigue
module. In particular overlapping braces of K joints are not handled by the wind fatigue
application implementation.
FRAMEWORK SESAM
Program version 2.8 12-1
12.0 THE HOT SPOT STRESS TRANSFER FUNCTION FROM POINT FORCING
A finite element model of a structure is characterized by its mass and stiffness matrices M
and K . The structural damping may be modelled as an imaginary component of the stiffness
matrix. The forced response of a structure to a force vector G( ω ) at frequency ω, where ω is
in radians/sec is given by
When Guyan reduction is employed to give reduced mass and stiffness matrices M M and
K M with an equivalent force vector G M applied at the master degrees of freedom to produce
a displacement vector U M at these degrees of freedom; and the exp( jwt ) term is dropped
then
−1 ω2
M M KM − I U M (ω ) = 0 Equation 12.2.1
(1 + j γ )
λ S = ω 2S / (1 + jγ )
Let φ$ be a matrix of eigenvectors φ S . This set of vectors is not unique because any linear
multiple of an eigenvector is also an eigenvector. If there are coincident eigenvalues the
situation is more complicated. Any linear combination of the eigenvectors corresponding to
the coincident modes is also an eigenvector. To establish a unique matrix φ$ of vectors φ S , φ$
is constrained to follow the relationship
T
φˆ M M φˆ = I Equation 12.2.2a
FRAMEWORK SESAM
12-2 Program version 2.8
Equation12.2.1 holds for each eigensolution, and may therefore be written as a set of
simultaneous eigensolutions.
K φˆ − diag (λ ) M φˆ = 0
M S M
S
T
Premultiplying by φ$ gives
φˆ K M φˆ − diag (λ S )φˆ M M φˆ = 0
T T
S 1424
3
Term1
φˆ K Mφˆ = diag (λ S )
T
()
−1
Left multiplication by φˆ
T −1
followed by right multiplication by φˆ gives
−1
K M = φˆ diag (λs )φˆ
T −1
Equation 12.2.3a
S
−1
M M = φˆ (I )φˆ
T −1
Equation 12.2.3b
T −1 −1
φˆ diag ( (1 + jλ ) λ S − ω )φˆ U M = G M
2
S
T 1
Successively left multiplying by φˆ , diag and φ$ gives
2
S (1 + j γ ) λ S − ω
1 ˆT
U M = φˆ diag φ G M
2
S (1 + j γ ) λ S − ω
1
Q = φˆ diag
2
S (1 + j γ ) λ S − ω
S = φˆ G M (ω )
T
[
Q rs = φˆ rs / (1 + jγ )λ S − ω 2 ]
S s = ∑ φˆ KS G M K
K
where the r and s subscripts refer to the rth row and sth column of their respective matrices.
Hence Equation 12.3.1 may be expressed as
U M = QS
φˆ rs ˆ
U M r = ∑ Q rs S S = ∑ 2 ∑ φ KS G M K
S S (1 + jγ )λ S − ω K
φˆ rsφˆ KS
U M r = ∑ ∑ 2 GMk
S (1 + jγ )λ S − ω
K
This gives the displacement at the rth degree of freedom as a summation of the displacements
from a force at each degree of freedom k. Hence if G M (k ) is the displacement at master
freedom r from an equivalent force G M at master freedom s then
k
φˆ KSφˆ rsG M k
U M r (k ) = ∑
S (1 + jγ )λ S − ω 2
FRAMEWORK SESAM
12-4 Program version 2.8
which may be extended to include the displacement vector for all master freedoms U M (k )
resulting from an equivalent force G M at master freedom k
k
[
∑ φˆ KSφˆ1S / (1 + jγ ) λ S − ω 2
S
]
S KS 2 S
[
∑ φˆ φˆ / (1 + jγ ) λ − ω 2
S
]
.
U Mk = GMk
.
.
ˆ ˆ
[
∑ φ KSφ MS / (1 + jγ ) λ S − ω
2
]
S
this may then be re-expressed in terms of the normalized eigenvectors φ$ S
φˆ KS
UM = ∑ ⋅ φˆ Equation 12.3.2
S (1 + jγ )λ S − ω 2 SG M k
φˆ KS
N σ (U M (k ) ) = ∑ ( )
σ ˆ
2 N φ S GMk
Equation 12.4.1a
S (1 + jγ ) λ S − ω
The stresses given by the above relationship are the stresses derived from the eigenvector
displacements factored by the appropriate stimulated amplitudes. To find the HSS at a joint
requires application of the appropriate SCFs to these stresses. This then allows HSS pk (ω ) ,
the hot spot stress at joint p from point forcing at degree of freedom k, to be written in terms
of a function and the value of the point forcing.
Section 12.0 established a transfer function between a forcing point load and the hot spot
stress at a selected joint. In this section the resultant hot spot stress power spectrum at a joint
from the ensemble of forcing point loads will be described.
To derive the power spectra the hot spot stress given by Equation 12.4.1 needs to be re-
expressed in the time domain. This is achieved by Fourier transform inversion using the
convolution theorem:
(
hss pk (t ) = ∫ h pk (t1) g k t − t dt1 )
∞
M 1
Equation 13.1.1
−∞
By summing this equation over all the forcings at each master dof, the total hot spot stress
∑ hss pk (t ) at any joint p from the ensemble of equivalent point forces in the time domain
k
may be found. The auto-correlation function of the joint's hot spot stress is then given by
From Equation 13.1.1, and by taking the summations outside the integrals
and noting that the time averaging operator ⋅ only applies to functions of time
∫ ∫ h pr (ε ) h ps (ε ) Mg r(t − ε ) Mg (t + t − ε 1 )
∞ ∞
C hss p hss p (τ ) = ∑ ∑ 1
s
dεdε 1
r s −∞ −∞
1 ∞ − jwτ
S hss p hss p (ω ) = ∫e C hss p hss p (τ ) dτ
2π − ∞
may be expressed as
FRAMEWORK SESAM
13-2 Program version 2.8
∑ ∑ ∫ ∫ h pr (ε ) h ps (ε )
1 ∞ − jwτ
(t + τ − ε )
∞ ∞
S hss p hss p (ω ) = (t − ε ) Mg dε dε 1 dτ
1 M 1
∫e gr
2π − ∞ r s −∞ −∞
s
S hss p hss p (ω ) = ∑ ∑ ∫ ∫ h pr (ε ) h ps (ε )
1 1
(t + τ − ε )
∞ ∞ ∞
∫e
− jωτ M
(t − ε ) Mg 1
dτ dε dε 1
r s −∞ −∞ 2π
−∞
gr s
dτ 1 = dτ
(
e − jωτ = e − jω τ
1+ 1− ε
ε ) = e − jωτ 1 e − jωε 1 e jωε
t +τ + ε1 = t +τ − ε
so that
1 1
( ) (t − ε ) Mg r (t + τ 1 − ε ) dτ 1 dε dε 1
∞ ∞ ∞
S hss p hss p (ω ) = ∑ ∑ ∫ ∫ h pr (ε ) h ps ε − jωτ 1 − jωε 1 jωε M
∫ e e e
2π − ∞
gr
r s −∞ −∞
S hss p hss p (ω ) = ΣΣ ∫ e
∞
−∞
jωε
∞
h pr (ε ) ∫ e
−∞
− jωε
1 ∞ ε − jωτ 1
h ps (ε ) ∫
1
−∞ 2π
M
gr (t − ε )Mg s
(t +τ 1 −ε ) dτ dεdε
1 1
For a stationary random process the time average with respect to t − ε should be the same as a
time average with respect to t, so that
M
gr (t − ε ) Mg s
(t +τ 1
−ε ) = M
gr (t ) Mg s
(t +τ )
1
and by definition
MM
S gg (r , s : ω ) = 1 ∞ − jωτ 1
∫ e
2π − ∞
M
gr (t ) Mg s
(t +τ )
1
dτ 1 Equation 13.1.2
r s −∞
∞ ∞
1 1
( )
S hss p hss p (ω ) = ∑ ∑ ∫ e jωε h (prε ) ∫ e − jωε h ps ε S Mg Mg (r , s : ω ) dεdε
1
−∞
FRAMEWORK SESAM
Program version 2.8 13-3
( )
∞ ∞
S hss p hss p (ω ) = ∑ ∑ S g (r , s : ω ) ∫ e jωε h pr ( ε ) dε − jωε 1
h ps ε dε
MM 1 1
g ∫e
r s −∞ −∞
H ps (ω ) =
1 ∞ − jωε 1
∫e
2π − ∞
( )
h ps ε dε
1 1
1 ∞ jωε
H *pr (ω ) = ∫ e h pr ( ε ) dε
2π − ∞
where H *pr (ω ) is the complex conjugate of H pr (ω ) . This second identity may be easily
verified by expressing the exponent as cosine plus imaginary sine terms. Hence Equation
13.1.2 may be re-written
S (hss
ω)
p hss p
= 4π 2 ∑ ∑ H ps (ω ) H *pr (ω ) S Mg Mg (r , s : ω ) Equation 13.1.3
r s
The hot spot stress power spectrum shown in equation 13.1.3 is frequency dependent in the
cross power spectra of the forcing and in the functions H pr (ω ), H *pr (ω ) .
At frequencies far below the modal frequencies the structure behaves in a quasi-static manner
and H pr (ω ) may be approximated by H pr (0) . If corrections are made for the variation in
drag coefficient with wind speed, the integral of the hot spot stress power spectrum may then
be evaluated by simple numerical integration.
For frequencies near the modal frequency, the variation of the forcing cross power spectra
with frequency is small compared to the variation in H pr (ω ) . Hence near a modal frequency
1
λ q2 the integral may be expressed as
FRAMEWORK SESAM
13-4 Program version 2.8
1 1
λ q2 + ∆2 λ q2 + ∆2
S hss p hss p (ω ) dω = 4π ∑ ∑ S gg (r , s : ω ) ∫ H ps (ω ) H *pr (ω )dω
2
∫ Equation 13.2.1
1 r s 1
λ q2 − ∆1 λ q2 − ∆1
Neglecting the affects of the SCFs, and substituting Equation 12.4.1a we get
1
λq2 + ∆ 2
1
λq2 + ∆ 2σ
φˆsm ⋅ p φˆ n σ
φˆrn ⋅ p φˆm ( ) ( )
dω
∫ H ps (ω ) H *pr (ω ) dω = ∫ ∑ ∑ ⋅
m n (1 + jγ ) λ m − ω (1 + jγ ) λ n − ω
1 1
2 2
λq2 − ∆1 λq − ∆1
2
1 1
( ) ( )
λq2 + ∆ 2 λq2 + ∆ 2
dω
H ps (ω ) H *pr (ω ) dω = ∑ ∑ φˆsm ⋅ p φˆm φˆr ⋅ p φˆn
σ σ
1
∫
λq2 − ∆1
m n 1
λq2 − ∆1
∫
[(1 + jγ )λ m − ω 2] [(1 − jγ )λ n − ω 2]
Equation 13.2.2
Now the integral within the summation on the RHS may be re-expressed as
1 1
λ q2 + ∆ λ q2 + ∆
dω dω
= ∫
1
∫
λ q2 − ∆
[(1 + jγ )λ m − ω ] [(1 − jγ )λ n − ω ]
2 2 1
λ q2 − ∆
(C m2 − ω 2 )(d 2n − ω 2)
where
1 1
( ) ( )
1 2
1 2
1+ 1+γ 2 2 2 −1
λ + j 1+ γ
2 1 1
Cm = 2 λ2
2 m 2 m
1 1
( ) ( )
12 1 2
1+ 1 + γ 2 2
λ 12 − j 1 + γ
2 2 −1
λ 12
dn =
2 n 2 n
1
λq2 + ∆ 2
ω − Cm ω − dn
( )
1
φˆsm ⋅ p φ m φˆrn ⋅ p (φ n )
log log
ω + d n
λq2 + ∆ 2 ˆσ σ e e
ω + Cm
∫ H ps (ω ) H pr (ω ) dω = ∑ ∑ −
*
1 m n C m − dn
2 2 C m dn
λq2 − ∆1 1
λq2 −∆1
back-substitution into Equation 13.2.1 with SCFs added then gives the integral of the hot spot
stress power spectrum.
FRAMEWORK SESAM
Program version 2.8 14-1
14.1 Assumptions
• The integral under these peaks (or frequency bands) is the variance of the stress
amplitude at the frequency associated with the peaks
• The stress amplitude within each frequency band has a Rayleigh distribution.
This is true for narrow band processes (Reference 3.8). The sub-modal section is
split into three portions, each of which is treated as having a Rayleigh distribution
• For each frequency band fatigue is directly related to the number of cycles
experienced in each stress range through the Palmgren-Miner relationship (Reference
3.6)
• The number of cycles to failure at any stress range (amplitude) may be found
from standard SN curves (Reference 3.7)
The Palmgren-Miner relationship (Reference 3.6), for the damage sustained due to stress
cycles, may be written in terms of the stress amplitude
D ( A) = n c ( A) / N c ( A)
The damage evaluated over all stress ranges is then obtained by integrating over all the
possible stress amplitudes, i.e.
∞
D = ∫ nc
( A) dA
− ∞ N c ( A)
FRAMEWORK SESAM
14-2 Program version 2.8
With an SN curve of the following form (θ is to allow for thickness corrections, where
appropriate)
m
3( m+2 ) −1 − (m + 2 ) 2 m2+ 2 m + 4 m + 4 Ao
2 2 K L θ 4 σ hss [ ] Γ Γp , 2 +
2 2 2σ hss
D = N ω ⋅ fn
2 32m K −1θ − m4 2 2 Γ m + 2 1 − Γ m + 2 , Ao
[ ]
m
σ hss p 2
2 2σ hss
2
2
where
∞
Γ(a ) = ∫ t a −1 e−t dt
−∞
1 x a −1 −t
Γ p (a, x ) = ∫ t e dt
Γ(a ) 0
2
and where σ hss is the variance of the stress amplitude, N w is the number of seconds for
which the parameterized wind-state forcing the tower is deemed to last within each year; and
fn is the frequency associated with the peak in the hot spot stress power spectrum.
The total annual damage is therefore the sum of the damages over all the frequency bands
and all the wind states. The estimated life is then the reciprocal of the total annual damage.
APPENDIX 1
Symbol/Subscript Description
(⋅ )
N identifies a tower member N to which (.) applies
FRAMEWORK SESAM
Appendix 1 -4 Program version 2.8
MATRIX LIST
The following matrix names are used within the text. Column vectors from these matrices are
not found within the VECTOR LIST, functions which are matrix elements are not described
elsewhere.
K Stiffness matrix
M Mass matrix
φ Matrix of eigenvectors
I Identity matrix
HSS Matrix of hot spot stresses element (r,s) is the hot spot stress at
joint r from a point force at dof k
VECTOR LIST
M
g
(t ) Fourier transform of time series
k
V1 Wind velocity
S yy (ω )
s
Single sided power spectral density function of function y
These are mappings which related to members rather than degrees of freedom or joints
g (t )
N a function: the wind load on a member
σ (U )
N a mapping between a displacement vector U and the set of
stresses in a member
FRAMEWORK SESAM
Program version 2.8 Appendix 1 -7
σ2hss integral under a peak within the hot spot stress power spectra; used to
define a Rayleigh distribution
FRAMEWORK SESAM
Appendix 1 -8 Program version 2.8
CONSTANTS
ρ air density
Cd drag coefficient
D member diameter
L member length
Er
UU
contribution to wind force at dof r from U U terms, normalised by 2U U
1 1
Er
UV
contribution to wind force at dof r from U V terms, normalised by 2U V
1 1
Er
UW
contribution to wind force at dof r from U W terms, normalised by 2U W
1 1
λs eigenvalue s
λs eigenfrequency
1
2
γ damping
π Pi
j square root of -1
INDEPENDENT VARIABLES
z height
t time
ω frequency (radians/sec)
f frequency (cycles/sec)
A stress amplitude
APPENDIX 2
C yy (τ ) = y (t ) y (t + τ )
1 ∞
S yy (ω ) = ∫ exp (− jωτ ) C yy (τ )dτ
2π −∞
1 ∞
= ∫ exp (− jωτ ) yr (t ) y (t + τ ) dτ
2π −∞
The power spectral density function is normally referred to as the power spectra of that
function. The integral of the power spectra with respect to frequency is the variance of
that process:
∞ ∞ 1 ∞
∫ S yy (ω )dω = ∫ ∫ exp (− jωτ ) C yy (τ )dτ dω
−∞ − ∞ 2π − ∞
1 ∞ ∞
= ∫ C yy ∫ exp(− jωτ )dω dτ
(τ )
2π − ∞ − ∞
∞
∫ exp(− jωτ )dω = δ (τ ) = 1 when τ = 0
−∞
= 0 when τ ≠ 0
it follows that
FRAMEWORK SESAM
Program version 2.8 Appendix 2 -3
∞ −1 ∞
∫ S yy (ω )dω ∫ C yy (τ )δ (τ )dτ = C yy (0)
−∞ 2π − ∞
where C yy (0 ) is the variance of the process. (See A2.1)
C yy (r , s : τ ) = y (t )y s (t + τ )
APPENDIX 3
This appendix deals with the algebraic manipulation of equations 10.3-10.5 to form equations
10.6.
The forced response of a structure to a force vector G(ω ) where damping is neglected, may
be expressed within a finite element model as;
where
(
U T = U TM ,U TS )
where
T
UM represents the displacements of the M master degrees of freedom
T
US represents the displacements of the S slave degrees of freedom
K MM K MS M M MS U M G M
− ω 2 MM =
K SM K SS M SM M SS U S GS
where G M and G S are the forcing vectors on these master and slave degrees of freedom
respectively; K ij and M ij are i × j matrices building up K and M .
(
K SM U M + K SSU S − ω 2 M SM U M + M SSU S = G S )
The reduction method assumes terms in the lower set of equations are small. By multiplying
−1
these lower terms by K SS an expression for U S may be obtained;
[ −1 −1
] −1
U S = − K SS K SM + ω 2 K SS M SM U M + ω 2 K SS M SS U M + K SS G s
−1
2
FRAMEWORK SESAM
Program version 2.8 Appendix 3 -3
~ −1
U S = − K SS K SM U M 3
back substitution into equation 10.2 yields the more accurate approximation containing first
order inertia and forcing terms.
~
~ [ −1 2 −1
{−1 −1
U S = − K SS K SM + ω K SS M SM − K SS M SS K SS K SM }]U M
−1
+ K SSG S 4
( )
K MMU M + K MSU S − ω 2 M MMU M + M MSU S = G M
which may now be expressed in terms of U M , by using the approximation for the inertia
terms, and the U%% S approximation otherwise, so that
[K MM
] ~
− ω 2M MM U M + K MSU~ − ω 2M MSU~ S ≈ G M
S
5
K M − ω
2
U = 6a
M M M G M
where K and M are respectively the reduced stiffness and mass matrices and G M , is
M M
the equivalent force vector, as below
−1
K M = K MM − K MS K SS K SM 6b
−1 −1 −1 −1
M M = M MM − M MS K SS K SM − K MS K SS M SM + K MS K SS M SS K SS K SM 6c
−1
G M = G M − K MS K SSG S 6d
The spectral method of fatigue analysis requires a dynamic model. With Guyan reduction this
means using the equivalent mass and stiffness matrices M M and K M along with the
equivalent force vector G M . To convert the point force time series of Section 8.0 into
equivalent forces at master degrees of freedom each component in 10.6c is Fourier
transformed into a time series, so that
FRAMEWORK SESAM
Appendix 3 -4 Program version 2.8
M
−1
g = g M − K MS K SS g S
M
where g , g M and g S are the vectors of the transformed components of G M , G M and G S
M
−1
respectively. Multiplying out the matrices K MS and K SS gives the rth component of g , as
M
gr = M
g ( )−∑ r
∑ ( K MS )rq ( K −SS1 )qp ( g S )q 7
p =1 q =1
S
is the sth component of g S .
(K MS ) rq
is the rth row and qth column of K MS .
(K −SS1 ) qp
is the qth row and pth column of K SS ( ) −1
As an input requirement, all dof that will attract significant fluctuating wind loading are to be
master dof. The dynamic forcing at slave degrees of freedom for dynamic response is
ignored. e.g. within the context of a symmetric tower, rotational and vertical degrees of
freedom are assumed to attract no fluctuating load.
g r ≈ (g M )r
M
ω ≠0
M
g r ≅ (g M )r − p∑= r ∑
over slaves
( K MS ) ( K −SS1 ) ( g S ) ω = 0
rq qp q
q
This allows the power spectra of the forcing at master dof's to be directly evaluated in terms
of the power spectra of the wind, from the equations 8.3.3 and 8.3.4.
r E s S UU (ω ) +
E UU UU
α α
zs zr
S g g ( r , s : ω ) = 4 v(10, t )
2
coh (r , s : ω ) E UV
r E s S UV (ω ) + ω ≠ 0
UV
MM
10 10 UW UW S
E r E s W ′W ′ (ω )
α α
2
over slaves
= v(10, t E s z s − ∑ ∑ K MS ( )sq (K −SS1 )qp E q z s
8
10 p q 10
over slavesα α
× E r z r − ∑ ∑ K MS ( )rq (K −SS1 )qp E q z r
ω =0
10 p q 10
This is a large gain in simplicity but normally means that extra masters are required on top of
those required to model the kinetics of the tower.
FRAMEWORK SESAM
Program version 2.8 Appendix 3 -5
Equation 3 is
~ −1
U S = K SS K SMU M
Equation 4 is
~
~ [−1 −1
{ −1 −1 −1
}]
U S = − K SS K SM + ω K SS M SM − K SS M SS K SS K SM U M + K SS G S
2
and Equation 5 is
[ ] ~ ~
G M = K MM − ω 2 M MM U M + K MS U~ S − ω 2 M MS U S
Let
−1
A = K SS K SM
and
−1 −1 −1
B = K SS M SM − K SS M SS K SS K SM
~ ~
[ ]
U S = − AU M & U~ S = − A + ω B U M + K SS G S
2 −1
[ ] {[ −1
G M = K MM − ω 2 M MM U M + K MS − A + ω 2 B U M + K SSG S] } −ω 2
M MS [− AU M ]
{ −1
}{
G M = K MMU M − K MS AU M + K MS K SS G S }
{
+ − ω 2 M MM U M + ω 2 K MS BU M + ω 2 M MS AU M }
Separating forcing terms to the LHS & taking the & function outside the brackets then gives
−1
[{ } {
G M = K MS K SS G S + K MM − K MS A − ω 2 M MM − K MS B − M MS A U M }]
Let
FRAMEWORK SESAM
Appendix 3 -6 Program version 2.8
K M = K MM − K MS A
M M = M MM − K MS B − M MS A
−1
G M = G M − K MS K SS G S
to give
[
GM = K M −ω 2 M M U M ]
where back substitution of A and B into A3.1 & A3.2 gives
−1
K M = K MM − K MS K SS K SM
−1 −1 −1 −1
M M = M MM − M MS K SS K SM − K MS K SS M SM + K MS K SS M SS K SS K SM
FRAMEWORK SESAM
Program version 2.8 Appendix 4 -1
APPENDIX 4
This appendix details the two SCF schemes. For each SCF scheme, there are validity ranges
to the geometric parameters. For joints with parameters outside the validity range SCFs are
calculated with the actual parameter and with the parameters set to the broached limit. The
greater SCF is used.
NOMENCLATURE
INDEX
FIGURE A4.1
FRAMEWORK SESAM
Program version 2.8 Appendix 4 -3
Chordside SCFs
Axial Load
(
SCF CS = γτβ 6.78 − 6.42 β
0.5
)sin (
1.7 + 0.7 β
3
)θ
SCF CC = K ′C + K O K ′′C
where {
K ′C = 0.7 + 1.37 γ
0.5
}(
τ (1 − β ) 2 sin 0.5θ − sin 3 θ )
τ (β − τ / (2γ ))(α / 2 − β / sin θ )sin θ
KO =
(1 − 3 / (2γ ))
NOTE: This applies for the simply supported condition only and represents the
overall bending in the chord.
Out-of-Plane Bending
2
SCFCS = γτβ (1.6 − 1.15 β 2) sin (1.35 + β ) θ
In-Plane Bending
Braceside SCFs
For the above modes of loading the SCF on the braceside of the weld may be
estimated from the following equation:
Validity Ranges
{
SCF CS = γ τ Aβ A 1.6 − 1.15 β( 5
A )}
sin (1.35 + β 5A ) + (0.016 γ β ) (0.45 + g / D ) ( / ) 0.3sin (1.35 + β 5B )
θA B θA θB θ B
{1 − 0.1 (1.0 + 2 g / R )
}
SCF BS = 1 + 0.63 × SCF CS
0.38 0.06
SCF CHORD = 1.822 γ τ 0.94 β sin 0.9θ
0.54
SCF CHORD = 1.83 γ τ 1.068β 0.12
sin θ 0° < θ ≤ 90°
Central Brace
{ (
SCF CS = γ τ B β B 1.6 − 1.15 β
5
B )}
sin (1.35 + β 5B )
θB
+ (θ A /θ B ) 0.3 (0.016 γ β A)
(0.45 + g AB / D )sin (1.35 + β 5A )
θA
+ (θ B /θ C )0.3 (0.016γ β C )
(0.45+ g BC / D ) ( 5
)
sin 1.35 + β C θ C }
{1−1.0 (1.0 + ( g AB + g BC )/ R )} 2
FRAMEWORK SESAM
Appendix 4 -6 Program version 2.8
{ (
SCF CS = γ τ A β A 1.6 − 1.15 β
5
A )}
sin (1.35 + β 2A )
θA
+ (θ A /θ B ) 0.3 (0.016 γ β B )
( 0.45 + g AB / D )sin (1.35+ β 2B )
θB
+ (θ A /θ C ) 0.3 (0.016 γ β C )
( 0.45+ g AC / D )sin (1.35 + β C2 )
θ C
0.38 0.94
SCF CHORD = 1.822 γ τ β 0.06 sin 0.9θ
Chordside SCFs
Axial Load
SCF CS = 0
SCF CS = 0
In Plane Bending
T-Joints
The geometrical parameters are as shown in Figure A4.1. C fix is an end fixity
parameter taken as 0.7. befar = 0. for our geometries. The output SCFs are as follows:
Axial SCFs
1
(
f = 1 − 0.83β − 0.56 β 2 − 0.02 γ ) 0.23
e − 0.21γ
−1.16
α 2 .5
(
f 2 = 1 − 1.43β − 0.97 β 2 − 0.03 γ ) 0.04
e − 0.71γ
−1.38
α 2 .5
C1 = (C fix − .5) ∗ 2
C 2 = C fix / 2.0
C 3 = C fix / 5.0
xstif = 1.0
( )
t1 = γ ∗ T 1.1 1.11 − 3 (β − 0.52 ) sin (θ )
2 1.6
(1 − β ) f ⋅ xstif
CC = t 1 + C1 (0.8α − 6 )Tβ 2 2
SSCFB = τ − 0.54
γ − 0.05
(0.99 − 0.47 β + 0.08β ) T 4
10
K-Joints
Axial SCFs
K1 = τ 0.9
a γ 0.5
(0.67 − β 2
a + 1.16 β a )sin θ a
KC1=K1
KC2=K1
KB1=K2
KB2=K2
Inplane Bending
τ =τ a, β = β a, θ =θa
θ max = max(θ a ,θ b )
θ min = min (θ a ,θ b )
β max = max(β a ,β b )
β min = min (β a ,β b )
T8 = 1.45β τ 0.85
γ (1− 0.68 β ) sin 0.7θ
T9 = 1 + 0.65 β τ γ (1.09 −.77 β ) sin θ
0.4 0.06
γ −1.16
k 9= T 9 ∗ (0.9 + 0.4 β )
β mx = max(β a ,β b )
f4= {
1 α ≥ 12
1 − 1.07 β 1a.88 e − 0.16 γ
−1.06
α 2.4
α ≤ 12
(
t10 a = γ τ a β a 1.7 − 1.05 β )sin
3
a
1.6
(θ a )
t10b = γ τ b β b (1.7 − 1.05 β )sin
3
b
1.6
(θ b )
x = 1 + ζ ∗ sin θ a / β a
+ . 1 − 0.08 (β ⋅ γ ) 2 e − 0.8 x
1
S b t10b
a
1
−1.3 x
∗ 2.05 β mx e
2
CHSA = K 14 ∗ f 4
BRSA = τ − 0.54 − 0.05
a γ (0..99 − 0.47 β a + 0.08 β 4
a )K14
CHSA is chordside SCF saddle
BRSA is braceside SCF saddle
FRAMEWORK SESAM
Appendix 4 -12 Program version 2.8
KT Joints
β max = max (β a , β b )
β min = min (β a , β b )
θ max = max (θ a , θ b )
θ min = min (θ a , θ b )
K1 = τ 0.9 0.5
b γ (0.67 − β 2
b + 1.16 β b )sinθ b
(1.64 + 29 β − 0.38
b atan (8ζ ) )
(
K 2 = 1 + K1 1.97 − 1.57 β 0.25
b )τ − 0.14
b sin 0.7 (θ b )
+ Cgapot β 1b.5 γ 0.5
τ −1.22
b sin 1.8 (θ a + θ b ) ∗
(0.131 − 0.084 ∗ ATAN / (143 + 4.2 β b )
KC 2 = KC1 = K1
KB 2 = KB1 = K 2
β max = max(β b ,β c )
β min = min (β b ,β c )
θ max = max(θ b ,θ c )
θ min = min(θ b ,θ c )
K1 = τ 0.9 −0.5
b γ (0.67 − β 2
b + 1.16 β b )sin θ b
FRAMEWORK SESAM
Program version 2.8 Appendix 4 -13
0.3 0.3
sin θ max β max
×
sin θ min β
min
(
× 1.64 + 0.29 β − 0.38
b tan −1 (8ζx ))
(
K 2 = 1 + K1 1.97 − β 0.25
b )τ −0.14
b sin 0.7 θ b
+ Cgapot β 1b.5τ −1.22
b sin1.8 (θ b + θ c )
(
× 0.131 − 0.084 ∗ tan −1 (14 ζ bc + 4.2 β b ) )
For both cases
S a = 1
S b = 1 in our use of the code
S c = 1
β mx = max (β a , β b , β c )
f 4 = 1 unless α ≤ 12when
1.88 1.06 2.4
f 4 = 1 − 1.07 β b exp − 0.16 γ α
(
t10 a = γ τ a β a 1.7 − 1.05 ∗ β )sin 3
a
1.6
θa
t10b = γ τ b β b (1.7 − 1.05 ∗ β )sin 3
b
1.6
θb
t10c = γ τ c β c (1.7 − 1.05 ∗ β )sin 3
c
1.6
θc
xab = 1 + ζ ab sin (θ b ) / β b
xbc = 1 + ζ bc sin (θ b ) / β b
powa = (β a / β b ) 2
powc = (β c / β b )
2
FRAMEWORK SESAM
Appendix 4 -14 Program version 2.8
(
K t10 = t10b 1 − 0.8 ( β a γ ) 0.5 e −0.8 x ab
) powa
∗
( S (1 − 0.8 (β γ )
b c
0.5
e − 0.8 xbc ) powc
(
+ S a t10 a 1 − 0.8 (β b γ )0.5 e − 0.8 x ab
(
× 2.05 β
1
mx e
2
−1.3 x ab
(
+ S c ∗ t10c 1 − 0.08 (β b γ ) 0.5 e − 0.8 x bc
(
× 2.05 β 0.5 −1.3 xba
mx e
Reference brace is a diagonal brace. a and c refer to the two diagonal braces.
K1 = τ 0.9 + 0.5
a γ ( 0.67 − β 2
a + 1.16 β a )sin θ a
β mx = max (β a , β b , β c )
f 4 = 1 unless α ≤ 12 when
1.06 2.4
1.88 − 0.16γ α
f 4 = 1 − 1.07 β a e
(
t10 a = γ τ a β a 1.7 − 1.05 β ) sin 3
a
1.6
θa
t10b = γ τ b β b (1.7 − 1.05 β ) sin 3
b
1.6
θb
t10c = γ τ c β c (1.7 − 1.05 β ) sin 3
c
1.6
θc
x ab = 1 + ζ ab sin (θ a ) β a
x ac = 1 + ( ( ζ ab + ζ ac + β b / sin (θ b ) )∗ sin θ a ) / β a
(
K t10 = t10 a 1 − 0.08 (β b γ )
1
2 e − 0.8 xab S a )
(
∗ 1 − 0.08 (β c γ )
1
2 e − 0.8 xab )
(
+ S b ∗ t10b ∗ 1 − 0.08 (βγ ) 0.5 )e − 0.8 xab
∗ 2.05 β 0mx.5 e −1.3 xab + S c t10c ( 1 − 0.08 (β a γ ) )e − 0.8 xac
0.5
1
2 −1.3
∗ 2.05 ∗ β mx e xac
ChSAD = K t10 ∗ f 4
(
BRSAD = τ −a 0.54 γ − 0.05 0.99 − 0.47 × β a + 0.08 β 4a ChSAD )
ChSAD is chordside saddle out of plane bending SCF
BrSAD is braceside saddle out of plane bending SCF
X-Joints SCF
Axial SCFs
f2 = 1 − (1.43β − 0.97 β 2
− 0.03 )γ 0.04 − 0.71γ −1.38 α 2.5
e
(
X 1 = 3.87 ∗ γ ∗τ ∗ β 1.1 − β 1.8 sin 1.7 θ . f )
X2 =γ 0.2
(
τ 2.65 + 5 ∗ ( β − 0.65 ) 2 − 3τ β sin (θ ) )
X 3 = 1 + 1.9 γ τ 0.5
(
β 0.9 1.09 − β 1.7 )sin 2.5
θf
X 4 = 3+γ 1.2
( 0.12 e −4β
+ 0.011 β − 0.045 2
)
Inplane Bending
f 3 = 1 unless α ≤ 12 when
− 0.49 γ −0.89 α 1.8
f 3 = 1 − 55 β 1.8 γ 0.16
e
(
CSA = γτβ 1.56 − 1.34 β 4
)sin 1.6
(θ ) ∗ f 3
BSA = τ − 0.54
γ − 0.05
( 0.99 − 0.47 β + 0.08 β )∗ CSA 4
REFERENCES
FRAMEWORK SESAM
Program version 2.8 Appendix 4 -17
APPENDIX 5
LLOYD'S REGISTER OF SHIPPING FORMULAE (ISOPE 1991)
FRAMEWORK SESAM
Appendix 5 -2 Program version 2.8
A5.1 Introduction
This appendix gives the SCF parametric equations for unstiffened tubular joints defined by
Lloyd's Register of Shipping in the ISOPE conference paper of 1991.
The equations are derived in terms of functions that may be factored for short chord effects.
The short chord factors are F1 to F3. There are also factors that isolate the localized bending
at saddle positions from the overall beam bending effects. The beam bending terms are B0
and B1. Typically the short chord correction factors are only applicable to the localized
bending terms. Framework automatically takes the appropriate action for the supplied L/D
ratio.
The stiffening effects of unloaded braces in the vicinity of the loaded brace under
consideration are also taken into consideration. These are characterized by the S1 and S2
factors. Similarly the effects of loaded braces in the same area are considered by the
application of influence factors. These are factors IF1 to IF8.
The equations consist of basic relations for T and X joints. These are then amalgamated,
using the appropriate factors for joint geometry and complexity, to build equations for K and
KT joints.
FRAMEWORK SESAM
Program version 2.8 Appendix 5 -3
NOMENCLATURE
FIGURE A.5.1
FIGURE A.5.2
FRAMEWORK SESAM
Appendix 5 -4 Program version 2.8
Axial Load
SCFCC = 1.2T2 + B0 B1
SCFBC = 1.23T4
Note: The short chord correction factor F1 is applied for fully fixed ends (C ≤ 0.7). F2 is
applied for pinned ends (C > 0.7). The F factors and the chord stress factors B0 and
B1 are defined in a later section of this appendix.
Out-of-Plane Bending
In-plane Bending
SCFC = 1.15T7
SCFB = 1.18T8
T Joint Factors
NB Apply the modified values of the β parameter when predicting the SCFs at the saddle for
joints that have values of β close to 1.0 and are under axial load or out-of-plane bending.
This affects the equations for T1, T3, T5, and T6.
3
T7 = 1.22τ 0.8 βγ (1− 0.68 β ) sin (1− β ) θ
SCFCC = 1.33 X 2 + B0 B1 )
SCFBC = 1.13 X 4
Note: The short chord correction factor F1 is applied for fully fixed ends (C ≤ 0.7). F2 is
applied for pinned ends (C > 0.7). The F factors and the chord stress factors B0 and
B1 are defined in a later section of this appendix.
SCFC = 1.23 X 7
SCFB = 1.12 X 8
X Joint Factors
NB Apply the modified values of the β parameter when predicting the SCFs at the saddle for
joints that have values of β close to 1.0 and are under axial load or out-of-plane bending. This
affects equations X1, X3, X5, and X6.
FRAMEWORK SESAM
Appendix 5 -6 Program version 2.8
1.5 0.5
X 2 = (0.36 + 1.9τγ 0.5e( − β γ )
(sin θ + 3 cos 2 θ )
X 3 = 1.0 + 0.6 X 1
2 0.5
X 4 = (1.3 + 0.06τγe( − β γ )
(sin θ ) −1
2
X 6 = 1.0 + τβγ 1.5 (0.19 − 0.185β 3 ) sin 7 (1− β ) θ
−0.5
X 7 = τ 0.8 βγ ( 0.5 β )
(1.0 − 0.32 β 5 ) sin 0.5 θ
Note: The short chord correction factor F1 is applied for fully fixed ends (C ≤ 0.7). F2 is
applied for pinned ends (C > 0.7). The F factors, the chord stress factors B0 and B1,
the stiffening factors, S1 and S2 and the influence functions IF1 to IF8 are defined in a
later section of this appendix.
Note: The expression T1A implies that the equation for T1 should be evaluated for the
geometric parameters associated with brace A. Brace A is the brace under
consideration and brace B is the other brace of the K joint.
SCFCC = 1.13T2 A S 2 AB + B0 A B1 A
SCFBC = 1.23T4 A S 2 AB
SCFC = 1.15T7 A
SCFB = 1.17T8 A
Note: The short chord correction factor F1 is applied for fully fixed ends (C ≤ 0.7). F2 is
applied for pinned ends (C > 0.7). The F factors, the chord stress factors B0 and B1,
the stiffening factors, S1 and S2 and the influence functions IF1 to IF8 are defined in a
later section of this appendix.
FRAMEWORK SESAM
Appendix 5 -8 Program version 2.8
Note: The expression T1A implies that the equation for T1 should be evaluated for the
geometric parameters associated with brace A. Brace A is the brace under
consideration unless otherwise stated.
where S 2 B = Max( S 2 BA , S 2 BC )
SCFC = 1.15T7 A
SCFB = 1.17T8 A
The central brace SCFs are evaluated by taking the maxima of the pairs of values that result
from considering brace B and its interaction with the two outer braces (A and C).
where
S 2 A = max(S2 AB , S2 AC )
S 2 B = max(S2 BA , S2 BC )
S 2C = max(S 2CA , S2CB )
For brace A
For brace B
The equation for S1AB gives the effect at the saddle of brace A due to the loads on brace B at
the joint. This appears as a reduction in the saddle SCF. Similarly S2AB gives the effect at the
crown of brace A due to loads on brace B at the joint. This increases the crown SCF. Note
that the reduced value of the β parameter should be applied for joints where β approaches 1.0
for the saddle SCF reduction factor, S1AB.
2
S1 AB = {1.0 − 0.4 exp(−30.0 x AB ( β A / β B ) 2 (sin θ A / γ ))}
2
S 2 AB = {1.0 + exp(−2.0 x AB /(γ 0.5 sin 2 θ B ))}
Equation IFAB gives the influence upon brace A of the applied loading in brace B.
Note that the reduced value of the β parameter should be used for joints where β approaches
1.0 for saddle influence function calculations. This affects equations IF1, IF3, IF5 and IF6.
FRAMEWORK SESAM
Program version 2.8 Appendix 5 -11
IF4 AB = (−1.5β Α )e − x AB
where C is the chord end fixity parameter (0.5 ≤ C ≤ 1.0). For fully fixed chord ends C = 0.5,
for pinned ends C = 1.0.
The short chord correction factors account for the reduction of chord ovalization that occurs
due to the presence of chord end restraints close to the joint. Care must be taken in applying
short chord factors in structural analysis. A chord length to diameter (L/D) of 20.0 implies
that short chord effects need not be considered. Accordingly SCFs will be calculated
conservatively; that is they may be higher than the true values, leading to shorter fatigue lives
being predicted.
For joints which satisfy the short chord criteria regarding effective support conditions the
following factors are applicable.
• For α ≥ 12.0 there are no short chord effects. Note that α = 2(L/D), where (L/D) is the
length to diameter ratio.
• For α < 12.0 the following expressions are used in the formulae quoted above for SCF
values.
( −1.16 )
α 2.5 )
F1 = 1.0 − (0.83β − 0.56 β 2 − 0.02)γ 0.23e( −0.21γ
( −1.38 )
α 2.5 )
F2 = 1.0 − (1.43β − 0.97 β 2 − 0.03)γ 0.04e( −0.71γ
( −0.89 ) 1.8
F3 = 1.0 − 0.55β 1.8γ 0.16e( −0.49γ α )
Validity Range
The equations quoted in this appendix are generally valid for joint parameters within the
following limits
0.13 ≤ β ≤ 1.00
10.0 ≤ γ ≤ 35.0
0.25 ≤ τ ≤ 1.00
30.0° ≤ θ ≤ 90.0°
4.00 ≤ α
0.0 ≤ ζ ≤ 1.00 (For K and KT joints)
FRAMEWORK SESAM
Program version 2.8 Appendix 6 -1
APPENDIX 6
This attachment outlines the recommendations contained in the Lloyd's Register of Shipping
Recommendations issued in the January 1996 ("Lloyd's Register of Shipping Recommended
Parametric Stress Concentration Factors", document number OD/TN/95001, January 1996).
The equations consist of a mix of those of Efthymiou (see Attachment 4) and those from the
Lloyd's 1991 ISOPE paper (see Attachment 5). Note that for the purposes of clarity the
terminology must be clearly defined. The Lloyd's (1991) equations refer to the ISOPE paper.
These equations are not to be confused with either of the two sets of Lloyd's Recommended
equations, i.e. those of 1988 or 1996.
For K and KT joints, with gaps greater than or equal to zero, Lloyd's Register recommends
the use of the Lloyd's (1991, ISOPE) equations.
For K and KT joints with overlaps (or equivalently, negative gaps) the equations due to
Efthymiou are recommended by Lloyd's Register.
For in-plane bending the Lloyd's Recommendations require the use of the Efthymiou
equations with balanced loads. This contrasts with the Shell recommendations that require
the use of the equivalent equations for unbalanced loads. Accordingly different SCFs will be
generated, for overlapped K or KT joints if the Efthymiou equation set (E) is used instead of
the Lloyd's Recommended equation set (R).
The Lloyd's recommendation for T, Y, and X joints is that the Efthymiou equations are to be
used.
Full details of the Lloyd's (1991, ISOPE) equations are to be found in Attachment 5.
FRAMEWORK SESAM
Program version 2.8 Appendix 7 -1
APPENDIX 7
S uu (r , s, f ) := coh(r , s, f ) ⋅ S uu ( f )
where
Lu ⋅ f
4.0 ⋅ k ⋅ V ⋅ V
V
S uu ( f ) := 5.0
Lu ⋅ f
2 6.0
2.0 + ⋅f
V
− c ⋅ z (r ) − z (s ) ⋅ f
coh(r , s, f ) := exp
V
and
S uu (r , s, f ) := coh(r , s, f ) ⋅ S uu ( f )
where:-
2
Lu ⋅ f
4.0 ⋅ k ⋅V ⋅ V
V
S uu ( f ) := 4.0
Lu ⋅ f
2 3.0
1.0 + ⋅f
V
z (r ) − z (s )
− c ⋅ 2 − ⋅ R ⋅ f
r
coh(r , s, f ) := exp
0.5 ⋅ (V (z (r )) + V (z (s )))
and
S uu (r , s, f ) := coh(r , s, f ) ⋅ S uu ( f )
Lu ⋅ f
15.0 ⋅ k ⋅ V ⋅ V
V
S vv ( f ) := 5.0
Lu ⋅ f 3.0
1.0 + 9.5 ⋅ ⋅ f
V
z (r ) − z (s )
− c ⋅ 2 − ⋅ R⋅
f
R
coh(r , s, f ) := exp
0.5 ⋅ (V (z (r )) + V (z (s )))
R := x(r )2 − x( s )2 + y(r )2 − y( s )2 + z (r )2 − z( s )2
and
S uu (r , s, f ) := coh(r , s, f ) ⋅ S uu ( f )
where
Lu ⋅ f
3.36 ⋅ k ⋅ V ⋅ V
V
S vv ( f ) := 5.0
Lu ⋅ f 3.0
1.0 + 10.0 ⋅ ⋅ f
V
z (r ) − z (s )
− c ⋅ 2 − ⋅ R ⋅ f
R
coh(r , s, f ) := exp
0.5 ⋅ (V (z (r )) + V (z (s )))
and
APPENDIX 8
The S-N curve for high amplitude stress cycles (number of cycles < 107) may be expressed as
m
NU ( A, m ) = K 2 ⋅ (2 ⋅ A) −m
⋅θ 4
The S-N curve for low amplitude stress cycles (number of cycles > 107) may be expressed as
m+2
− (m + 2 )
NL( A, m ) = K L ⋅ (2 ⋅ A) ⋅θ 4
For a narrow band process, the number of cycles at stress amplitude A is given by
A − A2
nc( A, σ ) = Time ⋅ freq ⋅ ⋅ exp
σ 2 2
2 ⋅σ
nc( A) ∞ nc( A)
∫ ∫
AO
D( A) := Time ⋅ freq ⋅ dA + dA
O NL( A, m ) AO NU ( A, m )
where AO is the stress amplitude at 107 cycles, taken from the S-N curve.
− ( m+ 2 )
− A2 A
∫
AO
DN ( A) := K L ⋅ (2 ⋅ A)
−1 m+ 2
⋅θ ⋅4
exp ⋅ dA
2
O σ2 2 ⋅σ
−m
∞ A − A2
∫ K 2 ⋅ (2 ⋅ A) ⋅θ 4 ⋅ 2 ⋅ exp
−1 m
+ 2
dA
AO σ 2 ⋅σ
FRAMEWORK SESAM
Program version 2.8 Appendix 8 -3
− ( m+ 2 )
A m +3 − A2
∫
AO
DN ( A) := 2 m+2 ⋅ K L ⋅θ
−1 4
⋅ ⋅ exp dA
2
O σ2 2 ⋅σ
−m ∞ Am+1 − A2
∫
−1
+ 2 ⋅ K 2 ⋅θ
m 4
⋅ ⋅ exp dA
2
AO σ2 2 ⋅σ
A p+1 − A2
∫
b
l (a, b, p ) := ⋅ exp dA
2
a σ2 2 ⋅σ
t ( A) :=
A2
2 ⋅σ 2
A(t ) := 20.5 ⋅ σ 2 ( ) 0.5
⋅ t 0.5 dA(t ) := 2− 0.5 ⋅ σ 2 ( ) 0.5
⋅ t − 0.5 ⋅ dt
This gives
( )
p +1
20.5⋅ 2 0.5⋅t 0.5
σ
[ ( ) ]
b
l (a, b, p ) :=
∫ ⋅ exp(− t ) ⋅ 2−0.5 ⋅ σ 2
2⋅σ
2 0.5
a
⋅ t −0.5 dt
σ 2
2⋅
σ
2
( ) ∫
p
l (a, b, p ) = 2 2 ⋅ σ 2 2 ⋅ t 2 ⋅ exp(− t )dt
P 2 p
2⋅σ
a
2⋅ 2
σ
−( m + 2 ) m+ 2 AO
2 ⋅ K L ⋅θ 4 ⋅ 2 2 ⋅ (σ ) 2 ⋅ ⋅σ 2 t 2 ⋅ exp(− t )dt...
m+2 m+ 2
−
D( A) := Time ⋅ freq ⋅
m+ 2 1 2
0
2
∫
m m
+ 2 ⋅ K 2 ⋅θ 4 ⋅ 2 2 ⋅ (σ ) 2 ⋅ AO t 2 ⋅ exp(− t )dt
− m m ∞ m
∫
−1 2
2⋅σ 2
FRAMEWORK SESAM
Appendix 8 -4 Program version 2.8
3⋅(m+ 2 ) −( m + 2 ) m+ 2 AO
2 2 ⋅ K L ⋅θ 4 ⋅ (σ ) 2 ⋅ ⋅σ 2 t 2 ⋅ exp(− t )dt...
m+ 2
∫
−1 2 2
Let
1
G (a, x ) := ∫ t a −1 ⋅ exp(− t )dt
x
⋅
Γ(a ) 0
Then
3⋅ ( m + 2 ) − (m + 2 ) m+2
m + 4 m + 4 A0
2 2 −1
⋅ K L ⋅θ 4 ⋅σ ( )
2 2
⋅ Γ ⋅ G ,
2 2 2 ⋅σ 2
...
D( A) := Time ⋅ freq ⋅
3⋅ m −m
m + 2 A0
( )
m
−1 2 2 m + 2
+ 2 ⋅ K 2 ⋅ θ
2 4 ⋅σ ⋅ Γ ⋅ 1 − G ,
2 2 2 ⋅ σ 2
FRAMEWORK SESAM
Program version 2.8 Appendix 9 -1
Appendix 9
DAMS
Standard Axis System
Z'
( X T′ , YT′ , ZT′ )
dZ′
Member
dY′ unit direction
vectors
dX′
Y'
W ( X B′ , YB′ , Z B′ )
V
U θ Wind direction
X'
Wind velocity
vector
FIGURE A9.1
In the ( X ' , Y ' , Z ' ) coordinate system the wind force vector acting on a member is
1
F= ⋅ ρ ⋅ Cd ⋅ D ⋅ L ⋅ u′n ⋅ un′
2
where ρ = mass density of air (1.225kg/m3 = 0.002377lbf·sec2/ft4),
Cd = the member's drag coefficient,
D = the member's diameter (including ice thickness as appropriate),
L = the length of the member within the specified wind profile,
u′n = the magnitude of the wind velocity vector normal to the member, and
un′ = the wind velocity vector normal to the member.
U
With reference to Figure A9.1, let the wind velocity vector u = V and the member unit
W
dX′
direction vector m = dY ′ .
d
Z′
FRAMEWORK SESAM
Program version 2.8 Appendix 9 -3
(d Z ′ ⋅ sin θ ⋅ U + d Z ′ ⋅ cosθ ⋅ V − dY ′ ⋅ W )2
= + (d X ′ ⋅ W − d Z ′ ⋅ cosθ ⋅ U + d Z ′ ⋅ sin θ ⋅ V )2
+ (dY ′ ⋅ cosθ ⋅ U − dY ′ ⋅ sin θ ⋅ V − d X ′ ⋅ sin θ ⋅ U − d X ′ ⋅ cosθ ⋅ V )2
= a12 4444
⋅U 2 + b ⋅U
2⋅4444 W + d 2 ⋅V 2 + e ⋅V ⋅ W + f 2 ⋅ W 2
V + c ⋅ U ⋅3
14444244443
dominant terms non-dominant terms
for V << U and W << U for V << U and W << U
(
where a 2 = dY2′ ⋅ cos 2 θ − 2 ⋅ dY ′ ⋅ cosθ ⋅ d X ′ ⋅ sin θ − dY2′ − d X2 ′ ⋅ cos 2 θ + 1 )
(
b = 2 ⋅ d X2 ′ ⋅ sin θ ⋅ cosθ + dY ′ ⋅ d X ′ − dY2′ ⋅ cosθ ⋅ sin θ − 2 ⋅ dY ′ ⋅ cos 2 θ ⋅ d X ′ )
c = −2 ⋅ d Z ′ ⋅ (sin θ ⋅ dY ′ + cosθ ⋅ d X ′ )
(
d 2 = d X2 ′ ⋅ cos 2 θ + 2 ⋅ dY ′ ⋅ sin θ ⋅ d X ′ ⋅ cosθ + 1 − dY2′ ⋅ cos 2 θ − d X2 ′ )
e = 2 ⋅ d Z ′ ⋅ (sin θ ⋅ d X ′ − cos θ ⋅ dY ′ )
(
f 2 = d X2 ′ + dY2′ )
FRAMEWORK SESAM
Appendix 9 -4 Program version 2.8
Define u′n = A ⋅ U + B ⋅V + C ⋅ W .
B=
b
=
( )( )
d X ′ ⋅ dY ′ ⋅ sin 2 θ − cos 2 θ + d X2 ′ − dY2′ ⋅ sin θ ⋅ cosθ
2⋅ A 1 − (dY ′ ⋅ sin θ + d X ′ ⋅ cosθ )2
c − d Z ′ ⋅ (sin θ ⋅ dY ′ + cosθ ⋅ d X ′ )
C= =
2⋅ A 1 − (dY ′ ⋅ sin θ + d X ′ ⋅ cosθ )2
U
u′n = H ⋅ V
W
and ( )
hUX ′ = cosθ ⋅ 1 − d X2 ′ − dY ′ ⋅ d X ′ ⋅ sin θ
( )
hVX ′ = − sin θ ⋅ 1 − d X2 ′ − dY ′ ⋅ d X ′ ⋅ cosθ
( )
hVY ′ = cosθ ⋅ 1 − dY2′ + dY ′ ⋅ d X ′ ⋅ sin θ
hVZ ′ = −d Z ′ ⋅ (dY ′ ⋅ cosθ − d X ′ ⋅ sin θ )
hW X ′ = − d X ′ ⋅ dZ ′
hW Y ′ = − dY ′ ⋅ dZ ′
h W Z ′ = d 2X ′ + dY2′
From the above the wind force vector F may now be written as
1
F = ⋅ ρ ⋅ Cd ⋅ D ⋅ L ⋅ ( A ⋅ U + B ⋅ V + C ⋅ W ) ⋅ H ⋅ u
2
FRAMEWORK SESAM
Appendix 9 -6 Program version 2.8
The wind velocity vector u may be considered to be the sum of both mean velocities and
time-varying velocities, ie
U U mean + U ′
u = V = Vmean + V ′
W W
mean + W ′
For a fatigue analysis the steady-state terms may be ignored. Ignoring also the second-order
time-varying terms gives
2 ⋅ A ⋅ U mean ⋅ U ′ + B ⋅ U mean ⋅ V ′ + C ⋅ U mean ⋅ W ′
1
F = ⋅ ρ ⋅ Cd ⋅ D ⋅ L ⋅ H ⋅ A ⋅ U mean ⋅ V ′
2
A ⋅ U mean ⋅ W ′
( )( )
B d X ′ ⋅ dY ′ ⋅ sin 2 θ − cos 2 θ + d X2 ′ − dY2′ ⋅ sin θ ⋅ cosθ
where B′ = =
A 1 − (dY ′ ⋅ sin θ + d X ′ ⋅ cosθ )2
FRAMEWORK SESAM
Program version 2.8 Appendix 9 -7
C − d Z ′ ⋅ (sin θ ⋅ dY ′ + cosθ ⋅ d X ′ )
C′ = =
A 1 − (dY ′ ⋅ sin θ + d X ′ ⋅ cosθ )2
The Framework application requires the wind force vector to be passed as three separate
vectors representing the force due to, respectively, U mean ⋅ U ′ (the first load condition
generated by Wajac for each wind direction), Vmean ⋅V ′ (the second load condition generated
by Wajac for each wind direction) and Wmean ⋅ W ′ (the third load condition generated by
Wajac for each wind direction), ie
2 B′ C ′ U mean ⋅ U ′
1
= ⋅ ρ ⋅ Cd ⋅ D ⋅ L ⋅ H ⋅ A ⋅ 0 1 0 ⋅ 0
2 0 0 1
0
2 B′ C ′ 0
1
+ ⋅ ρ ⋅ Cd ⋅ D ⋅ L ⋅ H ⋅ A ⋅ 0 1 0 ⋅ U mean ⋅ V ′
2 0 0 1
0
2 B′ C ′ 0
1
+ ⋅ ρ ⋅ Cd ⋅ D ⋅ L ⋅ H ⋅ A ⋅ 0 1 0 ⋅ 0
2 0 0 1 U
mean ⋅ W ′
The Framework application also requires U ′ , V ′ and W ′ to be set equal to 0. 5 ⋅U10 where
U10 is the wind velocity U at a reference height of 10m. Substituting 0. 5 ⋅U10 for U ′ , V ′ and
W ′ gives
SOFTWARE
DNV GL is the world-leading provider of software for a safer, smarter and greener future in the energy,
process and maritime industries. Our solutions support a variety of business critical activities including
design and engineering, risk assessment, asset integrity and optimization, QHSE, and ship management.
Our worldwide presence facilitates a strong customer focus and efficient sharing of industry best practice
and standards.