OTC 3378

FATIGUE ANALYSIS OF THE COGNAC PLATFORM

by R. K. Kinra, King-Wilkinson, Inc.;

and P. W. Marshall, Shell Oil Co .

., Copyright 1979, Offshore Technology Conference

This paper was presented at the 11th Annual OTC-In Houston, Tex" April30-May 3, 1979. The material Is subJect to correction by the author. Permission to copy is restricted to an abstract of not more than 300 words.

ABSTRACT INTRODUCTION

The overall Cognac project in 1025 ft. (310 m) Fatigue damage considerations played an im-
of water is more fully' described elsewhere in the portant role in the structural design of the
proceedings of this confe~ence (See OTC papers Cognac platform. Many of the heavy-wall joints
3493 through 35QO). This paper describes the in the structure were, in fact, controlled by
structural fatigue analysis and design process fatigue rather than by punching shear under the
in terms of generic preliminary analysis, an design wave. The reason, of course, is that the
allowable stress approach for initial design, natural period of the Cognac platform is in the four
and a final exhaustive fatigue check using prob- to five second range, whereas the natural periods
abilistic (spectral) analysis as outlined below. of other structures in the Gulf are generally less
than two seconds. Thus, the Cognac platform responds
Dynamic structural analysis is used to dynamically to everyday waves in the Gulf, which
generate local_stress data at each member end typically have mean zero-crossing periods between
in the structure as a function of wave frequency three to five seconds 1 • In comparison, other GuI£-
and direction. The dynamic analysis consists of of Mexico structures tend to respond statically. To
a complete static analysis including all degrees make matters more difficult, the hydrodynamic damp-
of freedom of the structure plus the dynamic con- ing due to low waves is very small. Indeed, recent
tribution of the first few modes of vibration. studies by Ruh1 2 indicate that the total damping
The local member stress data are used in con- including structural, foundation and hydrodynamic
junction with appropriate stress concentration damping contributions may only be about two per-
factors to determine transfer functions for hot cent of critical for platform response in low sea
spot stresses at each member end. The basis states. Low damping values result in large dynamic
of the stress concentration factors used is amplifications when the period of the forcing
described, including the results of model tests function (the wave period) is close to the natu-
and finite element analyses of typical grouted ral period of the structure. The combination of
and ungrouted joints in the structure. low damping and a natural period close to that
of frequently occurring Gulf of Mexico sea states
Using the member stress transfer functions results, not surprisingly, in a structure that is
generated, probabilistic spectral techniques unusually fatigue-sensitive.
are applied to evaluate short-term stress sta-
tistics at the hot spots for each sea state in The steps taken to assure the integrity of the
the annual directional wave scatter diagrams for structure against fatigue damage consisted of a
the platform location. Directional spreading of preliminary fatigue analysis which served as a
wave energy is included in the analysis. The basis for the design of the structure, and a detailed
calculated short-term statistics are combined member-by-member check of the final design using
with the wave scatter diagram data to yield long- probabilistic techniques as described herein.
term stress statistics at each hot spot. Final-
ly, Miner's cumulative damage law is applied to PRELIMINARY FATIGUE CONSIDERATIONS
evaluate local fatigue damage. The paper includes
fatigue analysis results for selected membgrs. The overall strategy of fatigue design con-
In addition, the sensitivity of the results to sists of a two stage approach:
changes in the major parameters is investigated
and discussed. (1) Initial design of members and joints using
allowables derived from generic analysis.

(2) A final hot spot by hot spot fatigue analy-

sis, which requires that all details of
References and illustrations at end of paper.
169
 the structure be established a priori. design to conform to the corresponding sensitivity.
Of necessity, the detailed analysis is done The first round analysis was performed using
after all other design work has been completed, program TOWER, developed by Penzien, et a1 5 , using
often after fabrication has begun. If any areas spectral techniques. The TOWER program performs
with undesirably short calculated fatigue lives a spectral dynamic analysis using a lumped-mass
are found, these can be upgraded--either by grinding flagpole model of the structure. A damping value
welds to improve fatigue performance, locally of 3.5 percent was used.
building up the section thickness to reduce cy-
clic stress, or making other minor modifications. Results for a trial design with an allowable hot
spot stress of 58 ksi for braces responding to base
This general approach has now been used on shear, are shown in Table 2. Similar analyses, car-
several platforms in the North Sea, in the Gulf ried out for braces responding to overturning mo-
of Mexico, and offshore Southern California. ment, and for waterline braces responding in part to
local wave pressure, indicated that these cases were
Allowable Fatigue Stresses less critical. These analyses, based on a prelimi-
nary representation of wave climate and the AWS-X-
The initial fatigue design was done in terms Modified fatigue curve 6 ,led to two primary conclu-
of reduced allowable stresses, to be used with the sions:
peak design wave load, as shown in Table 1. Note
the sensitivity ratio (ksi hot spot stress per (1) A reduced allowable hot spot stress of 46 ksi
foot of wave height) given in the last column (320 N/mm 2 ) would be required to meet the tar-
of the table. This ratio has a value of unity get fatigue life of 100 years.
for traditional Gulf of Mexico designs. To off-
set the sensitivity of the deepwater Cognac struc- (2) Over 80% of the fatigue damage was done by non-
ture to everyday waves, it was necessary to reduce hurricane sea states (e tllrough ~ in Table 2).
the allowable stress to be used with the design For these lower sea states, damage rates are
wave, resulting in a sensitivity ratio of 0.6. sensitive to wave period (compare g vs. h).
This is comparable with what has been used for
North Sea structures. Concurrent with the initial fatigue evaluation
studies, an analysis of the non-hurricane wave cli-
These allowable stresses were derived from mate data collected during the Ocean Data Gathering
preliminary generic fatigue analyses and used in Program] (ODGP) at station South Pass-62 was per-
proportioning tubular joints, principally in formed by Evans & Hamilton. l Of the ODGP stations,
selecting joint can thickness, and occasionally South Pass 62 is the closest to the Cognac site and
in modifying brace stub end dimensions. the wave environments at the two locations are gen-
erally expected to be similar. Based on this pren~
A post-processor to Shell's static wave ise, the directional wave scatter diagrams developed
force computer program was used to apply these from the South Pass 62 data were used for all further
allowable stresses in detailed design checks Cognac fatigue studies.
of each tubular joint, incorporating criteria
as generally described in References 3 and 4. Compared to the earlier wave climate basis of
Table 2, the ODGP-SP62A data showed an 8.6% higher
Generic Fatigue Analysis annual energy level, and - more significantly -
a shift toward shorter periods.
A gene~ic fatigue analysis is one which can
be performed early in the design process, when the When the South Pass 62 data analysis became
general structural geometry and wave climate are available, a second round of preliminary fatigue
known, but not with sufficient detail to permit analyses was carried out. Considering broadside
an exhaustive hot spot by hot spot final analysis. waves only, TOWER analyses were run for each of the
Several cycles of such analyses were performed as 58 sea states in the annual wave scatter diagram,
data became available and the changing design shown in Figure 1. End-on waves were not considered
evolved. since TOWER can only analyze planar structures, and
previous work suggested that end-on waves were less
The basic assumption is that, for typical critical for Cognac fatigue behavior than broadside
braces, influence coefficients can be found (or waves. From the TOWER analyses, values of dynamic
specified) relating local stress to gross structural base shear and average response period were computed
response, such as base shear or overturning moment. for each of the 58 sea states, with dynamic am-
For braces at ot near the waterline, local stress plifications reaching as high as 10 for the 3 to
is related to a linear combination of gross response 4 second waves. Making the assumption that member
and local wave pressure. stresses are proportional to base shear and using
the AWS-X-Modified fatigue curve resulted in a
Several trial designs are assumed', each with a predicted fatigue life of only 2 years, based
specified local stress for the design wave load, on the previously established maximum design wave
corresponding to a given influence coefficient or peak response hot spot stress of 46 ksi.
sensitivity (e.g. ksi hotspot stress per kip base
shear). For each trial design, a simplified dy- The Final Round
namic structural analysis and cumulative fatigue
damage calculation is performed. From the results, A fatigue life of 2 years was clearly unaccepta-
a trend of fatigue life versus specified local ble! Our attention, therefore, focused first on the
stress is established. The local stress values drawbacks of the TOWER-based analysis and on how best
giving adequate fatigue life are then imposed as to compensate for these. Recall that the analysis
the allowable stresses, forcing the final structure included broadside waves only. Thus, wave direction-

170
ality was not considered. Also, directional spread- shown in Fig. 4(a),·broadside waves with periods
ing of the wave energy was not included. In order close to the natural period of the structure would
to incorporate these normally beneficial effects crest simultaneously at the Row 'A' and Row 'B'
into the the analysis and determine where we stood legs. To make matters worse, half the conductors
before pressing the "panic button", Marshal18 were located near Row 'A' and half near Row 'B'.
developed a spectral dynamic fatigue analysis pro- One possibility of relieving the situation and de-
gram including directional spreading and direction- tuning the structure to some extent was to move
ality. The wave force model consists of a three the conductors from the edges of the structure to
dimensional array of vertical cylinders (totem the middle, as shown in Fig. 4(b). This was tried
poles), calibrated against applied force transfer and the results, shown in Table 3, indicated a
functions derived from the full 3-D frame. The four-fold improvement in predicted fatigue life.
program is designed for preliminary fatigue studies, The design of the structure was already at such
and is particularly useful for quick parameter an advanced stage that changing the locations of_
studies to investigate the effects of variations the legs at the waterline would have caused a
in damping, directional spreading, etc. The pro- lengthy delay. Hence, a change in the conductor
gram, however, requires as input some quantities locations to the locations shoivu in Fig. 4(b) was
(such as participation factors) which require a suggested. Since the change did not create any
previous dynamic analysis. For Cognac, the TOWER fabrication or operational problems, it was adopted.
dynamic modes served as a basis for the input.
Marshall's program was further calibrated by As shown in Table 3, a parameter study was
matching TOWER base shear (within 3%) for uni- carried out to investigate the effects of changes
directional broadside waves, and by performing in the damping value, the natural period of the
fatigue hindcasts on existing structures. structure and the fatigue curve. As expected, a
change in the fatigue curve from the AWS-X-Modified
Like many other preliminary fatigue analysis curve to the curve proposed by the British Welding
programs, Marshall's program makes the assump- Institute (BS 153 Class F,9 with a further safe-
tion that local member stresses can be related ty factor of 3), resulted in a drastic reduction
to an overall response quantity such as dy- in fatigue life. The AWS-X-Modified curve, however,
namic base shear or overturning moment. This is considered appropriate for Gulf of Mexico
type of analysis is adequate for preliminary structures fabricated with particular attention to
design purposes only. weld profiles which merge smoothly with the adjoining
base metal. The effect of increasing or decreasing
Using Marshall's program, with a simultaneous the natural period by 0.5 seconds resulted in an
pruning of the waterline member volume to 74 per- increase of almost 150 percent in the predicted fa-
cent of the original volume, resulted in a pre- tigue life. This, of course, is a result of shifting
dicted fatigue life of 9 years for_broadside waves the natural period away from the transfer function
only with zero spreading. The reduction in water- peak. The effect of reducing the damping value from
line member volume was achieved by reducing 3.5 to 2.0 percent resulted in a four-fold reduction
the diameter of the conductors, by reducing in fatigue life. Studies by Ruh1 2 were already be-
the barge fender and boat landing sizes and ginning to indicate that damping values for platform
by a careful review of the simulated wave response in low sea states could well be as low as
force members at the waterline. Since fa- 2 percent.
tigue wave forces are predominantly inertial, the
volume displaced by the members near the waterline In summary, the preliminary fatigue analysis
is a controlling parameter. showed that using the new conductor locations, with
a reasonable estimate of wave spreading (16 degrees
Considering the directionality of the waves, RMS) and damping (2%) and the AWS-X-Modified fatigue
the predicted fatigue life improved from 9 years curve, we could expect a fatigue life of 161 years,
to 41 years. With 32 degrees RMS wave spreading with the joints designed on the basis of an allowable
included, there was a further big jump to 405 hot spot stress of 46 ksi for the design wave.
years. These results suggest that the extreme This being well in the range of desired fatigue
fatigue sensitivity was limited to perfectly life, we-proceeded with the final design; followed
broadside waves. by the detailed member-by-member check of the final
design as described below.
We were, however, still searching for addi-
tional ways to improve the predicted fatigue life DETAILED FATIGUE ANALYSIS
and reduce the directional sensitivity. Examina-
tion of the base shear transfer functions generated An overall block diagram of the probabilistic
for broadside and end-on waves, shown in Figs. 2 dynamic analysis system used for the detailed fa-
and 3, revealed that for broadside waves the struc- tigue analysis of the Cognac platform is shown in
ture natural period of 3.9 seconds (obtained from Fig. 5. The following discussion of the analysis
the TOWER analysis using soil springs suitable for includes a brief description of the procedure,
normal waves) coincides almost exactly with a peak the changes made in the existing computer programs
in the transfer function. For end-on waves, to enable the analysis, and the development of a
on the other hand, the structure natural period new computer program for the evaluation of member
was at a valley in the transfer function. This end stress concentration factors. This is fol-
explains why the structure is so sensitive to lowed by a discussion of the data used for the
broadside waves and not to end-on waves. The analysis and by a discussion of the results.
shape of the static transfer functions is con-
trolled by the distribution of the members at Brief Description of Procedure
the waterline, and depends mainly on the spacing
of the legs and the conductor locations. Due The detailed fatigue analysis is based on
to the dimensions of the structure at the waterline, the probabilistic procedure described by Vugts

171
and KinralO. Using a dynamic time-step structural stress range ui and Ni is the permissible number
analysis, as described below, the transfer functions of occurrences of stress range ui. The permissible
for hot spot stresses at four points around each number of occurrences is defined by a suitable
member end in the structure are generated as a fatigue curve - e.g. the AWS-X-Modified S-N curve.
function of wave frequency and direction. The wave
environment is described by directional wave Dynamic Time-Step Structural Analysis ~ .
scatter diagrams developed from the ODGP datal.
The wave data were analyzed and grouped into Since fatigue damage is a highly localized phe-
one ft. increments of significant wave height, Hs~ nomenon, it is important to obtain an accurate and
0.5 second increments of mean zero-crossingperiod, detailed local stress analysis. In previous ap-
Tm, and 8 compass directions. Hs and Tm were ob- plications of the probabilistic fatigue analysis
tained directly from analysis of the wave staff method10 to structures having first mode na~ural
records, while direction had to be inferred from periods of 2.5 seconds or less, dynamic amplification
wind data. The resulting collection of 285 indivi- was neglected and a static stress analysis computer
dual sea states together represent the wave cli- program was applied in a quasi-static.stepwise mode
mate. to obtain the required detailed stress response to
waves.
For each sea state in the wave scatter dia-
grams, the wave spectrum SL(U,8), is represented For Cognac, a dynamic structural analysis was
by imperative. Shell’s existing regular wave time-do-
main dynamic analysis computer program, DYNAL, is a
Sc(u,e) =fn (0) . SC(M).................(l) modal analysis program. The Rayleigh-Ritz procedure
is used to reduce the complete three-dimensional
where SC(U) is the Pierson-Moskowitz frequency spec- structure with six degrees of freedom at each node
trum for fully developed seas and fn(6) is a di- to a maximum of 39 degrees of freedom. Application
rectional spreading function given by of dynamic loads and time-step integration of the
equations of motion is done in terms of these dy-
fn(e) = Cncosne. .....(2)
namic degrees of freedom (or modes). At designated
7/2
time steps, the modal response is expanded into a
with Cn such that ~fn(o)de = 1 ............(2a) full displacement vector, from which detailed member
-VI2 forces and stresses can be extracted.

Note that n = 2 in Eq. 2 is approximately equiva- Unfortunately, however, while the DYNAL approach
lent to 32 degreees RMS wave spreading and n = 10 adequately reflects gross platform response in a way
is approximately equivalent to 16 degrees RMS wave that is fully compatible with local stress-strain
spreading in the spreading function used in Mar- behavior, localized response modes may be constrained
shall’s preliminary analysis. out unless they are anticipated by the user and a
corresponding force pattern is specified. Near the
For each stress location of interest, the waterline, where significant dir~ct wave action oc–
response spectrum, SD (w), is computed for each curs, such local response modes have led to fatigue
aea state in the wave scatter diagrams from failures. Other types of local response could in-
clude prying (or breathing), out-of-plane bending,
j{:t(w,ew’ SC(u,Q)do ...........(3) and drumming of horizontal framing and non-structural
SO(6J)=
appurtenances. These are automatically considered
-rj2 in a complete static solution, which includes all six
degrees of freedom at each node.
where ft(~,o) is a linear transfer function repre-
senting the atress which results at the given loca-
A more accurate analysis, especially for-water-
tion, in ksi per foot of wave .amplitude due to
line members is obtained using the static-plus-imer-
wave energy having radian frequency w and direction
tial approachl~. The dynamic structural response is
e. The response spectrum is used to generate short-
separated into static and inertial parts as follows:
term stress statistics applying a modified Rayleigh
distributionl-”which considers spectral width
in addition to RMS response amplitude, ~[where {x(t)} ‘{x(t) }static+ [+l~y(t)-ys(t)}“””””””””””-”(5)
m. is the area under the spectral curve S.(0)] ,
where {x(t)} is the desired time-step displacement
and mean zero-crossing period of the atresa re-
response,
sponse, Tma[which is usually a composite of
the wave period Tm and the structure natural
{X(t)lstatic is the complete static solution at
period]. When spectral width is neglected, this output time step t,
reduces to the well known Rayleigh distribu-
tion. The short-term stress statistics for each [$1 is an array..containingthe dynamic
of the 285 sea states”in the.wave scatter diagrams mode shapes, normalized such that
are accumulated to obtain the long-term (annual) [$]T[M][$1=1.
stress statistics. Finally, fatigue life esti-
mates are obtained using Miner’s law of linear {y(t)} contains the modal amplitudes of the
accumulation of damage, which statea that the cal- dynamic response as computed by time-
culated fatigue life is exhausted when step integration, and

{yS(t)} contains the modal amplitude represen-

~$ = 1.0 ...........................(4) tation of the static solution at time
i step t.

where ni is the actual number of occurrences of

.—-
1/2
This is obtained by solving Table 4 shows some of.the stress concentration
factor formulas used for the various types of joints
h:J{y5(t)} = [(j]
T{F(t)} .:...........(6) in the Cognac platform. Stress concentration factors
are given for K-joints, Y-joints, unreinforced cross
in which[’mr~] contains the modal frequencies, joints and for grouted joints. These factors vary
and with the type of connection, the geometry of the
connection, and the relative sizes of the brace “and
{F(t)} is the applied wave force on
chord members.
the nodes at time t.

The member end stresses are recovered from {x(t)} Several comments need to be made regarding
using a-stress recovery algorithm which includes these stress concentration factors. The formulas
the four fixed-end moments for each member due to presented in Table 4 are those in use three years
direct local wave forces. ago, at the time Cognac was being designed. These
were based on in-house parameter studies using
Note that for higher modes (say, *periods Greste’s finite element program for K-joints, as
much shorter than 2.5 seconds) subjected to wave described in reference 3, and formulas consistent
loading (periods longer than 2.5 seconds), y(t) with AWS curve T punching shear criteria for T and
Y joints. It was recognized that cross joints pre-
approaches ys(t) and their full.effect is in- -.
eluded in the static solution. Therefore, only sent yet a more severe local stress situation. Some-
the first few modes need be included in the what mo e comprehensive results are available to-
day12,1S .
inertial term, reducing the chances for numeri-
cal errors to accumulate.
However, thin shell finite element analysis
The”DYNAL program was modified to include tends to systematically overestimate the brace end
the static–plus-inertial approach described SCF. In reference 3, a reduction factor Qr was
above and the modified program, named DYNFAT, introduced to account for this. Figure 6 comparea
was used for the Cognac fatigue analysia. Addi- the theoretical midplane intersection stress given
tional changes were made in DYNFAT to start runs by thin shell analysis, with the hot spot stress
from previously saved mode shapes and frequencies, adjacent to the toe of the weld in the brace end
to write the data on magnetic tapes in the format (consistent with AWS fatigue curve X) given by
required for program FATIGUE, and to loop on the experimental strain gage readings or by more ap-
value o.fdamping so that analyses for two or more propriate thick shell isoparametric finite element
analysis. It appears that some adjustment to thin
damping values could be obtained in the same run.
shell results is mandatory, either using the ex-
Program FATIGUE Modifications ponential decay factor Qr proposed originally, or
the simple 5/8 factor shown in the figure.
For use in the Cognac analysis, Shell’s
existing probabilistic fatigue analysis pro- Table 4 also suggests a treatment for grouted
gram, FATIGUEIOj which generates hot spot stress joints, based on the notf.onthat the grout disbonds,
transfer functions and computes fatigue lives and acts only as a spacer between joint can and pile;
while punching shear still must be carried entirely
using spectral analysis, was modified to
include wave spreading and the use of direc- by the outer joint can, the stiffening effect shows
tional scatter diagrams. The ability to up in the Y ratio.
use different scatter diagrams for different
wave approach angles was implemented to facili- Special tubular joint studies carried out in
tate the use of the SP 62 A wave scatter diagrams. the course of desi ning Cognac include isoparamet-
The original FATIGUE program was restricted to ric finite element ~4 (Fig.7) and experimental (Fig.8)
analyses of specific complex Y-joints near the
the use of a single scatter diagram with dif-
ferent probabilities of occurrence associated outrigger skirt piles; and studies of the effect
with each wave direction applying to the entire of grouted “big pins” inside the main jacket legs
scatter diagram. using a common K-joint geometry using both analy-
sis and experiment (See Figs. 9 and 10). The ex-
perimental results appear to be most consistent with
Development of Stress Concentration Factor Program
the finite element “sandwich” analysis in which
To assist in the onerous task of evaluating extensive areas of grout disbanding were specified
axial, in-plane bending and out-of-plane bending a priori. For braces loaded in compression,grouting
stress concentration —..=
factors yields significantly lower local stresses.
.— (SCF’S) for each mem-
ber end—in the structure, a computer-program was
developed implementing our current (1976) thinking A subsequent check of criteria developed for
on the subject. The SCF program evaluates empirics Cognac, against a wider data base can be seen in
hot spot stress concentration factors which recog- Figure 11. Using the SCF formulae of Table 4
nize separately the effects of axial load, in-plane together with curve “X” fatigue criteria yields
ns of actual fatigue tests
bending and out-of-plane bending in the branch mem- conservative Predicti?
ber at a tubular connection. While it would be ap- in the WRC data base , with bias and scatter
propriate to simultaneously consider the influence comparable to the original AWS correlation
of chord member stresses in the joint can, neither using measured hot spot strains.
the existing format of theempirical stress concen-
tration factors, nor the one-member-at-a-time logic Discussion of Data Used for Analysis
of the statistical bookkeeping can accommodate this
Microscale stress concentration effects at the toe The following discussion briefly describes
of the weld are included in the fatigue criteria. the evaluation and choice of important parameters

173
for the fatigue analysis. The importance of this erated for wave directions 1 through 8 were also
phase of “getting ready for the analysis” cannot used for directions 9 through 16 assuming for-
be overemphasized. The final fatigue lives wards-backwards symmetry of linear response.
computed after what appears during the analysia Thus, the fatigue analysis actually considered
to be an endless series of computer runs, depend a total of 16 wave directions at 22.5 degree
to a great extent on the data set used to start increments around the platform. The large number
the stress analysis. of increments used was appropriate to the demon-
strated directional sensitivity of the Cognac
The structural model of the platform for structure.
fatigue analysis consisted of 2500 members and
800 joints. The modeling of the structure in- For each wave direction, the structure was
eluded a careful review of the waterline members. analyzed for 12 different wave periods. The 12
The wave force members near the waterline were wave periods used are shown in Table 5. These
modeled accurately with regard to size and range from 12.5 sec. to 2.5 sec., corresponding
location. The conductors were treated as 16 to a frequency range of 0.5 to 2.5 radians. The
groups of four conductors each. choice of wave periods was based on the shapea of
the broadside and end-on transfer functions, in
The soil springs representing pile stiff- order to represent all the peaks and valleys.
nesses used for extreme wave analysis are gen-
erally not appropriate for fatigue analysis, The wave heights associated with each wave peri-
because the pile load-deflection curve ia non- od were computed on the basis of a wave steepness “of
linear. Hence, the appropriate spring stiff- 1/20, with a maximum height of 14 ft., equal to the
nesses for the low load levels associated with maximum significant wave height in the SE’62 A wave
fatigue waves are generally much greater than scatter diagrams. The choice of a wave steepness of
the spring stiffnesses appropriate for extreme 1/20 was based on the dashed line plotted in Fig. 1,
wave loads. The Cognac fatigue analysis soil which shows that a 1/20 wave steepness is fairly
springs were evaluated using a dynamic baae representative for everyday waves in the ODGP data.
shear of 4000 Icipa,which ia about 30% of de- This choice of wave height is significant in that
sign but somewhat greater than the maximum shear it determines the point at which non-linear effects
for any of the fatigue waves. Between load levels of drag and immersion are linearized.
of zero and 4000 kips,the springs remain essen-
tially linear. Therefore, the same set of Steady-state values of dynamic axial and bending
springs is appropriate for all the fatigue wavea. stresses computed by I)YNFATfor every tubular struc-
tural member in the jacket were saved for 12 wave
In order to ensure a conservative analysis positiona corresponding to-wave steps at L/12 incre-
and allow for extreme variations in the founda- ments in a complete wave cycle. In the FATIGUE pro-
tion atiffnesa, the structure stiffness and the gram, the stress data saved for the 12 wave positions
dynamic mass of the structure, the first natural are interpolated to obtain the stress cycle from
periods of the structure for broadaide and end-on which the stress range is determined. The transfer
motions were adjusted to match the peaks of the function ordinate is then.defined as stress range
corresponding base shear transfer functions. divided by wave height.
This was accomplished by modifying the deck masa
for the desired result. The structure waa analyzed for 2 percent
and 4 percent damping, which were considered
The wave kinematics were computed using linear sufficient to bracket the possible range of damping
Airy theory considering the full water depth. values. The stresses for each damping value were
This is consistent with linearity assumptions in stored on separate DYNFAT output tapes for each
spectral analyais. wave direction.

Morison’s equation was used for the evaluation RESULTS OF DETAILED ANALYSIS
of wave forces, with the drag coefficient, Gd, and
the inertia coefficient, Cm, assumed to be 0.6 and Using the DYNl?ATproduced member stress
2.0, respectively. These values are considered ap- data, reordered into the required format by pro-
propriate for fatigue analysis. .No elaborate wave gram PREJ?AT,fatigue life estimates for both enda
force calibration such aa described in Reference 10 of approximately 1400 members were obtained using
waa performed to determine the drag coefficient. the FATIGUE program; The eight CIDGPwave scatter
This was because, for the non-hurricane waves used diagrams at 45 degree increments were interpolated
in the fatigue analysis, drag forces contribute leas to obtain 16 scatter diagrams at 22.5 degree incre-
than 3Z of the total cyclic range of base shear, ments, and the resulting scatter diagrams were used
as indicated by the position of these aea states in for wave directions 1 through 16, as discussed
Figure 12. earlier. Care must be taken to match each wave
direction with the appropriate wave scatter diagram,
The fact that the nonlinear drag force term ia keeping the orientation of the structure in mind.
almost negligible eliminates a principal difficulty The Cognac structure is oriented such that zero
with linearized spectral analysis for this applica- degree end-on waves approach the structure from
tion. the SW direction.

The structure was analyzed using DYNFAT for Basic Fatigue Lives
eight different wave approach angles at 22.5 de-
gree increments aa shown in Fig. 13. ‘The struc- The first set of member fatigue lives,
ture was analyzed for w“avesapproaching from hereafter referred to as “basic fatigue livea”,
direction 1 (= 0.0 degrees) through direction 8 was computed using member stress data saved by
(= 157.5 degrees). The transfer functions gen- DYNFAT for 2 percent damping. Appropriate member
.-.
I{4
end stress concentration factors were evaluted for Effect of Wave Sprexding
each member end using the SCF program. Zero wave
spreading was assumed, and the AWS-X-Modified The effect of including wave spreading waa in-
S-N curve was used. vestigat,edby reanalyzing the skirt pile bracing mem-
bers with basic fatigue lives less than 1000 years.
The results of the analysis show that,of the The results, shown in Table 6, were obtained for 2
2800 member ends analyzed,only 181 member ends have percent damping, using the AWS-X-Modified curve and
predicted fatigue lives less than 1000 years. For assuming the grout to be ineffective. The results
these 181 member ends, the predicted fatigue lives show that 16 degrees RMS wave spreading increases
are distributed as follows: the fatigue lives by approximately 50 percent,
and 32 degrees RMS wave spreading increases the
Fatigue Life, years No. of Members Ends fatigue lives by approximately 100 percent. These
percentage increases are much smaller than those
o-1oo 8- obtained in the preliminary analysis. The reasons
100-200 21 for the difference are still unclear.
200-300 27
300-500 42 Effect of Damping
500-1000 83
The effect of an increase in the damping from_
Setting a minimum required fatigue life of 2% to 4,%was investigated by reanalyzing 120 mem-
100 years for zero wave spreading and 2% damping, ber ends with low basic fatigue lives. The results
the above table showa that only 8 member ends show a wide variation in the predicted fatigue life
required attention to improve their predicted increase for 4% damping. The ratio of fatigue
fatigue lives. Of these, 4 member ends were lives for 4% and 2% damping varies from a mini-
at cross.joints in the midsection. The joints mum of 1:08 to a maximum of 6.67. In general,
were assumed to be unreinforced in computing the it is found that the members affected by broad-
basic fatigue lives. Since the midsection fa- side waves have lower increases than those af-
brication was yet to start, the fatigue lives of fected by end-on waves. The reasons are very
these 4 member ends were easily improved by rein- likely related to the adjustment of the struc-
forcing the joints with internal diaphragms. ture natural periods to match the peaks of the
The other four member ends with basic fatigue broadside and end-on base shear transfer func-
lives under 100 years were in the base section tions, and the presence of more wave energy in
at skirt pile joints in Rows 1 and 4. Note that the sea at the higher end-on natural period.
the fatigue lives for these member ends are based
on SCF’S computed using Y-branch SCF formulas, This study reaffirms the great influence
since the horizontal members at the joints are of damping on predicted fatigue lives for struc-
very low-stressed and, therefore, provide negli- tures with relatively high natural periods, such
gible support. Since these member ends were in as Cognac, in wave environments in which the
the base section, at joints that were already fa- periods of normal everyday waves are close to
bricated, it was elected to improve their fattgue the natural periods of the structure. The
lives by improving the weld profile. These welds influence of damping reduces for structures with
were built up such that the effective thickness low natural periods at which there is little
of the joint was increased by 0.25t, where t is wave energy in the sea.
the brace thickness. This reduced the stress to
the point where calculated fatigue lives in ex- CONCLUSIONS
cess of 100 years were indicated.
An extensive preliminary fatigue analysis
The modifications described above were sub- served as a basis for the fatigue design of the
jetted to a cost-benefit analysis, and found to Cognac platform. This was followed by a detailed
be justified in terms of reduced risk of future member-by-member check of the designed structure
underwater repairs. using probabilistic techniques. Fatigue life
estimates were computed for approximately 1400
Effect of Grout members in the structure. The lowest predicted
fatigue life obtained using 2 percent damping,
In computing the basic fatigue lives given zero wave spreading and the AWS-X-Modified S-N
above, the stress concentration factors for members curve was 52 years. Suitable structure modifica-
at grouted joints were computed assuming that the tions (addition of internal diaphragms and the
grout remains effective as a spacer, even in the use of built-up weld profiles) were made to
event of diabonding. In case the grout becomes increase the minimum predicted fatigue life to
completely ineffective, higher stress concentration over 100 years. This represents a factor of
factors will-result. In order to investigate the safety of 4 on the expected useful life of the
effect on predicted fatigue lives, the braces at platform.
grouted joints with basic fatigue lives less than
1000 years were reanalyzed assuming that the grout NOMENCLATURE
becomes ineffective. In general, the 52 members
reanalyzed are at the interfaces between the top and Cd = drag coefficient
midsection, the mid and base section and at the bot- ~
= inertia coefficient
tom level. All 52 members were treated as Y-branch
members. The results obtained show that the ratio 5: = damage rate for sea state i
of fatigue lives with ineffective grout and with
fn(e) = directional spreading function of
effective grout varies from 1.01 to 1.72.
wave energy around a mean direction

.-l?
 ft(u,e) = linear stress transfer function Design of Tubular Connections”, Proceedings
of the First International Conference on the
F(t) = applied wave force vector at time t Behavior of Offshore Structures, Trondlieim,
H~ . significant wave height Norway, 1976.
L = wave length
4. Marshall, P.W., “A Review of Stress Concen-
. actual number of occurrences of
‘i tration Factors in Tubular Connections”,
stress range a~ Shell CE-32 Report, April 1978.
Ni . permissible number of occurrences
5. Penzien, J., Kaul, M.K. and Berge, B., “Stochas-
of stress range Oi
tic Response of Offshore Towers to Random Seas
Q, . stress reduction factor and Strong Motion Earthquakes”, Computers and
s~(u) . stress response spectrum as a func- Structures, Volume 2, Pergamon Press, New York,
tion of wave frequency 1972.

s~(fd) . uni-directional wave spectrum as

6. Marshall, P.W., “Basic Considerations for Tubu-
a function of wave frequency lar Joint Design in Offshore Construction”,
s~(u,q . directional wave spectruxn WRC Bulletin 193, April 1974.

Tm . mean zero-crossing wave period 7. Ward, E.G., “Ocean Data Gathering Program: An
Tmcr . mean. zero-crossing period of stress Overview”, Offshore Technology Conference Pro-
response ceedings, OTC 2108B, 1974.
Tn . natural pertod of structure
8. Marshall, P.W., “Dynamic and Fatigue Analysis
x(t) . dynamic displacement vector at time t Using Directional Spectra”, Offshore Technology
X(t)static= static displacement vector at time t Conference Proceedings, OTC 2537, 1976.

y(t) = modal amplitude vector of dynamic 9. Gurney, T.R. and Maddox, S.J., “A Reanalysis
response at time t of Fatigue Data for Welded Joints in Steel”,
The Welding Institute, Research Report
Ys(t) = modal amplitude vector of static
E/44/72, Cambridge, England, Jan. 1972.
response at time t

‘t . ratio of chord radius/thickness 10. Vugts, J.H. and Kinra, R.K., “Probabilistic Fa-
tigue Analysis of Fixed Offshore Structures”,
[$1 = mode shape array
Offshore Technology Conference Proceedings, OTC
[’or.] = modal frequencies 2608, 1976.

ACKNOWLEDGMENTS . ._._ ______ _ .. ... ... .... .. 11. Marshall, P.W. and Kinra, R.K., “Dynamic and
Fatigue Analysis for Deepwater Fixed Platforms”,
The writers are indebted to R.&f.Barrington Eo: Second Annual ASCE-EMO Specialty Conference;
performing the TOWER dynamic response studies used Raleigh, N.C., May 1977.
in the preliminary fatigue assessment; to J.H. Vugt
for providing the extensive program modifications 12. Kuang, J.G., Potvin, A.B., et al, “Stress
to FATIGUE required to consider wave spreading; and Concentration in Tubular Joints”, SPE
to PMB Systems Engineering and Southwest Research Journal, August 1977.
Institute for tubular joint analyses.
13. Smedley, G.P., “Peak Strains at Tubular Joints”
REFERENCES . presented at the Inst. of Mech. Engrs. Seminar
..-.
—......
on Corrosion Fatigue in Offshore Installations,
1. Hamilton, R.C., “Analysis of ODGP Data from September 1977.
Station No. 1 - SP 62 A,” Evans-Hamilton, Inc.
Feb. 1975. 14. Reimer, R.B., et al, “Improved Finite Elements
for Analysis of Welded Tubular Joints”, Offshor~
2. Ruhl, J.A., “Measurements of Response and Technology Conference Proceedings, OTC 2642,197e
Damping of Deepwater Platforms,” paper No. 911
1978 Offshore Engineering Conference, Shell 15. Rodabaugh, E.V., “Review of Data Relevant to th<
Oil.Company, Houston, Texas, Feb. 1978. Design of Tubular Joints for use in Fixed Off-
shore Platforms”, Battelle draft report for WRC,
3. Marshall, P.W. and Graff, W.J., “Limit State January 1978.

176
 TABLE 1

EXAMPLES OF ALLOWAISLEFATIGUE STRESSES KSI

T~

DESIGN NOMINAL HOTSPOT PUNCHING SHEAR SENSITIVITY

LOCATION WAVE fa + fb AWS-X K
—— T&Y ~ ksilft.
Gulf of Mexico 60 ft. 20 60 9 5 4 1.0
T<3 sec.

Gulf of Mexico 75 ft. 18 46 7 4 3 0.6 I

North Sea 100ft.

Waterline 34
Elsswhere ::* 58 a 0.6

South California 49 ft.

Waterline** 15 5 0.7
Elsewhere 19 i; 7.5 5 1.0

1 ft. = 0.3m fa = axial stress

1 ksf = 7 N/mm2 fb = bending stress

* using high strength steel and thickened stub ends

** includjng effect of 6 inches (150mm] Irarine9rwth

TA3LE 2
——

SAMPLE PRELIMINARY FATIGUE CALCULATION

Tf BASE SHEAR HOT SPOT ALLOWABLE OAt.lAGE YEARLY

Z;;o Hours Pe~ Signif. STRESS CYCLES — RATE —DAMAGE
SEA STATE KS Tm Crossings Per Dynamic % of N.
‘TR[@
i feet sec. Per Hour J&&r +Statj~ Design cycIes hi ‘f IJiTi

a 40 12 462 .01 1.22 54 45 1.7X107 .027 nil

b 29 10 507 1.0 33 27 1.4X105 .0034 .003

c 26 10 529 3.4 27 22 4X105 .00133 .004

d 21 9 571 4.8 20 17 ,.6 .00057 .003

e 16 8 621 34 16 13 3X106 .00021 .007

f 11 7 679 240 9.7 8.0 2X107 .000034 .0G3

9 7 7“ 679 408 4.9 4.1 4X108 .000002 .001

h 7 5 818 408 11.1 9.2 1.5X107 .000054 .022

i 4 6 750 1540 3.7 3.1 1.5X109 .000001 .001

j 4 4 857 1540 5.8 4.8 2.2X108 .000004 .006

k 2 5 818 1540 1.8 1.5 3X101O nil nil

1 2 3-4 878 3070 2.1 1.7 2X101O nil nil

= total hot spot stress range For Design Wave (75 ft.): Zfii Ti = ‘.055
‘TR
= peak hot spot stress used as Ful1 Cycle OTR= 83 ksi
*HS Life = 18 yrs.
design allowable Peak OH5= 58 ksi
For Design Sea State (a) Spectrum:
Significant UTR = 45 ksi
Prob. Max. = 83 ksi
‘TR
 TABLE 3

PRELIMINARY ANALYSIS PARAMETER STUDY RESULTS

Shear [Iirrction Natural Fatigue

Stresses Are Wave wave Percent Period. Fatigue Life,
Description of Structure Related To Oircction Spreadinq Ikaapinq Sec. Curve yea rs

Old conductor locations IJroadside Rroads id? 7Qro 3.5 3.9 AWS-X-Modified 9

Waterline vol .=74% of

,,
original vol. Broads ide Directional Zero 3.5 3.9 41

(New conductor s f.?es. no

,,
barge bunpers ) Resultant Directional 32” RMS 3.5 3.9 405

New Conductor 1ocat ions Broadside Broadside Zero 3.5 3.9 AWS-X-!kdi fied 31

Water-line vol. s 74% of

,,
original vol. End-on Directional Zero 3.5 3.9 14087

“
Broadside Directional Zero 3.5 3.9 135

,,
Broadside Directional 32” RMS 3.5 3.9 1579

Resultant Directional Zero 3.5 3.9 ANS-X-h’mdiFied 116

,, ,, ,,
16° RMS 3.5 3.9 613

,, ,, 16“ RMS 3.5 3.9 B.S. ’F’ 17

New conductor locations

,! ,, Zerm 2.0 3.9 AWS-X-Modified 2B

Waterline vol. w 77% of
original vol. ,,
,, ,, 2.0 3.9 161
16“ RMS
(New conductor sizes, .
t, ,, Zero 3.5 3.4 334
4 snnl 1 barge bumpers)

,, ,, ,!
Zero 3.5 4.4 i’8B

,, t! . 1BB6
16” RMS 3.5 3.4

,, ,t ,,
16° RMS 3.5 4.4 805

TABLE 4

SCF FORMULAS

AXIAL LOAD IN-PLANE BENDING OUT-OF-PLANE BENDING

I

r, K 1.8TsinGfi

14 T Sin O for 7 ~ 25

T&Y { 1.5 T sin EI’Y0”7 for 7>25

OR KUANG FORMULA FOR LARGE B

18.6 TSin(3+~fOr7~25

x { 2.0 T sin 137 0“7+~for T > 25 1.6-r sinefi 3.6T sin @@

OR SNEDLEY FORMULA FOR LARGE 8

u.!
v
~

g
m
m
ALL ~ ‘~(scF)c,ord)]
[l~oUr(l but
nOt
less‘ban18+s‘r 125

w I-M +t
T = t/T FOR GROUTED JOINTS USE

= ts/t -r= D/2Teff

‘eff
T3 + TP3
T- but not more than 2T
eff =
r T
7= (D-T)/:!T
-.———. _-— —
T remains as before
Qr=ex’o(-&ts)
 TABI_E 5 TABL.E 6

WAVES U! D TO GEN lATE

STRESS TI NSFER FU ~NS EFFECT OF WAVE SPREADING —
dave Frequenc~ Iave Period ‘aveHefgh Fatigue life, a rs

4=
ltembe
radians sec ft ocation ~ro
Numb@ RNs 32° RMS
0.5

0.75
12.57

8.38
14.0

14.0
1836

1838
Row A

Row A
S readtn

348

360
““ )
n=10 in Eq.2

6B5

620
n=2 in Eq.2)

791

717
1.0 6.28 10.11
1860 Row B 313 547 638
1.15 5.46 7.64
1862 Row B 306 572 664
1.30 4.83 5.98
1874 Row 1 366 487 751
1.45 4.33 4.81
1884 Row 1 64 86 126
1.60 3.93 3.95
1886 Row 1 53 69 102
1.75 3.59 3.30
1895 Row 1 403 533 B22
1.90 3.31 2.80
1898 Row 4 361 475 732
2.05 3.07 2.41
1908 Row 4 54 71 104
2.25 2.79 2.00
1909 Row 4 65 B4 125
2.50 2.51 1.62
1919 Row 4 426 567 872

NtlRH!7LIZED DISTF!lBUTICN OF

L13CRTI13N- ‘5P13ZR
TUTRL NW?EE!3 W NC!RWLIZEC! LICCIJREhCfS - 978

‘ 4
1/’
I 3, z.
3 3/ 1

/
5 Ii 7/S2

2 16 ~ /15 6 2
/
13 S2 /U2 16 7 2

E
3- ~ 6 ,./4 32 ,

2 UU,4 WI Is 6

El
&

q
1
=iLrm 1.UO
J

6.W3 7.00
+
8.00
MEfli”yEFiW3%%NG &%O [S~#N051

Fig. 1 - Annual SP62 A wave scatter diagram.

 m 50(
n &STR. NAT. PERIOO
!2 [Tn * 3.9SEC.)
. I
u
n
3
I
40(
l-r I
-1
n I
E
<
w
I I
I I
z 30C
x

&
o
L
I ‘k
I
OyNAM’c
I
m
Id
n
20C I I
LJJ
v
I \
z /
c
L
Ioc ——- /’
L3L
0 0.5

12.57
I.0

6.28 4.19
I.5
4ZC 2.0

3.14
2.5

2.51
w,RAOJSEC
T SEC

Fig. 2 - Preliminary analysis broadside shear force transfer function.

Ln
L STR. NAT. PERIOD
= 2oop [Tn =3.9 SEC.)
t-
.
g
=
!-
l~\
/,
J
a I50 - /\
x
<
l\
5
l\
s
I
!- 100 -
0
0
L
-\
=
UJ
n
w
v 50 -

z
L
STATIC

1
0 0.5 I.0 1.5 2.0 2.5 w .RAO./SEC.
12.!57 6.2.9 4.19 3.14 2.51 ‘T SEC,

Fig. 3 - Preliminary analysis end-on shear force transfer function.

 Y

n Y
ROW BO
VO *****O
O.. ..*O
O.. .**O
Ob . . ..o
L
0
n n
Q
X? b
t-
y:::::
0 ***9*0
::7
h
●
● ****a*
-—
*******
. – . L. - ● *e e***
.—.
● bob bboo x
fro. ..o. o ● **e**** @*e *e***
ROW A~ . I ● ******* ● ****m**
1

o
A I n
b
52.28’ 60’ 52.28’
- w
(b) CONDUCTORS AT NEW LOCAT~OllS
(Dimensions shown are at the waterline)

(a) CONDUCTORS AT O!-D LOCATIONS

Fig. 4 - Plan views of structure showing conductors at 01d and new 1ocations.

T
STRUCTURE

NO. OF WAVE
DIRECTIONS

/“:s/ ::@=”

7
SORTING OF MEM8ER
PREFAT END STRESSES

t
SCF’S FROM SCF PROGR4M
PROBABILISTIC
SN CURVES
FATIGUE FATIGUE
WAVE SPREADING +
ANALYSIS
WAVE SCATTER
DIAGRAMS

PREDICTED
FATIGUE
LIVES
5

Fig. 5 - Block diagram of probabilistic dynamic fatigue analysis system.

 / I I , 1 1

0 2 4 6 c1 10. 12

THEORETICALMIDPLANE INTERSECTIONSTRESS

THEORETICAL AWS
INTERSECTION HOT SPOT
JOINT STRESS STRESS
—. ..

PMB-76 3.5 2.0 HOT SPOT = ❑

ISOPARANETRIC
SRI-78 8.6 ‘- 5.3
1

SRI-KX 3.5 1.7

KY 2.4 HOT SPOT =
Kz 3.8 EXPERIMENTAL o
K1 3.5
K2 3.5 2.0
ORNL-T 2.8 1.7
SRI-T 10.6 5.6
OOVER-TZ 8.8 6.2
UKOSRP-TX 4.1 3.3
TY 2.1 -1.7
FILLET-TX 9.1
TY 3.1 arrowhead denotes estimated
TZ 6.1 3.2 stress at toe of minimum weld

KSEPL X 9.o.– _ 4,4

,- ...
FTG. ..—————
7 - THREE-DIMENSIONAL MODELING OF
COMPLEX TUBULAR JOINT c
FIG, 6- COMPARISON OF THEORETI CAL VS , ACTUAL HOT SPOT STRESS s

FIG, 8 - OVERALL VIEW OF TEST FRAME WITH MODEL OF COMPLEX JOINT INSTALLED.
 60 1 +$~wELD + “’ELD

(a) AXIAL. STRESS, CHORD SURFACE

10-3/4” @ .250” 180k

180k
‘\

Ii HELO
20’ o .500” _——

18” ~ .500”
—

(DISBONDED CASE)
(BA5E cAsO

: 10
I&l

F
c
-lo

-20 EXPERIMENTAL :
i

-30 1 (b) CIRCULAR BENDING STRESS, CHORD SURFACE

●
A
T’CIP
BOTTOM

Fig. 9 - Effect of grout on chord stress results near tension branch.

 GROUTEO NONLINEARBASE CASE
GROUTEO NONLINEAR OISBONOEO CASE
LO
WELO $
10 7 T
●
E —A

-10
BRANCH f
z
x ..20

u) /
g -30
~
-40

-50
(a) AXIAL STRESS, CHORD SURFACE
-60

10-3/4” O .250” 1130k

180k , /
\

18” 0 .500”

20
-J
OWELO

10
I
i=

--10 NOHIHAL
W/GROUT - LINEAR
~ NONLINEARBASE CASE - GROUTEO
--20 - STRESS
w NONLINEAROISBONOEDCASE - GROUTED
~
~ -30

-40 I
-50 1
1!
k-
(b)
W/O GROUT

CIRCULAR BENOING STRESS. CHORO SURFACE

EXPERIMENTAL:

● Top

‘BOTTOM
-60

FIG, 10 -EFFECT OF GROUT ON CHORD STRESS RESULTS NEAR COMPRESSION BRANCH,

(scF)chord
= 1.8 rfi Siin8

(SCF)hrace = 1.0 + 0.6 Qr

[
1.0+
{-
~ (SCF)choPd
1
*
I
+&+
AW5 -X
MO171FIE0 %% SURVIVAL

lt.ltTIA-IWbi
WWRO @~ Mt&w 12U+btT)
(J
- GAPMINTS

t
II 1.,,,1 1. ,,.1 1..!.1 1,,,,1 1...,1 1,,,.1
al
[....1 1..
D 10’ I& lo+ I& 10b I0s

CYCLES T~ FAILLIRE

FIG, 11 - MRC DATA BASE: K JOINTS WITH CHORD DIA z 8“, THICKNESS z 3/16”.
 “1 \ 1I
0000
INERTIA FORCES PREDOMINATE
DRAG FORCE MAY BE NEGLECTED
o 0000000
. -
ml t \ .
0
Sea States Used in Fatigue Study
Hs Plotted for D = 4 ft.
b 00,,
000900 0

r-. o 000** 7
0 00**0
In o ooeo solid Points &?
00000
g zoo Most Damaging
Ogaa TOWER Study

ml
/\\

1
BOTH DRAG & INERTIAL FORCES NEEDED
IN MORRISON’S EQUATION
g
c-)

+
DRAG FORCES o
PREDOMINATE

L Deep Water Waves

W
.001 .01 .1 1-

d/gT2 DEPTH PARAMETER

Fig. 12 - Nath wave force classification.

90°
67.5° 112.5°
45°

5
n v
22;5° ROW B
Y
Wave direction
number
0° *x + ——— 1g~o
9
16
/a ~o\ \
0
337.5” M ROW A n n % 202.5°
$“;”4
‘5/ 14/ 13 12\ ll\
/
\
315° ‘ I ~ \ \
\ 225°
/
247.5°
292.5° 27;0

Fig. 13 - Plan view of structure showing wave directions analyzed.