You are on page 1of 18

METHODOLOGY F O E ASSESSING RELIABILITY

OF TENSION L E G PLATFORM T E T H E R S
Downloaded from ascelibrary.org by "Indian Institute of Technology, Bhubaneswar" on 10/29/21. Copyright ASCE. For personal use only; all rights reserved.

By H . Banon, 1 Associate M e m b e r , A S C E , and S. J . H a r d i n g 2

ABSTRACT: The tension leg platform (TLP) is a large, floating offshore platform
for producing oil and gas in deep water. The TLP is moored to the ocean floor
by vertical tethers. Loads in the tethers have several components that are time-
variant and non-Gaussian. Therefore, conventional methods of reliability analysis
typically applied to fixed offshore platforms are not applicable to TLP tethers. This
paper presents a method for probabilistic combination of time-variant (Gaussian
and non-Gaussian) and time-invariant tether load components. The procedure is
applied to a Gulf of Mexico TLP design where the environmental hazard is due
to hurricane events. Two modes of failure are considered for the TLP tethers: (1)
The maximum lifetime tension exceeds the yield load; and (2) the minimum life-
time tension is less than zero, i.e., the tethers experience the loss of tension. The
resulting TLP tether reliability is compared to the calculated reliability of conven-
tional fixed offshore platforms, which form the basis of industry experience.

INTRODUCTION

The tension leg platform (TLP) is one of several structural concepts used
for oil and gas drilling and production in deep waters. The only operational
TLP is on the Hutton field (Curtis 1984) in the British North Sea. A second
TLP is being fabricated for the Jolliet field (Langewis 1987) in the Gulf of
Mexico and a third TLP is being designed for the Snorre field ("Snorre TLP"
1987) in the Norwegian North Sea. In addition, a TLP has been proposed
to develop the Heidrun field ("Deepwater TLP" 1987) in the Norwegian
North Sea.
Because of relatively limited experience with TLPs, the offshore industry
is learning to use theoretical reliability assessment to verify the adequacy of
the structural design. Confidence in the adequacy of a design is improved
by comparing the calculated probabilities of TLP structural failure with those
of fixed platforms that form the basis of offshore industry experience.
The work presented here focuses on the development of a method to com-
bine probabilistically time-variant and time-invariant, Gaussian and non-
Gaussian load components. This method was then applied to the structural
reliability of a Gulf of Mexico TLP design for extreme environmental loads.
The modes of failure considered were: (1) Overload (tension exceeding yield)
under the maximum tension load; and (2) loss of tension. While previous
work (Brekke and Gardner 1986) has shown that failure does not necessarily
occur when tension is lost, failure was defined as loss of tension in this study
for simplification. Fatigue is an important mode of failure that was not con-
sidered in this study. For this TLP, the tether cross-sectional design was
controlled by the maximum stress requirements.
TLPs are inertially dominated platforms, i.e., they respond dynamically
'Res.
2
Specialist, Exxon Production Res. Co., P.O. Box 2189, Houston, TX 77001.
Sr. Res. Engr., Exxon Production Res. Co., Houston, TX.
Note. Discussion open until February 1, 1990. To extend the closing date one
month, a written request must be filed with the ASCE Manager of Journals. The
manuscript for this paper was submitted for review and possible publication on Sep-
tember 12, 1988. This paper is part of the Journal of Structural Engineering, Vol.
115, No. 9, September, 1989. ©ASCE, ISSN 0733-9445/89/0009-2243/$1.00 +
$.15 per page. Paper No. 23865.
2243

J. Struct. Eng., 1989, 115(9): 2243-2260


to the ocean waves. For this study, frequency domain analyses were used
to obtain the TLP tether loads. TLP tether loads have several components
Downloaded from ascelibrary.org by "Indian Institute of Technology, Bhubaneswar" on 10/29/21. Copyright ASCE. For personal use only; all rights reserved.

that are time-variant, some of which may be assumed to be Gaussian whereas


others are non-Gaussian. The methodology developed in this paper was used
to combine probabilistically the time-variant and time-invariant tether load
components. The probability distribution of the combined tether load was
then used to calculate the annual probability of tether failure for the maxi-
mum and minimum tension failure modes.
The results of this study indicate that the risk of tether failure for the
maximum tension mode of failure (tension exceeding yield) is four to five
orders of magnitude lower than the failure probability of a typical Gulf of
Mexico fixed jacket. The risk of loss of tension for a TLP tether was shown
to be one to two orders of magnitude lower than the probability of failure
for a typical fixed jacket. For the maximum tension failure mode, the un-
certainty in tether resistance controlled the calculated risk, whereas for the
loss of tension failure mode the randomness in the combined, time-variant
tether load controlled the calculated risk.
While the tether reliability for maximum and minimum tension failure modes
was shown to be more than adequate, it should be noted that certain phe-
nomena, such as waves impacting the deck, were not included in this study.
Inclusion of such phenomena will reduce the maximum tension tether reli-
ability to a level more consistent with the reliability of typical fixed jackets.
Additional studies of TLP tether reliability include Stahl and Geyer (1984,
1985), Cornell et al. (1984), and Diamantidis (1988).

TENSION LEG PLATFORM

The TLP is a production system for deepwater that has a semisubmersible


type of hull permanently moored to the ocean floor by vertical steel tubulars,
typically referred to as tethers or tendons. The tethers are kept in tension by
the excess buoyancy of the hull.
The TLP analyzed in this study, as shown in Fig. 1, was designed for
762 m (2,500 ft) in the Gulf of Mexico. The hull is a square, symmetric
design with four cylindrical columns and four square pontoons. There are
two tethers per corner for a total of eight tethers for the platform. The neu-
trally buoyant tethers are steel tubulars 1.105 m (43.5 in.) in outer diameter,
with a 33.2-mm (1.31-in.) wall, and are connected using threaded couplings.
The tethers are critical structural elements and are subjected to a variety
of random loads during the lifetime of the platform. For this study, maxi-
mum tension failure was assumed to occur when the tether load exceeded
the steel yield strength. This is a conservative assumption but is consistent
with previous reliability analyses performed on other types of offshore struc-
tures. While the TLP can survive a storm with a tether damaged or removed,
a raptured tether falling to the ocean floor might cause significant damage
to risers, adjacent tethers, or foundation and wall templates. Failure of the
TLP system was, therefore, defined as failure of one tether. For the mini-
mum tension case, failure was defined as occurring when a tether became
"slack," i.e., when the tether lost its tension. This is a conservative defi-
nition of failure. Previous studies (Brekke and Gardner 1986) confirm that
a brief period of tether tension loss does not lead to excessive tether load
2244

J. Struct. Eng., 1989, 115(9): 2243-2260


Downloaded from ascelibrary.org by "Indian Institute of Technology, Bhubaneswar" on 10/29/21. Copyright ASCE. For personal use only; all rights reserved.

FIG. 1. Tension Leg Platform

or TLP motion. Although the tethers are composed of many segments that
are welded or connected together, we shall assume that failure of one seg-
ment represents the tether failure; i.e., the resistances of all segments are
assumed to be perfectly correlated.

TLP TETHER LOADS

The following load components were accounted for when calculating max-
imum and minimum tension:

Time-Invariant
0
Pretension.
• Tide.
• Mean wind overturning moment.
• Mean current overturning moment.
• Mean wave drift overturning moment.

Time-Variant
" Wave-frequency, motion-induced.
• Setdown.
2245

J. Struct. Eng., 1989, 115(9): 2243-2260


Downloaded from ascelibrary.org by "Indian Institute of Technology, Bhubaneswar" on 10/29/21. Copyright ASCE. For personal use only; all rights reserved.

m////////////////////////////////////////////^
FIG. 2. TLP Response to Environmental Loads

• Slow drift inertia moment.


8
Bending.

In general, a load caused by initial mispositioning of the foundation tem-


plates should be included for tether design. However, zero mispositioning
error was assumed for this study.
A sketch of the environmental loads acting on a TLP and the resulting
TLP response is shown in Fig. 2. A brief description of each of the resulting
load components is provided as follows:

Pretension
The pretension is the static tension in the tethers in the installed condition
with no environmental loads. The pretension balances the excess buoyancy
of the hull and provides the horizontal restoring force when the TLP is offset
horizontally from its zero-environment position.
Tide
The tide referred to here is a combination of storm tide and astronomical
tide. Tide increases the draft, which increases the TLP buoyancy and leads
to increased tether tension. The storm tide was assumed to vary with storm
intensity.

Mean Wind, Current, and Wave-Drift Overturning Moments


The mean environmental forces acting on a TLP result in overturning mo-
ments that lead to changes in tether tension. Tension in upweather tethers
increases, while tension in downweather tethers decreases.

Wave-Frequency Motion-Induced
The wave-frequency component of tether load is due to wave forces on
the TLP hull. This load component is time-variant with a period range of
2246

J. Struct. Eng., 1989, 115(9): 2243-2260


5-30 sec and is typically one of the most important load components. Lin-
ear, frequency-domain diffraction analyses were performed to calculate the
Downloaded from ascelibrary.org by "Indian Institute of Technology, Bhubaneswar" on 10/29/21. Copyright ASCE. For personal use only; all rights reserved.

wave-frequency load component. Linearized, velocity-squared, drag forces


were also included in the analyses. Wave-frequency tension in the upweather
tether was used for the maximum tension case while tension in the down-
weather tether was used for the minimum tension case.

Setdown
As a TLP is offset from its zero-environment position, the hull is pulled
down, or set down, into the water (see Fig. 2). This setdown is due to the
extremely high vertical stiffness of the tethers and is a quadratic function of
offset. The setdown leads to an increase in buoyancy acting on the hull,
which is balanced by an increase in tether tension for both upweather and
down weather tethers.
Tether tension due to setdown is a time-variant load that is a nonlinear
function of offset. Therefore, the setdown load becomes increasingly sig-
nificant as sea-state intensity increases. Time-invariant offset was calculated
based on the mean, or time-invariant, environmental forces. Wave-frequency
and low-frequency time-varying, wave-induced, offset statistics were based
on the results of hydrodynamic diffraction analyses. The effect of low-fre-
quency wind energy on platform offset was also included. (Wave-frequency
response occurs at 5-30 sec. Low-frequency response occurs at the vessel's
natural surge period of 110 sec.)

Slow Drift Inertia Moment


A TLP oscillates at its natural surge period of approximately 80-130 sec
(110 sec for the subject TLP). When the hull reaches its maximum low
frequency offset, a couple is generated due to the TLP center of gravity
being higher than the cross-load bearing, which is the location at which the

rain offset mean max offset

(a) Oscillation at Natural Surge Period (b) Free-Body Diagram

FIG. 3. Tension Due to Slow Drift Inertia Moment

2247

J. Struct. Eng., 1989, 115(9): 2243-2260


horizontal component of restoring forces in the tethers acts (see Fig. 3). The
resulting increase in tension in one tether and reduction in the opposite tether
Downloaded from ascelibrary.org by "Indian Institute of Technology, Bhubaneswar" on 10/29/21. Copyright ASCE. For personal use only; all rights reserved.

due to this moment is called the slow-drift inertia component. It is a time-


variant load. For simplicity and because the slow-drift inertia load was a
very small part of the total tether load, the correlation between the slow-
drift inertia load and the setdown load due to low frequency offset was ig-
nored in this study.

Bending
Because of the stiffness of elastomerics at the cross-load bearing, the TLP
offset induces bearing pressure on the tethers. This load is usually referred
to as the tether bending. For this study, outer-fiber bending stress was as-
sumed to be a linear function of total offset.

DESCRIPTION OF ENVIRONMENT

The extreme design environment for offshore platforms in the Gulf of


Mexico is associated with the occurrence of hurricanes. The intensity and
direction of waves, winds, and currents at a given site depends on the hur-
ricane intensity (as measured by the storm pressure drop and its radius of
maximum winds), as well as its position relative to the site. For simplicity,
the intensity of hurricane waves at a site is usually gauged by the significant
wave height. The significant wave height Hs is defined as the mean height
of the highest one-third of the waves. Fig. 4 shows the annual probability
distribution of Hs for the 762-m (2,500-ft) Gulf of Mexico site that was
considered in this study. This distribution is referred to as the wave hazard
curve and accounts for the annual frequency of hurricanes expected at the
site.

1 1 1 1 1 1 1 1 1 1 1 1 1 1 T"
8 10 12 14 16 IB 20 22
Sig. Wave Height (m)
FIG. 4. Wave Height Exeeedance Curve (Wave Hazard Curve)

2248

J. Struct. Eng., 1989, 115(9): 2243-2260


J
- h
Downloaded from ascelibrary.org by "Indian Institute of Technology, Bhubaneswar" on 10/29/21. Copyright ASCE. For personal use only; all rights reserved.

;j L

f1
0
1 1
2
1 i1
A
i1 i1
6
i1 i1
8
i1 i1
10
i1 i1 -
12
Storm D u r a t i o n (hours)

FIG. 5. Storm Buildup and Decay Profile

In general, the environmental condition at a site can be represented by the


combination of a slow process and a fast process. The slow process would
model the storm buildup and decay and the fast process would model the
time-variant random nature of waves, winds, and currents. Given the slow
process, the environment at any given instant in time can be described by
the wave height, wave direction, wave zero-crossing period, wind speed,
wind direction, current speed, and current direction. For this study, the slow
process was characterized by the significant wave height. Fig. 5 shows the
assumed storm buildup and decay as a function of the maximum Hs during
the storm. This generic curve was obtained from the analysis of oceano-
graphic data for a typical deepwater site in the Gulf of Mexico (Haring and
Heideman 1978). The duration of each storm was assumed to be fixed at
13 hours with the maximum significant wave height occurring for one hour.
The uncertainty in storm duration has a negligible effect on the calculated
tether reliabilities.
To simplify the structural response analysis of the TLP, it was further
assumed that winds and currents and their directions are functions of the
significant wave height. Wind, waves, and current were assumed to act col-
linearly in the direction that leads to the maximum tension response. In ef-
fect, the random process representing winds, waves, and currents was con-
densed into a single process, namely, the wave process. This is a common
assumption in design and analysis of fixed offshore platforms. Winds and
waves dominate the tether tension response; current contributes very little to
tether tension. TLP dynamic behavior is mainly due to the wave action,
although there is dynamic response at the natural surge period due to low-
frequency wind energy.
Tether response statistics were generated for each of the 11 sea states shown
in Table 1. The RBK (Robinson et al. 1967) wave energy spectrum was
assumed for all sea states. For sea states different from those listed in Table
1, a nonlinear regression scheme was used to estimate the tether response
statistics.
2249

J. Struct. Eng., 1989, 115(9): 2243-2260


TABLE 1. Environmental Conditions for Deterministic Analysis

Significant Wave Height


Downloaded from ascelibrary.org by "Indian Institute of Technology, Bhubaneswar" on 10/29/21. Copyright ASCE. For personal use only; all rights reserved.

m ft
(1) (2)
6.1 20
7.6 25
9.1 30
11.0 36
12.5 41 (100-year)
13.7 45
15.2 50
16.8 55
18.3 60
19.8 65
21.3 70

COMBINATION OF TIME-VARIANT TETHER LOADS

This section describes the procedure that was used to combine the follow-
ing four time-variant tether loads: (1) Wave frequency; (2) slow-drift inertia;
(3) setdown; and (4) bending. The equations presented in this section for
combining wave frequency and slow-drift inertia loads apply to both the
maximum and minimum tension cases. The equations presented for com-
bining the setdown and bending loads apply to the maximum tension failure
mode. Equations for combining these two components for the minimum ten-
sion failure mode are presented in a subsequent section.
Wave-frequency and slow-drift inertia loads were combined using simple
combination rules because they are independent Gaussian processes. The set-
down and bending loads were combined using a closed form solution. Sim-
ple combination rules could not be used because, as described previously,
the setdown and bending loads are both nonlinear functions of offset. TLP
offset is, however, a Gaussian random process. Finally, the point crossing
method (Larrabee and Cornell 1981) was used to develop the probability
distribution of the combined time-variant tether load. The load probability
distributions described below are all conditioned on the storm significant
wave height. However, for clarity of presentation, conditioning on Hs is
implicit and will not be shown in the equations presented in this section.
If Z is a random process, the probability distribution of its maximum value
during time T can be written as:
P(ZTmm <r) = exp [-vz(r)T] (1)
where vz(r) = the upcrossing rate of the threshold r. Eq. 1 assumes that at
time T = 0 the process is below the threshold and also that the upcrossings
follow a Poisson distribution. A better approximation to the upcrossing rate
of the threshold r can be found in VanMarcke (1975).
If ZLx in Eq. 1 represents the maximum combined time-variant tether load
during time T, then we need to derive the upcrossing rate of Z for this study.
The notation is simplified from here on and we use Z instead of Z^ax.
For a Gaussian random process, the upcrossing rate of level r can be writ-
ten as:
£.c.\J\J

J. Struct. Eng., 1989, 115(9): 2243-2260


v(r) = v„ exp I ~ u2) (2)
Downloaded from ascelibrary.org by "Indian Institute of Technology, Bhubaneswar" on 10/29/21. Copyright ASCE. For personal use only; all rights reserved.

where v0 = the mean zero-upcrossing rate (or mean zero-frequency) of the


normalized process and u is the normalized process;
r — ix
«= ~ (3)
a
1 CT
"o = — " (4)
2ir cr
and u. andCT= the mean and standard deviation of the arbitrary point-in-
time marginal distribution of a load and & is the standard deviation of the
derivative process of that load.
Eqs. 2-4 apply to the wave-frequency and the slow-drift inertia loads (de-
noted here as load W and load S), which are Gaussian random processes. A
better approximation to the combination of these two loads, that is a bimodal
process, can be found in Toro and Cornell (1986). Since loads W and S are
independent random processes, their combination (denoted here as load X)
is also a Gaussian process with the following upcrossing rate:

^ (r)=wr)= ^vtT^ exp H M ' 2 ) (5)


where
, r - n,w - n,t
" = —, (6)
Vol + v]
TLP setdown and bending loads are both functions of the offset and, there-
fore, they are perfectly correlated. The offset 8 is a Gaussian process with
mean u.s and standard deviationCTS.In fact, the TLP dynamic offset is com-
posed of two components which are both Gaussian, i.e., a low-frequency
component and a high-frequency component. For a given offset 8, the com-
bination of bending and setdown loads (denoted here as load Y) is:
Y = a|8| + &52 (7)
The arbitrary-point-in-time marginal PDF of Y can be derived from the
probability distribution of 8 (Benjamin and Cornell 1970). Appendix I de-
scribes the derivation of this PDF, which may be written as:

<8>
^-^M^-sM-^sJj
where 4>s = the Gaussian density function of offset and

^l^ «
Obviously, the combination of setdown and bending loads Y is not a Gaus-
sian random process because of the nonlinear terms in Eq. 7. The upcrossing
2251

J. Struct. Eng., 1989, 115(9): 2243-2260


rate of this combined load process can be derived (see Appendix I) as fol-
lows:
Downloaded from ascelibrary.org by "Indian Institute of Technology, Bhubaneswar" on 10/29/21. Copyright ASCE. For personal use only; all rights reserved.

Va 2 + Abr - a\ f-Va2 + 4br +


Mr) = 1 +v (10)
*[ Tb ) '{—^-b )
where v5( •) denotes the upcrossing rate of the Gaussian offset process and
the positive and negative signs on v refer to upcrossing and downcrossing
respectively.
The last step in combining the TLP time-variant loads involves a com-
bination of random loads X and Y. This can be achieved using the "point
crossing" method (Larrabee and Cornell 1981). Thus, the upcrossing rate of
the combined time-variant load Z = X + Y can be written as:

vz = vx+r(r) = J MQvAr - Qdt + I MQvx(r - Qdl (11)

where individual terms within this integral were previously defined. The PDF
of the combined time-variant load Z can be found by differentiating Eq. 1.
d d
fz(r) = - [Fz(r)] = - - [vz(r)T\ exp [-vz(r)T\ (12)
dr dr
where, from Eq. 11,

- tvz(r)] = MQ y [vY(r - ©]«£ + fr(& ~ [vx{r - ©]d| (13)


dr }_„ dr J_„ dr
Appendix II describes the derivation of the derivatives in Eq. 13.
This formulation for combining the time-variant loads assumed that the
sea state is stationary and the probability distributions were all conditioned
on the significant wave height. For a real hurricane, the sea-state buildup
and decay can be represented by a number of quasi-stationary sea states (see
Fig. 5). Assuming that the load exceedances in each time interval Tt are
independent, the CDF of the maximum combined time-variant load during
a storm is given as:

Fz{r) = Fz(r) • F2z(r) ... Fz{r) = exp • E v>W' (14)

where / = 1, 2, ... n denotes the n stationary sea states within a storm.


The PDF of the combined load for a storm is derived by differentiating that
CDF:
n

fzO-) = E Y\fz(r) (15)

MAXIMUM TENSION TETHER RELIABILITY ASSESSMENT

The TLP tethers are subjected to a combination of time-variant and time-


invariant loads. The foregoing methodology allows for combining the time-
variant tether loads into a single load component. This load Z, which is the
ZZb2

J. Struct. Eng., 1989, 115(9): 2243-2260


expected maximum or trie combined time-variant load during a storm, may
now be viewed as a simple random variable. The conditional limit state of
Downloaded from ascelibrary.org by "Indian Institute of Technology, Bhubaneswar" on 10/29/21. Copyright ASCE. For personal use only; all rights reserved.

the tether for maximum tension may be written as:


G = R - P - e, ^ L,. - e2Z (16)

where R = tether resistance; P = pretension; and L, = time-invariant tether


loads, e, and e2 are unbiased (mean = 1) random variables which represent
subjective uncertainties in calculating the tether loads. The time-invariant
tether loads were anchored to the storm significant wave height. However,
as Eq. 16 suggests, they were assumed to be invariant during the storm
buildup and decay. Failure occurs when G < 0. The conditional probability
of failure for the tether, given the maximum significant wave height of the
storm, also referred to as the "fragility curve," was obtained from the fol-
lowing:

P(Failme\Hs) = pi R < P + 6 , ^ I i H - e 2 Z|#J (17)

The advanced, first-order, second-moment technique (Ang and Tang 1984;


Rackwitz and Fiessler 1978) was used to calculate the conditional probability
of tether failure given in Eq. 17. The annual probability of tether failure due
to the maximum tension was calculated from a convolution of the tether
fragility curve and the probability distribution of storm intensity at the site:

P(Failure) = I P(R < P + e , ^ U + *&fl. = h)fHs(h)dh (18)

The second term in Eq. 18's integral represents the wave hazard (risk)
curve for the site (see Fig. 4). This integral assumes that the maximum ten-
sion in a tether would be due to the annual hurricane with the highest in-
tensity; i.e., the contribution to the failure probability from annual sea states
less severe than the most intense storm in that year is not taken into account.
This assumption is fairly accurate in the Gulf of Mexico where the hurricane
events are rare.
Using the procedure described in this paper, the conditional probability
distribution of the combined time-variant tether load [F2(r) in Eq. 14] was
obtained. Fig. 6 shows the conditional CDFs of the combined time-variant
tether load for three different significant wave heights. The conditional CDFs
were calculated for Hs values presented in Table 1 and were found to have
coefficients of variation ranging from 0.045-0.06. These distributions (which
were obtained numerically) were also found to follow the Gumbel distri-
bution very closely.
Table 2 lists the characteristics of the time-invariant random variables used
in the tether limit state (Eq. 16). The TLP tether design considered in this
study has a nominal yield stress of 482 MPa (70 ksi) and a mean yield stress
of 517 MPa (75 ksi). The random variables ei and e2 in Eq. 18 representing
uncertainties in our load calculations were assumed to have coefficients of
variation equal to 0.10 and 0.15, respectively. Fig. 7 depicts the tether frag-
ility curve (conditional probability of tether failure given the storm signifi-
cant wave height) normalized to the probability of tether failure in the 100-
2253

J. Struct. Eng., 1989, 115(9): 2243-2260


H - 12
Downloaded from ascelibrary.org by "Indian Institute of Technology, Bhubaneswar" on 10/29/21. Copyright ASCE. For personal use only; all rights reserved.

0.8
18.3

0.6

(same HJ

0.4 minimum maximum


tension tension

T — i — r — i — i — i i—i—i—i i — i — i — i — i — r
10 15 20 25
Tether Load (MN)

FIG. 6. Conditional Probability Distributions of Combined Time-Variant Tether


Load; Maximum and Minimum Tension

year-design storm. The convolution of this curve with the wave hazard curve
(Fig. 4) revealed that most of the contribution to the probability of failure
was from storms much rarer than the 100-year-design value. The annual
probability of tether failure for the maximum tension failure mode was cal-
culated to be four to five orders of magnitude lower than calculated values
for typical fixed platforms in the Gulf of Mexico.
One reason for the higher TLP reliability compared with fixed platforms
is that TLP maximum tension increases less rapidly with increases in sea

TABLE 2. Time-Invariant Random Variables in Tether Limit-State Equation


Probability Mean
Variable Description distribution (kN/tether) GOV
(1) (2) (3) (4) (5)
R Resistance Lognormal 57,900 0.10
P Pretension Normal 16,400 0.05
a
Mean wind load Normal 0.10
u
L2 Mean current load Normal
3
0.10
3
L, Mean drift load Normal 0.10
a
U Tide Normal 0.10
Subjective uncertainty in
6l mean loads Lognormal 1.0 0.10
Subjective uncertainty in
^2 time-variant loads Lognormal 1.0 0.15
"Function of sea state.

J. Struct. Eng., 1989, 115(9): 2243-2260


10
maximum tension
Downloaded from ascelibrary.org by "Indian Institute of Technology, Bhubaneswar" on 10/29/21. Copyright ASCE. For personal use only; all rights reserved.

minimum tension
10"

10

1.0

10

10

10

10,
-a
Sig. Wave Height/100-year Sig. Wave Height

FIG. 7. Conditional Probability of Tether Failure Given Sea-State Intensity; Max-


imum and Minimum Tension

state than does fixed platform response. This trend is due to the TLP being
an inertia-dominated structure (loads increase linearly with wave height) as
opposed to drag-dominated fixed platforms (loads increase quadratically with
wave height). In addition, pretension is a significant percentage of the total
tether load. Since pretension is a highly monitored load with relatively little
uncertainty, the corresponding total load uncertainty is reduced and the re-
liability is increased.
In the second-order reliability method, importance factors (square of di-
rection cosines) provide a measure of the relative importance of each variable
to the calculated probability of failure (Ang and Tang 1984). For the max-
imum tension case, tether reliability is dominated by the resistance uncer-
tainty (yield strength for this study) and not by any one load variable. Time-
variant loads have a more significant effect on reliability than do the time-
invariant load components.

MINIMUM TENSION TETHER RELIABILITY ASSESSMENT

The tether load combination problem is slightly changed here to reflect


the tether slack condition. For this case, the pretension and setdown loads
always help to keep the tether in tension, whereas the other loads induce
compression in the tether. To develop the probabilistic combination of time-
variant tether loads, Eq. 7, which represents the combination of bending and
setdown loads, is modified:
Y = a 8 - fc82 AbY > 0. (19)
The condition of Eq. 19 implies that there is a positive tension limit to
the combination of setdown and bending loads. This limit on Y is determined
2255

J. Struct. Eng., 1989, 115(9): 2243-2260


by constants a and b in Eq. 19. The PDF and upcrossing rate of Y in this
case are the same as for maximum tension except for the sign of b changed
Downloaded from ascelibrary.org by "Indian Institute of Technology, Bhubaneswar" on 10/29/21. Copyright ASCE. For personal use only; all rights reserved.

m the previous equations. The sign is changed to account for the fact that
setdown increases tension and, therefore, is beneficial in the minimum ten-
sion case. The resulting load combination is then subtracted from the pre-
tension in the limit state function as shown in Eq. 20. The combination of
wave-frequency and slow-drift inertia loads X, as well as the total combi-
nation of time-variant loads (Z = X + Y), is the same as in the maximum
tension case. The conditional limit state of the tether for minimum tension
is written as:
G
= P-*^Lt-& (20)

For the purpose of this study, zero tension at the bottom of the tether was
chosen as failure. The tethers are neutrally buoyant. For non-neutrally buoy-
ant tethers, the weight per length of the tether would also have to be ac-
counted for.
Fig. 6 shows the conditional CDFs of the combined time-variant tether
load for the minimum tension failure mode. The conditional probability of
failure for the minimum tension normalized with respect to the probability
of failure in the 100-year storm is compared to the maximum tension case
in Fig. 7. Fig. 7 shows that the tether fragility curve (i.e., the conditional
probability of failure) for minimum tension increases less rapidly for higher
sea states when compared to the maximum tension. The reason for this be-
havior is that the setdown load, which is beneficial to the minimum tension
failure mode, increases more rapidly as the sea-state intensity increases (re-
member that the setdown load is a nonlinear function of the TLP offset).
The convolution of this tether fragility curve with the wave hazard curve
produces the annual probability of tether failure for the minimum tension.
The results indicate that the risk of failure for minimum tension is three
orders of magnitude higher than for the maximum tension. However, the
probability of tether failure was still shown to be one to two orders of mag-
nitude lower than typical values for steel-piled fixed platforms in the Gulf
of Mexico.
Importance factors (square of direction cosines) for the minimum tension
case indicate that the combined time-variant load and the subjective uncer-
tainty in that load dominate the reliability. Pretension is the next most im-
portant variable in determining the reliability for this failure mode.

CONCLUSIONS

A procedure for probabilistic combination of time-invariant and Gaussian


and non-Gaussian time-variant loads was presented. The procedure was ap-
plied to evaluate the reliability of a Gulf of Mexico TLP tether design for
maximum and minimum lifetime tension loads. The results indicate that the
probability of tether failure for the maximum tension mode of failure (tension
exceeding yield) is four to five orders of magnitude lower than the failure
probability of a typical fixed platform. For the minimum tension failure mode
(tension loss), probability of tether failure was shown to be one to two orders
of magnitude lower than probability of failure for typical Gulf of Mexico
fixed platforms.

22b6

J. Struct. Eng., 1989, 115(9): 2243-2260


The writers wish to thank Exxon Production Research Company for per-
Downloaded from ascelibrary.org by "Indian Institute of Technology, Bhubaneswar" on 10/29/21. Copyright ASCE. For personal use only; all rights reserved.

mission to publish this paper. Prof. C. Allin Cornell's review of the research
presented in the paper is greatly appreciated.

APPENDIX 1. COMBINATION OF SETDOWN AND BENDING LOADS

The combination of bending and setdown loads Y for maximum-tension


tether failure mode is written as: Y = a|8| + £8 2 ; where the first term in Eq.
7 represents the TLP bending loads and the second term represents the set-
down loads. The CDF of Y may be derived as follows:
P(Y < y) = P(a|B| + bh2 < y) (21a)

P(Y < y) = P Y>0, * y ±_a


151 + <y* yy = - + —;2 (21*)
2b b 4b
a (21c)
P(Y < y) = P\ - V y * + — < 8 < V y *
2b 2b

P(Y<y) = F.(V?-^)-FJ-V? + fb (21d)

where F s ( •) denotes the CDF of the Gaussian distribution of the TLP offset.
The PDF of the combined load is obtained from taking the derivative of Eq.
21:
d 1
fr = -d
y
(Fy) =
2b\ff
•= h y
Zh<-*+Z y ^ 0 (22)

The upcrossing rate of Y = G(8) = r is:


vY(r) = vs(G~l) (23a)
2
r = a|8| + bh {23b)
+a ± V a + 4for
8 = r>0 (23c)
2b
Therefore, the upcrossing rate of the combined load is written as:
Na2 + Abr - a\ I'-\'a2 + 4br + a\
Vy{r) = vU +v5-| ) r>0 (24)
2b 2b
where the positive and negative signs on v refer to upcrossing and down-
crossing, respectively.

APPENDIX II. DERIVATIVES O F UPCROSSING RATES

The upcrossing rate of level r — £ for a Gaussian process is:


1 (r - € - u..v\
v.v(r - © = v0 exp (25)
CTV

2257

J. Struct. Eng., 1989, 115(9): 2243-2260


The derivative of the upcrossing rate with respect to r is:
d
~ ( r ~ ^j -~~
Downloaded from ascelibrary.org by "Indian Institute of Technology, Bhubaneswar" on 10/29/21. Copyright ASCE. For personal use only; all rights reserved.

T rK (,r - ©]
*M
= — 121 ^
-vx(r - © (26)
ar rr:
The upcrossing rate of level r — £ for the non-Gaussian (combination of
bending and setdown) process was derived in Appendix I (see Eq. 25).
Na2 + 4b(r - © - a\
Mr - © = v6+( )

/ - V a 2 + 4b(r - © + a \
+ v (27)
i — T b — j
The derivative of this upcrossing rate is:
d d dq d dw
- [Mr - ©] = — [v,(q)]~ + — [vB(w)] • — (28)
dr dq dr dw dr
where q and w are defined as:
Va2 + 4b{r - © -a
(29fl)
* = ift"
- V a 2 + 4ft(r - © + a
w = — (29ft)

and

— = [a2 + Ab(r - ©]~ (1/2) (30a)


dr
dw dq
— = ~— (30ft)
dr dr
Also, the other derivatives needed for Eq. 29 are:
d [v8(a)] = ~{q 3— ix8) v8(a) (31a)
— [v,(?)] = - ^ *
dq crs
d ~(w — |x8)
— [v 8 (w)] =
lv»(w)] v,(w) (31ft)
aw 05

Substituting Eqs. 31 and 32 into Eq. 29, one can calculate the desired
derivative of the non-Gaussian upcrossing rate.

APPENDIX HI. REFERENCES

Ang, A. H-S., and Tang, W. H. (1984). Probability concepts in engineering plan-


ning and design, vol. II: Decision, risk, and reliability. John Wiley and Sons,
New York, N.Y,
Benjamin, J. R., and Cornell, C. A. (1970). Probability, statistics, and decision for
civil engineers. McGraw-Hill Book Company, New York, N.Y.
Brekke, J. N., and Gardner, T. N. (1986). "Analysis of brief tension loss in TLP

2258

J. Struct. Eng., 1989, 115(9): 2243-2260


tethers." froc, 13 th Int. Symposium and Exhibit on Offshore Mech. and Arctic
Engrg., American Society of Mechanical Engineers, 43-47.
Cornell, A. C , Rackwitz, R., Guenard, T., and Bea, R. G. (1984). "Reliability
Downloaded from ascelibrary.org by "Indian Institute of Technology, Bhubaneswar" on 10/29/21. Copyright ASCE. For personal use only; all rights reserved.

evaluation of tension leg platforms." Proc, Fourth Specialty Conf. on Probabi-


listic Methods and Struct. Reliability, ASCE, Berkeley, Calif., 159-166.
Curtis, L. B. (1984). "How Conoco developed the tension leg platform." Ocean
Industry, 19(8), 35-46.
"Deepwater TLP development on Norwegian Shelf taking shape." (1987). Ocean
Industry. 22(11), 39.
Diamantidis, D. (1988). "Experience gained from first-order reliability methods in
structural analysis." Proc, Third Int. Conference on CAM/CAD, Southfield, Mich.
Haring, R. E., and Heideman, J. C. (1978). "Gulf of Mexico rare wave return pe-
riods." Paper 3220, Proc. Offshore Tech. Conf., Houston, Tex., 1537-1550.
Langewis, C. (1987). "Jolliet field development in Green Canyon." Deep Offshore
Tech., 4th Int. Conf. and Exhibit, Monte Carlo, Monaco, G.5/44-G.5/50.
Larrabee, R. D., and Cornell, C. A. (1981). "Combination of various load pro-
cesses." J. Struct. Div., ASCE, 107(1), 223-239.
Rackwitz, R., and Fiessler, B. (1978). "Structural reliability under combined random
load sequences." Computers and Structures, 9, 489-494.
Robinson, R. J., Brannon, H. R., and Katawar, G. W. (1967). "Storm wave char-
acteristics." Soc. of Petroleum Engrs. J., Mar., 87-98.
"Snorre TLP set to establish Saga's North Sea credentials." (1987). Offshore Engr.,
4, Apr., 72-75.
Stahl, B., and Geyer, J. F. (1984). "Fatigue reliability of parallel member systems."
J. Struct. Engrg., ASCE, 110(10), 2307-2323.
Stahl, B., and Geyer, J. F. (1985). "Ultimate strength reliability of tension leg plat-
form tendon systems." Proc. Offshore Tech. Conf. Paper 4857, Houston, Tex.,
151-162.
Toro, G. R., and Cornell, C. A. (1986). "Extremes of Gaussian processes with
bimodal spectra." J. Engrg. Mech., ASCE, 112(5), 465-484.
VanMarcke, E. H. (1975). "On the distribution of the first-passage time for normal
stationary random processes." / . Appl. Mech., 42(1), 215-220.

APPENDIX IV. NOTATION

The following symbols are used in this paper:

a,b = constants;
d{-) = differential;
CDF = cumulative probability density function;
COV = coefficient of variation;
exp = exponential;
Fx = cumulative probability density function of x;
f* = probability density function of x;
G = general nonlinear function;
Hs = significant wave height;
Li = time-invariant load i;
P(-) = probability;
P = tether pretension;
PDF = probability density function;
R = tether resistance;
r = threshold level;
S = slow drift inertia load;
T = time;
u = normalized random process;
W = wave frequency load;

2259

J. Struct. Eng., 1989, 115(9): 2243-2260


X = combination of wave-frequency and slow drift inertia loads;
Y = combination of setdown and bending loads;
C ,Z = maximum combined time-variant tether load during time T;
Downloaded from ascelibrary.org by "Indian Institute of Technology, Bhubaneswar" on 10/29/21. Copyright ASCE. For personal use only; all rights reserved.

5 = TLP offset;
et = uncertainty in calculating time-invariant loads;
e2 = uncertainty in calculating time-variant loads;
u, = mean of process x;
v = upcrossing rate;
v0 = zero upcrossing rate;
£ = integration parameter;
<JX = standard deviation of process x;
&x = standard deviation of derivative process x; and
<(> = normal probability density function.

2260

J. Struct. Eng., 1989, 115(9): 2243-2260

You might also like