Wellstream Methodology of Fatigue Analysis of

Flexible Riser
April 2005, by John Zhang

1.0 INTRODUCTION
In offshore application of unbonded flexible pipe, generally, the pipe service life is
limited by the tensile or the pressure armoring layer of the pipe. Therefore, the fatigue
life analysis of flexible pipe is focused on the tensile and the pressure armoring layers.
For a flexible riser, especially for a Free-Hanging riser in deep water, the maximum
fatigue damage always occurs at the top connection or at the outlet of the guide tube (I-
tube) on the floating unit on the water surface. With more and more flexible risers used in
deep water, the fatigue life analysis of flexible riser at the top connection becomes more
important design procedure.

Compared to the steel pipe, the stress in both the tensile and pressure armoring layers of
flexible pipe has a more complicated relationship to the pipe tension and curvature.
Therefore, the stresses in the armoring layers cannot be obtained simultaneously with the
global dynamic simulation. That is, the stresses should be calculated in a separated
postprocessing after the global simulation. This makes the fatigue analysis of flexible
riser much more difficult and much more time consuming than the steel riser. To make
the fatigue analysis practicable, therefore, it is unavoidable to adopt some simplified
assumptions.

However a simplified assumption always introduces some additional conservatism into

the analysis and sometimes could make the analysis too conservative to obtain a
satisfactory fatigue life. Therewith the more simplifications assumed the more
conservatism introduced in the analysis. Improving the fatigue analysis, therefore, to
make the alternating stress calculation more reasonably simplified and without
introducing significant conservatism is under urgent necessity. In the paper some
improvements on the fatigue life estimation for the armoring layers of flexible pipe are
presented and discussed.

In addition, recent years more and more clients require the global dynamic simulation for
fatigue analysis to be performed with irregular wave instead of regular wave and using
the rainflow method to count the number of cycles instead of the average zero-crossing
method. Since the irregular wave simulation need much longer CPU time (1 or 3 hours)
than the regular wave simulation, this requirement leads the fatigue analysis to become
much more time consuming. If say this requirement did not make any significant
difficulty for steel pipe, there should be many difficulties for flexible pipe because the
stresses could not be directly calculated from the global dynamic simulation. The
possibility and necessity for using irregular wave and rainflow method in the fatigue
analysis of flexible pipe are discussed in the paper.

1
2.0 GLOBAL DYNAMIC SIMULATION
2.1 General
The fatigue life estimation of flexible riser must be based on the results of global
dynamic analysis which provides the loading conditions to the riser which should be used
for the fatigue stress calculation.

The global dynamic simulation for fatigue analysis of flexible riser is carried out with
regular wave. The reason for regular wave simulation is due to that the stress calculation
for each layer of pipe wall is complicated and to be done with some local analysis
software based on the global simulation results.

There are two methods for the wave environment (wave scatter diagram) discretization as
discussed below.

2.2 Load Case Definition

Generally the wave environment is given in the form of wave scatter diagram as shown,
for example, in Figure 2-1. It is often classified with the significant wave height, Hs, and
the spectrum peak period, Tp, or the zero-crossing period, Tz. The entries of diagram are
often one of the two: (a) numbers of records obtained during a long observation term and
(b) percentage of occurrence for a certain period. Therefore, with the diagram, we can
calculate the total time of occurrence for an individual sea state during the required
service life.

To save time and make the fatigue analysis feasible, it is necessary to group the diagram
into a number of wave blocks. Then each block is regarded as a load case of global
dynamic simulation. There are generally two methods for scatter diagram discretization
as discussed below.

2.2.1 Direct Discretization on Scatter Diagram

The most simply discretization is to group the entries directly on the scatter diagram as
shown for example in Figure 2-1 where 9 blocks are grouped.. To be conservative, the
maximum significant wave height in the block and the weighted wave period should be
used as representatives. The table in the figure summarizes the representative Hs and Tp
for the 9 load cases.

Another type of grouping directly on the diagram is shown in Figure 2-2 where the
grouping is made with significant wave height only. The table in the figure summarizes
the representative Hs and Tp.

If the former grouping is called as Block Grouping, the later could be called as Strip
Grouping. With the block grouping method the wave period effect could be better
considered, while if using the same number of load cases, the strip grouping method is
better for considering the wave height effect.

2
The calculation of the number of waves for each load case is dependent on what the
entries of the scatter diagram represent. For entries to be the number of observation
record which is generally 3 hours long, the number of waves for a load case can be
calculated as:

mi L
ni = (2.2.1)
Tzi M

with ni – the number of waves in Case i,

mi – the total number of records of Case i,
L – the required service life,
Tzi – the average zero-crossing wave period of Case i,
M – the total number of records of the scatter diagram.

There are several methods to calculate the average zero-crossing period. The following
method is suggested by DNV [1] for calculating the average zero-crossing period, Tz,
with the weighted peak period, Tp:

1/ 2
⎛ 5+γ ⎞
Tz = T p ⎜⎜ ⎟⎟ (2.2.2)
⎝ 11 + γ ⎠

where γ is the peakedness parameter of wave spectrum. If no particular values are given,
the peakedness parameter can be calculated [1]:

Tp
γ = 5 for ≤ 3.6
Hs
5.75−1.15T p / H s Tp
γ =e for 3.6 ≤ ≤5
Hs
Tp
γ = 1 for 5 ≤
Hs

3
 Tp (sec)
Hs (ft) 0 - <1 1 - <2 2 - <3 3 - <4 4 - <5 5 - <6 6 - <7 7 - <8 8 - <9 9 - <10 10 - <11 11 - <12 12 - <13 13 - <14 14 - <15 15 - <16 totals
0 - <1 1736 10004 320 1948 2148 2176 2480 1172 240 116 68 32 48 16 12 22516
1 - <2 5128 23516 8884 5872 2412 1036 460 184 12 0 16 8 0 0 47528
2 - <3 0 4976 26016 10376 2812 928 168 72 20 32 0 0 0 0 45400
3 - <4 0 72 10384 18532 4256 1568 308 60 16 0 0 0 0 0 35196
4 - <5 0 463 13768 8776 1616 336 24 12 0 0 0 0 0 24995
5 - <6 41 2542 11172 2611 489 63 46 1 0 0 0 0 16965
6 - <7 5 336 6686 5251 503 68 33 4 4 0 0 0 12890
7 - <8 1 59 1321 6506 940 157 15 5 0 0 0 0 9004
8 - <9 13 441 4608 1698 182 59 4 2 0 0 0 7007
9 - <10 1 53 1549 2645 235 58 4 3 2 0 0 4550
10 - <11 1 702 2304 438 73 7 5 1 0 0 3531
11 - <12 1 211 1361 764 42 5 1 3 1 0 2389
12 - <13 21 659 897 116 4 8 1 3 0 1709
13 - <14 2 257 709 153 15 8 1 4 0 1149
14 - <15 Hs (ft) Tp (s) 1 94 428 227 11 4 5 3 1 774
15 - <16 Case 1 4 4.97 9 195 191 28 2 2 2 1 430
16 - <17 Case 2 7 8.28 6 83 194 19 1 2 3 1 309
17 - <18 Case 3 7 13.08 3 33 119 65 0 1 12 1 234
18 - <19 Case 4 12 6.61 1 8 81 34 13 2 9 2 150
19 - <20 Case 5 13 8.62 0 3 41 66 14 2 5 1 132
20 - <21 Case 6 17 13.83 0 3 33 34 7 5 2 0 84
21 - <22 Case 7 17 10.21 0 2 21 34 7 10 0 1 75
22 - <23 Case 8 25 11.47 1 8 28 9 9 0 0 55
23 - <24 Case 9 25 14.57 2 7 13 5 9 1 1 38
24 - <25 9 9 2 1 0 1 22
totals 0 1736 15132 28884 47742 53647 40107 29090 13413 4851 1702 490 143 112 61 22 237132
Figure 2-1 Block Grouping of Wave Scatter Diagram

4
 Tp (sec)
Hs (ft) 0 - <1 1 - <2 2 - <3 3 - <4 4 - <5 5 - <6 6 - <7 7 - <8 8 - <9 9 - <10 10 - <11 11 - <12 12 - <13 13 - <14 14 - <15 15 - <16 totals
0 - <1 1736 10004 320 1948 2148 2176 2480 1172 240 116 68 32 48 16 12 22516
1 - <2 5128 23516 8884 5872 2412 1036 460 184 12 0 16 8 0 0 47528
2 - <3 0 4976 26016 10376 2812 928 168 72 20 32 0 0 0 0 45400
3 - <4 0 72 10384 18532 4256 1568 308 60 16 0 0 0 0 0 35196
4 - <5 0 463 13768 8776 1616 336 24 12 0 0 0 0 0 24995
5 - <6 41 2542 11172 2611 489 63 46 1 0 0 0 0 16965
6 - <7 5 336 6686 5251 503 68 33 4 4 0 0 0 12890
7 - <8 1 59 1321 6506 940 157 15 5 0 0 0 0 9004
8 - <9 13 441 4608 1698 182 59 4 2 0 0 0 7007
9 - <10 1 53 1549 2645 235 58 4 3 2 0 0 4550
10 - <11 Hs (ft) Tp (s) 1 702 2304 438 73 7 5 1 0 0 3531
11 - <12 Case 1 2 4.71 1 211 1361 764 42 5 1 3 1 0 2395.707
12 - <13 Case 2 4 5.60 21 659 897 116 4 8 1 3 0 1718.602
13 - <14 Case 3 6 6.74 2 257 709 153 15 8 1 4 0 1161.739
14 - <15 Case 4 8 7.69 1 94 428 227 11 4 5 3 1 789.693
15 - <16 Case 5 10 8.44 9 195 191 28 2 2 2 1 448.4391
16 - <17 Case 6 12 9.10 6 83 194 19 1 2 3 1 330.1015
17 - <18 Case 7 14 9.80 3 33 119 65 0 1 12 1 257.8027
18 - <19 Case 8 16 10.39 1 8 81 34 13 2 9 2 176.3937
19 - <20 Case 9 18 11.06 0 3 41 66 14 2 5 1 161.0571
20 - <21 Case 10 20 11.79 0 3 33 34 7 5 2 0 115.7943
21 - <22 Case 11 22 11.94 0 2 21 34 7 10 0 1 108.9371
22 - <23 Case 12 25 12.30 1 8 28 9 9 0 0 92.30435
23 - <24 2 7 13 5 9 1 1 38
24 - <25 9 9 2 1 0 1 22
totals 0 1736 15289 28993.57 47742 53647 40107 29090 13413 4851 1702 490 143 112 61 22 237398.6
Figure 2-2 Strip Grouping of Wave Scatter Diagram

5
 Hs Tp (s) Subtotal
(m) 1 3 5 7 9 11 13 15 17 19 21 23
wave wave number of
Case #
height (m) period (s) waves
0.25 0 1 9.06 9.10 10
0.50 0.01 0.01 2 8.14 8.62 100
0.75 0.1 0.01 0.08 0.14 0.07 0.15 0.01 0.56
3 7.14 8.08 1,000
1.00 0.04 0.22 0.63 1.36 0.86 1.89 0.26 0.11 0.04 5.41
4 6.07 7.45 10,000
1.25 0.14 2.03 4.74 3.79 4.15 1.05 0.4 0.06 0.01 16.37
1.50 0.23 0.67 7.77 9.34 7.62 2.29 0.75 0.04 0.01 0.01 28.73
5 4.96 6.74 100,000
1.75 0.19 0.67 4.67 9.87 6.73 2.63 0.71 0.19 0.01 25.67 6 3.88 5.96 1,000,000
2.00 0.01 0.26 1 5.56 3.94 1.51 0.63 0.11 13.02 7 2.85 5.10 10,000,000
2.25 0.03 0.25 1.62 1.78 1.2 0.37 0.04 5.29 8 1.81 4.07 101,032,197
2.50 0.14 1.03 0.78 0.68 0.23 0.01 2.87
2.75 0.32 0.4 0.32 0.1 1.14
3.00 0.01 0.18 0.19 0.07 0.45
3.25 0.01 0.16 0.06 0.04 0.01 0.28
3.50 0.01 0.06 0.01 0.01 0.09
3.75 0.03 0.01 0.06 0.1
4.00 0.01 0.01

L046 Fatigue Analysis Load Case Classification

1.E+09

1.E+08 individual wave height distribution of 25 years

8
Load Case Steps
1.E+07
Cumulative Wave Number

1.E+06
6

1.E+05
5
Figure 2-3 Sample 1.E+04
for Individual Wave 4

Height Distribution 1.E+03

3
Curve Method 1.E+02
2

1.E+01
1

1.E+00
0 1 2 3 4 5 6 7 8 9 10
Individual Wave Height (m)
6
2.2.2 Discretization on Individual Wave Height Distribution Curve
This method is based on the distribution theory of wave amplitude during a short term,
i.e. the Rayleigh distribution theory [2]. With the Rayleigh distribution we can obtain the
following formula:

⎛ H2 ⎞
⎜
P ( H ) = exp⎜ − 2 2 ⎟ (2.2.3)
Hs ⎟
⎝ ⎠

where P is the cumulative probability for wave height to be equal to or larger than the
given wave height of H during a shot term of observation. Then the cumulative number
of waves whose height is equal to or larger than H can be calculated for a given scatter
diagram as:

m ⎡ ⎛ ⎞⎤
kL H2
n( H ) = ∑ ⎢ i exp⎜⎜ − 2 2 ⎟⎥
⎟ (2.2.4)
⎢ Tzi M
i ⎣ ⎝ Hs ⎠⎦⎥

where m – the total number of entries on the scatter diagram,

ki – the value of the ith entry,
Tzi – the average zero-crossing period of the ith entry,
M – the sum of all entries on the diagram.

Figure 2-3 shows a sample for the individual wave height distribution curve method. The
figure includes:

• The wave scatter diagram,

• The individual wave height distribution curve corresponding to the scatter
diagram and the load case classification on the curve, and
• The load case summary table.

For conservatism purpose, the maximum individual wave height in a load case is defined
as the representative wave height. The wave period associated to the individual wave
height will be discussed in Section XXX.

2.2.3 Refinement of Load Case Classification

For both discretization methods described above, it is general that the finer the load case
classification the lower the conservatism imposed on the global dynamic simulation. This
is because in both method the maximum wave height in a load case is taken as the
representative. Therefore, when the fatigue damage is overestimated, the common
method for reducing the conservatism in global simulation is to refine the load case
classification.

For example, Figure 2-4 shows the decrease in calculated maximum cumulative damage
in tensile wires of a flexible riser with increase of the number of load cases. From the

7
example, we can see that when the number of load cases is enough large, the decrease in
calculated maximum fatigue damage becomes very small.

Cumulative Damage vs. Number of Load Cases in L046

Project

1.2

1
Cumulative Damage

0.8

0.6

0.4

0.2

0
0 10 20 30 40 50 60
Number of Load Cases

Figure 2-4 Fatigue Damage vs. Number of Load Cases

2.3 Wave Period Determination

In discretization of wave environment, there are at least two wave periods for each load
case to be determined:

• Average Zero-Crossing Period used in the calculation of number of waves

• Period Associated to Representative Wave Height used in the global dynamic
simulation

If the wave scatter diagram is given in terms of significant wave height and zero-crossing
period, the average zero-crossing period is the weighted period for each load case. If the
diagram is given in terms of significant wave height and peak period, the average zero-
crossing period can be calculated with the weighted peak period as described in Section
2.2.1.

A very difficult issue for using the regular wave method to do the fatigue analysis should
be how to select the wave period associated to the representative wave height assigned to
a load case. This is because there is no unique relation between wave height and wave
period. In addition, the fatigue analysis results are generally susceptible to the wave
period, especially when the vessel RAOs change with wave period rapidly. There are a
few options to solve the problem:

• Option 1 - using some kind of empirical formula to determine the wave period,
for example, using the following relation suggested by DNV:

8
 2.5
⎛ T − 1 .0 ⎞
H =⎜ ⎟ (2.2.5)
⎝ 4. 1 ⎠

where wave height in meter and period in second. But it is noted that the above
relation is an average relation, which means than the period would vary around
the value given by the relation. If period ranges are specified to the load cases, the
middle value of the range should be used as the wave period of the case.

• Option 2 – to ensure a sufficiently accurate calculation of the maximum response,

selecting several periods from a period range for a given wave height, i.e.
separating the load case into a several sub-cases which have the same wave height
but different wave period. If period ranges are specified to the load cases and for
two wave periods to be defined for each case, the two boundary values of the
period range should be used as the two wave periods for the case. For three wave
periods to be defined, the two boundary values plus the middle value of the range
should be used as the three periods for the case.

• Option 3 - the so-called frequency screening method which is a sensitive study of

wave period in frequency domain and the period at which the maximum RAO
amplitude of the parameter that is crucial to the fatigue analysis occurs is finally
suggested.

• Option 4 - a method should be regarded as a modified version of Option 3.

Option 1 is the simplest one, but could not cover the worst environmental conditions that
the riser would be subjected to. Therefore, it is often considered to be not enough
conservative. Option 2 has been more and more applied in the fatigue analysis of the
flexible riser because the method can consider the effect of wave period on fatigue life to
a certain extent. In the meantime, although it increases the number of load cases couple
times compared to Option 1, but still simple relatively to Options 3 and 4.

Option 3 has been used in some fatigue analyses of flexible riser. The major advantage
would be in that the period selected with the method for a regular wave was enough
conservative. The disadvantages are in that a frequency domain software (say
Freecom3D) is needed and that the wave periods given by the method sometimes are too
conservative to generate a reasonable estimation of fatigue life, especially when the
period range includes a resonent point.

The difference of Option 4 from 3 is in that the crucial parameter must be selected as one
motion amplitude of the hang-off point of the riser (e.g. the vertical motion amplitude,
the vertical acceleration amplitude or some rotation amplitude, et al). This means if the
riser top is connected to a fixed structure (e.g. a jacket platform), Option 4 cannot be
used. The remarkable advantage of the method should be in that as long as the crucial
parameter was reasonably selected, the wave period can be calculated quickly with a code
and the conservatism of the wave period selected can be adjusted arbitrarily to a certain
extent. The disadvantage is just a need of a sensitive study to determine the crucial

9
parameter which is time consuming if doing so with time domain software. But for a
certain kind of floating vessel the crucial parameter would be the same. For example, a
study had shown that the crucial parameter for a general FPSO with turret at bow could
be the amplitude of the horizontal motion at the hang off point. Some details of the
method are talked in the following.

With Option 4, the fatigue load cases should be defined also with the wave spectrum (e.g.
for JONSWAP spectrum, the significant wave height Hs, peak period Tp and peakedness
parameter γ, et al). Before applying the method, a study for determining the crucial
motion which has the highest correlativity to the fatigue life of the pipe should be
conducted. For the risers of Salema-Bijupira project in Brazil which will hang on the
turret at the bow of a FPSO, a sensitive study was carried out and it was found that the
resultant horizontal motion at the hang-off point was crucial. That is, the larger the
horizontal motion the larger the curvature variation range subjected to by the top end of
the riser. The procedure for calculating the wave period associated with the wave height,
H, specified for a load case can be illustrated as below:

1. Calculating the RAO curve of the crucial motion amplitude vs. period. For the
horizontal motion amplitude, the following equations should be used:

⎛ Surge * e iϕ surge ⎞
⎜ ⎟
⎜ Sway * e iϕ sway ⎟
⎛ Ax e iϕ x ⎞ ⎛1 0 0 0 z − y ⎞⎜ ⎟
⎜ ⎟ ⎜ ⎟⎜ Heave * e iϕ heave ⎟
⎜ A y e iϕ y ⎟ = ⎜ 0 1 0 − z 0 x ⎟⎜ ⎟ (2.2.6)
⎜ ⎟ ⎜ ⎟⎜ Roll * e iϕ roll ⎟
⎜ Az e iϕ z ⎟ ⎝0 0 1 y − x 0 ⎠
⎝ ⎠ ⎜ iϕ pitch ⎟
⎜ Pitch * e ⎟
⎜ Yaw * e iϕ yaw ⎟
⎝ ⎠

with Ax, Ay and Az – the amplitudes of the translational motions at the hang-off
point in X, Y and Z directions respectively,
X, Y and Z – coordinates originated at the vessel RAO reference point with
the positive X being horizontally forwards to bow, the positive Y being
horizontally to port and the positive Z being vertical upwards,
φx, φy and φz – the phase angles of the three translational motions,
x, y and z – the coordinates of the hang off point.

The amplitude of horizontal motion in XY plane for a certain wave frequency,

M(ω), should be the maximum of the resultant of the translational motions in X
and Y during a wave period, T , and written as:

{
M (ω ) = MAX [ Ax (ωt − ϕ x )]2 + [ Ay (ωt − ϕ y )]2 , for t = 0 to T } (2.2.7)

With numerical method, the relation curve of the amplitude of the resultant
horizontal motion vs. wave frequency can be obtained.

10
2. The significant value of the horizontal motion amplitude can be calculated with
the wave spectral density of S(ω):

∞
M sig = 2 * ∫ 0
S (ω ) * [ M (ω )]2 * dω (2.2.8)

The zero-crossing period of the resultant horizontal motion can be calculated as:

TZH =
∫
0
[ M (ω )]2 * S (ω ) * dω
(2.2.9)
∞
∫ 0
ω 2 * [ M (ω )]2 * S (ω ) * dω

3. The most possible maximum amplitude of the horizontal motion can be calculated
with the following formula [2]:

2 ⎛⎜ 0.2886 ⎞⎟
M max = ln( N ) + M sig (2.2.10)
2 ⎜⎝ ln( N ) ⎟⎠

where N is the number of horizontal motion cycles and calculated for a 3 hour
storm as:

10800
N=
TZH

4. The wave period that should be associated with the specified wave height finally
can be obtained by interpolation with the horizontal motion RAO curve, M(ω),
and the ratio of the wanted value of the horizontal motion amplitude (Msig, Mmax
or a value between them) to the specified wave amplitude H/2. Figure 2-5, as
example, shows the interpolation with the ratio of M/(H/2) being about 0.96
resulting in an associated wave period of about 11.4 s.

1.4
1.2
1
0.8
0.6
0.4
0.2
0
0 5 10 15 20 25
Period (s)

Figure 2-5 Interpolation for associated wave period with horizontal

motion RAO curve

11
From Step 4 above, one can see that the conservatism of the wave period can be
controlled by selecting the wanted of value of the crucial motion amplitude. For fatigue
analysis the significant value of the crucial motion amplitude should be the reasonable
one, while for the extreme dynamic analysis, the most possible maximum value should be
selected.

2.3 Special Considerations

There are some special things which sometimes should be considered in the global
dynamic simulation for fatigue analysis:

• Combination of high (wave) frequency (HF) motion with low frequency (LF)
motion,
• Combination of swell with local wave (sea),
• Simulation with irregular wave,
• VIV simulation.

2.3.1 Combination of HF motion with LF motion

The vessel slow drift is a kind of LF motion whose period could be from 50 s to more
than 200 s. When the vessel slow drift is specified, the global dynamic simulation should
consider the effect of combination of HF motion with the LF motion. For conservative
purpose, it is generally to assume the phase difference between HF and LF motions to be
the value with which the two motions simultaneously reach their maximums or
minimums whatever the worst results to the riser could be yielded.

2.3.2 Combination of Swell and Sea

When the wave environment is specified with both swell scatter diagram and the sea
scatter diagram, the total swell and sea spectrum may be developed by adding the
individual swell and sea spectra as:

S total ( f ) = S swell ( f ) + S sea ( f ) (2.3.1)

If the amount of energy in the two spectra is comparable, the resultant spectrum should
have two distinct peaks. The total significant wave height due to swell and sea is the
square-root of the sum of the squares of the swell and sea significant wave heights, i.e.:

H s ,total = H s2, swell + H s2, sea (2.3.2)

2.3.3 Irregular Wave Analysis

Recently, more emphasis has been placed on the use of the irregular wave approach to do
the global dynamic simulation for fatigue analysis. The benefit of using irregular wave
approach against the regular wave is in that it could make the global simulation much
closer to the real behaviors of the riser in the field, although much more CPU time is

12
needed compared to the regular wave approach. The irregular wave approach has been
applied to steel risers. For flexible risers, however, the approach is still in the stage of
investigation of this possibility because there are many difficulties for calculating the
fatigue stresses in flexible riser. Unlike steel riser, stresses in flexible riser cannot be
generated directly during the global dynamic simulation and must be separately
calculated with a local stress analysis by using some special software after the global
simulation, which makes the fatigue analysis not affordable in time.

2.3.4 Fatigue Induced by VIV

For deep water risers there is a concern on the fatigue damage that could be induced by
vertex induced vibration (VIV). A study for this issue had been carried out with a 5.625
inch ID flexible riser in water depth of 1090 m in Gulf of Mexico. The riser is a Free-
Hang riser. The commercial software Shear7 and some special codes for local stress
analysis were used. According to the current scatter diagram, 9 load cases were defined.
Figure 2-6 shows the distribution of the maximum fatigue damage induced by VIV in 20
years along the riser where the arc length ratio of zero denotes the touchdown point
(TDP) of the riser, while 1 denotes the top hang off point. From the figure we can see that
the maximum damage occurs near to the TDP and its value is negligible. In the study
many conservative assumptions are imposed, e.g. assuming the riser vibration to be in
single mode and the riser pinned on the seabed etc. Therefore, it could be concluded that
the fatigue damage induced by VIV generally should be not a concern for flexible risers
because:

• The fatigue damage induced by VIV is small, and

• The maximum damage occurred near to TDP where the wave induced damage in
the riser is very small.

9.00E-06

8.00E-06

7.00E-06

6.00E-06
Fatigue Damage

5.00E-06

4.00E-06

3.00E-06

2.00E-06

1.00E-06

0.00E+00
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

arclength ratio (x/L) from sea bed

Figure 2-6 Distribution of Fatigue Damage Induced by VIV along Riser

13
3.0 CALCULATION OF CURVATURE VARIATION IN
CYCLIC BENDING OF PIPE
The critical fatigue damage almost always occurs at the top connection for a flexible
riser. Although the features of the top connection can be simulated to a certain extent
with the common used commercial software, e.g. OrcaFlex and Flexcom3D, many details
of the top connection cannot be perfectly considered. For example, with a bending
stiffener being used, the gap between the pipe and the bore of the stiffener, the relative
slip between them and the nonlinear relationship of stress vs. strain of the stiffener
material cannot be exactly modeled in the commercial software. In the meanwhile, the
details sometimes could give considerable effect on the fatigue life estimation. It is better,
therefore, to develop a FE model for the top part of the riser with all important details of
the top connection for evaluating the curvature variation of the pipe. The local analysis
should be based on the global analysis results, generally the extreme top angles and the
associated effective tensions or the extreme top tensions and the associated top angles.
With the detailed local analysis, the top connection of the riser can be simply assumed as
pinned on the vessel in the global dynamic simulation.

3.1 Local Curvature Analysis Model

The critical fatigue damage typically occurs at the top connection of the flexible riser.
Although the features of the top connection can be simulated to a certain extent with most
commercial software, several details of the top connection cannot be represented
adequately. For example, when a bending stiffener is being used, the gap between the
pipe and the bore of the stiffener, the relative slip between the pipe and the stiffener, and
the nonlinear relationship of stress vs. strain of the stiffener material cannot be exactly
modeled in the commercial software.

However, these details sometimes have a considerable effect on the fatigue life
estimation. Therefore, a detailed local FE model for the top part of the riser is used for
evaluating the local curvature variation of the pipe. The local analysis is based on the
global analysis results, replicating the extreme top angles and the associated effective
tensions or the extreme top tensions and the associated top angles. Performing the
detailed local analysis subsequent to the global dynamic analysis allows the top
connection of the riser to the vessel to be simplified to a pinned connection in the global
dynamic simulation.

The most severe location for fatigue damage in the riser is usually in the top hang off
region. The riser is protected from over bending in this area by either a bend stiffener or a
bellmouth. The detailed local analysis for the curvature distribution along the riser at the
top stiffener or bellmouth is carried out using a 2D FE model. Figure 3-1 shows the basic
arrangement for the top stiffener or bellmouth.

The bend stiffener is modeled by a 2D tapered unsymmetric beam element and the pipe is
simulated by a 2D beam element. The interface between the pipe and the bend stiffener
is represented by a 2D general contact element. Both the non-linear stress strain curve of

14
the bend stiffener and the non-linear bending curvature hysteresis loop are considered in
the analysis. If a bellmouth is used, a steel 2D solid element is adopted.

Knowing the top tensions and angles for each load case from the global dynamic
analysis, the curvature distribution along the flexible riser can be found by applying these
tensions and angles at the bottom of the model as can been seen in Figure 3-1.

Figure 3-1 Models for Local Curvature Analysis about Bell Mouth and Stiffener

3.2 Maximum Bending Angle Range Calculation

With the top end of riser being assumed pinned on the vessel, the top angle and the
associated top tension time traces can be obtained from the simulation for a wave cycle.
Generally, the time traces of the last wave period from the simulation are taken by
considering the stable solution could be achieved just in the last cycle. If taking a unit
length of pipe at the top end of the riser, the lower end of the piece of pipe should draw a
closed cycle in a 3D space during a wave period. There are should be two special
orientations for the piece of pipe between which the bending range is larger than any
other two orientations, i.e. the maximum. Figure 3-2 schematically shows the above
thought. In the figure, V1 and V2 can be regarded as two vectors and represent the two
special orientations of the top piece of pipe. Coordinate Z stands for the center line of the
top stiffener or bell mouth (which ever used) and the positive direction is from the fixed
end to the tip end. The rotational vector, Rotation, denotes the total rotation between V1
and V2. The component of the total rotation in XY plane, Bend, should represent the
bending range, and the component in the negative Z direction, Twist, represent the twist
range between V1 and V2. The bending range can be expressed in the vectors of V1 and
V2 as:

⎟ sin ⎨arccos⎢ Z • (V 1 × V 2) ⎥ ⎪⎬
⎛ V 1 • V 2 ⎞ ⎧⎪ ⎡ ⎤⎫
BendingRange = arccos⎜⎜ ⎟ (3.2.1)
⎝ V 1 V 2 ⎠ ⎪⎩ ⎣⎢ Z V 1 × V 2 ⎦⎥ ⎪⎭

15
It is not difficult to develop a code to find the maximum bending range with the top angle
time traces resulting from the last wave period of the simulation.

Figure 3-2 Maximum Bending Range in a Wave Cycle

4.0 CALCULATION OF STRESS IN TENSILE ARMORING

LAYER
4.1 General
The axial tension and the internal pressure are the principal loads subjected to by the
flexible pipe in an offshore application, which both could induce very large tension in the
tensile armoring layers. In addition, all periodical loads induced by both HF and LF
motions of the floating vessel and the hydrodynamic forces should introduce alternating
tensions in the tensile armoring layer which could give rise to fatigue damage problem.
Therefore, the fatigue life of the tensile armoring wires is one of very important criteria
of the pipe design.

The alternating stresses in the tensile armoring wires are mainly generated from the
following two sources:

1. Effective tension variation in pipe induced by the HF and LF motions of the

vessel (or called as RAO motion and drift)
2. Curvature variation of pipe induced by the HF and LF motions of the vessel

The torque variation in pipe should also induce alternating stresses in the tensile wires.
But there are some difficulties to correctly calculate the torque in a flexible pipe with the

16
global dynamic simulation which generally was conducted by assuming there were no
torsional motion in the pipe. Therefore, the alternating stresses due torque variation are
generally assumed to be negligible.

The internal pressure variation also induces alternating stresses in the tensile wire. But, in
most cases of application of flexible pipe, the frequency of internal pressure variation is
very low compared with effective tension variation and curvature variation. Therefore, in
fatigue analysis, the alternating stresses due internal pressure variation are neglected. But
it should be noted that the internal pressure has important effect on the mean stresses in
the tensile wire.

Straightforward one part of alternating stresses in the tensile wire should be considered
directly resulting from the effective tension variation in the pipe. There should be some
available software developed by the flexible pipe manufacturers for calculating the
alternating stresses due to effective tension variation.

The alternating stresses in the tensile wires due curvature variation in the pipe can be
sorted into the following three kinds:

1. Axial stresses due to interlayer friction, σf, which are uniform distributed on the
cross section of the wire and can be tensile or compression depending on the
cross section considered being on the extrados or intrados region of the bent pipe
2. Bending stresses due to normal curvature change of the wire, σbn, which linearly
vary from minimum (negative) to maximum (positive) through the thickness of
the wire
3. Bending stresses due to transverse curvature change of the wire, σbt, which
linearly vary from minimum (negative) to maximum (positive) through the width
of the wire

Besides the above three kinds of alternating stresses due to pipe bending, there should be
torsional stresses induced by torsion change of the wire as pipe bending. But the torsional
stresses are generally very small compared to the stresses due curvature change.
Therefore, in the present paper, the alternating torsion stresses are ignored.

To sum up, the mean stress in the tensile wire in a certain loading condition should be
regarded as a function of:

• Average effective tension in the pipe

• Internal and external pressures of the pipe
• Average curvature of the pipe

The alternating stress in the tensile wire should be generally regarded as a function of:

• Variation range of effective tension in the pipe

• Variation range of curvature of the pipe

17
4.2 Stress in Tensile Wire under Axisymmetric Loads
The effective tension and the internal and external pressures of a flexible pipe result in a
pipe wall tension that can be written as the following equation:

Twall = Te + PiAi − PoAo (4.2.1)

with Twall – pipe wall tension,

Te – effective tension of the pipe,
Pi and Po – internal and external pressures,
Ai and Ao – internal and external cross section areas of the pipe.

Although the pipe wall tension is mainly taken by the tensile armoring layers, all of
layers of the pipe wall are considered as a combined structure having compatible
deformations under the pipe wall tension effect. The so-called plane assumption with that
any cross section of the unstrained pipe always remains in a plane during straining is the
basis of the model for calculating the stresses in the pipe wall layers. For tensile wires,
the Love’s theory for thin rod [3] are applied. In the theory, there are a total of 6
equilibrium equations. But for a tensile wire in straight pipe the 6 equations can be
reduced to two, which can be expressed as:

− Nτ + Tk + X = 0
(4.2.2)
− Gτ + Hk − N = 0

with N – shear force in tensile wire in the wide direction,

G – bending moment in tensile wire about the wide direction,
T – axial tension in tensile wire,
H – torque in tensile wire,
X – external distribution force per unit length in the thickness direction,
τ – torsion of tensile wire,
k – normal curvature of tensile wire.

In addition, there are 2 physical equations which can be expressed as:

G=EI*k, H=C* τ and T=EAεa (4.2.3)

with EI – normal bending stiffness of tensile wire,

EA – axial stiffness of tensile wire,
εa - axial strain of tensile wire,
C – torsion stiffness of tensile wire.

If assuming there are no residual stresses in the tensile wire of an as-built pipe, the above
equations 4.2.2 and 4.2.3 can be considered as incremental relations. The incremental
deformations of τ and k can be expressed in the incremental radius of the tensile armoring
layer, δr and the incremental lay angle, δα as:

18
 sin 2 (α + δα ) sin 2 α
k= −
r + δr r (4.2.4)
sin(α + δα ) cos(α + δα ) sin α cos α
τ = −
r + δr r

In turn the increments of δr and δα have the following relation:

L(1 + ε ) tan(α + δα ) = 2π (r + δr ) + ϕLr (4.2.5)

with L – pitch of tensile wire,

ε – axial strain of pipe,
α – lay angle of tensile wire (angle between wire and center line of pipe),
r – radius of tensile armoring layer,
δα – incremental lay angle,
δr – incremental radius of tensile armoring layer,
φ – torsion per unit length of pipe

The above relation can be interpreted schematically in Figure 4.2.1. From the plot it is
easy to find out the axial strain of the wire can be expressed as:

sin α
εa = [2π (r + δr ) + ϕLr ]2 + L2 (1 + ε ) 2 − 1 (4.2.6)
2πr

Finally, if the deformations of the pipe, ε and φ, are known, there are 10 unknowns for
the tensile wire:

• 1 external force of X,
• 4 internal forces of G, H, N and T and
• 5 deformations of δr, δα, k, τ and εa.

However, there are just 9 equations given in the above:

• 2 equilibrium equations, Eqs. 4.2.2,

• 3 physical equations, Eqs. 4.2.3,
• 4 geometry equations, Eqs. 4.2.4 to 4.2.6.

The needful 10th equation can be established using the compatibility condition between
layers in the pipe wall.

If similar works have been done for all layers of the pipe, eventually, the following
equilibrium equations for a loaded straight pipe can be obtained:

⎧ε ⎫⎧ K11 , K12 ⎫ ⎧Twall ⎫

⎨ ⎬⎨ ⎬=⎨ ⎬ (4.2.7)
⎩ϕ ⎭⎩ K 21 , K 22 ⎭ ⎩Tq ⎭

19
with Kij for i and j=1,2 are the stiffness which are the functions of structure properties of
all pipe wall layers and the pipe loading conditions. By contrast, if just the external loads
of the pipe, effective tension, internal and external pressures and torque, are known, the
pipe deformations of ε and φ can be also solved from the equations.

Figure 4.2.1 Geometry Relation of Deformation of Tensile Wire

4.3 Stress in Tensile Wire as Pipe Bending

4.3.1 Stress in Tensile Wire due to Interlayer Friction

The stresses in the tensile wire due to pipe bending can be decomposed into two kinds:

• Uniform axial stress induced by the interlayer friction as pipe bending, σf,
• Bending stress generated due to curvature change of the wire.

The very difficult and very uncertain stress calculation is for the stress in tensile wire due
to interlayer friction as pipe bending because:

• It is very difficult to assume the friction coefficients between the tensile wire and
the adjacent layers although the coefficients could be easily obtained in laboratory
and library. This is because the annulus conditions (moisture, lubrication and
corrosion) of the pipe could change with time and the friction coefficient is often
very sensitive to the conditions.
• The friction coefficient could change along the wire by considering the slip of the
wire relatively to the underlaying layer varies along the wire.

20
 • The normal pressure on the wire should periodically change along the wire from
the extrados to the intrados of the bent pipe and up to date there seems no
practicable method to evaluate it.

To overcome the above difficulties, a model for evaluating the equivalent bending
stiffness of the pipe in a bending cycle was develop based on some available bending test
data. As example, Figure 4.3.1 shows three hysteresis loops resulting from a bending test
with a 4-inch ID flexible pipe made at Wellstream [4]. The hysteresis loops can be
approximately regarded as a parallelogram and the slope of the longer diagonal of a loop
should represent average bending stiffness of the pipe in a bending cycle as shown in
Figure 4.3.2. In the meantime, it was found that the shape and dimensions of the
parallelogram have certain relations to:

• The pipe ID,

• The internal pressure Pi,
• The nominal bending stiffness EI which mainly the sum of the bending stiffness
of all plastic layers (regarded as separated cylinders) and the bending stiffness of
the tensile wires (regarded as a number of individual helixes),
• The curvature range in a pipe bending cycle ΔC.

With the model, the equivalent bending stiffness, EIq, for a bending cycle undergone by
the flexible pipe can be expected if the curvature range is known, i.e. EIq can be
expressed as:

EIq=EIq(ID, Pi, EI, ΔC) (4.3.1)

The available test data showed that the equivalent bending stiffness was always
considerably larger than the nominal bending stiffness that could be roughly considered
as the pipe bending stiffness without taking the contribution from the interlayer friction
into account. Therefore, the difference subtracting the nominal stiffness from the
equivalent stiffness, EIq-EI, could be regarded as the bending stiffness contribution from
the interlayer friction on all tensile wires.

21
 Bending Moment vs. Curvature of WS 4" PIPE

5
f = 0.1 Hz 4
3
2
1
0
-0.15 -0.1 -0.05 -1 0 0.05 0.1 0.15
-2
-3
-4
-5
curvature (Rad/m)

Pi=200 psi Pi=100 psi Pi=7 psi

Figure 4.3.1 Bending Moment vs. Curvature in Bending Test

10

9 from test
parallelogram
8 diagram for equivalent EI
Bending Moment (kN-m)

0
0 0.05 0.1 0.15 0.2
Curvature (Rad/m)

Figure 4.3.2 Parallelogram Model for Equivalent Bending Stiffness

If assuming all tensile wires completely bonded in the pipe wall, the axial stress in the
wire due to pure bending of the pipe, σ s , can be expressed as [5]:

⎛ R 2 + 2rR cos 2 α cosθ + r 2 cos 2 α cos 2θ ⎞

σ s = E⎜ − 1⎟ (4.3.2)
⎜ R ⎟
⎝ ⎠

with R – bending radius of the pipe,

θ – position angle of the cross section of wire considered measured from the
extrados of the pipe.

22
To propose a simple method for approximately evaluating the stress in wire due to
interlayer friction as pipe bending, σf, it is assumed that the stress has a linear relation to
σs with the ratio of the bending stiffness formed by all friction forces on the wires, EIq-
EI, over the bending stiffness by all completely bonded wires, EIs, i.e.

EIq − EI
σ f =σs (4.3.3)
EIs

The bending stiffness formed by all completely bonded wires, EIs, can be calculated with
the following equation:

m n
EIs = R∑∑ σ s ,ij t j w j r j cos α j cos θ ij (4.3.4)
j i

with R – the bending radius of pipe,

m – the number of tensile armoring layers,
n – the number of wires in the jth layer,
σs,ij - the axial bonded stress in the ith wire of the jth layer,
tj and wj – the thickness and width of wire in the jth layer,
rj and αj – the radius and lay angle of the jth layer,
θij – the position angle of the ith wire measured from the extrados of the pipe.

It should be noted that the stress in wire due to interlayer friction depends not only on the
pipe curvature at the moment but also on the curvature range in a bending cycle.
Therefore, the above method is just suitable to the fatigue analysis. The advantage of the
method is in that the difficulties for estimating the interlayer friction coefficients can be
avoided with a equivalent bending stiffness model developed based on bending tests.

4.3.2 Stress in Tensile Wire due to Wire Curvature Change

The tensile wires in a straight pipe can be regarded as a number of helixes. When pipe
bends, the wires could have two extreme positions:

• Bent helix position which means the wire does not slip at all relative to the
underlayer during pipe bending,
• Geodesic position which means the wire slip enough into the position at which the
deformation energy in the wire reaches the minimum.

In most cases, tensile wires would finally stay at a position just between the two extreme
ones as the pipe is bent. However, some fatigue test data show that the position of the
wire during cyclic bending is very close to the bent helix position [6]. In addition, the
bent helix position generally results in well larger bending stresses in the wire than the
geodesic position. To be conservative, therefore, the bent helix position should be
assumed in the fatigue analysis of tensile wire.

23
For bent helix, the maximum bending stress due to normal curvature change of the wire,
σbn, can be written in Equation (4.3.5) [5], while the maximum bending stress due to
transverse curvature change, σbt, can be written in Equation (4.3.6) [5]. It is noted that
here the cross section of the wire is assumed as a txw rectangle.

Et − R + 2 R cos 2 α + r cos θ cos 2 α

σ bn = cos θ cos 2 α (4.3.5)
2 2rR cos 2 α cos θ + r 2 cos 2 α cos 2 θ + R 2

Ew r 2 cos 2 α cos 2 θ + 2rR cos 2 α cos θ − R 2 cos 2 α + 2 R 2

σ bt = sin θ cos α 3
2
(r 2 cos 2 α cos 2 θ + 2rR cos 2 α cos θ + R 2 ) 2
(4.3.6)

To sum up, the maximum and minimum normal stresses on the cross section of the
tensile wire in a cyclically bending pipe can be expressed as:

σ max = σ a + σ f + σ bn + σ bt
(4.3.7)
σ min = σ a + σ f − σ bn − σ bt

where σa is the uniform axial stress in wire due to the axisymmetric loads (the effective
tension and the internal and external pressures).

5.0 CALCULATION OF STRESS IN PRESSURE

ARMORING LAYER
Alternating stresses in the pressure armor layer result from relative motion of adjacent
pressure armor rings under high contact pressure at the layer OD and ID. Since the
pressure armor cross-sectional profile is an irregular shape, methods of analytical
mechanics cannot be used reliably to predict the alternating stresses under this combined
loading. Therefore, a finite element approach is used to calculate the stresses.

To conduct the pressure armor fatigue analysis, an axisymmetric ring finite element
model of the actual pressure armor profile is used. The typical FE model is illustrated in
Figure 5-1.

Figure 5-1 Pressure Armor Fatigue Model

24
In a cross-sectional stress analysis, the contact pressure at the OD and ID of the pressure
armor layer is determined. The curvature range is used to calculate the axial displacement
of the pressure armor layer, assuming that the plane sections remain plane as the pipe is
bent. Only one wrap of the pressure armor layer is needed in the analysis, with a partial
wrap on either side to assure proper loading at the boundaries of the FE model.

The Z-shape cross-section will be forced to twist under internal and external pressure. It
is this twist of the section that causes the “hook” and “valley” of the section to come into
contact, which results in an increased bending stress at the internal corners of the Z-shape
section. The bending motion will cyclically change the contact pressure position which
then results in an alternating stress.

The alternating stresses in the pressure armor layer are determined by applying the
internal and external pressure with the relative axial position of adjacent rings located
based on the minimum and maximum curvature in a wave cycle. The maximum
alternating stresses have been determined to occur near the internal corners of the profile.

6.0 CALCULATION OF FATIGUE DAMAGE AND LIFE

6.1 Theory of Cumulative Damage Calculation
The cumulative fatigue damage is calculated based on the linear theory (Miner’s Rule):
k
ni
D=∑ (6.1)
i =1 Ni
where
D is the total damage
k is the total number of load cases
i is the load case number
ni is the number of waves in the i-th load case, and
Ni is the expected fatigue life under the conditions of the i-th load case.

6.2 Fatigue Life Calculation

The fatigue life is calculated with SN curve which results from fatigue tests with wire
samples in laboratory. The SN curve can be written in the form:

Log ( Sr ) = A * Log ( N ) + B (6.2)

where
Sr is the stress range in MPa,
N is the expected life in cycles,
A & B are the slope and interception of the SN curve, respectively.

Figure 6-1 for example shows a SN curve of tensile wire in 10% CO2 environment.

25
 3.5

design SN curve
mean SN curve
3
test data
no failure
log(Sr)
2.5

1.5
4 5 6 7
log(N)

Figure 6-1 SN Curve of Tensile Wire in 10% CO2 Water

When the total damagefor all load cases, D, is calculated (see Eq. 6.1), the fatigue life of
can calculated as:
L
Life( years) = (6.3)
D

where L is the required service life in year on which

7.0 FLOWCHART OF FATIGUE ANALYSIS PROCEDURE

The following flowchart summarizes the procedure of fatigue analysis for flexible riser.

26
 Wave Scatter Wave Direction
Diagram Information
Refine load case Refine load case
definition with wave definition with
height or/and period wave directionality
or

Direct discretization on Generate individual wave height

scatter diagram to define distribution curve and then
load cases classify load cases on the curve

or
Define wave period for each load case

Run global dynamic simulation

And/or
Post-process for time traces of Post-process for time traces of
top tension and angle at riser tension and curvature somewhere on
top connection the riser, e.g. sag bend

Local curvature analysis with bend stiffener Find the extreme

or bell mouth for extreme combinations of combinations of tension and
tension and curvature curvature in sag bend

Fatigue stress calculation in

Fatigue damage and life tensile wires and the pressure
calculation with Minor’s armoring layer
rule and SN curves

1st NO If estimated fatigue life satisfy the 2nd NO

required service life?

YES

STOP

27
REFERENCES:

[1] DNV, Environmental Conditions and Environmental Loads, March 1991

[2] S.K. Chakrabarti, Hydrodynamics of Offshore Structures, 1987
[3] Love, A.E.H., A treatise on the Mathematical Theory of Elasticity, 4th Edition, 1944,
Dover Publ., New York
[4] SINTEF Structural Engineering, Fatigue Testing of a Wellstream Flexible Pipe,
Report 2.2-14, 1992
[5] J.A. Witz and Z. Tan, Rotary Bending of Marine Cables and Umbilicals, Engineering
Structures, Vol. 17, No. 4, PP. 267-275, 1995
[6] Wellstream Report, Calibration of Wellstream Fatigue Life Analysis Model with
Respect to Troll C 15” Riser Dynamic Tests, A951m001, Rev.1 January 1999

28