You are on page 1of 13

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/306581287

Recent Advances in Pipeline Free Span Design: A New Revision of DNV-RP-F105

Conference Paper · June 2016


DOI: 10.1115/OMAE2016-55010

CITATIONS READS
2 757

3 authors:

Knut Vedeld Olav Fyrileiv


SWECO DVN GL
35 PUBLICATIONS   223 CITATIONS    46 PUBLICATIONS   287 CITATIONS   

SEE PROFILE SEE PROFILE

Håvar Sollund
Det Norske Veritas
30 PUBLICATIONS   178 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Advanced non-linear dynamic pipe-soil interaction in multi-spans View project

Fracture toughness View project

All content following this page was uploaded by Knut Vedeld on 25 August 2016.

The user has requested enhancement of the downloaded file.


RECENT ADVANCES IN PIPELINE FREE SPAN DESIGN: A NEW REVISION OF
DNV-RP-F105

Knut Vedeld Håvar Sollund


DNV GL DNV GL
Oslo, Norway Oslo, Norway

Olav Fyrileiv
DNV GL
Oslo, Norway

ABSTRACT F101 [4] and API RP 1111 [5], both recommend that free span
Pipeline free span design has evolved from basic avoidance design is performed according to DNV-RP-F105 [3]. Over the
criteria in the DNV ’76 rules [1], to fatigue and ultimate limit last decade, DNV GL have accumulated experience with how
state considerations in Guideline no. 14 [2]. Modern multi- DNV-RP-F105 has been applied in the industry, and also how
mode, multi-span free span design is predominantly performed the application of free span technology has gradually changed.
according to DNV-RP-F105 [3]. Changes to free span technology are influenced by advances in
In 2006, the latest revision of DNV-RP-F105 [3] was several areas such as global buckling, deep water technology
written as a direct result of extensive research, performed due to and increasingly efficient and comprehensive numerical
significant free span challenges in the Ormen Lange pipeline analysis tools.
project. DNV-RP-F105 was at the time, and still is, the only Since the last revision of DNV-RP-F105, significant work
pipeline design code giving contemporary design guidance for has been done on interpretations on free span dynamic
vortex induced vibrations (VIV) and direct wave loading design response, particularly on structural modelling. Improved
for pipelines in free spans. analytical and semi-analytical methods for modal response
The last revision of DNV-RP-F105 included a few, but calculations have recently been developed, allowing for more
highly important advances, particularly the consideration for accurate and reliable calculations of eigen frequencies and
multi-mode and multi-span pipeline dynamic response behavior. modal stresses in single- and multi-spans with little or no
In the 10 years that have followed, no breakthroughs of similar additional computational effort [6-8].
magnitude have been achieved for pipeline free spans, but a A significantly improved frequency domain calculation
large number of incremental improvements to existing method for direct wave loading has been developed, which
calculation methods, and some novel advances in less critical allows for accurate fatigue and extreme environmental stress
aspects of VIV understanding have been made. As a result, calculations for both isolated single spans and interacting multi-
DNV-RP-F105 has recently been revised to account for these spans with several responding modes [9]. The previous model
advances, which include improved frequency-domain analyses assumed that the main damage contribution was associated with
of wave-induced fatigue, a new response model for cross-flow the fundamental mode, and could not reliably account for
VIV in low Keulegan-Carpenter (KC) regimes in pure waves, multiple interacting free spans. In addition, only a single critical
new analytical methods for dynamic response calculations of location was conventionally considered. With the introduction
short spans in harsh conditions, and extensive guidance on how of a true multi-mode stress-response spectrum, the stress
to apply the recommended practice for assessment of fatigue distribution is now allowed to vary along the pipe length.
and extreme environmental load effects on curved structural More elaborate guidance has been given on the problem of
members such as spools, jumpers and manifold flexloops. distinguishing between isolated single spans and interacting
This paper gives an overview of most of the important multi-spans, compared to the previous approach which was
changes and updates to the new revision of DNV-RP-F105. considered hard to apply and too unreliable [10].
Case studies are used to demonstrate the importance and effects VIV and dynamic response due to pure currents is a vast
of the changes made, and to some extent how the revision of and ongoing field of research. Research groups focused on riser
DNV-RP-F105 can enhance its applicability and ease of use. technology are mainly concerned with pure currents since it is
normally considered to be a conservative assumption to
1. INTRODUCTION disregard the influence of waves. On pipeline free spans, this
The existing design codefor free spanning pipelines, DNV- assumption is not correct and waves must be considered. DNV-
RP-F105 [3]was last revised a decade ago. Two of the most RP-F105 has traditionally considered combined waves and
prominent standards for offshore pipeline design, DNVGL-OS- currents, but cross-flow VIV due to waves in low Keulegan-

1
Carpenter regimes was not covered in the 2006 revision [3]. In  The separate single- and multi-mode VIV response
the new revision, a novel approach to account for cross-flow models have been merged. The historic distinction
VIV in pure wave conditions in low Keulegan-Carpenter flow between single- and multi-mode analyses is considered
regimes has been included for both fatigue and ultimate limit
artificial and obsolete.
state considerations [11].
Jumpers, spools, piping systems and manifold flexloops  A separate response model for cross-flow VIV in wave
have traditionally been designed according to VIV avoidance dominated low KC (Keulegan-Carpenter) flow regimes
criteria. In recent years, it has been more common to reduce (LKCR) (based on [11]) has been introduced and
conservatism by allowing for VIV in design of such structures harmonized with the general multi-mode approach.
[12-16]. The implication of allowing for VIV in design is to  General guidance on how to apply DNV-RP-F105 to
account for fatigue accumulation and extreme environmental
spools, jumpers, flexible loops and piping systems has
loading due to VIV. Due to the lack of general guidance on VIV
for geometries which include bends, it has been observed that been introduced in a new Appendix.
DNV-RP-F105 increasingly is the preferred design document to  Highly accurate modal response quantity calculation
cover VIV analyses of such structures, even if the document procedures (based on [8]) have been introduced for
was not initially intended for that purpose. Recent progress in short spans, typically relevant for river crossings, inlets
this area has also been made on the empirical side, since and straits with extreme environmental conditions.
realistic physical experiments have been conducted [17]. To
 A new section has been included, to provide some
account for updated knowledge and the industrial application of
the recommended practice to curved geometries, DNV-RP-F105 guidance for how to apply and interpret sensor
has been updated with a new appendix, which describes all the technology on free spans, and how the application of
additional considerations the designer needs to make in order to sensors influences the safety factor format.
conservatively apply DNV-RP-F105 to fatigue and ultimate  Systematic revisions to clarify symbols, text,
limit state calculations of jumpers, spools, goosenecks, doglegs,
limitations and ranges of applicability.
piping systems and manifold flexloops.
There have been some recent failures due to VIV of
The new revision has improved a number of key physical
pipelines in river crossings where floods or erosive processes
models, introduced guidance to new aspects where no real
have dug up pipelines [18]. Similar problems have proven
alternative guidance is given elsewhere, and the document has
relevant for offshore pipelines in straits, or in areas where
also systematically been improved based on experience and
strong erosive processes may excavate buried pipelines. There
feed-back provided by users of the previous revision.
is relatively little guidance on how to treat very short spans in
Apart from the brief description of the motivations and
extreme environments with significant flow velocities. Recently,
changes to the recommended practice presented above, the
however, analytical methods have been developed, which are
paper will focus on two main areas:
capable of performing a large number of analyses quickly,
satisfying the need for simple, robust and accurate means of
1. How the novel frequency plane force model presented
dynamic response calculations for short spans in harsh
by Sollund et al. [9] can be incorporated in DNV-RP-
environments [8].
F105. The impact of the novel method on realistic
Presently, it is not common to apply sensor technology to
design scenarios is studied.
monitor fatigue exposure of pipelines in free spans due to VIV
2. The new cross-flow response model for LKCR has
and direct wave action. There are, however, isolated and special
been extensively described in [11], and is already
cases where such technology has been applied and it is expected
adapted to the format of DNV-RP-F105. However, the
that sensor technology will be more relevant in the future. In the
impact of the new model on typical design cases will
new revision of DNV-RP-F105, a novel section has been
be investigated further herein.
included to provide some basic guidance on specific challenges
and considerations pertaining to the use of real-time monitoring
Due to the extensive revision of the recommended practice, the
devices for assessment of wave- and current-induced damage.
present paper only includes a high level introduction to the
The main changes from the 2006 revision to the new
remaining changes.
revision of DNV-RP-F105 may be summarized as follows:

 An updated stochastic frequency domain approach for


2. IMPROVED FREQUENCY-DOMAIN DIRECT WAVE
direct wave action has been introduced (based on [9]), ACTION CALCULATIONS
which allows for multi-mode and multi-span analyses.
The model also allows fatigue to be calculated as a 2.1 Background
function of the position along the pipe axis.

2
A force model based on the Morison equation is described and, RD is a factor accounting for wave spreading and
in the 2006 revision of DNV-RP-F105 [3] for assessment of direction, b is a linearisation constant, gD is the drag force term,
load effects from direct wave action. A procedure is also gI is the inertia force term, G(ω) is the frequency transfer
detailed for a stochastic analysis of wave-induced fatigue in the function from sea surface elevation to wave-induced flow
frequency domain. The frequency-domain analysis is based on velocity at pipe level, Sηη is the single-sided wave elevation
the assumption that the main damage contribution comes from spectrum for the irregular sea state, ωj = 2πfIL,j/γf is the still-
the fundamental mode, and that higher-order modes can be water in-line angular natural frequency for the jth mode, me is
neglected, for both single- and multi-spans. The assumption the effective mass per unit length incl. added mass, N is the
seems reasonable for single spans, where the second lowest number of modes with non-negligible damage contribution and
mode generally will not contribute to fatigue damage (since it is ζT,j is the total damping ratio for the jth mode. The total damping
anti-symmetric), and the third mode has an eigen frequency ratio includes contributions from structural, soil and
typically more than 5 times higher than the fundamental hydrodynamic damping. The mode shape weighting factor λj in
frequency. For interacting multi-spans, however, the modes may Eq. (2) can be interpreted as a measure of how strongly a
be closely spaced and the fundamental mode will normally not particular mode will be responding to the assumed uniform, and
be sufficient to appropriately account for the dynamic response. thus perfectly symmetric, wave loading. It is given by
It has recently been documented based on time-domain
simulations that the third mode may contribute non-negligibly
L  j ( x)dx
j  2 (3)
(by an estimated 10 to 15%) to fatigue damage even for single-   j ( x)dx
spans [8]. In the paper by Sollund et al. [8], an extension of the L

frequency-domain methodology given in the new revision of where φj is the mode shape of the jth mode. ψj is the modal
DNV-RP-F105 is presented, which allows multiple modes to be stress amplitude for the jth mode, and may be expressed as
considered. This extended multi-mode representation of the
  2 j 
stress response spectrum was demonstrated to be highly  j x   1  CSF Er  2 
 x  (4)
accurate compared to sophisticated time-domain analyses, and  
differences between the two methods were found to be small
[8]. where E is the Young’s modulus and the radial coordinate r is
The novel frequency domain solution has been included in taken as Ds/2 on the outside of the pipe wall and as (Ds-2ts)/2 on
the electronic implementation of DNV-RP-F105 FatFree [19], the inside.
allowing fatigue lives and peak stresses from single- and multi-
mode calculations to be compared for a number of free span 2.3 Multi-mode assessments
scenarios. In the present context, a small study is conducted to Generally, modal stresses vary with position, and an
simplify the introduction of the multi-mode approach for single accurate frequency-domain assessment of fatigue requires a
spans, and a study of the implications to multi-spans is also critical location to be determined for each span based on fatigue
presented. calculations for a large number of locations along the pipe axis.
Such analyses will be quite time consuming. However, for a
2.2 Mathematical formulations single-span there will generally be only two critical locations,
The mathematical expressions for the fatigue damage as a namely in the middle of the span and on the shoulder.
function of the sea-state specific standard deviation of the stress It has previously been demonstrated that including the third
response σS in the previous revision of DNV-RP-F105 [3] are mode normally increases the fatigue damage on the shoulder
kept, but changes are introduced to the stress spectra where the modal stress for the third mode has a local maximum
themselves, and their related parameters. When non-negligible [8]. The third mode also has a maximum value in the middle of
damage contributions from higher order modes cannot be the span. However, the cross-spectrum between the first and
excluded, the one-sided stress response spectral density third modes is generally negative at the latter location, resulting
function SSS(x,ω) is given by in a reduction of the fatigue damage as illustrated in Figure 1. It
S SS x,    R D2  b 2 g D2   2 g I2 G 2  S   
can further be observed from the figure that the third mode
increases the damage on the span shoulders. It should be noted
N N  j j x  k k x  T jk

j 1 k 1 m 2

T  U 2jk
2
(1) that modes that are anti-symmetric with respect to the mid-span
location (i.e., the second and fourth modes) will not be excited,
e jk
since the wave loading is assumed uniform along the span
where length. This may also be inferred from Eq. (3), where the mode
 
T jk   2j  k2   2  2j   k2   4  4 T , j  T ,k  j  k  2 shape weighting factor becomes equal to zero for anti-

    2   symmetric modes.
U jk  2 T , j j
2
k 2 T ,k k
2
j 2 (2)

3
fatigue lives and increased peak stresses as compared to
conventional single-mode calculations. Thus, it is likely that
conventional single-span, single-mode calculations are biased in
non-conservative direction for spans where the shoulder
location is more critical in terms of fatigue damage. As a
possible means to reduce or altogether remove this bias, the
present study will investigate the effect of including a multi-
mode correction factor ψmm on the standard deviation of the
stress response σs.
As a result, the maximum environmental stress due to
direct wave action, σFM,max, given by Sec. 2.5.12 in DNV-RP-
F105 [1] becomes
 FM , max  k p  k M   s  mm ,
Figure 1 – Damage distribution along pipe axis for a single-
span with mid-span at x = 140m. The fatigue calculation k p  2 ln f v T  ,
was based on time-domain simulations with 1, 2 or 3 modes (5)
included [8]. 1  s 
kM  1   1
2   s , I 
In order to conservatively and efficiently estimate the effect of
where fv is the vibration frequency and ΔT is the time period
including multiple modes for a large number of single-spans,
used to calculate the significant wave height, typically 3 hours
the following approach has been adopted for multi-mode
in modern metocean studies. σs,I is the standard deviation for the
calculations:
stress response when drag loading is disregarded. It is assumed
1. Eigenfrequencies, mode shapes and associated modal that σs and σs,I are affected in similar manners by including
stresses are calculated by a semi-analytical method, multiple modes, thus explaining why ψmm is not included in the
Ref. [6], and given as input to a revised FatFree. expression for kM. This assumption will also be tested in the
2. Mode shape weighting factors are calculated for each present study.
mode. For fatigue calculations the stress range S, given by Sec.
3. The maximum modal stress for the fundamental mode 5.2.2 in DNV-RP-F105 [3], is modified to
on the shoulder is identified. S  2 2   s  mm   s (6)
4. The maximum modal stress on the shoulder is
determined also for the third and fifth modes. where γs is a safety factor on the stress range.
5. The three stress maxima are conservatively assumed to
act at the same location, and fatigue and peak stress
calculations are carried out based on the total stress 2.5 Results for maximum environmental stress
response spectrum corresponding to this critical calculations
location. The maximum bending stresses σFM,max corresponding to 1-
6. If the maximum stress for the fundamental mode is at year, 10-year and 100-year events were calculated for each free
mid-span, fatigue and peak stress calculations are also span using FatFree with both single- and multi-mode
carried out for this location. All symmetric modes calculations. The ratio rULS was calculated for each case
generally have a peak value at this location, but the according to the formula
effect of including higher order modes is then  FM ,max multi  mode
rULS 
generally beneficial as exemplified in Figure 1.  single  mode
FM , max
(7)
Multi-mode calculations have been carried out according to the
methodology outlined above and compared to corresponding Thus, a value of rULS > 1 indicates increased stress due to
single-mode calculations for 51 different free spans. The spans inclusion of more modes. The values for 1-year, 10-year and
are listed in Annex A. The selected spans cover a range of pipe 100-year estimates were generally similar, and only the 100-
geometries, effective axial forces, span lengths and soil stiffness year results will be discussed in the following.
values. A histogram of the values recorded for the 51 cases
described in Annex A is shown in Figure 2 below.

2.4 Single-mode calculation with a multi-mode


correction factor
The added fatigue damage at the critical location on the
span shoulder is expected to result in consistently reduced

4
Figure 2 – Histogram of observed rULS ratios. Figure 3 - Histogram of observed rFLS ratios.

The mean value of rULS was found to be 1.053 and the standard The mean value of rFLS was found to be 0.833 and the standard
deviation was 0.016. The minimum value was 1.023 and the deviation was 0.058. The minimum value was 0.749 and the
maximum value was 1.084. maximum value was 0.982.
A 5% increase in peak stresses is quite modest. However, In cases where only a few sea-states contribute with fatigue
safety factors are not included in ULS calculations according to damage, the ratios may be somewhat uncertain and should thus
DNV-RP-F105. Hence, a non-conservative bias is undesired be given less weight. When all cases with a fatigue life higher
and should be corrected for. Based on the observed results it is than 104 years for calculations without safety factors are
suggested to include a correction factor ψmm = 1.07, excluded, 38 cases remain in the study (note that the reported
corresponding roughly to the mean value of rULS plus one results are based on calculation with safety factors according to
standard deviation. safety class normal for pipes 1 and 2, and safety class high for
pipe 3). The mean value and standard deviation of rFLS then
2.6 Results for fatigue calculations change negligibly to 0.834 and 0.060, thereby demonstrating
The fatigue life Tlife was calculated using single-mode and that the results reported above are not affected by uncertainty
multi-mode approaches for all 51 cases, and for each case the due to small damage accumulation.
ratio rFLS was calculated according to A 16.7% reduction in fatigue lives is not very large, but
Tlife multi  mode since the reduction is quite consistent it is desirable to correct
rFLS 
T single  mode
life
(8) for it. A correction factor ψmm on the stress range, as described
in Eq. (4), was therefore tested as part of the present study. In
A value of rFLS < 1 hence indicates a smaller fatigue life in the order to set a reasonable value, results for Case 45 were
multi-mode calculation, implying that the single-mode estimate extracted for every combination of significant wave height Hs
is non-conservative. A histogram of the values recorded for the and direction θ (of co-linear mean current and wave-induced
51 cases described in Annex A is shown in Figure 2 below. flow velocity) as specified in the tables with Weibull parameters
for Pipe 3 in Annex A. For each of the 480 combinations, the
ratio of multi-mode to single-mode results were calculated for
the standard deviation of the stress response σs and the standard
deviation of stress response with inertia loading only σs,I.
The ratios were found to consistently increase with Hs and
peak period Tp. This is not surprising, since an increase in Tp
reduces the peak wave frequency and hence also reduces
dynamic amplification of the fundamental mode, implying a
more pronounced influence of higher-order modes. The values
were observed to increase from 1.027 to 1.045, which
corresponded closely to rULS (= 1.047) for the selected case.
Moreover, it was found that the ratios for σs and σs,I were almost
identical, with a root-mean-square deviation of only 3∙10-4,
thereby justifying the omission of correction factors in the
expression for kM in Eq. (3).

5
Based on the case study, it appears that a value of ψmm close reducing the length of the intermediate span shoulder in Figure
to the mean value of rULS should be conservative in most cases. 4 a) to 3 diameters [8]. A simplified approach of treating the
The value was therefore set to ψmm = 1.055. longest span as a single-span is thus demonstrated to be
Note that all the fatigue lives reported in this section were inaccurate.
calculated for the critical location on the shoulder. In the (few) The presented case demonstrates that several contributing
cases where the modal stress associated with the fundamental modes must be accounted for in multi-span analyses, and
mode has a global maximum at mid-span, the inclusion of furthermore that the damage varies significantly along the pipe
additional modes have a beneficial effect. The correction factor axis.
ψmm can therefore be set to unity (i.e., disregarded) in such
cases. DNV-RP-F105 already contains an approximate method
for determining modal stresses both on the shoulder and at mid-
span, thereby allowing cases where the maximum stress occurs
at mid-span to be readily determined. Alternatively, the equation
suggested by Sollund et al. [7] may be applied:
 Seff
limit  4.25  (9)
30
where
 K L L4s  S L2
  log10  and  Seff  eff s
 (10)
 EI  EI

Whenever β calculated according to Eq. (10) is larger than βlimit


from Eq.(11), the maximum stress occurs at the shoulder. This
implies that cases where β < βlimit, a full multi-mode analysis
may be conducted in order to reduce conservatism in the design,
ref. Figure 1. In Eq. (10) EI is the pipe bending stiffness, KL is
the lateral dynamic soil stiffness, Ls is the span length and Seff is
the effective axial force [20-22].

2.7 Effects of multi-spans


A double span subjected to direct wave action was reported
by Sollund et al. [8], and the main results are presented here,
and discussed in further detail here. The double span in [8],
illustrated in Figure 4 a), was subjected to a 1-hour sea-state
with Hs = 7m at a water depth of 60m. The pipe was a 30 inch
concrete coated pipe, for which more details are given in [8].
The mode shapes and associated modal stresses for the first two
modes are shown in Figure 4 b) and c), while the annual fatigue
damage is plotted in Figure 4 d). Time-domain and frequency
domain solutions produced fatigue lives of 193 years for a
single mode approach. A significantly smaller fatigue life of Figure 4 – Mode shapes (b), modal stresses (c), distributed
44.2 years was obtained by including the second mode in the yearly fatigue damage (d) for an interacting double span (a)
calculations, and for 10 modes the solution converged at 36.5
years. Thus, the assumption that the main damage contribution A range of previous studies have investigated the effects of
comes from the fundamental mode does not hold for multi- interacting multi-spans on VIV fatigue damage [23-30]. Based
spans. Fatigue lives may be particularly non-conservative if the on the results presented by Sollund et al. [8] and discussed
fundamental mode has an anti-symmetric mode shape, resulting herein, it is clearly demonstrated that span and modal
in a small mode shape weighting factor λ1. interaction is equally important for the study of multi-spans
A frequency domain calculation treating the longest span as when considering direct wave action as when considering VIV.
a single-span, and with only the fundamental mode included,
gave Tlife = 48.0 years, which is 31% higher than the 3. A NEW RESPONSE MODEL FOR CROSS-FLOW
corresponding time- and frequency-domain multi-span VIV IN LKCR
estimates. The single-span fatigue life was also 79% higher than
the multi-span fatigue life of 26.8 years that was obtained when 3.1 Presentation of the model

6
The theoretical background for the cross-flow response
, j x 
2
m  LKCR
S CF 
model for LKCR is extensively documented in [5]. In this f LKCR
x     f CF , j LKCR 
paper, the model is merely reproduced and adapted to the multi- cyc ,CF
j 1  S comb,CF  x  

(16)
mode format of DNV-RP-F105. The range of validity for the
presented model is 2 ≤ KC < 10, and α ≤ 0.5, where KC is the The response model given in Figure 5 has been thoroughly
Keulegan-Carpenter number and α = Uc/(Uc+Uw) is the current documented, including a discussion of its limitations [5]. The
to wave flow velocity ratio. Uc is the current induced flow multi-mode response calculations, i.e. equations (15) and (16),
velocity and Uw is the wave induced flow velocity. In Figure 5, are, however, without experimental foundation. Multi-mode
the response model for cross-flow VIV in LKCR is presented. response in pure currents has been extensively studied, with
particular relevance for pipeline free spans in relation to the
Ormen Lange project [31-34]. In wave conditions, however,
multi-mode response due to synchronization between the eigen-
frequency of a span and the wave lift load frequency has not
been tested with respect to potential multi-mode response. For a
single span, the notion of multi-mode response becomes fairly
exotic since the spacing between two consecutive cross-flow
frequencies is relatively large. For multi-spans, however, any
number of cross-flow modes may have similar frequencies and
potentially be excited by the same wave. For that reason, it has
been assumed that the mathematical expressions for multi-mode
interaction in wave dominated LKCR flow conditions are equal
to the typical multi-mode response in pure currents. Due to the
Figure 5 – Response model for cross-flow VIV in LKCR lack of experimental data, any reduction in VIV amplitude due
to competing modes has conservatively been disregarded.
The KC dependent coordinate, fr,KC, is defined by:
KC  6 3.2 Case studies
 2.68 ,
f r , KC   For the present study, the algorithm given in Section 3.1
2.68  0.08 KC  6  , 6  KC  10 (11)
was implemented in FatFree [19]. A considerable amount of
The frequency ratio frd is the ratio between the jth cross-flow calculation testing has been performed (see case study
frequency fCF,j and the wave frequency fw, reduced by the safety description in Annex A) using three different pipe cross-section,
factor γf. With nCF modal frequencies excited by cross-flow VIV with varying water depths, span lengths, effective axial forces
according to the LKCR response model, the associated modal and gaps. For pipes 1 and 2 (cf. Table 1), no triggering of the
stress response SLKCRCF,j for the jth mode is: new LKCR model was achieved in the relevant cases. It was
only possible to trigger cross-flow VIV in LKCR when fatigue
 Az 
, j  x   2  ACF , j  x      KC  Ks  
LKCR
S CF lives were predicted down to matter of days, but these cases
(12)
 D j were considered irrelevant. For pipe 3, however, KC numbers
are lower due to the higher diameter, and hence LKCR were
where ACF,j is the unit diameter stress amplitude for the jth cross- achieved combined with VIV excitation according to the
flow mode and the reduction factor ψKC is a function of KC: response model presented in Section 3.1. At around 80 m water
 KC  0.4  0.075KC  2 . (13) depth, a balance was struck where the occurrence of LKCR
regimes was obtainable for cases where it may actually have an
The reduction factor ψKs is a function of the stability parameter, influence on the design.
or Scruton number, Ks [19]: Cross-flow fatigue life as a function of span length for pipe
1  0.1K s , Ks  4 3 is shown in Figure 6. From the figure, it is observed that the
 Ks   (14)
 0.6 , Ks  4 LKCR response model has a significant impact on the cross-
flow fatigue life, starting at about 95 m span lengths. The
The combined response model cross-flow induced stress range response model in Figure 5 requires a range of different
at a given pipe axial position x, SLKCRcomb,CF (x) is given as: combinations of variables to match at the same time, since the
2

 S x 
nCF KC-number, the frequency ratio and the current flow velocity
,CF  x  
LKCR LKCR
S comb CF , j (15) ratio all must be within specific intervals. Therefore, the effect
j 1 of cross-flow VIV in LKCR is an on/off effect.
Similar plots to Figure 6 were generated for a wide range of
and the cycling frequency fLKCRcyc,CF at a given location x is water depths, for pipes 1 through 3. All of the cases are in
defined by: shallow waters and wave dominated flow conditions since the
wave conditions are generally more severe than the current

7
conditions (as may be observed from the tables of  Empirical correction factors have been introduced for
environmental statistics in Annex A). Only pipe 3, for specific single spans, taking into account the effect of multi-
water depths, encountered any significant effect of cross-flow mode response in the frequency domain solution.
VIV in LKCR.

REFERENCES

[1] DNV’76
[2] DNV Guideline No. 14
[3] DNV-RP-F105, “Free spanning pipelines”, February
2006.
[4] DNVGL-OS-F101, “Submarine pipeline systems”,
2012.
[5] API RP 1111, “Design, Construction, Operation and
Maintenance of Offshore Hydrocarbon Pipelines
(Limit State Design)”, 2009, API publishing service.
[6] Vedeld, K., Sollund, H.A., Hellesland, J., 2013. Free
vibrations of free spanning offshore pipelines.
Engineering Structures 56, 68-82.
[7] Sollund, H.A., Vedeld, K., Fyrileiv, O., 2014. Modal
Figure 6 – Cross-flow fatigue life as a function of span response of free spanning pipelines based on
length for pipe 3, at 80 m water depth with various levels of dimensional analysis. Applied Ocean Research 50, 13-
effective axial force 29.
Based on the case studies presented herein, it is concluded [8] Sollund, H.A., Vedeld, K. and Fyrileiv, O. “Modal
that the occurrence of cross-flow VIV in LKCR is likely to be response of short pipeline spans on partial elastic
rare. Larger diameter pipes are more susceptible to such foundations”, Ocean Engineering, 105:217-230, 2015.
excitation due to the inverse proportionality of the KC number [9] Sollund, H.A., Vedeld, K., Fyrileiv, O., Hellesland, J.,
to the pipe diameter. For large diameter pipes, there will 2015. Improved assessments of wave-induced fatigue
therefore be some rare cases when the response model for for free spanning pipelines. In: Dynamic response of
cross-flow VIV in LKCR may be dominating for the fatigue life offshore pipelines in free spans, PhD thesis, University
prediction. In most cases, however, it appears to be difficult to of Oslo.
achieve flow conditions where cross-flow VIV in LKCR is [10] Sollund, H.A., Vedeld, K., 2013. A semi-analytical
relevant unless the standard response models for in-line or model for structural response calculations of subsea
cross-flow VIV dominate the fatigue life. pipelines in interacting free spans. In: Brinkmann, B.,
and Wriggers, P., editors. Proc. of Vth Int. Conf. on
4. CONCLUSIONS Comp. Methods in Mar. Eng., May 29-31, MARINE
2013, Hamburg, Germany.
The following conclusions may be drawn from this paper: [11] Vedeld, K., Sollund, H., Fyrileiv, O., Nestegard, A. “A
response model for vortex induced vibrations in low
 A new revision of DNV-RP-F105 has been developed. KC number flows“, 2016, 35th Int. Proc. Conf. OMAE,
 A novel response model for cross-flow VIV in LKCR July 19-24, Busan, Korea.
has been introduced. Some combinations of pipe [12] Vedeld, K., Sollund, H.A. and Fyrileiv, O. “Fatigue
geometries, span characteristics and flow conditions and environmental loading of large-bore manifold
may excite cross-flow VIV in such a way that design piping”, Proc. of 30th Interenational Conference on
may be influenced, but these cases appear to be rare. Ocean, Offshore and Arctic Engineering, OMAE 2011,
 A novel frequency-domain method for calculation of Rotterdam, The Netherlands, Jun. 19-24, 2011
fatigue and extreme environmental load effects from [13] Vedeld, K., Sollund, H.A. and Fyrileiv, O. “Fatigue
direct wave actionhas been introduced, allowing for and environmental loading of small-bore manifold
multi-mode and multi-span calculations. It has been piping”, Proc. of 30th Interenational Conference on
demonstrated that consideration for multi-mode Ocean, Offshore and Arctic Engineering, OMAE 2011,
behavior is necessary, particularly for multi-span Rotterdam, The Netherlands, Jun. 19-24, 2011
configurations. [14] Bruschi, R., Bartolini, L., Molinari, C., Vignati, G.C.,
Vitali, L. “VIV basics for subsea spool/jumper

8
design”, 2015, 34th Int. Proc. Conf. OMAE, May 31st – [26] Kristiansen, N.Ø., Tørnes, K., Nystrøm, P.R.,
June 5th, St. John’s, Newfoundland, Canada. Damsleth, P., 1998. Structural modelling of multi-span
[15] Nair, A.R., Sharma, P., Grytøyr, G., Fyrileiv, O. and pipe configurations subjected to vortex induced
Velded, K. (2013): “VIV Assessment of Rigid Jumper vibrations. In: Proc. of 8th Int. Offshore and Polar
Systems- A Comparative Study on Jumper Shapes”, Eng. Conf., May 24-29, ISOPE 1998, Montreal,
Proc. of the 21st Int. Offshore and Polar Eng. Conf., Canada.
Maui, Hawaii, USA, June 19-24, 2011 [27] Wang, J., Wang, S.F., Duan, G., Jukes, P., 2009. VIV
[16] Nair, A., Kadiyala, R., Whooley, A., Eltaher, A. and analysis of pipelines under complex span
Jukes, P. (2010). “Vortex Induced Vibrational conditions. In: Deepwater Offshore Specialty Symp.,
Assessment of Multi-planar Jumpers”, Proc. Of the 5th Jan. 16-19, DOSS-71, Harbin, China.
International Offshore Pipeline Forum IOPF 2010, [28] Xiao, Z.-G., Zhao, X.-L., 2010. Prediction of natural
Houston, USA, Paper No. 2010-3001 frequency of free spanning subsea pipelines. Int.
[17] Wang, H., Huang, J., Gioielli, P., Kan, W., Spencer, D. J. Steel Struct. 10(1), 81-89.
and Islam, M. (2013), “VIV Response of a Subsea [29] Soni, P.K., Larsen, C.M., 2005. Dynamic interaction
Jumper in Uniform Current”, Proc. of the ASME 2013 between spans in a multi span pipeline subjected
32nd International Conference on Ocean, Offshore and to vortex induced vibrations. In: Proc. of 24th Int.
Artic Engineering, OMAE2013, OMAE2011-11417 Conf. on Offshore Mech. and Arctic Eng., Jun. 12-
[18] Heggen, H.O., Fletcher, R., Fyrileiv, O., Ferris, G. and 17, OMAE 2005, Halkidiki, Greece.
Ho, M., 2014. “Fatigue of pipelines subjected to [30] Sollund HA, Vedeld K, Hellesland J, Fyrileiv O.
vortex-induced vibrations at river crossings”, Proc. of Dynamic response of multi-span offshore pipelines.
the Rio Oil & Gas Expo and Conference, Rio de Marine Struct 2014;39:174-97.
Janeiro, Brazil, Sep. 15-18, 2014 [31] Holden, O.M., Paulsen, G., Bryn, P., Marthinsen, T.,
[19] FatFree, “Fatigue of free spans”, 2015, software 2005. Ormen Lange pipelines – Routing and cost-
program, v. 12.0 effective seabed preparation. In: Proc. of 15th Int.
[20] Vedeld, K., Sollund, H.A., Hellesland, J. and Fyrileiv, Offshore and Polar Eng. Conf., Jun. 19-24, ISOPE
O. “Effective axial forces in offshore lined and clad 2005, Seoul, South Korea.
pipes”. Engineering Structures 66:66-80, 2014. [32] Søreide, T., Paulsen, G., Nielsen, F.G., 2001. Parameter
[21] Fyrileiv, O., Collberg, L., 2005. Influence of pressure study of long free spans. In: Proc. of 11th Int. Offshore
in pipeline design – effective axial force. In: Proc. of and Polar Eng. Conf., Jun. 17-22, ISOPE 2001,
24th Int. Conf. on Offshore Mech. and Arctic Eng. Jun. Stavanger, Norway.
12-17, OMAE 2005, Halkidiki, Greece. [33] Fyrileiv, O., Mørk, K., Chezhian, M., 2005.
[22] Fyrileiv, O., Aamlid, O., Andreassen, E., 2010. Experiences using DNV-RP-F105 in assessment of free
Pipeline free spans – influence of internal pressure. In: spanning pipelines. In: Proc. of 24th Int. Conf. on
Proc. of 29th Int. Conf. on Ocean, Offshore and Arctic Offshore Mech. and Arctic Eng., Jun. 12-17, OMAE
Eng., Jun. 6-11, OMAE 2010, Shanghai, China. 2005, Halkidiki, Greece.
[23] Sollund, H.A., Vedeld, K., 2015. Effects of seabed [34] Fyrileiv, O., Mørk, K., Chezhian, M., Sigurdsson, G.,
topography on modal analyses of free spanning 2006. Updated design procedure for free
pipelines. In: Proc. of 25th Int. Ocean and Polar Eng. spanning pipelines DNV-RP-F105 – Multi-mode
Conf., Jun. 21-26, ISOPE 2015, Kona, Hawaii, USA. response. In: Proc. of 25th Int. Conf. on Offshore
[24] Vitali L, Marchesani F, Curti G, Bruschi R., 1993. Mech. and Arctic Eng., Jun. 4-9, OMAE 2006,
Dynamic excitation of offshore pipelines resting on Hamburg, Germany.
very uneven seabeds. In: Moan, T., et al., editors. Proc.
of 2nd European Conf. on Struct. Dynamics, Jun.,
EURODYN‟93, Trondheim, Norway.
[25] Anfinsen, K.A., 1995. Review of free spanning
pipelines. In: Proc. of 5th Int. Offshore and Polar Eng.
Conf., Jun. 11-16, ISOPE 1995, The Hague, The
Netherlands.

9
ANNEX A

CASE STUDY DESCRIPTIONS

In the case studies, a representative variation with respect to steel diameter Ds, steel wall thickness ts, effective axial force Seff, relative
span length Ls/Ds and dynamic soil stiffness KL have been emphasized. A total of 54 cases were originally chosen and the specific
combinations of variables are given below in Table 1.
Three different pipe cross-sections were selected, corresponding to a 6-inch oil pipeline (Pipe 1), a 16-inch flowline (Pipe 2)
and a 40-inch gas pipeline (Pipe 3). The three pipes have Ds/ts-ratios of 15, 25 and 34, respectively.
The effective axial force has for each pipe been varied from 0.35 Pcr in compression to 0.2 Pcr in tension, where Pcr is the critical
buckling load calculated according to Sec. 6.7.2 in DNV-RP-F105 [1]. For each level of effective axial force, the relative span length
Ls/Ds has been varied from 60 to 150, and for each span length calculations have been performed with soil stiffness corresponding to
both very soft clay and dense sand.
Note that three of the cases (case nr. 1, 7 and 13) in the original case matrix of 54 cases have been omitted from the study, thus
reducing the number of cases to 51. The reason for this is that the shortest spans on very soft clay had mode shapes that extended far
into the span shoulders, and the full mode shapes were for this reason not recorded when the modal analyses were conducted (mode
shapes were recorded for a total pipe axis length of 3Ls, where Ls is the span length).
The water depths were chosen to be 45m, 60m and 80m for Pipe 1, 2 and 3, respectively. Weibull parameters for current velocity
and significant wave height corresponding to different North Sea locations are specified for each pipe, given in the tables below.

Table 1 – Input parameters for the case studies


Case nr. Pipe Ds ts Seff/Pcr Ls/Ds KL
(mm) (mm) (-) (-) (kN/m2)
1a 5.834·105
60
2 1.955·107
3 5.834·105
0.2 110
4 1.955·107
5 5.834·105
150
6 1.955·107
7a 5.834·105
60
8 1.955·107
9 5.834·105
1 168.3 11.2 0 110
10 1.955·107
11 5.834·105
150
12 1.955·107
13a 5.834·105
60
14 1.955·107
15 5.834·105
-0.35 110
16 1.955·107
17 5.834·105
150
18 1.955·107
19 5.780·105
60
20 1.937·107
21 5.780·105
0.2 110
22 1.937·107
23 5.780·105
150
24 1.937·107
25 5.780·105
60
26 1.937·107
27 5.780·105
2 408.3 16.3 0 110
28 1.937·107
29 5.780·105
150
30 1.937·107
31 5.780·105
60
32 1.937·107
33 5.780·105
-0.35 110
34 1.937·107
35 5.780·105
150
36 1.937·107

10
37 9.109·105
60
38 3.053·107
39 9.109·105
0.2 110
40 3.053·107
41 9.109·105
150
42 3.053·107
43 9.109·105
60
44 3.053·107
45 9.109·105
3 1027.2 30.3 0 110
46 3.053·107
47 9.109·105
150
48 3.053·107
49 9.109·105
60
50 3.053·107
51 9.109·105
-0.35 110
52 3.053·107
53 9.109·105
150
54 3.053·107
a
Cases nr. 1, 7 and 13 have later been excluded from the study because the mode shapes were not recorded.

Table 2- Weibull parameters for current velocity Uc for Pipe 1. Pipe heading is 60 deg. relative to N. Gap e = 4 m, reference
height zr = 3 m and turbulence intensity Ic = 0.04.
Direction Sector Weibull parameters Statistics Return period (years)
relative to probability F(x)=1-exp(-((x-)/a)^) mean CoV 1 10 100
geographic N Shape () Scale (a) Location () value (m/s) (m/s) (m/s)
0 0.2175 2.159 0.143 0.000 0.126 0.488 0.36 0.41 0.45
30 0.1523 1.819 0.120 0.000 0.106 0.569 0.35 0.41 0.46
60 0.0777 1.350 0.075 0.010 0.079 0.651 0.31 0.39 0.46
90 0.0577 1.300 0.068 0.008 0.070 0.688 0.28 0.36 0.43
120 0.0626 1.500 0.089 0.000 0.081 0.679 0.31 0.38 0.44
150 0.0837 1.397 0.086 0.012 0.091 0.627 0.34 0.42 0.50
180 0.0736 1.600 0.101 0.001 0.091 0.636 0.32 0.39 0.45
210 0.0486 1.606 0.095 0.000 0.086 0.638 0.29 0.36 0.42
240 0.0363 1.458 0.090 0.001 0.082 0.693 0.30 0.38 0.45
270 0.035 1.421 0.084 0.000 0.077 0.714 0.29 0.36 0.44
300 0.0532 1.473 0.092 0.002 0.085 0.675 0.32 0.39 0.46
330 0.1016 1.975 0.132 0.000 0.117 0.529 0.35 0.40 0.45

Table 3 – Weibull parameters for significant wave height Hs for Pipe 1. Peak period Tp = 7.5(Hs)0.3 and peak parameter γ is
calculated according to Sec. 3.3.3 in DNV-RP-F105. Spreading constant s = 8. Water depth is 45 m.
Direction Sector Weibull parameters Statistics Return period (years)
relative to probability F(x)=1-exp(-((x-)/a)^) mean CoV 1 10 100
geographic N Shape () Scale (a) Location () value (m) (m) (m)
0 1.90E-01 1.172 1.495 0.850 2.27 0.53 8.06 10.25 12.35
30 2.02E-02 1.083 0.738 0.810 1.53 0.43 3.51 4.89 6.24
60 1.04E-02 1.076 0.779 0.840 1.60 0.44 3.28 4.78 6.23
90 1.47E-02 1.163 1.061 0.650 1.66 0.52 3.96 5.65 7.24
120 8.23E-02 1.611 2.898 0.290 2.89 0.57 8.62 10.65 12.46
150 7.14E-02 1.771 2.862 0.910 3.46 0.43 8.28 9.93 11.38
180 7.32E-02 1.597 2.373 0.220 2.35 0.58 7.01 8.72 10.24
210 8.38E-02 1.509 2.338 0.460 2.57 0.55 7.70 9.58 11.29
240 8.42E-02 1.499 2.278 0.500 2.56 0.55 7.61 9.47 11.16
270 1.33E-01 1.440 2.505 0.650 2.92 0.55 9.30 11.51 13.53
300 1.08E-01 1.239 1.836 0.700 2.41 0.58 8.24 10.59 12.82
330 1.28E-01 1.329 2.177 0.600 2.60 0.58 8.91 11.23 13.40

11
Table 4 - Weibull parameters for current velocity Uc for Pipe 2. Pipe heading is 90 deg. relative to N. Gap e = 4 m, reference
height zr = 3 m and turbulence intensity Ic = 0.04.
Direction Sector Weibull parameters Statistics Return period (years)
relative to probability F(x)=1-exp(-((x-)/a)^) mean CoV 1 10 100
geographic N Shape () Scale (a) Location () value (m/s) (m/s) (m/s)
0 1 2.159 0.143 0.000 0.126 0.488 0.40 0.44 0.48

Table 5 - Weibull parameters for significant wave height Hs for Pipe 2. Peak period Tp = 7.5(Hs)0.3 and peak parameter γ is
calculated according to Sec. 3.3.3 in DNV-RP-F105. Spreading constant s = 8. Water depth is 60 m.
Direction Sector Weibull parameters Statistics Return period (years)
relative to probability F(x)=1-exp(-((x-)/a)^) mean CoV 1 10 100
geographic N Shape () Scale (a) Location () value (m) (m) (m)
0 1.00E+00 1.193 1.790 0.660 2.35 0.60 10.87 13.28 15.61

Table 6 - Weibull parameters for current velocity Uc for Pipe 3. Pipe heading is 60 deg. relative to N. Gap e = 4 m, reference
height zr = 3 m and turbulence intensity Ic = 0.04.
Direction Sector Weibull parameters Statistics Return period (years)
relative to probability F(x)=1-exp(-((x-)/a)^) mean CoV 1 10 100
geographic N Shape () Scale (a) Location () value (m/s) (m/s) (m/s)
0 0.2175 2.159 0.143 0.000 0.126 0.488 0.36 0.41 0.45
30 0.1523 1.819 0.120 0.000 0.106 0.569 0.35 0.41 0.46
60 0.0777 1.350 0.075 0.010 0.079 0.651 0.31 0.39 0.46
90 0.0577 1.300 0.068 0.008 0.070 0.688 0.28 0.36 0.43
120 0.0626 1.500 0.089 0.000 0.081 0.679 0.31 0.38 0.44
150 0.0837 1.397 0.086 0.012 0.091 0.627 0.34 0.42 0.50
180 0.0736 1.600 0.101 0.001 0.091 0.636 0.32 0.39 0.45
210 0.0486 1.606 0.095 0.000 0.086 0.638 0.29 0.36 0.42
240 0.0363 1.458 0.090 0.001 0.082 0.693 0.30 0.38 0.45
270 0.035 1.421 0.084 0.000 0.077 0.714 0.29 0.36 0.44
300 0.0532 1.473 0.092 0.002 0.085 0.675 0.32 0.39 0.46
330 0.1016 1.975 0.132 0.000 0.117 0.529 0.35 0.40 0.45

Table 7 - Weibull parameters for significant wave height Hs for Pipe 3. Peak period Tp = 7.5(Hs)0.3 and peak parameter γ is
calculated according to Sec. 3.3.3 in DNV-RP-F105. Spreading constant s = 8. Water depth is 80 m.
Direction Sector Weibull parameters Statistics Return period (years)
relative to probability F(x)=1-exp(-((x-)/a)^) mean CoV 1 10 100
geographic N Shape () Scale (a) Location () value (m) (m) (m)
0 1.37E-01 1.13 1.43 0.54 1.91 0.64 7.56 9.91 12.19
30 3.50E-03 1.161 0.910 0.410 1.27 0.59 2.29 3.81 5.23
60 1.80E-03 1.608 1.230 0.190 1.29 0.54 1.88 3.09 4.04
90 1.80E-03 1.075 0.870 0.480 1.33 0.59 1.87 3.61 5.28
120 4.20E-03 1.287 1.100 0.440 1.46 0.55 2.69 4.17 5.49
150 7.70E-02 1.764 3.070 0.180 2.91 0.55 8.18 9.96 11.52
180 1.35E-01 1.507 2.270 0.380 2.43 0.57 7.81 9.61 11.24
210 1.01E-01 1.568 2.420 0.190 2.36 0.60 7.52 9.30 10.89
240 1.03E-01 1.434 2.270 0.270 2.33 0.63 7.92 9.96 11.83
270 1.34E-01 1.300 2.090 0.330 2.26 0.66 8.59 10.95 13.16
300 1.11E-01 1.137 1.630 0.420 1.98 0.69 8.05 10.66 13.19
330 1.91E-01 1.160 1.670 0.470 2.06 0.67 8.66 11.17 13.59

12

View publication stats

You might also like