Professional Documents
Culture Documents
1.
INTRODUCTION
The analyses have been carried out using a suite of flexible riser
analysis programs developed by the author over the past decade.
These programs are capable of analysing the response of all
common flexible riser configurations in the frequency and time
domains. The development and verification of these programs have
been detailed in Patel & Seyed (1989) and Seyed et al. (1990). The
validation studies presented in these references have included
comparisons against independent and in-house model tests as well
as other programs. Recently, a comprehensive verification was
carried out against results from 10 independent commercial
programs. The findings of the study which are reported in Larsen
(1992) indicate encouraging agreement amongst most of the
programs and discrepancies which highlight the present uncertainties
in the analysis of flexible risers.
2.
PARAMETRIC STUDIES
2.1
Description of Analysis
Preliminary results from one such study were presented in the earlier
paper by the authors - see Seyed and Patel (1991). This paper
presents a second complementary set of findings from the study
and, ideally, should be studied in conjunction with the earlier paper.
These works do not attempt to provide a starting point to the
designer. Rather, they attempt to furnish the analyst, who is using a
commercial flexible riser analysis package, with some degree of
expectation towards the final results and with an ability to spot check
the results and make shortcuts saving time and cost in the process.
Due to shortage of space only illustrative findings have been
presented and no attempt has yet been made to generalise. The
presented data may only be considered a summary of analysis
experience.
2,2
- Numerical Difficulties with
Flexural Rigidity
Figu.re (2) s~~ws th.e changes in bending moments for the same
loadmg condmon. Little change from the static results are obtained
from the method with tensile non-linearity seeming to be the least
accurate.
Pipes of Low
The same analysis was then repeated for the original pipe with
(E:A~I=34000). All other parameters including the convergence
cntenon ~ere ~ept constant. Figures (3) and (4) show the results of
the analYSIS. DlsI?lacement~ are seen to be indistinguishably similar.
A complet~ly different picture emerges in the case of bending
moments. Figure (4) shows large disagreements between the results
from th~ different methods. It should be noted that the magnitudes
?f bendmg. moments are very small and 1000 times less than those
m the prevIous case. The observed deviations are solely a numerical
featur~ and are caused by the finite precision of the computer used.
Reducmg the toleranc~ proved successful in bringing the solutions
toge~~er and cu~es Similar to those in Figure (2) were obtained.
A~~lllonally, us~ng a ~on-standard, quadruple precision feature of
DigItal Corporallon Mlcro-VAX FORTRAN compiler reduced the
to~er!lnc~ level, improved the rate of convergence as well as
ehml,natlng t~e problem. !t should be. n.oted that the existing
algo~~ms are Implemented m double precISIon and the use of single
preclSlon would be guarantied to produce numerical problems. The
fo:e~oing observations confirmed the problem to be of a numerical'
ongm and showed that increased computational precision was one,
met~~d of overcoming the difficulties. In the absence of high
preclsl<;>n computational facilities, it would be necessary to decouple
the aXial degrees ?f freedom or use a cable type solution and'
calculate the bendmg moments from the curvatures in the finaL
geometry.
With the direct method the iterations are carried out for equilibrium
of tensile forces. This involves a shooting approach about the mean
geometry where for each estimate of displacements, tensile forces
are. cal.culated and used to re-estimate the system stiffness matrix
which IS then used to re-calculate the force vector. An estimate of the
error is found from the ratio of the sum of squares of residual force
to total !orce. The geometry is only updated at the final stage after
calculation of forces. The second method is based on the incremental
approach where the loading is calculated for the original equilibrium
geometry. This is then applied in a user prescribed number of
increments and the geometry and internal forces are updated at each
st~ge. One equilibrium c<;>rrection is carried out per load step. The
third method uses a hybnd approach where the direct approach is
employed with two levels of iterations one for tensile forces and the
other for geometry. This amounts to placing the total load on the
mean geometry, calculating the resulting tensile forces and
displacements and computing the system stiffness matrix for the
updated geometry and tensile forces. The residual force is then
calculated as previously and the root mean squared error is used to
chec.k f~r convergence. Having established convergence, the total
loadmg IS re-calculated for the updated geometry and the difference
with that applied to the original geometry is used to replace the total
load vector and repeat the iterations. This correction constitutes an
outer incremental analysis loop for the displacement dependence of
loading. The convergence criterion for this section is the r.m.s of the
ratio of the residual force to the original total force. The hybrid
method represents an exact approach where the only non-linearity
excluded is that of the pipe structural properties. Details of these
methods may be found in Seyed et al.(1989).
The l?s.s of precision was traced within the program and was found'
to ongmate from local rotations of the element. These rotations
served to increase the curvature of the element and bring the ends
cl?ser together. These rotations have little axial strains associated
:-vlth them. However, the simple beam finite element model,
mterprete~ the apparent reduction in element lengths as a net
compressive stram. The large apparent compressions served to
sever~ly de-~tabil.ise the algori~hm through erroneous negative
effec~lv~ tenslO~S m the geometric stiffness matrix. It was possible
t~ ehmmate thiS effect by correction of the resulting first order
displacements before calculation of element forces This was carried
ou~ in accordance with the large deflection theory and was found
qUIte successful and accurate. The source of numerical illcon~i~ioning whic~ require~ t.he use of low tolerances and high
precIsion c<;>mPl!tatIon was.slmllar~y traced to the multiplications of
small r~tallon-m~uced a~IaI strams by the pipe axial rigidity in
calculallon of aXial tenSile forces. This constitutes an operation
between. very small and very large numbers and causes the system
geometric sllffness to undergo lengthy oscillations which damp out
very gradually.
The case study uses the pipe with properties indicated in Table (1).
A catenary riser of 250m length with horizontal and vertical
separations of 100 and 150 metres, respectively, was used. A
simple geometry was deliberately selected to simplify comparison of
results. A steady current of 0.5 mls of uniform distribution with
depth was applied to the riser in the direction of the positive x-axis.
This case study is used to demonstrate the effects of the calculation
method as well as the numerical problems with high (EAIE!) ratios.
C?n~
Figures (1) and (2) show the calculated geometries and bending
~oments respectively.
It is noted from Figure (1) that
displacements calculated using tensile non-linearity and incremental
loading, despite being close to each other, are larger than those
pred.icted using t.he exact non-linear approach including geometric,
tenSile and loadmg non-linearities. Additionally, the incremental
methods are seen to approach the fully non-linear version with
increasing number of steps. The validity of the presented results is
confirmed noting:
2,3
301
body motions of the upper riser which tend to be linear, periodic and
at the excitation frequency. Nevertheless, the apparent damping of
the non-linearities remains an interesting feature.
in-plane with the riser and in opposing directions whist a third was
perpendicular to the vertical plane of the riser. The remaining two
directions were chosen to be oblique at 45 degrees to the plane of the
riser. The co-ordinate axes were defined such that the positive x-axis
was horizontal along the line joining the riser base to the projection
of the riser top and the y-axis was vertical. With reference to the
riser geometry in Figure (5), the z-axis points towards the reader
(out of paper). The wave angles are measured from the z-axis using
the right hand screw rule such that a wave at an angle of 0 degree
propagates along the z-axis whilst a 90 degree wave is directed
along the positive x-axis. The riser is given a 30m freeboard over
the still water level (SWL). To avoid confusing the picture vessel
motions are set to zero. The simulations were carried out at 0.5
second intervals (V32nd. of excitation period) and for a total period
of 200 seconds. The remaining analysis parameters are given in
Table (2).
Figure (8) shows the envelopes of tensile forces for different wave
headings. It is noted that little dynamics are present which is partly
due to the low amplitudes of motion and partly a result of the riser
resisting lateral loads through the pendulum effect. However, it
should be stressed that this feature is not typical and should be
considered specific to this case. Figure (9) shows the out of plane
bending moment envelope for the riser. The magnitudes of bending
caused by all out of plane waves seem to be quite similar with large
magnifications noted at the SWL. Surprisingly, this comment was
seen to apply to the in-plane bending as well. It is seen in Figure
(10) that the bending envelopes produced by in-plane waves are
rivalled very closely by the out of plane and oblique waves. In
selected positions, such as the top connection and the SWL, the
latter are seen to dominate. These observations illustrate that out of
plane loading is of significance and is capable of exciting greater
responses than those induced by in-plane loadings. Whilst this
single case study cannot assert a general rule, it clearly provides an
example to prove the need for this type of analysis in design.
302
.~----
--------------------
Inner radius
Outer radius
Dry mass per unit length
AxiaJ rigidity (EA)
Flexural rigidity (EI)
Torsional rigidity (GJ)
Polar inertia (Ip)/ X-sect Area
Pipe drag coefficient
Pipe inertia coefficient
Water depth
Water density
Internal fluid density
Effective weight/length
Effective weight/length including buoyancy
Additional dry mass/length of buoyancy
foam
Effective outer diameter of buoyancy foam
In these studies the mean internal fluid density was 520 kg/~3 and
the amplitude of density variations was set at 500 kg/m 3 . This
represents a transition from heavy slug to gas. The effect of such
changes on the tensions and bending moments along the pipe are
quite drastic with the worst loadings occurring in the areas of high
curvature. The reason for the large motions is due to the large
fluctuations that changes in the internal fluid density of the pipe
impart on the flexible pipe stiffness which is predominantly derived
from the weight of the pipe. The influence of these large motions on
the internal forces in the pipe are quite severe.
The formulation of the internal flow induced loads is an
approximation whose range of validity may be assessed from the
earlier publication, Patel & Seyed (1989). Accurate quantification of
internal flow dynamics is presently not possible. However, the
observed trends from model tests - see Seyed et al. (1990) - and
their agreement with simulations suggest that these dynamic effects
do, indeed, have major implications in design. Further, the
interaction of internal flow with the pipe wall tension is a feature
without which fatigue and wear calculations may yield overoptimistic service lives.
3.
100.0 m
380.0 m
180.0 m
95.0m
230.0 m
20
20
20
REFERENCES
90.0m
110.0 m
200.0 m
70.0m
20.0m
11O.0m
10
15
5
20
5.
0.406 m
CONCLUSIONS
4.
0.0508 m
0.0762 m
26.7 kg/m
115MN
3400Nm2
2600Nm2
0.05
1.0
2.0
350m
1025 kg/m3
1025 kg/m3
160N/m
-425 N/m
45 kg/m
100.0 m
250.0 m
242.0 m
268.5 m
40.0m
8.0m
1100 kN
1.0
1.5
200.0 kg
2.929 m3
5.309 m2
350m
10
15
5
20
TABLES
303
!------
150.0
350.0
+I-
---- ~~:::;~I:f~':.~!i!.~~+
300.0
U
11)
'0'
....
i5
E
60 eleme~ds pinned
BAlBI:
U, :O.Sm/s
250.0
-.au
.S
>
U, =0.5m/s
"
+":'"
+"
+.1
/"
,I
0.0
50.0
100.0
150.0
200.0
250.0
Arclength (m)
400.0 - , - - - - - - - -_ _ _ _ _ _ _ _--,
-.------,+I-..,--------------~___,
xlO'
~""!,e~~ds pinned
1,'
'I
-f-:i .
60.0
150.0
100.0
50.0
1,1
1/\ ' .
+-----.----r-----,----1
0.0
III
I"
30.0
-50.0
I
I I
1111
200.0
150.0
+'I
'"Cl
10:
11)
,:Q
11)
Inaemental, 10 steps
"
I I
I I
90.0
'E
inaem"'ltal, 20 steps
1 correcbon per step
-tt:
bI)
geometric non-liiiearity
"
"
I,
11)
p...
"
Z
'-"'
incremental, 10 steps
1 correction per step
- - Direct iteration
tensile non-lineanty
I
t
120.0
. inaem"'ltal, 20 $Ieps
1 cotrecbon per step
.9
-LBGEND-
+ +
tensile non-1inearity
----=Irfc~::,'::~=~+
12.00
e
e
i5
,-..
inaemental, 20 steps
1 correction per step
Incremental, 10 steps
1 correction per step
9.00
300.0
.
u
11)
~""!,,~s pinned
U,:O.Sm/s
200.0
6.00
-LBGEND--90deg
--- 4Sdeg
100.0
-45 deg
3.00
---- -90deg
0.0
0.0
50.0
100.0
150.0
250.0
200.0
~:~~ U:1fo:l;: l~
-t-----t-----,----..,------1
-100.0
0.0
200.0
300.0
Arclength (m)
Fig 2 - Effect of non-Iinearities on static bending moments under current
350.0
100.0
T;::=========:----,----.
- - - - Dieecr it~ration,.tensi)e +
gcomemc non-uncanty
,-.. , 300.0
Inaemef!,tal, 20 steps
1 cotrccbon per step
Increm"'ltal, 10 steps
1 cotrccbon per step
.9
U
11)
.=
,-..
g
c
.9
U
11)
60 elementslBnds pinned
BAlBI: 34000
U,: 0.5 m/s
250.0
300.0
-.au
200.0
'E
>
100.0
---- 45 deg
--Odeg
-45 deg
150.0
--r----,-----r------r---_-l
-50.0
0.0
50.0
100.0
0.0
-t------;r-----+-----r-----l
-8.00
150.0
-4.00
0.00
4.00
8.00
304
110'
UOO-,-------------------,
8.00
-LEGEND---- 45deg
-LEGEND- - 90deg
- - Odeg
4Sdeg
",
-45 deg
4.00
Wave Height
Wave Penix!
!:
-_ ...... -
C1.)
.,, ,,
,, ,,,
,
,
=20m
=168
.8
,,
... :.:.,
a::
.....
.......... :--':_-" ,
,,,
I
I
I,
'5
-4.00
---- -90deg
80
0.300
"0
!:
-0.300
C1.)
t:Q
"
C1.)
~
-8.00
-0.900
+----,-----,------;,----1
0.0
30.0
60.0
90.0
- - - -45 deg
::E
bIl
.S
,,
,,, .
0.900
Z
'-"
i:!
C1.)
,,
0.00
C1.)
"""
8
-1.500
120.0
-+-----.----.----...----,------l
100.0
0.0
200.0
300.0
400.0
500.0
Arclength (m)
Fig 10 - In-plane bending envelopes for different wave headings
30.00
18.00
200.0 - , - - - - - - - - - - - - - - - - - - - ,
6.00
,,
,
,,
,,,
-6.00
",
"
45deg
-45 deg
-18.00
---- -90deg
MeanStatic
-30.00
a::
+-----,,-----.-----.----.-----1
0.0
110.0
220.0
330.0
440.0
-160.0
.El
5500
Arclength (m)
Fig 8 - Tensile force envelope for different wave headings
1.500 - , - - - - - - - - - - - - - - - - - - - ,
s
e
30.00
45.00 -
60.00
. -4S deg
90.00
105.00
-LEGEND- - -45deg
C1.)
75.00
Time (s)
--Odeg
0.900
:E
-280.0
-LEGEND----45deg
Odeg
- - - - 45 deg
0.300
90deg
+-----r----r-----.----.----\
0.0
100.0
200.0
300.0
400.0
500.0
Arclength (m)
Fig 9 -Out-of-plane bending envelopes for different wave headings
305
------
---------------------------------
120.00
~.OO
-LEGEND-
- - -45deg
Odeg
- - - - 45 deg
e0
90deg
OIl
I
I
,,
-6.00
~
Q)
r'~~
~
,
I
,.,.1
!
,
,
I
Ii
;./11
1'"
"
""
r1:!,i
"i
:j
Ii
;
'i
:.;
,
I.,
,,
!
I :\1,/
,i
AI, !
i5:::
...... -21.00
.;
'I,
Q)
IV'
I'
I=:
';/,
,, .'" ,
, .,
,
,
I,
. ,! ~,
,,
1W.0 - , - - - - - - - - - - - -______________________-,
",
/.,'
"
!i
I;
"
!,
\hi ~
9.00
'.
II
i::
. - - - - - - - - - - -________________~----.,
'5
-36.00
+----.,----...-----r-----.------"r-----1
30.00
45.00
W.OO
75.00
90.00
105.00
120.00
Time (8)
'5
-280.0
+---.,---,----r----;,---r-,-30.00
45.00
W.OO
75.00
90.00
105.00
,," ,,
:..,',~
,I 'I
-480.0
120.00
i::
Time (8)
I '
!,
"
OIl -640.0
.5
"2
Q)
~
Q)
3W.0 , - - - - - - -__________________________- ,
i5:::
-800.0
.s
-9W.0 +--;-----,--~--,.....--;__-___1
30.00
45.00
W.OO
75.00
90.00
105.00
120.00
16.00 , - - -______________..:....________- ,
',,I'
I
i5:::
8.00
126.0
"
.,j
;,
:j
30.00
45.00
W.OO
75.00
90.00
105.00
,
.,,
120.00
Time (s)
,;
"
"
:t
0.00
48.0
,, !
';
"
"
"
"
.!
"
.s
,~
,',
.! .:
,
,
,, ,,,
"
',',
{,
i,'1
.,,,
306
400.0
120.0
-LEGENDlntcnnediate Profiles
-LEGENDRiser Profile
Slug: 230m, 25s
Wave: 2.6m, 5s
320.0
960
.g
.g
72.0
ti
<1.)
ti
<1.)
.~
'0'
.....
,:l.
,:l.
""@
u
240.0
""@
u
480
'E
'E
>
>
160.0
Slug
80.0
24.0
0.0
Wave +
Vessel
Wave +
Vessel +
<1.)
<1.)
+--""'-=~--"---r----,:-----1
-50.0
-10.0
-30.0
10.0
30.0
50.0
0.0
-l--.t:..--.,.-----,.-.....::..--,----.,------l
-100.0
70.0
0.0
200.0
300.0
400.0
Figure
100.0
J 1- - Lazy-wave response in
12m, 20s regular waves
.,~", - , - - - - - - - - - - - - - - - - - - - - - - ,
400.0 - . - - - -_ _ _ _ _ _ _ _ _ _ _ _
-LEGEND-
-~~,1if~2ss
Wave: 12m, IS.
/ /
,/
"
,,/ / /
- - - - -..,,,/
'
320.0
\
....
,,"/
/
........
!':
;'/
.g
ti
<1.)
240.0
E
<a
Arolenglh (m)
--,....
.Lt!CE<D
160.0
'E
Wavc+
>
80W11J10 (R..dS-..... C _ )
Vessel
Slug
Wave +
Vessel
80.0
Figure
\10-'
.~
-r---------------------,
0.0
'10 -
,I
_--J
---. ....... /\
I
I
...
..
.,,,, - + - - - - - - , r - - - - - - , - - - - - - , - - - - , . - - - - - - j
I,.
It'"
"'10
An:length (m)
Figure
100.0
200.0
300.0
400.0
0.0
.<-10 -
+-""--.,------.---'--"T"----.-----I
-100.0
307