You are on page 1of 8

Mo t i o n res pons e o f f l oat i ng s t ruct ures t o

regul ar waves
B. J. NAT VI G* and J. W. P E NDE RE D
Aker Enqineerin 9 A/S, Tjurholmen, Oslo, Norway and Pell Frischman Ojfshore, Oslo, Norway
Li neari zed and non-l i near met hods for anal ysi ng mot i on response of floating st ruct ures to regular
waves are discussed in this paper. The linearized met hod is based upon a t radi t i onal frequency
domai n appr oach, whereas the non-l i near met hod is based upon the time i nt egrat i on met hod as
pr oposed by Newmark. In addition, a new appr oxi mat e met hod is proposed. This met hod, which
empl oys a Newt on Raphson sol ut i on of the equat i ons of mot i on, is an extension of the linearized
met hod and is t herefore valid for near linear systems. These met hods are applied t o a t aut line
moor ed structure; the Aker Tet her ed Pr oduct i on Pl at f or m and to a convent i onal semi-submersible,
the Aker H3.
I NTRODUCTI ON
Spectral met hods for the analysis of mot i on response and
st ruct ural st rengt h of floating structures, have been widely
used in recent years (e.g. ref. 2). However, spectral met hods
assume t hat the mot i on response can be adequat el y
represent ed by linearized equations. It is common practise
to assume t hat convent i onal semisubmersibles behave in
a linear manner, but it is not appar ent t hat this assum-
pt i on can also be applied to a t aut line moor ed structure.
Dur i ng the devel opment of the Aker Tet her ed Fl oat i ng
Pr oduct i on Pl at form, it was found necessary to devel op a
comput er based analysis of linearized and non-l i near
mot i on response so t hat the assumpt i ons of linearity
coul d be tested before entering i nt o a spectral analysis.
Spectral analysis, for the cal cul at i on of short and long
t erm ext remes of rigid body mot i on response and moor -
ing line forces are easily performed due to t he few (6)
degrees of freedom required. St ruct ural st rengt h analysis
by the use of spectral met hods are, however, i mpract i cal
for or di nar y design work owing to the high costs and
shoul d hence onl y be used as a final check on the
st ruct ural integrity. This appr oach was previously adop-
ted for the semisubmersible Aker H3 2.
A pract i cal and popul ar met hod of st rengt h analysis is
the design wave met hod. Thi s has obvi ous appl i cat i on to
fixed structures, but can also be applied t o floating
st ruct ures if mot i on i nduced forces are included in the
analysis. It was found t hat this met hod coul d be made
mor e real i sti c by using non-l i near mot i on response when
cal cul at i ng the st ruct ure mot i on forces and by including
wave non-linearities when calculating the wave forces
before these forces are applied t o a st ruct ural space frame
analysis.
By using ' the non-l i near design wave met hod' duri ng
the design phase and by runni ng a final check on the
st ruct ural st rengt h by a linearized spectral met hod, it is
possible to perform two i ndependent checks on st ruct ural
integrity by t wo met hods of which the short comi ngs and
features are to some extent compl i ment ary.
These ideas were devel oped duri ng a t hree year re-
* Present address: Det Norske Veritas, PO Box 300, N-1322 Hovik,
Oslo, Norway.
search proj ect at Aker Engineering, in col l abor at i on with
t he Depar t ment of Mechani cal Engineering, Uni versi t y
College London.
Whilst the work on the t et hered pr oduct i on pl at form
provi ded the initial mot i vat i on for this project, the
resulting comput er pr ogr am will also handl e semisub-
mersibles in a similar manner. The comput er pr ogr am
named SEARESPONS was used for all dynami c analysis
of the Aker TPP. This included: mot i on response, moor -
ing line design, l oads for the st ruct ural analysis and offsets
for mari ne riser design.
LI NEARI ZED ANALYSI S
LinearJbrces
The first order (Airy) wave t heor y is implicit in linear
analysis for the cal cul at i on of fluid particle accelerations
and velocities and dynami c pressures. The cor r espondi ng
wave forces are calculated by the well known Mor i son
equat i on 1-3 which is assumed to be valid as long as t he
cross-sectional dimensions are less t han one-fifth of the
wave length. This corresponds to wave peri ods down to
5-6 seconds for most current floating structures.
The relative velocity between fluid and st ruct ure is a
second power t erm in the Mor i son equat i on. An explicit
and linear solution to the equat i ons of mot i on is t herefore
onl y possible if the st ruct ure velocity and the wave
particle velocity are separat ed and linearized. This is
usually achieved by t he subst i t ut i on u/u/=(8/3rc) u,u,
where u is either the wave particle velocity or st ruct ure
velocity, u a is the velocity ampl i t ude and 8/3~ is the
linearization const ant which makes the energy absorbed
duri ng one wave period equal for bot h assumed forms in
the above aquat i on.
Equations of motion
The simplified equat i ons of mot i on 2'3 result from
equat i ng the ext ernal forces on the st ruct ure (assumed to
be prevent ed from moving) to t he reactive mot i on forces
acting on the st ruct ure (which is assumed to be movi ng in
a calm fluid). This results in a second or der differential
equat i on which may be expressed in mat ri x form as:
0141-1187/80/0300994)852.00
1980 CML Publications Applied Ocean Research, 1980, Vol. 2, No. 3 99
Mol i ot l rcspotL~e O[ f l oat i ml structurc,s to re qul ar waives: B . . I :\;atcit a m / J . |11 Pemh' r vd
I M + A j ) t i J + B '~ " "
3rc~Pd)+(Ct +C_~I~/~--f: cxp (i,)t) (l)
where M =s t r uct ur e mass mat ri x, A I =a dde d mass mat -
rix, B =d a mp i n g matrix, C I = hydr ost at i c rest ori ng mat -
rix, C2 =a n c h o r restoring matrix, ~, , ~/J,=complex
vect ors of st ruct ure acceleration, velocity and displace-
ment, ~p,,=real velocity ampl i t ude vector, F=c o mp l e x
forcing vector. All the mat ri ces are symmet r i cal except for
the dampi ng mat ri x which is unsymmet r i cal due to the
way in which the velocity ampl i t ude must be included.
A special anchor rest ori ng mat ri x for t aut line si-
t uat i ons has been devel oped by the writers. Thi s rest ori ng
mat ri x assumes st rai ght moor i ng lines and includes the
rest ori ng effect due to the change in di rect i on of the
pret ensi on force duri ng mot i on as well as t hat due to the
mat eri al stiffness of the moor i ng lines.
In or der to reduce the above second or der differential
equat i ons to a set of linear equat i ons it is necessary to
assume har moni c mot i on. Thus, the velocities and accele-
rat i ons are expressed in t erms of the compl ex displace-
ment ampl i t ude:
= exp (i~J)t)
= i~o~ exp (iv)t)
~)= -oo2~P exp (io)t) (2)
the equat i on of mot i on can hence be solved for the
compl ex response ~P.
Since the dampi ng forces are linearized as dependent on
the velocity ampl i t ude, an i t erat i ve sol ut i on t echni que is
required. The sol ut i on using this i t erat i ve t echni que is
also onl y valid for a given wave height and is hence agai n
not strictly a linear solution. In practice, however, the
mot i on dampi ng force is onl y significant at r esonance and
the wave induced dr ag force is onl y significant at the
peri od of cancel l at i on of vertical added mass and dynami c
pressure forces: leaving the response basi cal l y linear with
respect to wave height out si de these regions.
N O N - L I N E A R A N A L Y S I S
Non- l i ne ar Jorces
Apar t f r om l i neari zat i on of the viscous dr ag force,
several ot her non-l i near effects have to be omi t t ed in the
linearized anal ysi s in order to achi eve an explicit sol ut i on
to the equat i ons of mot i on. In a non- l i near analysis,
however, it is possible to include any non- l i near force as
l ong as it is possible to f or mul at e this force. In the present
study, the following non-l i neari t i es were assumed to be
i mpor t ant enough to be included in the force calculations:
(a) wave forces are cal cul at ed at the di spl aced and r ot at ed
st ruct ure posi t i on and not at the mean posi t i on as for the
linearized analysis; (b) the dr ag force is cal cul at ed using
the t rue relative velocity bet ween fluid and st ruct ure
(including the current velocity) before squari ng; (c) all
fluid forces are cal cul at ed up to the wave surface and not
to the still wat er level as for the linearized analysis; (d)
non-l i near moor i ng forces are included.
By a sui t abl e choi ce of the added mass and dr ag
coefficients, the linear wave t heor y can be made to yield
wave forces close to t hose obt ai ned using the mor e
compl ex higher order wave theories. Thus first or der
t heor y is ret ai ned here.
Thi s first order wave t heory is inconsistent in the splash
area when applied to finite waves. Thi s i nconsi st ency was
o~er come by assumi ng thc wave surface as the dat um for
the wave forces instead of the still wat er level which is
usually done. Thi s assumpt i on makes the sum of the
hydr ost at i c and dynami c pressure at the surface equal to
zero and also makes the wave particle velocity and
accel erat i on have a dept h at t enuat i on of unity at the
surface.
Non- l i ne ar equat i ons q] mot i on
The non-l i near component s of the forces are, in genera l,
not of simple forms capabl e of concise mat hemat i cal
description. Hence, every posi t i on of interest in a wave
must be exami ned separat el y and a numeri cal eval uat i on
and s ummat i on of all the vari ous force cont r i but i ons
act i ng on all submer ged part s of the st ruct ure must be
carried out.
The non-l i near equat i ons of mot i on can be written:
ZA2t ~ --(M + A~)q) + ZB(t] + C,, - ~1l~1 + C. - ~l -
ZCl(t): r / ) - EC2 ( ~) + W+ F , , = 0 (3)
where Y, A2 q =s u m of all wave inertia forces, M
= st ruct ure mass mat ri x, A~ = i nst ant aneous added mass
mat ri x, ZB(// . . . . sum of all dr ag forces due. to the relative
fluid velocity, ZCI ( . . . . sum of all buoyancy and dy-
nami c pressure forces, ZC2( 6 . . . . sum of all moor i ng line
and pret ensi on forces, W= st ruct ure weight, F, . =wi n d
forces and moment s, r/,//, ~/= fluid particle di spl acement s,
velocities and accelerations, and C, , =c u r r e n t velocity.
Ti me series sol ut i on
The most obvi ous appr oach to the sol ut i on of non-
linear equat i ons of mot i on is t hr ough an i nt egrat i on of the
accel erat i on and velocity curves in the t i me domai n. The
well known Ne wma r k met hod 4 is widely used. In the
offshore cont ext , this met hod has been appl i ed to fixed
j acket t ype st ruct ures by Burke and Ti ghe 4, but no
reference is found on the appl i cat i on of this met hod t o
fl oat i ng structures.
The Newmar k met hod describes the ki nemat i c re-
l at i onshi p in the form:
@( t , +, ) = @(t,)+(1 - 7) ~j ( t , ) At +" / ( O( t , +OAt
, l t + 1 ) = ~ ( t . ) + ( ~ ( t . ) A t +
( - fl)(O(t,)At 2 + fl@(t, + OAt 2 (4)
where ~0(t,) denotes, the responses at t i me t, and ~(t , + 1)
denot es the responses at a small t i me i nt erval ahead of t,.
The value of 7 is usually set to 1/2 f r om dampi ng
consi derat i ons and the value of the coefficient/3 is chosen
in the range 1/4, 1/6 or 1/8 to give good convergence
propert i es.
Subst i t ut i ng for the vect or of velocities and di spl ace-
ment s at t i me t,+ 1 into the equat i ons of mot i on (3), the
vect or of accel erat i ons at t i me t,+ 1 are found thus:
4) = (M + A) - ' . ( Z A z # + Z B ( i l + C . - ~ b ) I # + C . - @ -
~ C l ( ~ ; t I ) - Y C 2 ( ~ ) + F,~,+ W') (5)
100 Appl i e d Ocean Res ear ch, 1980, Vol. 2, No . 3
6[ ,1
Motion response of floating structures to reyular waves: B. J. Natvi9 and J. W. Pendered
3'ok/
Figure 1. Newton-Raphson iterations
If this is repeat ed for a sufficient number of time steps
using sufficiently small time i ncrement s At, a convergent
sol ut i on for the displacements, velocities and accele-
rat i ons will be obt ai ned.
The time step met hod is often st art ed from zero
response. Thi s will result in an initial t ransi ent sol ut i on
which will be damped out after a few cycles. The met hod
t o be present ed in the next section, however, will pr oduce
a good estimate of the st art values for the responses and
consi derabl e saving in comput er costs is t herefore
possible.
Ti me series sol ut i on met hods are also well suited for
cal cul at i ng the mean offset. This is simply the area under
the di spl acement curve over one response cycle divided by
the wave period.

A Newton-Raphson technique
Several aut hor s 3 have shown t hat linearized met hods
will give good cor r espondence with results obt ai ned from
model tests or full scale measurement s. The writers aim
was t herefore t o devel op a cheap and simple met hod for
solving the non-l i near equat i ons of mot i on assuming first,
t hat the mot i on is near linear, and secondl y t hat the
linearized sol ut i on is known.
A feat ure of this met hod is t hat the response is found for
a specified posi t i on in a wave, whilst time i nt egrat i on
met hods will find the sol ut i on over the compl et e wave
cycle. This met hod is mai nl y i nt ended t o be used with a
non-l i near design wave st rengt h analysis and the response
is t herefore onl y requi red for a single position.
It was shown 8 t hat the linearized sol ut i on to the
velocities cor r esponded well with t hose calculated by t he
Newmar k met hod, whilst the linearized accel erat i ons and
di spl acement s gave less good agreements. This is, of
course, due t o the fact t hat the degree of non-l i neari t y
increases with the displacement. For a near linear system,
the di spl acement s are small at large velocities which
explains why the linearized sol ut i on is mor e accurat e for
velocities t han for di spl acement s and accelerations.
Therefore, anot her assumpt i on made in this met hod is
t hat the linearized velocity can be ret ai ned so t hat the
non-l i near equat i ons of mot i on need onl y be solved with
respect to the di spl acement s and accelerations.
Before the ki nemat i c rel at i onshi p bet ween displace-
ment s and accelerations is discussed, it is necessary t o split
t he di spl acement s i nt o a mean and a dynami c part:
~ = ~ d y n + IPmean ( 6 )
The mean response qJ . . . . . which must be calculated for all
degrees of freedom, is due to wind, current and wave drift
forces. An approxi mat e, but simple met hod to calculate
I ] / . . . . was found: (a) the linearized sol ut i on t o t he displace-
ments, velocities and accelerations are subst i t ut ed into the
non-l i near equat i ons of mot i on for a number of positions
al ong the wave length (8 positions was found to be
adequate); (b) since the moor i ng line stiffness and t he
hydrost at i c stiffness are non-linear, the mean response
vectors . . . . for all degrees of freedom are t herefore found
by the Newt on- Raphs on met hod. The gradi ent mat ri x to
yield zero residual force is t he sum of the t angent moor i ng
line stiffness and the t angent hydrost at i c stiffnesses. The
Newt on- Raphs on met hod, which is also used t o solve the
equat i ons of mot i on, will be discussed later.
Since this met hod for solving the non-l i near equat i ons
of mot i on is restricted to near linear systems, it can be
assumed t hat the ki nemat i c rel at i onshi p for simple har-
moni c mot i on remains valid:
I//dy n ~ - I / / C O S ( 0 - - 0 ) t )
~b= _0)2q, cos (O-0)t) (7)
where qJ is the response ampl i t ude, and 0 is the phase
angle, found from the linearized analysis.
The acceleration can hence be written in t erms of the
dynami c part of the displacements:
~ - - - 0 ) 2 I//dy n ( 8 )
Choosi ng a posi t i on in a wave at which the response is
requi red and substituting for the assumed values of ~dyn
and the linearized velocity ~b~m into the non-l i near equa-
tions of mot i on yields.
A ( I P d y n ) = Y~A2rj + (M + A 1 ) 0 ) 2 ~ d y n -~-
ZB(F/+ C. - ~,i=)10 -+- Ca - ~linl +
Y C 1 ( ~ / d y n -~- IP . . . . ; ~ ] ) - -
E C 2 ( I / / d y n "1- ~ / . . . . ) + I 4 1 + Fw (9)
A ( ~ t d y n ) is the unbal anced force vect or due t o an i naccurat e
di spl acement 0y,. Starting with the linearized solution,
the vect or ~bdy n can be modified so as t o reduce the
unbal anced force A(dyn) below a specified minimum.
Once the true Cdyn is found accel erat i ons can be deri ved
from expressi on (8) above.
The iterative process for the response ~9 is of t he form:
i]tk + 1 = i~tk __ ( 1 ) ( l ~ t k ) - 1 . A ( ~ / K )
(10)
where k denot es the i t erat i on cycle number, and O(~b) is
the gradi ent mat ri x requi red t o zero the bal anced force.
The iterative sol ut i on is illustrated in Fig. 1 for one degree
of freedom.
The gradient mat ri x is a 6 6 mat ri x of part i al de-
rivatives, defined as:
(I)() = 0-A~p) = (M + A00) 2 +C~ + C 2
(11)
Applied Ocean Research, 1980, Vol. 2, No. 3 101
Molio#t rcspon.se ~#ljloatim! ,s/ruclU/-cs to re~tular lvatc.~: B. .I.
h
Fi gure 2. A k e r - T P P
where the mat ri ces M, A~, C~ and C 2 a r e the t angent
matrices, i.e. the mat ri ces are cal cul at ed for the di spl aced
posi t i on of the fl oat i ng structure.
It is possible t hat the initial st art value, t aken f r om the
linearized sol ut i on, coul d be t oo far f r om the final sol ut i on
to pr oduce a conver gent iterative process. Thi s mi ght be
the case for ext reme wave heights since the linearized
sol ut i on is based on small wave heights. The non-
linearities are, however, increasing with the wave height.
To over come this, an ot herwi se non- conver gent sol ut i on
is made conver gent by first finding the sol ut i on for a
smaller wave height and by using this sol ut i on as the st art
value for a l arger wave height. Thi s is repeat ed until the
requi red wave height is reached.
COMPARISON OF METHODS
The objectives in devel opi ng the Ne wt on- Ra phs on me-
t hod present ed in the previ ous section were t hat it woul d
be cheaper and si mpl er to use t han t he Ne wma r k met hod
and t hat it woul d give a bet t er sol ut i on t han the linearized
met hod. In this section the usefulness of this met hod will
be tested agai nst each of these mor e general l y accept ed
al t ernat i ves.
Owi ng t o the different mot i on charact eri st i cs of t aut
and slack moor ed structures, results are present ed here for
bot h types of structures. The confi gurat i on of such
st ruct ures vary consi derabl y, but the Ake r - TPP and the
Aker - H3 may be consi dered represent at i ve of many
7V(uriq a#ld .I. 141 I'cndcrc~t
cxisling st ruct ures or later generat i on pr oposed designs.
The response met hods havc been lcslcd for several
headi ngs and wave condi t i ons, but in Ihe intercsl ol
brevi t y a few sets of heave and surge curves are present ed
here. These are, however, for a typical sea condi t i on and
the results presented are represent at i ve of all the tested
condi t i ons.
Aker T P P
Fi gure 2 shows a version of the Aker Tet hered
Pr oduct i on Pl at f or m for which the analysis is discussed
below. The parallel moor i ng lines are kept t aut by
sufficient excess buoyancy to avoi d snatching. Thi s moor -
ing line confi gurat i on and the necessary pret ensi on result
in heave, roll and pitch nat ural peri ods of the or der of 2 4
sec and surge, sway and yaw nat ural peri ods of about 60
s ec.
The first anal ysi s present ed below is for a wat er dept h of
180 m, a wave height of 20 m, a head sea (parallel to
pont oons) with a wave length of 200 m which cor r esponds
to a period of 11.3 sec.
Fi gure 3 shows the surge response (hori zont al mot i on)
in t erms of the di spl acement , velocity and accel erat i on
curves for each of the linearized met hod, the Ne wma r k
met hod and the Newt on Raphson met hod.
The surge di spl acement curves for the Newt on
Raphson and the Newmar k met hods are r easonabl y close
to each other. The linearized met hod, however, onl y gives
good cor r espondence when the drift offset is added (since
this is not cal cul at ed by the linearized method). Thi s par t
of Fig. 3 al so shows t hat the drift offsets from the t wo
met hods are r easonabl y close.
The linearized and the Newmar k met hod also give
virtually identical results for the surge velocity. It was
not ed earlier t hat the linearized velocity solution is used in
the Newt on Raphson technique, due to the good agree-
ment with the non-l i near solution.
DISPL A MNT
z ~
f ._..Li _ _ . /
VELOCITY
4' " ~
f
I \ .
. ~ . , ~ ~ , ~ =
d gC L R A l /O N O ~, . , , , m . e ~ 1 / 5 0 m . N
,~""=" . 8 "~
200 m 150 m I . 4 " ~
Fi pure 3. A k e r - T P P surge response. Wave height, 20 m;
wave period, 11.3 sec; no wind and current. Wave length,
200 m, water depth, 180 m, , Newt on Raphson:
........... , Newmark: , linearized
102 Appl i ed Ocean Research, 1980, l~)/. 2, No. 3
Mo t i o n response o f f l oat i n9 st ruct ures t o regul ar waves: B. J. Nat vi 9 and J. W. Pendered
2)/5PLACEMENT
\ /
X X x ~ ~ l . 0 5
\ ~ ! I --' DRI FT" OFFSET 1
~ICCELERA T/ON
I \
/ /
" /5 E
Fi gure 4. A k e r - T P P heave response. Wave hei ght 20 m;
wave peri od, 11.3 sec; wave l engt h, 200 m; wat er dept h,
180 m; no wi nd and current . , Ne wt o n - Ra p h s o n ;
~ - - , Ne wma r k ; - - , l i neari zed
The surge accel erat i on curves for t he Ne wma r k and t he
Ne wt o n - Ra p h s o n sol ut i ons are agai n al most identical
but differ significantly f r om the sol ut i on given by the
linearized met hod. I t is i nt erest i ng t hat the Ne wt o n -
Raphs on and t he Ne wma r k met hod shoul d give good
cor r espondence for this case since the accel erat i on de-
pendent forces are the pr edomi nant forces for surge
mot i on.
The heave r esponse curves are shown in Fig. 4 and it
shoul d be not ed t hat this degree of f r eedom is highly
stiffness domi nat ed since the exci t at i on peri od is much
hi gher t han the heave nat ur al period. The heave di spl ace-
ment curves show this mot i on to be highly non- l i near and
t hus one shoul d exercise caut i on in appl yi ng the assum-
pt i on of near l i neari t y in the Ne wt o n - Ra p h s o n met hod.
Despi t e this, t he Ne wt on- Ra phs on met hod is seen t o
cor r espond well wi t h the Ne wma r k met hod. It shoul d be
not ed, however, t hat whilst the heave mot i on is highly
non- l i near t he di spl acement s are small. I t is i nt erest i ng t o
not e t he presence of the t wo downwar d humps on the
non- l i near heave curves. These humps occur at the
posi t i ons of ma xi mum hor i zont al di spl acement s and are
due to the i nvert ed pendul um t ype mot i on of the st urc-
ture. I t is also i nt erest i ng to not e t hat the non- l i near
sol ut i on gives a much hi gher upwar d di spl acement at t he
wave crest t han the linearized solution. Thi s is due t o t he
wave- s t r uct ur e i nt ersect i on effects which the linearized
met hod cannot handle.
The heave velocity curves (not shown) are widely
different. But agai n these velocities are small and hence
will have little influence on the mot i on of t he st ruct ure.
The Ne wt o n - Ra p h s o n heave accel erat i on curve is
qui t e different in magni t ude f r om t hat of t he Ne wma r k
met hod, but t he general shape is the same.
The second condi t i on present ed represent s the 100 year
s t or m condi t i on which cor r esponds t o a wave height of 30
m. The wave peri od is 16.06 sec, the cur r ent is set to 1.40
m/ sec and the one mi nut e sust ai ned wind velocity is 56
m/sec. The wat er dept h is 180 m.
For this condi t i on, the surge response curves, whi ch are
not shown, gave even bet t er cor r espondence with t he
t hree met hods t han for the 200 m wave length case.
The heave di spl acement and accel erat i on curves are
shown in Fig. 5. Thi s sol ut i on will be even closer to a static
sol ut i on t han for the previ ous condi t i on due t o the
difference bet ween the wave peri od and the nat ur al heave
period. Despi t e the high non-l i neari t i es, the heave dis-
pl acement s f r om the Ne wma r k and Ne wt on- Ra phs on
met hods are agai n al most identical whilst the linearized
met hod gives a compl et el y different result. I t is i nt erest i ng
to not i ce how the wind and the current affects the heave
di spl acement curve. The ma xi mum hor i zont al displace-
ment occurs at the 200 m wave posi t i on resul t i ng in a large
down- pul l of the structure. At one hal f wavel engt h ahead
in the wave, the hor i zont al dynami c di spl acement is
count er act ed by the mean offset and t here is hence no
downpul l for this posi t i on. It is also i nt erest i ng t o not e
t hat the onl y posi t i on where the linearized and non-l i near
sol ut i ons cor r espond is at the wave crest and wave t rough.
The dampi ng and inertia heave forces are small com-
par ed t o t he domi nat i ng stiffness force. I t is still, however,
of interest to st udy the accel erat i on curves since t hey
illustrate how the Ne wma r k met hod is convergi ng. The
Ne wma r k and the Ne wt on- Ra phs on heave accel erat i on
curves are shown at the bot t om of Fig. 5. The Ne wt o n -
Raphs on sol ut i on gives a s moot h curve whilst the
Ne wma r k sol ut i on oscillates about some mean near to t he
Ne wt on- Ra phs on sol ut i on, i.e. t here is a t i me lag before
an i naccur at e accel erat i on is det ect ed and this will make
the accel erat i on oscillate about the correct sol ut i on rat her
t han t o converge t owar ds it. By reduci ng the t i me
i ncrement s this effect will reduce in magni t ude, but such
i mpr ovement in the accel erat i on curve does not appear t o
al t er the cor r espondi ng di spl acement curve. It is also the
experi ence t hat this oscillation is mor e appar ent for a stiff
t han for a soft dynami c system.
The moor i ng line forces due to the mot i on in the 400 m
wave case are shown in Figs. 6 and 7. Fi gure 6 represent s
the anchor force in a moor i ng line on the wi ndwar d side of
the st ruct ure and bot h the linearized and Newt on
Raphs on sol ut i ons are shown. I t is not i ced t hat whilst t he
Ne wt on- Ra phs on curve is si nusoi dal in shape, its ampl i -
t ude is about twice t hat of the linearized curve. It was not
unexpect ed to find a di scr epancy bet ween linearized and
non-l i near anal ysi s for this moor i ng line. Model tests of a
si mi l ar st ruct ure showed the same ki nd of di scr epancy
bet ween the tests and linearized cal cul at i ons.
,DI,,,ePLAGEPIf.,NT
0 m tO0 en 200 en
ACCL ERA T/ON
,~OOm
- ~a . o E
I
J.O
. 5
Fi gure 5. A k e r - T P P heave response. Wave hei ght , 30 m;
wave peri od, 16.06 sec; current , 1.40 m/ sec; wave l engt h, 400
m; wat er dept h, 180 m; wind, 56 m/ sec.
Ne wt o n - Ra p h s o n ; . . . . . , Ne wmar k ; . . . . , l i neari zed
Appl i ed Ocean Research, 1980, Vol. 2, No. 3 103
Mot i on re.sponse ?t.lloatinq st ruct ures to re,qtdar wate.s. B. ,I.
STATIC+ LINE,4RIZEI.~ ,DYN,4M/CFORCE
NON- L INE,4R FORCE
$OOm
Fi gure 6.
structure.
O,O
$00 m 200 m I O0m
Anchor Jbrce o f leg on wi ndward side o f
400m 300m ZOOm lOOm
Fi gure 7. A nchor. l brce o f leg on lee side o f structure.
0.0
Fi gure 7 shows the anal ysi s carri ed out for a moor i ng
line on the lee side of the structure. Consi deri ng the highly
non-l i near heave response for this condi t i on, it is surpri s-
ing to find such a good cor r el at i on bet ween linearized and
non-l i near moor i ng forces. Yet, good cor r el at i on was
agai n in conf i r mat i on of model tests which also gave good
correl at i on with linearized cal cul at i ons for this moor i ng
line.
Aker H3
The Aker H3 (see Fig. 8) is a semi -submersi bl e oil drilling
rig with convent i onal moori ng. 28 rigs of this design are
built. The heave nat ur al peri od for this design is about 22
sec, the roll and pitch of the order of 35 sec and the nat ur al
peri ods for the r emai ni ng degrees of freedom are higher
t han 60 sec.
The surge linearized response curves for the Aker - H3
were even closer to the non-l i near curves t han for the
Ake r - TPP and t hey are t herefore not shown here.
Heave response in the case of semi -submersi bl e rigs has
pr oved to be difficult to anal yse using the Newmar k
met hod. Consi der abl e effort has been made to opt i mi ze
the values of the coefficient [] and the st ep length, and the
cal cul at i ons have been carri ed out for up to 6 cycles
N~lll'i~,t t l l l d d. [~i P e ml e r e d
st art i ng from bot h the linearized and the Newt on
Raphson solution. But despite this. a truly repeat i ng
sol ut i on to the heave mot i on response has not been
obt ai ned. The general trend is t hat the sol ut i on al most
repeat s itself every two cycles and the results from t wo
consecut i ve cycles woul d lie on either side of the Newt on
Raphson solution. Thi s is illustrated in Fig. 9 where thc
Newmar k heave accel erat i on curves are alrnost repeat i ng
whilst the heave di spl acement curves are much furt her
apart . Similar difficulties were experienced with the pitch
and roll curves, but to a lesser extent t han for the heave
c ur ve s .
In the writers opi ni on, these difficulties are due to the
implicit f or mul at i on of the response par amet er s in the
Fi gure 8. Aker H3
DISPL ACEMENT
f
ACCELERATION
$200m0
m IO0 m 5 0 ~ f
I .O
Fi gure 9. Ak e r - H3 heave response. - - .... , Ne wt on
Raphson," " . , Ne wmar k
104 Appl i ed Ocean Research, 1980, ~l . 2, No. 3
Mot i on response of floatin9 structures to reyular waves: B. J. Natvi9 and J. W. Pendered
j f
I f
0 m 50 m /O0)'m~ . . / 50 m ~ / " 200 m
200 rn 250m JOO'~h~ 350m 1 1 0 0 m
o o ,,, 5o o 7, ,
" ~ ~ ..... . . . . . . . . . . i j "
Figure 10. Surye response. , Newmark (start-
in9 f rom zero); . . . . , Newmark (convergent)
Newmar k met hod whereby an i naccurat e ext r apol at i on
of velocities and displacements at one time i nst ant will be
correct ed by means of a deri ved accel erat i on at the next
time instant. If the stiffness forces are small, which is
usually the case for a semisubmersible, a consi derabl e
er r or in the displacement curve can build up before this is
det ect ed by the acceleration. For a floating structure, this
f or mul at i on is made even more implicit by the fact t hat
the under wat er geomet r y and hence the exciting forces are
dependent on the combi ned heave, roll and pitch displace-
ments. (Not e t hat the hori zont al mot i on (surge) does give
a repeat i ng solution.) Runs in which the heave and pitch
stiffnesses were increased artificially showed this oscil-
l at or y effect to a much lesser degree.
CONCLUSI ONS
A linearized and t wo non-l i near met hods for finding
mot i on response for t aut or slack moor ed floating
st ruct ures have been presented. The results for the t hree
met hods have been compar ed and are present ed for
typical sea condi t i ons and for a t aut and a slack moor ed
structure. Based upon the experi ence of appl yi ng these
met hods t o dat e the following concl usi ons are drawn.
(1) In general, linearized met hods give good results
part i cul arl y when pr oper account is t aken of the mean
offset. The velocity was found t o be the most accurat e of
the t hree response paramet ers. The linearized met hod
does not give good results for heave mot i on for t aut
moori ng, but in this case the di spl acement s are small
anyway.
The main advant ages of a linearized analysis is t hat it is
cheap on comput er time, the compl et e mot i on for a wave
cycle is described by an ampl i t ude and a phase angle and it
is readily i ncor por at ed within spectral methods.
(2) The time i nt egrat i on met hod (the Newmar k me-
t hod) was i nt ended to be the st andard against which the
ot her t wo met hods coul d be compared. This met hod
worked well for most cases except for heave, roll and pitch
for the slack moor ed system. Whilst the Newmar k
met hod is a general met hod, it was found t o be expensive
on comput er time even when st art ed from an appr oxi mat e
solution. This met hod will find the response for a whole
wave cycle which is an advant age if the whole mot i on is
required, but is a positive di sadvant age if results are onl y
requi red at discrete time instants.
(3) It may be not ed t hat the Newmar k met hod was
initiated from the Newt on- Raphs on solution which again
was initiated from the linearized solution. This is an
obvi ous and simple pr ocedur e when all the t hree met hods
are included in the same comput er program. As a result, in
most cases, onl y a few wave cycles were requi red before a
repeat i ng solution was obtained. In or der to check the
efficiency of this procedure, this met hod was also st art ed
by setting the accelerations, the velocities and the dis-
placements to zero. When this process was t ermi nat ed
after 300 time steps (which cor r esponded to 3 wave cycles)
the sol ut i on was still far from its stable solution. This is
shown in Fig. 10.
(4) A new appr oxi mat e met hod has been present ed
here in which the response can be found directly at a given
position in a wave. This met hod uses the linearized result
for displacements and accelerations as t he st art value for a
Newt on- Raphs on type iteration, whilst the linearized
solution for velocity is ret ai ned t hr oughout . The Ne wt on-
Raphson appr oach was found to be cheap on comput er
time, to have good convergence propert i es (it would find a
sol ut i on to heave mot i on of slack moor ed systems) and in
general the results woul d compar e well with t hose of the
Newmar k met hod. This met hod in its present form
cannot , however, be used at resonance.
It is seen t hat all the t hree met hods for finding mot i on
response enj oy cert ai n merits whilst also suffering from
several limitations. The met hods share several common
links and features in such mat t ers as st art values and by
the fact t hat the same wave force and st ruct ure mot i on
force subrout i nes can be used for bot h the Newmar k and
Newt on- Raphs on methods. It is t hus nat ural to include
all met hods in the same comput er pr ogr am as was done in
the pr ogr am SEARESPONS. This makes it possible t o
choose the met hod which is most appr opr i at e to the t ype
of structure, the type of analysis and the degree of
accuracy required.
R E F E R E N C E S
1 Hogben, N. Fluid Loading on Offshore Structures, a St at e of the Art
Appraisal: Wave Loads, Royal Institution of Naval Architects,
London, 1974
2 Pedersen, et al. Calculation of long term values for motions and
structural response of mobile drilling rigs, Offshore Technol. Conf.
1973, OTC 1881
Applied Ocean Research, 1980, Vol. 2, No. 3 105
Mot i ml response of j l oat i n. q st ruct ures to re,qular wat~es: B. J. Nat l ' i g amt J. 1 4 : Pendered
3 Hooft , J. P. Hydrodymm~ic Aspects ol Semi-Suhmersihle Pla(lbrm.s,
Wageni ngen, 1972
4 Newmar k, N. M. A me t hod for c omput a t i on for s t r uct ur al dynami cs ,
Proc. ASCE Mech. Emt. Div. Jul y 1959
5 Burke, B. J. and Ti ghe, J. T. A t i me series model for dyna mi c
behavi our of offshore st r uct ur es, O[f~hore Technol. Cot~l11971, OTC
1403
6 Bergan, P. G. and Soreide, T. A compar at i ve s t udy of different
numer i cal sol ut i on t echni ques as appl i ed to a non- l i near st r uct ur al
pr obl em, Computer Methods Appl. Mech. Eng. 1973
Ti l l erson, J. R., Stricklin, J. A. and Haisler, W. E. Numer i cal me t hods
for t he sol ut i on of non- l i near pr obl ems in st r uct ur al anal ysi s, ASMF,
Nov. 1973
Nat vi g, B. J. and Pender ed, J. W. Non- l i near mot i on r esponse of
fl oat i ng s t r uct ur es to wave exci t at i on, (?[].~;hore Technol. ('ot{fi 1977.
OTC 2796
1 0 6 Appl i ed Ocean Research, 1 9 8 0 , Vol. 2, No. 3

You might also like