Proceedings of the Eleventh (2001) hlternational Offshore and Polar Engineering Conference Stavanger, Norway, June 17-22, 2001 Copyright © 2001 by The International Society of Offshore and Polar Engineers ISBN 1-880653-51-6 (SeO; ISBN 1-880653-52-4 (Vol. I); ISSN 1098-6189 (SeO Effect of Ramp Duration on the Dynamic Response of Spars Iftekhar Anam and Jose M. Roesset Texas A&M University College Station, TX, USA ABSTRACT conditions to get the total solution. This is often a more efficient method of solution, particularly for linear systems. However, even the steady state f-d approach is not an exact representation of many real prob|ems. Among other limitations, the f-d method can only provide solutions to '!i_n.earized' problems or reproduce nolmineari_fies_ (e.g., those involving forces) only up to a certain order using a perturbation approach. Therefore, the t-d approach is still widely used in solving dynamic problems and is often preferred because of the complexities it can handle. A common approach adopted by several researchers, studying the response of flexible offshore s m a ~ e s is to ~ e a ramp fi~ction (wbAch increases gradually from 0 to 1) at the initial stage of the t-d simulation to avoid initial transient effects that may give unreasonably large responses. Ramp functions are also used in wave model basins testing offshore structures to simulate the real sea situations and to get steadT state r~°ponses "~i~,hin~e limited time of the experiment. Larger ramp durations are generally believed to W e steadier structural responses. But many studies on flexible offshore stmct~es like spars (with surge natural periods of around 300 seconds) have been conducted to simulate iaboratolT conditions, taking very small ramp durations; e.g., only 50 seconds proto~pe scale in Ran etal. 1995, Mekha et al. 1995, Cao and Zhang 1996, In ._someo~_es._~.the d y n ~ c re._spo~e within the ramp duration was also considered as part of the 'recorded' results, which is not the correct numerical approach, nor even representative of the way the data are recorded in the tests. Such anomalies show a lack of clarity about the possible effect of ramp duration on the results of t-d sim~ations. Mekha and Roesset (1998) showed the effect of ramp duration on the response of spars with some numerical results, concluding that a 'sufficiently large' ramp was necessary to obtain steady state response. Tl-fis paper tries to explain tl~ese res~dts a~alytically m~d support the main conclusions numerically. The pro'pose of this work is to develop a better understanding of the effect of the ramp ftmction in the numerical analyses using the t-d approach. The analytical work in the first part of the paper is followed by some numerical results for a monochromatic wave excitation acting on a spar platform. By considering this very simple excitation and only a linear response one can show clearly the transient effects with the frequency spectrum of the response. To e..... ':~"m,,~e ~+ the free -dbrafion terms and to simulate actual field conditions, it is common practice in wave model basins to multiply the excitation by a ramp function and to disregard the portion of the response corresponding to the application of the ramp. The effect of ramp duration on the dynamic response of a spar platform subjected to irregular waves was shown numerically in a previous paper° This topic is inves_ti_gated again in this paper both analytically and numerically, using a simp|e monochromatic excitation and a finear system. It is shown that free vibration will result from discontinuities in the forced response both at the time of application of the rarap and wilen it is removed_. Both discontinuities decrease in Lmportance as the duration of the ramp increases. To reach the steady state condition one can thus increase the duration of the ramp and/or neglect the initial part of the response unti! a certain time after the end of the ramp. For normal, low values of damping the ramp duration is the predominant factor. K E Y W O R D S : Spar, ramp, free vibration, steady state, time domain, frequency domain. INTRODUCTION In the numerical analysis of dynamic systems, the more direct approach is to fommiate the problem ha the time domain and to solve the governing differential equations numerically using a time-step inte~ation scheme. The time domain method (called t-d here) accounts for the initial conditions of the system (e.g., displacements and velocities) and represents properly in general the nonlinearities of the system. However this method can result in unrealistic transient effects, which may not represent the real-life situations. This is particularly true for flexible offshore structures (i.e., structures having large natural periods of vibration) subjected to ocean wave loads that develop ~adually. The large transient effects in t-d solutions due to a sudden application of fully developed sea-conditions can therefore be artificial. To solve this problem, two approaches can be taken. The first is to use a combined time-frequency domain approach that converts the steady state frequency domain-(f-d) results to t-d and applies appropriate init/al 427 01-JSC-308 I. Anam EFFECT OF RAMP FUNCTION .'. The steady response just after the ramp is applied, is x(0+) = 2: Pi {cos(qgi0)/ki0 - 0.5 cos(tpi_)/ki_- 0.5 cos(qoi+)/ki+}/2 while just before, i.e., initially, it was x(0-) = 0 (5) Similarly, just before the ramp is withdrawn, the forced response is, x(Tr'/Wr) = ~ Pi {cos(oiTt:/~r+q)i0)/ki0+ 0.5 COS(r.Oi~//gr+~i.)/ kit- 0.5 cos(oin/~ +9i+)/k~+}/2 while just after withdrawal, i.e., it is x(Tg+/~i~r)= ~2 Pi cos(f-°i~/UIr+(Di0)/ki0 (6) Equations (5) and (6) show that the forced responses go through sudden discontinuities both at the beginning and at the end of the ramp durafior,, as was logically concluded earlier. To see the effect of ramp duration on these d_iscx~ntimfifies,one may consider the case when the ramp duration becomes very large; i.e., Tr --~ oe ~ ~r--~ 0. • Clearly, both ki. and ki+--~ ki0;also q0i. and q~i+-+ q0i0,in this case. " x(0 +) -+ X Pi {cos(~0)/ki0 - 0.5 cos(qoi0)/ki0- 0.5 cos@pi0)/ki0}/2 = 0, which is x(0); and x(xT~0 -~ 2 Pi {cos(~iTrJ~Jr+CPi0)/kio+ 0.5 COS(OiTr,/'UJr+q~iO)f~O+0.5 COS(OiT~//~Jr-q-(1)i0)/rki0}/2 = 2 Pi cos(°)iTg/~Jr+~i0)/ki0, which is x(x+/~0 (7) Equation (7) gives an improved insight on how long the ramp should be in order for it to become 'sufficiently' iarge. This ~411always remain a qu~,ditafive statement, bm tim essential fact is that the transient effects will be diminished iXthe ~scontku,,fifi~ of the motions on eider side of ~ e ramp duration can be reduced. So, the ramp duration has to be 'sufficiently' large compared to (half of) the longest period which contributes 'significantly' to the displacement. For flexible offshore structures like spar platforms, the smallest ffequen.cies (i.e., longest periods) coombuting si~ifica_ntly to the structural motions are the surge natural frequencies. Therefore, 'sufficiently' large ramp duration for surge has to be 'sufficiently' large compared to halt" of the surge natural period, it will not need to be that large for the pitch resportse of a typical spa,-. On the other hand, the longest 'significantly-excited' period of motion for rind structures is within the range of the wave periods themselves, which is much larger than the structures' natural periods of vibration. Theretbre, the 'adequate' ramp durations tbr rind structures should be much smaller. Tiffs is part of the reason why the work done ruth ramp functions h ~ focused on flexAbte smacmres like spars and Tension Leg Platforms (TLPs), and not on rind structures like gravity-based structures or steel-jacket plaffom~s. The ramp function is applied to enrage a continuous ~ d gradual transition of wave loads from an initial zero value to a fully developed stage. It is applied as a multiplying factor to the wave-load and should therefore increase from zero to one. A ramp function often used in the literatm-e is framp = (1" cos(nt/T~))/2, for 0 < t < Tr (1) where Tr is the ramp duration. It is a cosine function which does increase from zero to one, with the additional fact that it has zero slopes both at the beginning and at th,, end of the ~,,~,,~,~,l If it res~ted in continui~, of steady state response, there would be no transient effects and the results from the t-d solution would represent the steady state. However, the continuity of the force does not result in similar continuity of the steady state response due to it. The continuity of the total response (steady + ~ausient) will be maintained however, wlfich requires a discontinui~ ~ in the transient response as well. The application of the ramp results m transient effects in two ways. First, it will still produce a transient response when applied initially on a structure at rest; then it will create an additional transient effect when it is withdrawn. The decay of these two transient effects will depend on the damping of the system and the time they are given to die out. Clearly, the Fast transient effect will decay during the entire period of appfication of the ramp; the~,e~or., a longer rmnp ,.m help its decay. The second transient effect will only beg~n after the ramp is withdrawn; therefore its decay will depend on the time given after the withdrawal of the ramp. Thus, qualitatively it will take a 'sufficiently' long ramp as well as a 'waiting time' after its withdrawal for the transient effects to die out stffficieml_y and for the response to become more ' steady'. These concepts are easy to visualize from basic concepts of structural dynamics. However it remains to be seen if the ramp dm-ation has any effect on the discontinuities introduced in the forced responses. As is shown later, a large ramp does decrease them too. Thus, the effect of ramp in decreasing transient response is twofold. This is now shown analytically. For simplicity, a single-degree-of-freedom (SDOF) system is investigated here. The main conclusions can still be presented in general terms. The force from i~egular waves is the stun of its different Fourier components acting at different frequencies (~i), each having different magnitudes (Pc) and phases (q~c), The ramp function within the ramp duration is = (I- cos(x'ffTr))/2 = 0.5 (1- COS~rt), where Igr = ~/Wr (2) Here, w,. = x/Tr = 0.5 (2x/T0 = Half the natural frequency corresponding to the ramp-period. .'. During the ramp duration, the dynamic force F(t) = (1- cos ~rt)/2 22 Pi cos(Oit+g)i) = {E Pi cos(°it+(pi)- 0.5 ~2 Pi COS(Oit-~rt+(Pi)- 0.5 ~] Pi cos(toit+Wrt+9~}/2 (3) Thus, the force is reduced to a summation of three sets of cosines. The forced pm-t of flae dynamic response (i.e., particttlm" solution of equations of motion_) can be easily found to be, x(t) = Z Pi {cos(c0it+~i0)/ki0 - 0.5 cos(oit-_m_~t+gi3/ki.- 0.5 cos(oit+~t+(pi+)/ki+}/2 (4) where ki0 = ~/{(k-oi~m)2~ o i c);}, the dynamic stiffness at frequency oi ki-- Dynamic stiffness at frequency ( o i - ~ ) , and ki+ = Dynamic stiffness at frequency (Oi+'rJr). Also, q0i0= tan-1 {oi c/(k-oi2m) }, the phase angle at frequency oi The phase angles q)i.and (Pi+are defined likewise. N~TvIERI CAL RESULTS The main conclusion from the theoretical work of the previous section was that a longer ramp helps to suppress transient effects, and thereby should provide responses closer to the steady state (forced vibration) solutions. Another aspect was the "waiting period' after the ramp is withckawn; a bigger waiting period would help also to minimize transient effects. These conclusions are now studied numerically for a spar platform with a draft of 198 meters, a diameter of 40.5 meters and long periods of vibration [332 seconds in surge and 66 seconds in pitch]. This particular spar operates at a water depth of 3i8.5 meters and has a mass = 259xi06 kg, spring stiffness = 188 kN/m and radius of gyration = 62.3 meters. A detailed work has akeady been conducted showing the possible effect of ramp duration on the dynamic response of spar platforms (Mekha and Roesset 1998) where the above conclusions were shown to be true even for random waves. This section deals with the shnpler case of a monochromatic wave (called 1'v~,lI here) with mnplimde 3.0 m~ers and period !4 seconds. This way, it is possible to show the ramp's effect purely with the frequency specmun for a linear 428 01-JSC-308 I. Anam 2 5 formulation. The response should converge to a single peak at the wave-frequency and be zero everywhere else. Another point of interest is the difference between the 'adequate' ramp duration for surge and pitch motions, with their very different natural periods. According to the analytical conclusions of the previous sections, the adequate ramp durations for the two motions (to suppress transient response) should be different. The results of the t-d analyses with ramps of different duration are compared to those of a steady state f-d solution. The frequency-spec~a obtained from the t-d and f-d simulations are shown in Figs. 1-12. For the four ramps chosen here, the results for surge mad pitch are shown alternately between Figs. 1-10. Finally, the f-d results are shown in Figs. 11 and 12. First, a ramp duration of 20 seconds was used. This is very small compared to the surge natural period (332 s~onds), and also with respect to the pitch natural period (66 seconds). It results in transient response amplitudes of 0.341 m for surge (Fig. 1) and 0.265 ° for pitch (Fig. 2) which are about half of the amplitudes at the excitation frequency. Obviously, these components would not be there if the response were steady. However, the responses at the wave-frequency (0.696 m, 0.60! °) agree very well with the f-d solution (0.712 m, 0.615 °) [Figs. 11 and 12]. The ramp was increased next to 50 seconds (a ramp value sometimes used in the literature). This decreases the transient effects, mad the corresponding amplitudes are now 0.233 m (Fig. 3) and 0.017 ° (Fig. 4). This ramp, being small compared to tile surge period, cannot suppress the transient surge response adequately. But now the transient pitch response is diminished to almost zero; therefore, it is adequate for pitch. In all these cases, the 'waiting' period after the ramp is taken as zero. To demonstrate the effect of the waiting period, the next simulation was performed with a long waiting period of 950 seconds. The transient respotkse amplitudes decrease slightly as a result, this time to 0.186 m (Fig. 5) and 0.011 ° (Fig. 6) respectively. This decrease is caused by the decay of the responses due to damping. The ramp duration was then increased to 400 sec, with a waiting period of 600 sec. This way the same 'initial' (ramp + waiting) period of 1O00 sec (= 400 + 600) is maintained. But the transient responses are much decreased this time. The transient surge amplitude is now 0.082 m (Fig. 7), while there is no transient pitch amplitude (in degrees) even up to three decimal points (Fig. 8). Finally, the ramp duration was increased to 900 sec, with a waiting period of 100 sec again maintaining the same 'initial' period of 1000 sec (= 900 + 100). The transient responses are almost completely removed this time. The transient surge amplitude is only 0.015 m (Fig. 9), while there is obviously no observable transient pitch amplitude (Fig. 10). This not only shows the steadying effect of ramp duration on the dynamic responses, but also the trade-off between the rennp duration and the waiting period. In this case, the ramp duration turns out to be more important. However, increased damping would favor the waiting period. No matter how big a transient response is induced in the ramp period, the waiting period should be able to suppress it more effectively in that case. However, the effect of damping ratio on the required waiting period is beyond the scope of this work and has not studied here. Although here the forces are calculated using a "linearized' form of Morison's equation, the conclusions are valid for any dynamic problem. Similar results could be obtained here by any combination of loads equal to the resultant forces calculated in this case, and with representative damping ratios. CONCLUSIONS The main conclusions from this work can be stmunarized as: 1. Application of a ramp function results in discontinuities in the steady state response of a structure subjected to dynamic loads. These discontinuities occur both at the beginning and the end of the ramp duration. 2. TILe discontinuities cause transient responses, both at the time the ramp is applied and when it is withdrawn. 3. A 'sufficiently' long ramp helps to decay the initial transient effect. A 'stffficiently' long 'waiting time' helps to decay the second transient effect. Therefore, in order to get the steady state response, the t-d 27 simulation should wait for awhiie to allow this second transient effect to die out. 4. A large ramp also helps to reduce the discontinuities themselves. That is, the longer the ramp duration,, the smaller the transient effect 'induced" by the ramp. 5. For flexible structures with natural frequencies much smaller than the wave frequencies, different modes of vibration need different ramp durations to reach steady state. It depends on their respective natural periods of vibration. 6. A large 'waiting period' can offset the large transient effect caused by a small ramp. The trade-off between ramp duration and waiting period favored a longer ramp in this case. But in general, this may depend on the amount of damping. It must be mentioned here that this work mainly concentrates on the effect of ramp duration on the free vibration of flexible structures. Some other effects of the ramp duration; e.g., increasing numerical stability of the t-d algorit~-n, are not studied here. ACKNOWLEDGEMENTS The financial support provided by the Offshore Teclmology Research Center (OTRC) through the National Science Foundation Grant # CDR 8721512 is gratefully recognized. REFERENCES Cao, P and Zhang, J (1996). "Slow Motion Responses of Compliant Offshore Structures." Proc. 6 a' Int. Symp. Offshore and Polar Engrg., ISOPE-96, Los Angeles, CA, Vol 1, pp 296-303. Me!d,.a, BB, Johnson, CP and Roesset, ~A (1995). "Nonlinear Response of a Spar in Deep Water: Different Hydrodynamic and Structural Models." Proc. 5 th Int. Syrup. Offshore and Polar Engrg., ISOPE-95, The Hague, The Netherlands, Vol 3, pp 462-469. Mekha, BB and Roesset, JM (1998). "Effect of Ramp Duration on the Response of Spar Platforms to irregular Waves." Proc. 8 ~hint. Syrup. Offshore and Polar Engrg., ISOPE-98, Montreal, Canada, Vol 1, pp 270-277. Ran, Z, Kim, MH, Niedzwecki, JM and Johnson, RP (1995). "Responses of a Spar Platform in Random Waves and Currents (Experiment vs. Theory)." Proc. 5 th Int. Syrup. Offshore and Polar Engrg., ISOPE-95, The Hague, The Netherlands, Vo! 3, pp 363-371. 429 01-JSC-308 I. Anarn 0.8 0.8 0.6 0.6 0.4 a= 0.4 0.2 0.2 g - 0 +.a 0 , . i 0.1 , ,~ 0,2 .,-, 0.3 1 t 0.4 0.5 , 0.4 0 0.5 O.l 0.2 0.3 freq (rad/s) freq (rad/s) Fig. 1 MN1 Surge: Ramp = 20 sec Fig 2 MN1 Pitch: Ramp = 20 sec Wait = 0 Wait = 0 0.8 0.8 0.6 0.6 ,,._., 0.4 0.4 O 0.2 0.2 , 0.1 0.2 0.3 0.4 . I 0 0.5 0.1 0.2 0.3 0.4 freq (rad/s) freq (rad/s) Fig. 3 MN1 Surge: Ramp = 50 sec Fig. 4 M N I Pitch: Ramp = 50 sec Wait = 0 Wait = 0 0.8 0.8 0.6 0.5 - 0.6 ,,,--,. g 0.4 ,= 0.4 . ,...~ 0.2 0 0.2 ~ 0.1 , , ~ , 0.2 , , i 0.3 , ,I , 0.4 ] , 0,5 0 freq (rad/s) 0.2 0.3 0.4 freq (rad/s) Fig. 5 MN1 Surge: Ramp = 50 sec, Wait = 950 sec Fig 6 M N ] P i t c h Ramp = 50 sec, Wait = 950 sec 430 01-JSC-308 0,1 I. Anam 0.5 0.8 0.8 - 0.6 0.6 ~D 0.4 0.4 O +.o 0.2 0.2 0 0.1 0.2 0.3 0.5 0.4 0.1 0.2 freq (rad/sec) 0.3 0.5 0.4 freq (rad/sec) Fig. 7 MNI Surge: Ramp = 400 sec, Fig. 8 MN1 P i t c h Ramp = 400 sec, Wait = 600 sec Wait = 600 sec 0.8 0.8 0.6 0.6 0.4 0.4 O 0.2 0.2 -.~ 0 i ,i . . . . 0.1 0.2 i , 0~ : 0.3 0 0.5 0.4 ,I 0. t , i 0.2 . . . . T ,~ , 0.3 0.4 0.5 freq (rad/s) freq (rad/s) Fig. 10 MN1 Pitch: Ramp = 900 sec, Fig. 9 MN1 Surge: Ramp = 900 sec, Wait = 100 sec Wait = 100 sec 0.8 0.8 0.6 0.6 0.4 "-" 0.4 0.2 0.2 ~D . . . 0 , 0.1 .......... , 0.2 ~ 0.3 . i . . . . . 0.4 r 0 0.5 0.2 0.4 freq (rad/s) freq (rad/s) Fig. 12 MN1 Pitch: Frequency Domain Fig. 11 MN1 Surge: Frequency Domain 431 01-JSC-308 I. A n a m 5 5